United States Office of Underground Storage
Environmental Protection Tanks
&EPA
Standard Test Procedures For
Evaluating Release Detection
Methods: Automatic Tank
Gauging Systems
May 2019
-------
Acknowledgments
The U.S. Environmental Protection Agency's Office of Underground Storage Tanks contracted
with Battelle under Contract No. EP-C-10-001 to revise EPA's 1990 Standard Test Procedures
for Evaluating Release Detection Methods. Individual members of the National Work Group on
Leak Detection Evaluations, as well as Ken Wilcox and Associates, reviewed this document and
provided technical assistance. A stakeholder committee, comprised of approximately 50
representatives from release detection method manufacturers and various industry associations,
also commented on this document.
-------
Contents
Acknowledgments ii
List Of Acronyms And Abbreviations v
Section 1: Introduction 1
1.1 Background 1
1.2 Objectives And Applications 2
1.3 Evaluation Approach Summary 2
1.4 Organization Of This Document 3
Section 2: Safety 4
Section 3: Apparatus And Materials 5
3.1 ATGS Release Detection Method 5
3.2 Tanks 5
3.3 Product 6
3.4 Leak Simulation Equipment 7
3.5 Water Ingress Sensor Test Equipment 8
3.6 Miscellaneous Equipment 8
Section 4: Test Procedures 9
4.1 Environmental Data Records 10
4.2 ATGS Evaluation Test Procedures For Release Detection Mode 11
4.2.1 Induced Leak Rates, Temperature Differentials, and Product
Volume 11
4.2.2 Testing Schedule 14
4.3 Testing Problems And Solutions 17
4.4 ATGS Evaluation Test Procedures For Water Ingress Detection 17
4.5 ATGS Alternative Evaluation Procedures For Release Detection Mode 20
Section 5: Calculations 23
5.1 ATGS Release Detection Mode Performance Parameters 23
5.1.1 Basic Statistics 23
5.1.2 False Alarm Rate, P(fa) 25
5.1.3 Probability Of Detecting A Leak Rate Of 0.20 gal/hr, P(d) 27
5.1.4 Other Reported Calculations 27
5.2 ATGS Water Ingress Detection Mode Performance Parameters 28
5.2.1 Minimum Detectable Water Level 28
5.2.2 Minimum Water Level Change 29
5.2.3 Water Ingress Detection With Liquid Level Measurements
(Optional) 31
5.2.4 Water Ingress Detection After Water Ingress Alarm, Mixing, Then
No Alarm (Optional) 31
5.2.5 Time To Detect A 0.20-Gal/hr Water Incursion (Optional) 31
iii
-------
5.3 Supplemental Calculations And Data Analyses (Optional) 32
5.3.1 Minimum Threshold 33
5.3.2 Minimum Detectable Leak Rate 33
5.3.3 Test For Adequacy Of Stabilization Period 34
5.3.4 Test For Adequate Temperature Compensation 36
5.3.5 Test For Effect Of In-Tank Product Volume 39
5.3.6 Option To Test The Evaluator-Determined Minimum Fill Height
Level 42
5.4 Outline Of Calculations For Alternative Approach 44
5.4.1 Calculation of P(fa) and P(d) 45
5.4.2 Limitations On The Results 46
Section 6: Interpretation 47
6.1 Release Test Function Evaluation 47
6.2 Water Level Detection Function 47
6.3 Minimum Water Level Change Measurement 48
6.4 Limitations 48
Section 7: Reporting of Results 51
Appendices
Appendix A. Definitions And Notational Conventions A-2
Appendix B. Reporting Forms B-2
Figure
Figure 1. Student's t-Distribution Function 26
Tables
Table 1. Analytical Methods For Bio-Component Determination 7
Table 2. Product Volume Leak Rate And Temperature Differential Test Design 12
Table 3. Notation Summary 24
Table 4. Notation Summary For Water Sensor Readings At The jth Replicate 30
Table 5. Organization Of Data To Test For Temperature Effects 37
Table 6. Organization Of Data To Test For Product Volume Effect 40
iv
-------
List Of Acronyms And Abbreviations
ANOVA analysis of variance
ASTM International American Society for Testing and Materials (ASTM) International
ATGS
automatic tank gauging system
B
bias
°C
degree Celsius
CFR
Code of Federal Regulations
CITLDS
continuous in tank leak detection system
df
degrees of freedom
EPA
U.S. Environmental Protection Agency
°F
degree Fahrenheit
gal/hr
gallon per hour
K
tolerance coefficient
1/hr
liter per hour
LR
leak rate
MLC
minimum water level change
MSE
mean squared error
P(d)
probability of detecting a leak
P(fa)
probability of false alarm
SD
standard deviation
SE
standard error
SIR
statistical inventory reconciliation method
Th
threshold
-------
tolerance limit
underground storage tank
-------
Section 1: Introduction
1.1 Background
The federal underground storage tanks (UST) regulation in 40 Code of Federal Regulations
(CFR) Part 280 specifies performance standards for release detection methods. UST owners and
operators must demonstrate that the release detection method they use meets the U.S.
Environmental Protection Agency's (EPA) regulatory performance standards. This document
provides test procedures for evaluating the release detection category automatic tank gauging
systems (ATGS).
This automatic tank gauging systems document is one of four EPA standard test procedures for
release detection methods. The test procedures present performance testing approaches to
evaluate various release detection method categories against the federal UST regulation in 40
CFR Part 280, Subpart D. To provide context for the four test procedure documents, EPA
developed General Guidance Usins EPA's Standard Test Procedures For Evaluating Release
Detection Methods. The general guidance provides an overview of the federal UST regulations,
methods, and testing that may demonstrate that release detection methods are compliant with the
regulatory performance standards. The general guidance is integral; it must be used with the test
procedures.
The ATGS method must be capable of detecting a leak of 0.20 gallon per hour (gal/hr) with a
probability of detection (P(d)) of (at least) 95 percent while operating at a false alarm rate of 5
percent or less. The federal regulation requires that for ATGS, the automatic product level
monitor test must be able to detect a 0.20 gal/hr leak from any portion of the tank that routinely
contains product and that its automatic inventory control function meets federal requirements for
inventory control.
EPA's 1990 test procedures for ATGS intentionally did not address each aspect of the inventory
control function of ATGS. In addition to the leak test function of the ATGSs, the 1990 test
procedures only evaluated the water sensing function that is used for water measurement within
the inventory control function of the ATGS. The leak test function and water sensing function
are historically considered the primary leak detection modes of ATGS.
Through EPA's 1988 UST regulations Technical Compendium EPA, in 1993 allowed ATGS
which met the performance standards for leak rate, probability of detection P(d), and probability
of false alarm (P(fa) to be used without inventory control as another method under 40 CFR §
280.43(h). EPA updated this guidance in the 2015 federal UST regulation's Technical
Compendium to state that ATGS must meet the water measurement requirement of the inventory
control procedures. The 1990 test procedures for ATGS maintained the requirement for
evaluation of the water sensing function
These revised test procedures for ATGS continue to evaluate the water sensing function of
ATGSs by testing for minimum detectable water level and minimum detectable change in water
level.
1
-------
1.2 Objectives And Applications
The objective of the standard test procedures is to evaluate ATGS in a consistent manner. These
test procedures only evaluate the leak test function and the water sensing function of ATGS since
these functions are considered the primary modes to detect releases. UST owners and operators
are required to demonstrate that the method of release detection they use meets the EPA
performance standards of operating at (no more than) a 5 percent probability of false alarm
(P(fa)) while having a P(d) of (at least) 95 percent to detect a leak of 0.20 gal/hr. These test
procedures describe how this level of performance can be proven. In addition, by using lower
leak rates with the same test design, an ATGS may be evaluated at a more stringent 0.10 gal/hr
performance level.
The application of these test procedures evaluates methods that are installed in the tank and
monitor product volume changes on a continuous basis during the test period. The evaluation
will estimate the performance of the method's test mode and compare it with the EPA
performance standards. These procedures provide tests to determine the threshold of water
detection for the ATGS. The procedures also test the water ingress detection function to measure
changes in the water level. The test results are compared to the EPA performance standard of
0.125 inch and are evaluated over a range of a few inches vertically from the bottom of the tank.
The threshold and height resolution of the water detector are converted to gallons using the
geometry of the tank.
The evaluator should determine whether a method is different enough from the originally tested
model to warrant retesting. Some changes such as housing, cosmetic, or user interface are minor
and would not warrant retesting. Other changes to the method that suggest improved
performance or changes in the algorithm or equipment configuration should be retested.
Although safety is a consideration while conducting testing, these test procedures do not address
the issue of safety specific to detection methods and their operating procedures, merely basic
laboratory safety concerns and procedures. The vendor is responsible for conducting the testing
necessary to ensure that method equipment is safe for operation and capable of being used with
the intended product.
Ultimately, you can use the results from these procedures to prove that the method meets the
requirements of 40 CFR Part 280 and is subject to the limitations listed on the EPA's standard
evaluation form in Appendix B.
1.3 Evaluation Approach Summary
The ATGS is installed in the test tank and measures a leak rate under the no-leak condition
(0.0 gal/hr) and with three induced leak rates of 0.10, 0.20, and 0.30 gal/hr for verification at
0.20 gal/hr. For the optional evaluation at 0.10 gal/hr, the ATGS will be used to measure a leak
rate under the no-leak condition and with three induced leak rates of 0.05, 0.10, and 0.20 gal/hr.
A total number of at least 24 tests are to be performed for both the 0.20 gal/hr and 0.10 gal/hr
evaluation options.
2
-------
The tank must be 50 percent full for half of the tests. It is refilled to about 90 to 95 percent full
for the other 12 tests. A third fill level may be tested to demonstrate the lowest volume of
product in the tank where the performance of the method still meets the regulatory performance
requirements. The evaluator, using probe geometry, decides the fill level to test and thus
establishes the low-level limitation.
When filling the UST, use product at least 10 degrees Fahrenheit (°F) (5.6 degrees Celsius [°C])
warmer than that in the test tank for one third of the fillings, product at least 10°F (5.6°C) cooler
for another third of the fillings, and product at the same temperature for the third filling. The
ATGS's ability to track volume change is determined by the difference between the volume
change rate measured by the test method and the actual, induced volume change rate for each
test. These differences are then used to calculate the performance of the method. Performance
results are reported on the Results of U.S. EPA Standard Evaluation Form included in
Appendix B.
The ability of the method to measure water entering a tank is tested by placing the water sensor
in a standpipe containing a test product. Measured amounts of water are added to the tank and
the sensor's ability to detect the water either at the bottom of the tank or entrained in the fuel is
evaluated. The evaluation approach for where the water is detected at either of these two
locations is similar except with what independent measurement device is used for comparison.
Whether to monitor the bottom of the tank or the bulk fuel will depend on the miscibility of
water with the test product. These results are also reported on the standard forms in Appendix B.
1.4 Organization Of This Document
The evaluation approach is presented in the following sections of this document:
Section 2 presents a brief discussion of safety issues during testing
Section 3 discusses the apparatus and materials needed to conduct the evaluation
Section 4 provides step-by-step test procedures
Section 5 describes the data analysis
Section 6 provides the interpretation of results
Section 7 describes how the results are to be reported. Two appendices are included in
this document.
Appendix A includes definitions for some technical terms
Appendix B contains the forms for the data collection and reporting
o Standard reporting form for the evaluation results
o Standard form for describing the detection method
o Data reporting forms
o Individual test logs
3
-------
Section 2: Safety
These test procedures only address the ATGS ability to detect leaks; they do not address testing
the release detection method for safety hazards. The vendor should test and determine that the
ATGS release detection method is safe for the intended products. Each release detection method
should have a safety protocol provided by the vendor as part of its standard operating procedures.
The protocol should specify requirements for safe installation and use of the method. In
addition, all facilities hosting an evaluation of an ATGS should provide the safety policy and
procedures to evaluating personnel on site. All safety requirements must be followed to ensure
the safety of those performing the evaluation and those near the evaluation.
At a minimum, the following safety equipment should be available at the site:
Two class ABC fire extinguishers
One eyewash station (portable)
Adequate quantity of spill absorbent
Appropriate Safety signage such as No Smoking, Authorized Personnel Only, and Keep
Out.
Personnel at the UST facility should wear safety glasses when working with product and steel-
toed shoes when handling heavy pipes or covers. After the safety equipment has been placed at
the site and before any work can begin, the area should be secured with appropriate signage.
All safety procedures appropriate for the product in the tanks and test equipment should be
followed. The vendor should address key safety hazards such as fire, shock, intrinsic safety,
product compatibility, etc. according to construction standards. Before testing, the evaluator
should determine what safety procedures will be followed to ensure the test operation will be as
safe as possible.
4
-------
Section 3: Apparatus And Materials
3.1 ATGS Release Detection Method
The vendor will supply the ATGS. In general, an ATGS consists of a method for monitoring
fuel volume or level, for compensating for temperature, and for detecting and monitoring water
in the fuel. It will also typically include instrumentation for collecting and recording the data
and procedures for using the data to calculate a leak rate and interpret the result as a passed or
failed test.
The ATGS being tested should be operated by the evaluator personnel after the customary
training. These test procedures evaluate methods of release detection and the water ingress
detection separately.
Some vendors combine traditional release detection methods into hybrid methods. For example,
some methods combine the automatic data collection features of ATGS with the sophisticated
statistical data analysis used in statistical inventory reconciliation (SIR) methods. This allows
the new methods to monitor the tank continuously, using data collected continually that are
reviewed for adequacy. These methods then can operate without interfering with normal tank
operation, whereas a traditional ATGS requires a shut-down period to conduct a leak test. These
new technologies are collectively referred to as continuous in-tank leak detection systems
(CITLDS). These hybrid methods may be evaluated using alternative test procedures. The
National Work Group on Leak Detection Evaluations maintains a list of acceptable test
procedures for reviewing third-party evaluations of equipment and methods to verify established
performance standards are met.
3.2 Tanks
The standard test procedures require an UST known to be tight. A second tank or a tank truck is
required to store product for the cycles of emptying and refilling. As previously stated, the tank
must have been tested and shown to be tight. The tank should not have any history of problems.
In addition, the test procedures call for an initial trial run with the test method under stable
conditions. This test should confirm that the tank is tight; if it does not, there may be a problem
with the tank or the test method that should be resolved before proceeding with the evaluation.
The tank facility used for testing must have at least one monitoring well, to determine the
groundwater level. The presence of a groundwater level above the bottom of the tank would
affect the leak rate in a real tank, that is, the flow of product through an orifice. The flow would
be a function of the differential pressure between the inside and outside of the tank. However, in
a tight tank with leaks induced to a controlled container separate from the environment, the
groundwater level will not affect the evaluation testing. Consequently, it is not necessary to
require that testing against the evaluation test procedures be done in a tank entirely above the
groundwater level.
Testing may be conducted in any size UST. The results of the evaluation will be applicable to all
smaller tanks; therefore, the larger the test tank used in the evaluation, the broader the
5
-------
applicability of the evaluation. The results are also applicable to larger tanks with the condition
the tanks be no more than 50 percent larger in capacity than the test tank. This is a conservative
and previously approved approach since there is no standard scaling method to apply here.
Because the probe sensitivity to height is dependent on the tank geometry and volume,
establishing a rule on volume is relatively arbitrary.
The test procedures call for filling or emptying the tank several times. Therefore, a second tank
or a tank truck with the associated pumps, hoses, or pipes to transfer the product is needed to
hold reserve product.
3.3 Product
The most common products in USTs today are motor fuels, particularly non-alcohol blended
gasoline, alcohol-blended gasoline, diesel, and biodiesel fuels. These test procedures, using at
least 24 tests, are designed primarily to evaluate the methods with these widely marketed
products
The evaluator and the vendor choose the test product, but it must be capable of being used with
the release detection method. Products with similar physical and chemical characteristics may be
used and results may, in some instances, be inferred to represent typical responses. Evaluating
the method with a specific product verifies its performance with that product. Caution must be
exercised in inferring that results represent typical responses across products with similar
physical and chemical characteristics. The evaluator must justify the applicability of results to
other products. However, since alcohol-based fuels and bio-blended fuels are appreciably
dissimilar to petroleum-based fuels, the evaluation must specifically test a representative product
under reasonable conditions likely encountered in the field, such as the presence of water from
common sources like tank top sumps. Considerations such as water miscibility with the fuel
blend, especially with alcohol-blended fuels, will require testing the functionality of the water
ingress detection method used on the ATGS.
Because water is essentially immiscible in petroleum-based fuels, a very small addition of water
to an UST storing petroleum-based fuels will cause a water phase to settle in the bottom of the
tank. This makes it relatively simple to determine the presence of water in USTs storing
petroleum-based fuels. However, low alcohol-blended fuels can hold approximately 0.5 percent
of water before phase separation occurs. As fuel temperature is lowered, the amount of water
needed before phase separation occurs is also lowered. Because water alters the solubility of
alcohol in gasoline, when phase separation occurs in E-10, for example, the separated phase
consists of an ethanol - water mixture with a density greater than ethanol but less than water. If
water entering an UST does not mix into a low ethanol-blended fuel, a separated aqueous phase
will collect at the bottom of the UST. However, once the UST receives a fuel delivery mixing
the contents, the water is absorbed into the fuel until it reaches saturation. This phenomenon has
been shown to render traditional water detection floats unreliable unless the float composition
density is adjusted in comparison with the density of the separated phase. Another alternative
would be for the ATGS console to be programmed to recognize this reoccurring pattern of
detected water followed by no detectable water.
6
-------
As mentioned previously, water absorbed into the blended fuel will also increase the density and
other physical parameters of the blend, thus making proper selection of volumetric correction
factors difficult. In addition, a certain amount of water can be absorbed in alcohol without an
increase in volume and without separating at the bottom of the tank. In a large volume of stored
fuel, the amount of water absorbed into the alcohol fraction of an alcohol-blended fuel could be
appreciable and undetected. Therefore, an ATGS may be unreliable in detecting absorbed water,
because the product volume will not accurately reflect the total volume of water that has entered
a tank. An alternative means of determining water ingress may be required.
Given the variability of the proportion of bio-components in fuels, during testing the true
proportion of ethanol, for example, or biodiesel to fuel needs to be determined and reported with
the test results. Following the ASTM International standard methods presented below, or
another national voluntary consensus code, an aliquot of the fuel must be analyzed for the biofuel
content in order to characterize the fuel for listing the method. Table 1 below specifies the
methods that may be used for bio-component analysis by fuel type.
Table 1. Analytical Methods For Bio-Component Determination
Method
Designation
Method Title
Fuel Product
ASTMD7371
Determination of Biodiesel (Fatty Acid Methyl Esters)
Content in Diesel Fuel Oil Using Near Infrared
Spectroscopy
Biodiesel
ASTMD4815
Standard Test Method for Determination of MTBE,
ETBE, TAME, DIPE, tertiary-Amyl Alcohol and CI to
C4 Alcohols in Gasoline by Gas Chromatography
Alcohol blend
up to 20%
ASTMD5501
Standard Test Method for Determination of Ethanol and
Methanol Content in Fuels Containing Greater than 20%
Ethanol by Gas Chromatography
Alcohol blend
over 20%
3.4 Leak Simulation Equipment
The method of inducing the leaks must be capable of being used with the release detection
method and the product used during testing. Simulating leaks can be done by removing product
from the tank at a constant rate, measuring the amount of product removed and the time of
collection, and calculating the resulting induced leak rate. The experimental design described in
Section 4 provides the nominal leak rates that are to be used.
Establishing the simulated leak may be achieved using a variety of equipment; however, a
method that has been successfully used for inducing leaks in previous testing is based on a
peristaltic pump. In this approach, an explosion-proof motor is used to drive a peristaltic pump
head. The sizes of the pump head and tubing are chosen to provide the desired flow rates. A
variable speed pump head is used so different flow rates can be achieved with the same
equipment. The flow is directed through a rotameter so that flow can be monitored and kept
constant. One end of the tubing is inserted into the product in the tank. The other end is placed
in a container. Typically, volatile products are collected into a closed container in an ice bath.
7
-------
The time of collection is monitored, the amount of product weighed, and the volume at the
temperature of the tank is determined to obtain the induced leak rate. While it is not necessary to
achieve the exact nominal leak rates, the induced leak rates should be within ±30 percent of the
nominal rates. The induced leak rates should be carefully determined and recorded. The leak
rates measured by the ATGS will be compared to the measured induced leak rates.
3.5 Water Ingress Sensor Test Equipment
A vertical standpipe is used to test the water sensor. The standpipe diameter should be large
enough to accommodate the water sensor part of the ATGS and the height must be 8 inches or
more. Minimizing waste is a consideration in determining the size of the water testing
standpipe. The water sensor test setup needs to accurately measure the water phase height level
to ±0.001 inch. The ATGS should be mounted so the water sensor is in the same relation to the
bottom of the standpipe as it would be to the bottom of a UST. In addition, a means of adding
water to the standpipe is needed. This can be accomplished by using a pipette or a peristaltic
pump. Dispose of product miscible with water after the test. For water sensors used in alcohol-
based fuels, an alternative means of determining water ingress is required. See Section 4.4 for
suggested alternatives.
3.6 Miscellaneous Equipment
As noted, the test procedures require the partial emptying and filling of the test tank. One or
more large capacity fuel pumps will be necessary to fill the tank in a reasonably short time.
Hoses or pipes will be needed for fuel transfer and containers will be necessary to hold the
product collected from the induced leaks. In addition, a variety of tools are necessary for making
the necessary equipment connections.
Measuring the temperature of the product consistently is very important. One suggested way to
measure the temperature of the product in the tank is to use a probe with five temperature sensors
spaced to cover the diameter of the tank. The probe is inserted, or installed permanently, in the
tank and the temperature readings of those sensors are used to obtain an average temperature of
the product. The temperature sensors can be spaced to represent equal volumes or the
temperatures can be weighted with the volume each represents to obtain an average temperature
for the tank.
8
-------
Section 4: Test Procedures
The test procedures for ATGS consist of two parts. The first evaluates the release detection
function of the ATGS. The second evaluates its water ingress detection function and the
method's resolution of sensing water ingress.
The overall performance of the ATGS is estimated by comparing the method's measured or
detected leak rates to the actual induced leaks. Performance is measured over a variety of
realistic conditions, including filling effects, such as potential loss of product from nozzle drips
during filling operations. Extreme conditions, not represented in the testing, can cause any
method to give misleading results. If the method performs well overall, then it may be expected
to perform well in the field. The test procedures have been designed so that additional analyses
can be done to determine whether the method's performance is affected by the product
properties, the amount of product in the tank, or the size of the leak.
The test procedures introduce three main factors that may influence the method's test results:
size of the leak, amount of product in the tank, and temperature variation. An additional factor is
the method's ability to deal with groundwater level effects. This factor is evaluated when
determining the method's water sensing threshold and resolution at the bottom of the tank or
alternative means of determining water ingress is required.
The primary factor is the size of the leak. The method is evaluated on its ability to measure or
detect leaks of specified sizes. If a method cannot closely measure a leak rate of 0.20 gal/hr or if
the method demonstrates excessive variability on a tight tank, then its performance is not
adequate. The ability of the method to track the leak rates can be compared for the different leak
rates.
The second factor is the amount of product in the tank. Since ATGSs work at different levels of
product in the tank, the required monthly test may be done at various levels. Two main levels
have been chosen to represent these product levels:
Half-full, which requires the most sensitive level measurement; and
90 to 95 percent full, which produces the most head pressure and the largest volume
change given product temperature differentials.
For tanks that may routinely be operated at low volumes, an optional analysis is included for a
third product level for vendors to demonstrate the lowest product level a test method can detect.
This entails at least one test at each nominal leak rate and one at the 0.0 gal/hr leak rate, which
total four additional tests. Details are provided in the supplementary analysis section
(Section 5.3.6).
The third factor is temperature variation. The method is evaluated at ambient temperature and
set temperature differentials to determine effects of temperature changes on the product.
9
-------
In addition to varying these factors, environmental data are recorded to document the test
conditions. These data may explain one or more anomalous test results.
The groundwater level is a potentially important factor in tank testing, and the method's means
of addressing it must be documented. A method that does not determine and account for the
groundwater level is not adequate. Groundwater levels are above the bottom of the tank at
approximately 25 percent of UST sites nationwide, with higher proportions in coastal regions.
The water sensing function of the ATGS is used to detect leaks in the presence of a groundwater
level above the bottom of the tank. If the groundwater level is high enough so there is an inward
pressure through most levels of product in the tank, then water will come into the tank if there is
a hole below the groundwater level. Since an ATGS must operate at normal operating levels of
product in the tank, it uses water incursion to detect leaks if there is a high groundwater level.
These test procedures evaluate two aspects of the method's water sensing function: the minimum
detectable water level and the minimum detectable change in water level. Together, these can be
used with the dimensions of the tank to determine the ability of the method's water sensor to
detect inflows of water at various rates.
4.1 Environmental Data Records
The test procedures, as referenced in 4.2.1, require that physical and environmental conditions
and other test variables experienced or evaluated during the evaluation be reported. The
following additional measurements should be reported (see the Individual Test Log in
Appendix B):
Ambient temperature, monitored at the beginning and end of each test
Barometric pressure, monitored at the beginning and end of the test
General weather conditions such as wind speed; sunny, cloudy, or partially cloudy sky,
rain; snow; etc.
Groundwater level
Any special conditions that might influence the results
Both normal and unacceptable test conditions for each method should be described in the
operating manual for the ATGS and provide a reference against which the existing test
conditions can be compared. The evaluation should not be conducted under conditions outside
the vendor's recommended operating conditions.
Pertaining to the tank and the product, the following items should be recorded on the Individual
Test Log (see Appendix B):
Type of product in tank
Bio-component in product
Tank volume
Tank dimensions and type
Amount of water in the tank (before and after each test)
10
-------
Temperature of product in the tank before filling
Temperature of product added each time the tank is filled
Temperature of product in the tank immediately after filling
Temperature of product in the tank at start of test
4.2 ATGS Evaluation Test Procedures For Release Detection Mode
The following presents the test conditions and schedule to determine the performance of the
ATGS.
4.2.1 Induced Leak Rates, Temperature Differentials, and Product Volume
Following a trial run in the tight tank, a minimum of 24 tests will be performed using a chosen
fuel product according to the experimental design exemplified in Table 2. The fuel product
tested must be a product that is expected to perform in a similar manner. Any product with
similar physical and chemical characteristics may be used and results may be inferred to
represent typical responses. The leak rates used will be randomized for each product volume. In
Table 2, four nominal leak rates will be induced during the procedures and will be assigned
randomly to the four leak rates LRx- LR4. These 24 tests evaluate the method under a variety of
conditions. An option to perform testing at the lowest product level will add four more tests to
the matrix and is described in more detail below.
11
-------
Table 2. Product Volume Leak Rate And Temperature Differential Test Design
Product Volume
(%)
Test No.
Pair No.
Set No.
Nominal Leak
Rate (gal/hr)
Nominal Temperature
Differential (°F)
Trial Run
-
-
0.00
0
Fill to 90-95% full
1
1
1
LRi
T2
2
1
1
lr2
t2
Empty to 50% full
3
2
1
lr4
t2
4
2
1
lr3
t2
Optional: Empty to
25
13
1
LR}
t2
lowest level*
26
13
1
lr2
t2
Fill to 90-95% full
5
3
2
LRi
Ti
6
3
2
lr4
Ti
Empty to 50% full
7
4
2
lr2
Ti
8
4
2
lr3
Ti
Fill to 90-95% full
9
5
3
lr4
T3
10
5
3
LRi
t3
Empty to 50% full
11
6
3
lr3
t3
12
6
3
lr2
t3
Fill to 90-95% full
13
7
4
lr3
t2
14
7
4
lr4
t2
Empty to 50% full
15
8
4
lr2
t2
16
8
4
LRi
t2
Optional: Empty to
27
14
4
lr3
t2
lowest level*
28
14
4
lr4
t2
Fill to 90-95% full
17
9
5
lr2
Ti
18
9
5
lr3
Ti
Empty to 50% full
19
10
5
lr4
Ti
20
10
5
LRi
Ti
Fill to 90-95% full
21
11
6
lr3
T3
22
11
6
lr2
t3
Empty to 50% full
23
12
6
lr4
t3
24
12
6
LRi
t3
*The evaluator determines the lowest fill level with consideration for geometry of the equipment. If this option is used, the
test numbers and pair numbers need to be updated.
12
-------
Leak Rates
There are two possible evaluations, 0.20 gal/hr and 0.10 gal/hr. The following four nominal leak
rates will be induced during the test procedures at the 0.20 gal/hr regulatory level:
English units Metric units
(gal/hr) (liters per hourll/hrl)
0.10 0.379
0.20 0.757
0.30 1.14
The following four nominal leak rates will be induced during the test procedures for a 0.10 gal/hr
evaluation:
English units Metric units
(gal/hr) (1/hr)
0.0 0.0
0.05 0.189
0.10 0.379
0.20 0.757
Temperature Differentials
In addition, three nominal temperature differentials between the temperature of the product to be
added and the temperature of the product in the tank during each fill cycle will be used. These
three temperature differentials are -5.6°, 0°, and +5.6 °C (-10°, 0°, and +10°F).
Product Volumes
The tests will be run in sets of two pairs, holding the temperature differential constant within a
set of four tests but changing the leak rate within each pair. The product volume will alternate
from pair to pair. The first pair of tests within a set will be run with the tank filled to 90 to 95
percent capacity. Then the tank will be emptied to 50 percent full and the second pair of tests in
the set will be run. A third fill level, at the lowest product level a method is expected to measure,
may be included by adding in a test at each of the four leak rates, increasing the number of tests
to 28.
Randomization
The standard evaluation of 24 tests will be performed by inducing the 12 combinations of the
four leak rates (LRi, LR2, LR3, and LR4) and the three temperature differentials (Ti, T2, and T3)
at the two product volumes (50 percent full and 90 to 95 percent full) as outlined in Table 2.
The evaluator is responsible for the randomization of the tests and achieves this by randomly
assigning the nominal leak rates of 0.0, 0.10, 0.20 and 0.30 gal/hr to LRi, LR2, LR3, and LR4 and
nominal temperature differentials of 0.0°, -5.6°, and +5.6 °C to Ti, T2, and T3, following the
sequence of 24 tests as shown in Table 2. In addition, the evaluator will randomly assign the
13
-------
groups of four tests to the set numbers 1 to 6, without disturbing the order of the four tests within
a set.
The vendor will install the ATGS and train the evaluator to operate it. After the trial run, the
ATGS will be operated as it would be in a commercial facility. The evaluator will operate the
ATGS and record its data. Note that since an ATGS operates automatically, it is not necessary to
keep the induced leak rates blind to evaluator. The evaluator merely starts the release detection
function of the ATGS at the appropriate time and records the results. The randomization is used
to balance any unusual conditions and to ensure the vendor does not have prior knowledge of the
sequence of leak rates and conditions to be used.
In summary, each test set consists of two pairs of tests. Each pair of tests is performed using two
induced leak rates, one induced temperature differential (temperature of product to be added -
temperature of product in tank), and one in-tank product level. Each pair of tests indicates the
sequence in which the product volumes (in gal/hr) will be removed from the tank at a given
product temperature differential.
Notational Conventions
The nominal leak rates, that are 0.0, 0.10, 0.20, and 0.30 gal/hr, after randomizing the order, are
denoted by LRi, LR2, LR3, and LR4. It is clear that these values cannot be achieved exactly in
the field. Rather, these numbers are targets that should be achieved within ±30 percent.
The leak rates induced for each of the tests will be measured during each test. They will be
denoted by Si, S2, .... S24. These are the leak rates against which the leak rates obtained by the
vendors performing their tests will be compared.
The leak rates measured by the ATGS during each of the 24 tests will be denoted by Li, L2,...,
L24and correspond to the induced leak rates Si, S2,..., S24.
The subscripts 1,..., 24 correspond to the order in which the tests were performed (see Table 2).
That is, for example, S5 and Ls correspond to the test results from the fifth test in the test
sequence.
4.2.2 Testing Schedule
The vendor should be aware that the first test is a trial run, conducted with a tight tank in stable
condition. The results of the trial run will be reported along with the other data but are not
explicitly used in the calculations estimating the performance of the method.
The trial run has three purposes. One is to allow the vendor to check out the ATGS and provide
instructions to the evaluator before starting the evaluation. As part of this check, any faulty
equipment should be identified and repaired. The second purpose is to ensure that there are no
problems with the tank and the test equipment. Practical field problems such as leaky valves or
plumbing problems should be identified and corrected with this trial run. Finally, the trial run
results provide verification that the tank is tight and a baseline for the induced leak rates to be
run in the later part of the evaluation.
14
-------
The evaluator performs the tests using a randomized arrangement of nominal leak rates,
temperature differentials, and in-tank product levels as shown in Table 2. The time lapse
between the two tests in each pair should be kept as short as practical. The date and time of
starting each test are to be recorded and reported in the test log. Twelve pairs of tests will be
carried out. After each pair of tests, the test procedure starts anew with either emptying the tank
to half full or filling it up to 90 to 95 percent capacity, stabilizing, etc. Specific details of the
testing procedures are presented in sequential steps in the following sections.
Step 1: Randomize test variables. For the 0.20 gal/hr evaluation, randomly assign the
nominal leak rates of 0.0, 0.10, 0.20, and 0.30 gal/hr to LRi, LR2, LR3, and LR4.
For the 0.10 gal/hr evaluation, randomly assign the nominal leak rates of 0.0,
0.05, 0.10, and 0.20 gal/hr. Also, randomly assign the temperature differentials of
0°, -10°, and +10°F to Ti, T2, and T3. Randomly assign the groups of four tests to
the six sets.
Step 2: Setup. The vendor-installs the ATGS and leak simulation equipment in the tank,
making sure the leak simulation equipment will not interfere with the ATGS. The
vendor also performs any calibration or operation checks needed with the
installation of the ATGS and leak simulation equipment.
Step 3: Trial run. Following the vendor's standard operating procedure, fill the tank to
50 percent full, or any level within the operating range of the ATGS release
detection mode, and allow for the stabilization period (or longer) as called for by
the method. If product is added it should be at the same temperature as that of the
in-tank product. Conduct a test on the tight tank to check out the tank system
(tank, plumbing, etc.) and the ATGS. Perform any necessary repairs or
modifications identified by the trial run.
During the trial run, record the temperature of the product in the test tank and that
of any product added to fill the test tank. After the product has been added to fill
the test tank, record the average temperature in the test tank.
Step 4: Begin release detection testing. Establish the tank fill height at 50 percent, if the
product volume was above or below that level during the trial run.
Step 5: Fill the tank to between 90 and 95 percent capacity with product at the
temperature required by the randomized test schedule. The temperature
differential will be T2 (Table 2, Test No.l). Record the date and time at the
completion of the fill. Allow for the vendor-stated stabilization period, but not
longer.
Step 6: Continue with the vendor's standard operating procedure and conduct a test on the
tank, using the method's recommended test duration. Record the date and time of
starting the test. This test will be performed under the first nominal leak rate of
the first set in Table 2. This nominal leak rate to be induced is LRi.
15
-------
When the first test is complete, determine and record the actual induced leak rate, Si, and the
ATGS measured leak rate, Li. Also record the data used to calculate the leak rate and the
method of calculation. Retain all data sheets, computer printouts, and calculations. Record the
dates and times at which the test began and ended and the length of the stabilization period.
Report the data and environmental conditions for each test in the T Individual Test Log Form in
Appendix B
Step 7: Change the nominal leak rate to the second in the first set, that is LR2 (as in
Table 2). Repeat Step 6. Note that there will be an additional period (the time
taken by the first test and the setup time for the second test) during which the
tank may have stabilized. When the second test of the first set is complete, again
record all results (dates and times, measured and induced leak rates,
temperatures, calculations, etc.).
Step 8: Empty the tank to 50 percent capacity (to within ± 5 percent of the tank
midpoint). The temperature of the in-tank product will remain unchanged.
Step 9: Change the nominal leak rate to the third in the first set, that is LR4. Repeat
Step 6. Record all results.
Step 10: Change the nominal leak rate to the fourth in the first set, that is LR3. Repeat
Step 7. Record all results.
Optional step: Empty the tank to the evaluator-determined lowest fill level (to within ± 5
percent of the target height). The temperature of the in-tank product will remain
unchanged.
Change the nominal leak rate to the first in the first set, that is LRi. Repeat Step
6.
Change the nominal leak rate to the second in the first set, that is LR2. Repeat
Step 7.
Step 11: Repeat Step 5. The temperature differential will be changed to Ti.
Step 12: Repeat Steps 6 through 10, using each of the four nominal leak rates of the
second set, in the order given in Table 2.
Steps 5 through 10, which correspond to a fill and empty cycle and one set of two pairs of tests,
will be repeated until all tests are performed. After two neutral temperature test pairs, the tank is
emptied to the evaluator-determined lowest product level and an additional pair of tests may be
conducted. This entails at least one test at each nominal leak rate and one at the zero-leak rate,
which totals four additional tests. Testing at low product levels involves reduced static head
pressure. The low-level testing provides additional performance data of the method's ability to
determine a leak under low product conditions.
16
-------
4.3 Testing Problems And Solutions
Some tests may be inconclusive due to broken equipment, spilled product used to measure the
induced leak rate, or other events that may interrupt the testing process. It is assumed that the
evaluator would be able to judge whether a test result is valid. If a test is judged invalid during
testing, then the following rules apply.
Rule 1: If a test is invalid, it needs to be rerun. The total number of tests must be at least
24. Report the test results as invalid with the reason and repeat the test.
Rule 2: If the method fails during the first test of a set of four tests and if the time needed
for fixing the problem(s) is short (less than 20 percent of the average stabilization
period or less than 1 hour, whichever is greater), then repeat that test. Otherwise,
repeat the empty and fill cycle, the stabilization period, etc. and record all time
periods.
Note: The average stabilization period is defined as the average time from filling
to start of the test. The average, along with the range (shortest and longest
periods), can be reported on the results of the EPA Standard Evaluation Form
under Optional Test Results in Appendix B. If the delay would increase this time
noticeably, then the test set should be redone.
Rule 3: If the method fails after the first test in a set of four has been completed
successfully, and if the time needed for fixing the problem(s) is less than 8 hours,
then repeat the test. Otherwise, repeat the whole cycle of empty and fill,
stabilization, and test at the stated conditions according to applicable step.
4.4 ATGS Evaluation Test Procedures For Water Ingress Detection
The ATGS probe typically has a water sensor near the bottom of the tank. A standpipe to test the
function of the water sensor consists of an independent height measurement capability accurate
to ±0.001 inch. The ATGS probe is mounted so the water sensor is in the same relation to the
bottom of the standpipe as to the bottom of a tank. Enough product is put into the standpipe so
the liquid level sensor is high enough so as not to interfere with the water sensor. The vendor
determines product or products used for testing based on the desired performance listing. The
testing approach is the same regardless of the product's miscibility in water; however, the
independent measurement must be appropriate for comparison to the method. For water
detection at the bottom of a tank, a metered amount of water (equivalent to approximately l/5th
inch height increase per minute) is added to the standpipe until the water sensor detects it, at
which time the water phase level is measured and recorded both independently and with the
ATGS. Additional amounts of water are added to produce a measurable water phase level
change of lesser than 1/16 inch or half of the vendor-stated resolution. Again, the independently-
measured level change and the level change measured by the ATGS are recorded. This is done
over the range of the water sensor or 6 inches, whichever is less. When testing is complete, the
17
-------
product and water are removed, separated or wasted (depending on the product water
miscibility), and the process is repeated.
Depending on the water miscibility of the product, water entering a UST may or may not collect
at the bottom of the tank. In this instance, ATGS vendors may adapt their water ingress
detection methods to include detecting water ingress by monitoring the change in the total liquid
level or a change in another characteristic of the bulk fuel, such as conductivity. An appreciable
change in liquid level height can be interpreted as the detection of the increase of the total liquid
volume in the tank. In the absence of a fuel delivery, that increase can be determined to be water
and trigger an alarm. Again, in the absence of a fuel delivery, an appreciable change of another
monitored fuel characteristic can be determined to be water and trigger an alarm. If an ATGS
vendor claims the ability to detect water either entrained in the fuel or collecting at the bottom of
the UST using liquid level sensor measurements or another bulk fuel sensor, follow the test
procedures for water ingress detection and collect parallel independent measurements during the
replicate tests. If the water ingress tests at the bottom of the tank do not use the 6 inches of
height to detect, it may be necessary to continue the testing for this option up to 6 inches to
observe the liquid level float or the bulk fuel monitor to detect the change and subsequently
alarm.
Collect these data and the ATGS alarms associated with detecting water from the liquid level
sensor measurements or another bulk fuel sensor. Note that a water sensor at the bottom of the
tank and detecting water using liquid level measurements or another fuel characteristic may be
tested simultaneously if the ATGS has both capabilities and differentiates the alarms for the
operator. If the alarm does not differentiate the signals, the sensors would need to be evaluated
separately as opposed to simultaneously.
The testing setup may need to be altered to accommodate the ATGS sensors and independent
detectors. A larger and more rugged standpipe may be needed to gather the liquid level
measurements by securing or burying the standpipe in a way that simulates the underground
environment. Considerations such as material of construction of the standpipe need to be
considered if glass is not strong enough to withstand simulating the underground environment.
Finally, the method of independently measuring the characteristics of interest may need to be
monitored using different technologies, for example if the sides of the standpipe are not visible
for measurement with a ruler.
Another challenging operating condition with water detection due to water miscibility is when an
ATGS water sensor detects the presence of water at the bottom of the tank, then a fuel delivery is
received. Because the fuel delivery mixes water into the fuel, water is no longer detected at the
bottom of the tank and the alarm stops. This alarm history could be repeated many times with or
without the ATGS recognizing this pattern as an unusual operating condition that needs to be
investigated. Again, in normal operations, all alarms indicating water detection should be
investigated. If the ATGS can interpret this pattern and respond with an alarm, this capability
can be evaluated as optional testing of the ATGS water detection mode. An additional step of
simulating a fuel delivery is taken at the end of the water ingress test replicates (assuming the
water sensor is in alarm from detecting a separated water phase).
18
-------
To conduct the fuel delivery simulation, the standpipe needs to be large enough to hold the
additional volume of fuel product to be added. The amount of fuel product to entrain the
separated water should be mathematically calculated and will vary with the size of the standpipe,
ethanol concentration of the product, and amount of water introduced during the test. The fuel
volume calculated should be increased by 20 percent to ensure the water phase mixes during this
portion of the test and to keep fuel use to a minimum. After the completion of all other water
ingress testing for a replicate, dump the fuel into the standpipe and record the ATGS reaction.
Repeat this simulation with all 20 replicates or until the ATGS recognizes and responds to the
pattern of water being detected, mixing, and then not being detected.
The testing procedure for typical water ingress detection is given in detail below.
Step 1: Install the water sensor temporarily in the test standpipe with a diameter large
enough to accommodate the water sensor. The water sensor test setup needs to be
able to accurately measure the water phase level to ±0.001 inch.
Step 2: Fill the bottom section of the standpipe with the product (typically this will
require a gallon or less). Enough product needs to be added so the product level
is high enough not to interfere with the water sensor.
Step 3: Add water in increments or at a metered rate to the standpipe until the water
sensor detects the presence of the water. Record the water phase level, the
volume of water added and the water sensor reading until the sensor responds.
The water sensor readings will be zero until the first sensor response. At that
point, measure the water phase height, Xi, of water detected. Record all data on
page 1 of the Reporting Form for Water Sensor Evaluation Data in Appendix B.
Step 4: Add enough water to the standpipe to produce a height increment (h) measured to
the lesser of 1/16 inch or half of the claimed resolution. At each increment,
record the water height (denoted by Wy in Table 4 of Section 5.2) measured
independently and by the water sensor. Use pages 2 and 3 of the Reporting Form
for Water Sensor Evaluation Data in Appendix B as necessary. Repeat the
incremental addition of water at least 20 times to cover the height of about 6
inches (or, the range limit of the water sensor, if less).
Step 5: Empty the standpipe, refill with product and repeat Steps 2 and 3 20 times to
obtain 20 replications.
Collection of the additional data to evaluate other water detection capabilities can be
simultaneous with the 20 replicate tests of water ingress testing. When the alarm for the
detection of water ingress signals using the liquid level sensor, the height measurements are
recorded on the data logs for evaluation. If this alarm is not triggered during a test or after 6
inches of water has been added, it is recorded as a false negative. When the alarm for the
detection of an alarm pattern related to fuel delivery signals, the number of simulated fuel
deliveries is recorded each time it alarms over the 20 replications.
19
-------
Record all data using the reporting forms for ATGS water sensor data in Appendix B. The 20
minimum detectable water levels are denoted by Xj, j=l,..., 20. The water sensor reading at the
ith increment of the jth test is denoted by Wy as described in Table 4 and Section 5.2.
4.5 ATGS Alternative Evaluation Procedures For Release Detection Mode
Sections 4.1 to 4.4 provide test procedures that can be accomplished in about three calendar
weeks. The standard approach described requires a tank that can be fully devoted to testing,
which may be a difficult requirement. The following alternative approach uses in-service tanks.
Only a limited amount of work is required that would prohibit using the tank for dispensing
product.
The alternative approach consists of installing the ATGS in several tanks. Since the ATGS
operates automatically, it can be programmed to perform a test whenever the tank is out of
service for a long enough period, typically each night. With several available tanks, a large set
of tests could be performed in a relatively short time. By selecting tanks in different climates or
observing tanks over the change of seasons, tests can be performed under a wide variety of
conditions. Thus, with little effort, a large database of test results on tanks assumed to be tight
can be readily obtained.
This alternative approach will provide test data under a variety of actual conditions. In selecting
the sites and times for the data collection, the evaluator should attempt to obtain a wide variety of
temperature conditions and to conduct the tests at a wide variety of product levels in the tank as
well as a variety of times after the tank receives a product delivery. This approach will produce
data under conditions as observed in the field. The primary difference between the standard and
alternative procedures is how the test conditions are attained. Both approaches attempt to
conduct the evaluation testing under conditions representative of the real world. The standard
approach does this by controlling the test conditions, while the alternative tests under a variety of
situations and records the test conditions.
Supplement the database of ATGS test results on tight tanks with a limited number of tests using
an induced leak. This demonstrates that the method can track an induced leak adequately and
will respond to and identify a loss of product from the tank of the magnitude specified in the
EPA performance standard. The combined data sets can then be analyzed to estimate the
performance of the ATGS. If the resulting performance estimate meets the performance standard
for an ATGS, that would constitute demonstration that the method meets the EPA standard.
The alternative approach will result in many tests on tight tanks, and relatively few tests under
induced leak rate conditions. A suggested sample size is 100 tight tank tests and 10 induced leak
rate tests. Larger numbers of either type of test can be used. It should be easy to collect tight
tank tests; however, some work will be needed to prepare the database, recording the ancillary
data. It will also be necessary to exclude some tests, for example those that were started, but had
a delivery or dispensing operation during the test period.
20
-------
The following steps provide an outline of the alternative approach to tank evaluation.
Step 1: Identify several tanks for installation of the ATGS. These tanks must be tight.
The tanks can be of varying sizes, but the sizes used will limit the applicability of
the results. The tanks should be at several sites, with a suggested minimum of
five different sites and 10 different tanks.
Step 2: Install identical ATGSs in the tanks. Collect and record ancillary data to
document the test conditions. The data needed are:
Average in-tank product temperature prior to a delivery
Time and date of each delivery
Average in-tank product temperature immediately after a delivery
Amount of product added at each delivery
Date, time, and results of each test
Product level when the test is run
Tank size, type of tank, product contained, etc. (see the Individual Test
Log for a form to record these data)
Step 3: Conduct tests in each tank for at least a two-week period. Tests should be run
approximately nightly or as frequently as practical with the tank's use. Report the
start and end dates of the test period. Record the test result along with the data
listed in Step 2. The data above define the conditions of each test in terms of the
time since the last fill (stabilization period), the product level, and the difference
between the temperature of the product added and that of the product in the tank.
Report all test results, even if some tests must be discarded because of product
delivery or dispensing during the scheduled test period. Identify and report the
reason for discarding any test data on the test log.
Step 4: Conduct tests with an induced leak at the rate between 0.10 and 0.20 gal/hr.
These induced leak tests will generally require a person on site to monitor the
induced leak rates and measure the rates achieved. A minimum of 10 tests is
required, with some conducted shortly after a fill with a nearly full tank, and
others conducted when the tank is about half full. The induced leak tests should
be conducted on the largest available tanks to demonstrate the performance on the
largest tank for which the ATGS is intended.
Step 5: At some time during the evaluation period, evaluate the performance of the water
sensor function. This can be done at a separate site and does not require a tank.
Follow the procedure described in Section 4.4.
Step 6: Using the resulting data, analyze the differences between the leak rate measured
by the ATGS and the induced leak rate achieved (zero for the many tests on tight
tanks) for each test to estimate the performance.
21
-------
The resulting data can also be used to investigate the relationship of the error size (the leak rate
differences) to each of the variables measured for the tests. These include tank size, length of
stabilization period, temperature differential, product level, and presence of induced leaks.
Multiple regression techniques can be used for these analyses, most of which would fall under
the category of optional analyses. However, the data should be analyzed with the two groups of
tight tank tests and induced leak rate tests separately to demonstrate that the method can
determine the leak rates. Otherwise, it would be possible to have many tight tank tests with
small errors that would obscure large errors on the small number of induced leak rates tests. An
outline of the data analysis approach is given in Section 5.4.
22
-------
Section 5: Calculations
From the results obtained after all testing is completed, the evaluation the method's performance
will be calculated.
The evaluation of the ATGS in its release detection mode is presented first. These calculations
compare the method's measured leak rate with the induced leak rate under a variety of
experimental conditions. The P(fa) and P(d) are estimated using the difference between these
two numbers. If the overall performance of the ATGS is satisfactory, analysis and reporting of
results could end at this point. However, the experimental design has been constructed so the
effects of stabilization period, product level, and temperature can be tested to provide additional
information to the vendor.
A separate section (Section 5.2) presents the calculations to estimate the minimum detectable
water level and the minimum water level change (MLC) the method can detect.
5.1 ATGS Release Detection Mode Performance Parameters
After all tests are performed according to the basic test design, a total of at least n = 24 data
points each (4 leak rates x 3 temperature differentials x 2 fill levels) of measured leak rates and
induced leak rates will be available. These data form the basis for the performance evaluation of
the method. The measured leak rates are denoted by Li,..., Ln and the associated induced leak
rates by Si,..., Sn. These leak rates are numbered in chronological order. Table 3 summarizes
the notation used throughout the test procedures, using the example test design in Table 2.
5.1.1 Basic Statistics
The number of tests is designated by n. Calculate the mean squared error (MSE), the bias (B),
and the variance of the method as follows.
Mean Squared Error, MSE
MSE
-I
71
(Li - Si)2
n
i=1
where Li is the measured leak rate obtained from the ith test at the corresponding induced leak
rate, Si, with i =1, ..., n.
Bias, B
B
= ^(Lj-50/n
B is the average difference between measured and induced leak rates over the number of tests. It
is a measure of the accuracy of the method and can be either positive or negative.
23
-------
Table 3. Notation Summary
Absolute
Pair
lVn
Set
lVn
Nominal
Nominal
Induced
Measured
Leak Rate
Test No.
Temperature
Leak Rate
Leak Rate
Leak Rate
Difference
llO.
llO.
UlIIciclllUl
(°F)
(gal/hr)
(gal/hr)
(gal/hr)
|L-S|
(gal/hr)
1
l
l
t2
LRi
Si
Li
di
2
l
l
t2
lr2
s2
l2
d2
3
2
l
t2
lr4
s3
l3
d3
4
2
l
t2
lr3
s4
l4
d4
5
3
2
Ti
LRi
s5
L5
d5
6
3
2
Ti
lr4
s6
u
d6
7
4
2
Ti
lr2
s7
L7
d7
8
4
2
Ti
lr3
s8
Ls
d8
9
5
3
t3
lr4
s9
L9
d9
10
5
3
t3
LRi
Sio
L10
dio
11
6
3
t3
lr3
s
Ln
dn
12
6
3
t3
lr2
Sl2
Li2
di2
13
7
4
t2
lr3
Sl3
Li3
di3
14
7
4
t2
lr4
Sl4
Li4
di4
15
8
4
t2
lr2
S15
L15
dis
16
8
4
t2
LRi
Sl6
Ll6
di6
17
9
5
Ti
lr2
S17
L17
dn
18
9
5
Ti
lr3
S18
Ll8
di8
19
10
5
Ti
lr4
Sl9
L19
di9
20
10
5
Ti
LRi
S2o
L2o
d2o
21
11
6
t3
lr3
S21
L2i
d2i
22
11
6
t3
lr2
s22
l22
d22
23
12
7
t3
lr4
s23
l23
d23
24
12
7
t3
LRi
s24
l24
d24
Optional Lowest Level Tests
25
13
8
t2
lr3
S25
L25
d25
26
13
8
t2
lr2
S26
L26
d26
26
14
9
t2
lr4
S27
L27
d27
28
14
9
t2
LRi
S28
L28
d28
24
-------
Variance And Standard Deviation
The variance is obtained as follows:
n
Variance = ^[(Lj S£) B]2/df
i=1
Standard deviation (SD) is the square root of the variance.
Note: The differences between the measured and induced leak rates can be plotted against the
time or the order in which they were performed. This data presentation detects any patterns that
might exist, indicating potentially larger differences in the results from the first test of each set of
tests, the three temperature differentials, or the in-tank product levels. The results could suggest
the method calls for an inadequate stabilization period after filling; the method does not properly
compensate for temperature differences between in-tank product and product to be added; or the
method is influenced by the product level. The differences between the measured and induced
leak rates by induced leak rate can also be plotted against each other. This data presentation
would graphically show the accuracy and precision of the ATGS at the various leak rates used
during testing. (See Sections 5.3.3, 5.3.4, and 5.3.5 for appropriate statistical tests.)
Test For Zero Bias
To test whether the method is accurate - that is, the bias is zero - the following test on the bias
calculated above is performed.
Compute the t-statistic
tB = y/nB/SD
From the t-table in Appendix A, obtain the critical value corresponding to a t with (n - 1) =
degrees of freedom (df) and a two-sided 5percent significance level. For 24 tests and 23 df, this
t-value is 2.07.
Compare the absolute value of tB, abs(tB), to the t-value. If abs(tB) is less than the t-value,
conclude the bias is not statistically different from zero, and the bias is negligible. Otherwise,
conclude the bias is statistically significant from zero.
5.1.2 False Alarm Rate, P(fa)
The normal probability model is assumed for the errors in the measured leak rates. Using this
model, together with the statistics estimated above, allows for the calculation of the predicted
P(fa) and the P(d) of a leak of 0.20 gal/hr.
The vendor will supply the threshold (Th) for interpreting the results of the ATGS test function.
Typically, the leak rate measured by the ATGS is compared to Th and the results interpreted as
indicating a leak if the measured leak rate exceeds the vendor stated Th. The P(fa) is the
probability the measured leak rate exceeds Th when the tank is tight. Note that by convention,
all leak rates representing volume losses from the tank are treated as positive.
25
-------
P(fa) is calculated by one of two methods, depending on whether B is statistically significantly
different from zero.
P(fa) With Negligible Bias
In the case of a nonsignificant B, compute the t-statistic
t± = Th/SD
where SD is the SD calculated above and Th is the method's threshold. Using the notational
convention for leak rates, Th is positive, P(fa) is then obtained from the t-table, using n-1 df.
P(fa) is the area under the curve to the right of the calculated value ta.
In general, t-tables are constructed to give a percentile, ta, corresponding to a given number of df,
df, and a preassigned area, alpha (a), under the curve, to the right of ta (see Figure 1). For
example, with 23 df and a = 0.05 (equivalent to a P(fa) of 5percent), ta = 1.714.
In this case, however, the area under the curve to the right of the calculated percentile, ta, with a
given number of df needs to be determined. This can be done by interpolating between the two
areas corresponding to the two percentiles in Table A-l on either side of the calculated statistic,
ta. The approach is illustrated next.
Suppose that the calculated ta = 1.85 and has 23 df. From Table A-l, obtain the following
percentiles at df = 23:
Calculate X by linearly interpolating between 1.714 and 2.069 corresponding to 0.05 and 0.025,
respectively.
f(t)
Figure 1. Student's t-Distribution Function
t Alpha (a)
1.714 0.05
1.85 X to be determined
2.069 0.025
X= 0.05-
(0.05 - 0.025)
(1.714- 2.069)
X (1.714- 1.85) = 0.040
26
-------
Thus, the P(fa) corresponding to a t« of 1.85 would be 0.040 or 4 percent.
A more accurate approach would be to use a statistical software package (e.g., SAS or SYSTAT)
to calculate the probability.
P(fa) With Significant Bias
The calculations are similar to those in the case of a nonsignificant B except the B is included in
the calculation. Compute the t-statistic including B as follows:
t2 = (C - B)/SD
P(fa) is then obtained by interpolating from the t-table, using n - 1 = df. P(fa) is the area under
the curve to the right of the calculated value t2. Note that Th is positive, but the B can be either
positive or negative.
5.1.3 Probability Of Detecting A Leak Rate Of 0.20 gal/hr, P(d)
The P(d) with a leak rate of 0.20 gal/hr is the probability the measured leak rate exceeds Th
when the true mean leak rate is 0.20 gal/hr. As for P(fa), one of two procedures are used in the
computation of P(d), depending on whether the B is statistically significantly different from zero.
P(d) with Negligible Bias
In the case of a nonsignificant B - that is, the B is zero - compute the t-statistic:
t3 = (C - 0.20)/SD
Next, using the t-table at n-1 = df, determine the area under the curve to the right of t3. The
resulting number is the P(d).
P(d) with Significant Bias
The calculations are similar to those in the case of a nonsignificant B except the B is included in
the calculation. Compute the t-statistic.
t4 = (C -B - 0.20)/SD
Next, using the t-table at n-1 = df, determine the area under the curve to the right of U. The
resulting number is the P(d).
5.1.4 Other Reported Calculations
This section describes other calculations needed to complete the Results of U.S. EPA Standard
Evaluation Form (Appendix B).
Size Of Tank
The evaluation results are applicable to tanks of up to 50 percent larger capacity than the test
tank and to all smaller tanks. Multiply the volume of the test tank by 1.5. Round this number to
27
-------
the nearest 100 gallons and report the result on page 1 of the results form (Appendix B). If the
alternative approach for release detection testing is used, reference Section 5.4.
Tested Temperature Difference
Calculate the SD of the six actual temperature differentials achieved during testing (these six
tests are the first in each of the six temperature sets) and reported with the testing details.
Average Waiting Time After Filling
Calculate the average of the time intervals between the end of the filling cycle and start of the
test for the six tests that started immediately after the specified stabilization period (first test in
each set). Note: If more than six tests are done immediately after the filling, use all tests.
However, do not use the time to the start of the remaining three tests in a set as this would give a
misleading waiting time. Report the average time as the waiting time after adding product on the
results form. At the discretion of the evaluator, the median may be used instead of the average if
there are atypical waiting times.
Average Data Collection Time Per Test
Use the duration of the data collection phase of the tests to calculate the average data collection
time for the total n. Report this time as the average data collection time per test.
5.2 ATGS Water Ingress Detection Mode Performance Parameters
Estimate two parameters for the water sensor: the minimum detectable water level or threshold
that the water sensor can determine, and the smallest change in water level the sensor can record.
These results specific to the product used during testing will also be reported on the Results of
U.S. EPA Standard Evaluation Form in Appendix B. Additional data analyses collected under
optional testing of the ATGS water detection is also presented below.
5.2.1 Minimum Detectable Water Level
The data obtained consist of 20 replications of a determination of the minimum detectable water
level (see Section 4.4). These data, denoted by xi,j=l,..., 20, are used to calculate the minimum
water level that can be detected reliably by the method.
Step 1: Calculate the mean, X, of the 20 observations:
20
Step 2: Calculate the SD of the 20 observations:
SD=\m^i
20-1
28
-------
Step 3: From a table of tolerance coefficients (K) for one-sided normal tolerance intervals
with a 95 percent probability level and a 95 percent coverage, obtain K for a
sample size of 20. The coefficient in this instance is K = 2.396. See Table A-2
for the appropriate K values for one-sided normal tolerance intervals.
Step 4: Calculate the upper tolerance limit (TL) for 95 percent coverage with a K of 95
percent:
TL = X + KSD,
or
TL = X + 2.396 SD
TL estimates the minimum level of water that can be detected by the method. That is, with 95
percent confidence, the ATGS should detect water at least 95 percent of the time when the water
depth in the tank reaches TL.
5.2.2 Minimum Water Level Change
The following statistical procedures provides a means of estimating the MLC the water sensor
can detect, based on the testing described in Section 4.4.
Denote by Wy the sensor reading (in inches) at the jth replicate and the ith increment (i=l,... ,nj,
with nj being 20 or more in each replicate). Note the number of steps in each replicate need not
be the same, so the sample sizes are denoted by nj.
Denote by h (measured to the lesser of 1/16 inch or half the claimed resolution) the level change
induced at each increment. Let m (greater than or equal to 3) be the number of replicates.
Step 1: Calculate the differences between consecutive test results. The first increment
will be Wi,i-Xi for the first replicate (j=l); more generally, Wij-Xj, for the jth
replicate. The second increment is W2,i-Wi,i for the first replicate; more
generally, W2j-Wij for the jth replicate, etc.
Step 2: Calculate the difference, at each incremental step, between h, the level change
induced during testing, and the difference obtained in Step 1. Denote these
differences by dy, where i and j represent increment and replicate numbers,
respectively. Table 4 summarizes the notations.
29
-------
Table 4. Notation Summary For Water Sensor Readings
At The jth Replicate
Increment
No.
Independently
Measured
Level Change,
h (inch)
A
Sensor
Reading (inch)
B
Measured Sensor
Increment (inch)
C
Increment
Difference
Calculated-meas.
(inch)
C-A
1
+ h
Wij
Wij-Xj*
dij
2
+ h
W2J
W2J - Wi j
dii
3
+ h
W3J
W; - W2j
dsj
11 +h ^" J c' -.j
* Xj is the water depth (inches) detected for the first time during the j"1 replication of the test.
Note the first result, Xj, may vary from replicate to replicate, so the number of differences di,j
will also vary. Let nj be the number of increments necessary during replicate j.
Step 3: Calculate the average, Dj, of the differences di,j, i=l,... ,nj, separately for each
replicatej,j=l,..., 20.
nj
Dj
- ^ di,j/nj
i=1
Step 4: Calculate the variance of the differences, Vary, i=1,..., nj separately for each
replicate j,j=l,..., m.
(
-------
Step 7: From Table A-3 of K values for two-sided tolerance intervals with 95%
probability and 95 percent coverage, obtain K for (£ rij m) df. A single set of
20 or more results can be used to estimate this SD, resulting in n - 1 df (19 if 20
are used, giving a K of 2.784.) For the suggested sample size of 20 increments,
one might want to use different starting levels to make sure that the starting level
does not adversely affect the ability to track level changes. Two starting levels
would result in 20 increments. The factor of K is based on the df for the pooled
SD. If two sets of 10 determinations are used, the df would be 18 and the
corresponding K would be 2.189.
Step 8: Calculate the MLC the water sensor can detect.
MLC = K SDp
or
MLC = 2.233 SDp
5.2.3 Water Ingress Detection With Liquid Level Measurements (Optional)
The 20 test results from this optional portion of testing identify water ingress using the liquid
level measurement increase. These results will be calculated and reported as a percentage of the
replicates where the increase was correctly identified as water ingress (qualitative).
5.2.4 Water Ingress Detection After Water Ingress Alarm, Mixing, Then No Alarm
(Optional)
The 20 test results from this optional portion of testing identify water ingress after a simulated
fuel drop. The results will be reported by the number of replicates of the cycle conducted until
the unusual operating condition was detected by the ATGS. If detected multiple times within the
20 replicates, report the average number of cycles to detect.
5.2.5 Time To Detect A 0.20-Gal/hr Water Incursion (Optional)
The minimum detectable water level and the MLC can be used to determine a minimum time
needed to detect a water incursion into the tank at a specified rate. This time is specific to each
tank size and geometry and to product-water miscibility. The calculations are illustrated for an
8,000-gallon steel tank with a 96-inch diameter and 256 inches long. Any figure derived would
also need to include the times that the given tank dimensions were measured.
Suppose there are x inches of water phase in the tank. The tank is made of quarter-inch steel, so
the inside diameter is 95.5 inches, giving a radius, r, of 47.75 inches. The water phase surface
will be 2d wide, where d, in inches, is calculated as
d = ^jr2 (r x)2
where x is the water phase depth. The area of the water surface at depth of x inches of water is
then given by 255.5 x 2d inch2. Multiplying this by the MLC and dividing the result by 231
inch3 per gallon gives approximately the volume change in gallons the water sensor can detect
reliably. This differs with the level of water phase in the tank. (For a somewhat more accurate
31
-------
approximation, calculate d at level x and at level x + MLC and average the two readings for the d
to be used to calculate the change in volume of water phase that can be detected.)
To determine how long the ATGS will take to detect a water incursion at the rate of 0.20 gal/hr,
divide the minimum volume change the water sensor can detect by 0.20 gal/hr. As a numerical
example, suppose the depth of the water was 1 inch and the MLC were 1/8 inch. In an 8,000-
gallon tank with inside diameter 95.5 inches and inside length 255.5 inches, the water surface
width, d, is calculated as
d = V(47.75)2 - (46.75)2 = 9.72 inches
The volume, in inch3, corresponding to a 1/8-inch increase is
V = 2(9.72) x 255.5 x (1/8) or
V = 620.94 inch3
In gallons, the volume is
620.94
V = = 2-688 gallon
The time the water sensor will take to detect water incursions at the rate of 0.20 gal/hr will be
2.688 gallons
time = ¦, = 13.44 hours
0.20
hr
Thus, the sensor would detect water coming in at the rate of 0.20 gal/hr after 13.4 hours, or about
half a day. The incursion of the water into the tank should be obvious on a day-to-day basis
under these conditions. The analysis presumes that water ingress rate is constant at 0.2 gal/hr,
which is unlikely given static head pressure changes caused by the tank being active and height
of the groundwater outside the tank. This calculation also assumes the ATG probe is mounted at
the midpoint of the tank, otherwise the tank tilt becomes a factor that has the potential to
drastically increase or decrease this result.
5.3 Supplemental Calculations And Data Analyses (Optional)
Other information can be obtained from the test data. This information is not required for
establishing the ATGS meets the federal EPA performance requirements but may be useful to the
vendor of the ATGS. The calculations described in this section are therefore optional. They
may be performed and reported to the vendor but are not required and are not reported on the
results form. These supplemental calculations include determining a minimum threshold, a
minimum detectable leak rate, and relating the performance to factors such as temperature
differential, stabilization period, and product level. Such information may be particularly useful
to the vendor for future improvements of their ATGS.
The experimental design tests the method under a variety of conditions chosen to be reasonably
representative of actual test conditions. The tests occur in pairs after each fill cycle. A
32
-------
comparison of the results from the first of the pair with the second of that pair allows one to
determine if the additional stabilization time improved the performance. Similarly, comparisons
among the tests at each temperature condition allow one to determine whether the temperature
conditions affected the performance. Finally, the performance under the four induced leak
conditions can be compared to determine whether the method performance varies with leak rate.
The factors can be investigated simultaneously through a statistical technique called analysis of
variance (ANOVA), The detailed computational formulas for a generalized ANOVA are beyond
the scope of these test procedures. For evaluators unfamiliar with ANOVA, equations to test for
the effect of stabilization period, temperature, and product volume individually are presented in
detail, although the evaluator may to use the ANOVA approach to the calculations if they have
the knowledge and computer programs available.
5.3.1 Minimum Threshold
The 24 test results can also be used to determine a threshold to give a specified false alarm rate
of 5 percent, for example. This threshold may not be the same as the threshold (Th) pertaining to
the method as reported by the vendor. Denote by Ths%, the threshold corresponding to a P(fa) of
5 percent. The following demonstrates the approach for computing Ths%. Solve the equation
. (" ThSo/o - B)
P{fa) = P [t > j = 0.05
for Th5%. If the bias is not statistically significantly different from zero (Section 5.1.1), then
replace B with 0. From the t-table with n-1 = df obtain the 5th-percentile. This value is 1.714.
Solving the equation above for Ths% yields
ThSo/o B
= 1.714
SD
In the case of a nonsignificant bias, this would be Ths% = 1.714 SD.
5.3.2 Minimum Detectable Leak Rate
With the data available from the evaluation, the minimum detectable leak rate, LRs%,
corresponding to a P(d) of 95 percent and a calculated threshold, Ths%, can be calculated by
solving the following equation for LRs%:
P(d(ifl5%)) = P {t > c5% ~ LR50/o - fij = Q 95
where Ths% is the threshold corresponding to a P(fa) of 5 percent as previously calculated.
At the P(fa) of 5 percent, solving the equation above is equivalent to solving
SD
or
Thso/o LRS% B^ _
33
-------
LRS% = 1.714 SD+Ths%-B
which, after substituting 1.714 SD for (Ths%-B), is equivalent to
^5% = 2Th5 0/o 2 B
Substitute 0 for B in all calculations when the B is not statistically significant. Otherwise, use
the value of B estimated from the data.
Thus, the minimum-detectable leak rate with a P(d) of 95 percent is twice the calculated
threshold, Ths%, determined to give a false alarm of 5 percent, minus twice the bias if the bias is
statistically significant.
In summary, based on the 24 pairs of measured and induced leak rates, the minimum threshold,
Th5%, and the minimum detectable leak rate, LRs%, are calculated as shown below.
If the bias is not statistically significant:
For a P(fa) of 5% Ths%= 1.714 SD
For a P(d(R)) of 95% LRs% = 2Cs%
If the bias is statistically significant:
For a P(fa) of 5% Ths%= 1.714 SD + B
For a P(d(LR)) of 95% LRs% = 2Cs%- 2 B
5.3.3 Test For Adequacy Of Stabilization Period
The performance estimates obtained in Sections 5.1.2 and 5.1.3 will indicate whether the method
meets the EPA performance standards. The calculations in this section allow one to determine
whether the method's performance is affected by the additional stabilization time the tank has
experienced by the second test after each fill cycle. These statistical tests are designed primarily
to help determine why an ATGS did not meet the performance standards.
The ATGS conducts the test after a specific stabilization period to ensure the temperature is
stable enough to perform the test. The rate of temperature change as the threshold for the
stabilization period may also be accounted for by ATGS. As such, the stabilization period would
be specific to the fuel being measured and not encompassed under a blanket stabilization period
used for all tests.
The procedures outlined in Section 4 allow time for the tank to stabilize after fuel is pumped into
the tank prior to the first test of each set. Thus, additional stabilization takes place between the
first and second tests of the first pair in each set. The length of the stabilization period following
refueling as well as the time between tests are specified by each ATGS vendor. The following
34
-------
statistical test is a means to detect whether the additional stabilization period for the second test
improves performance. If the stabilization period prior to the first test in each set is too short,
then one would expect larger discrepancies between measured and induced leak rates for these
first tests as compared to those for the second tests.
Step 1: Calculate the absolute value of the 12 differences, dj, between the measured (L)
and induced (S) leak rates for the first 2 tests in each set (second to last column in
Table 5).
Step 2: Calculate the average of the absolute differences for the first and second test in
each set separately.
D-^ (d-^ ~t~ d§ ~t~ dq ~t~ "I" "I" C?2l)/6
D2 = (d-2 + dw + + d18 + d22>)/£>
Step 3: Calculate the variances of the absolute differences from the first and second test
in each set separately.
Si = {(d, - DrfHds - DJ2 + - + (d21 - D!)2)}/5
Si = {(
-------
If the results are statistically significant, then the performance of the method is different for the
tests with the additional stabilization period. If the performance is better, that is, if the absolute
differences for the testing with additional stabilization are smaller than those for the tests with
the minimum stabilization period, then the method would show improved performance if it
increased its required stabilization period. If the method's overall performance did not meet the
EPA performance standard, performance estimates with the additional stabilization can be
calculated using only the 6 test results with the additional stabilization time. If the results
indicate the method does not meet the EPA performance standard but could meet the EPA
performance standard with the additional stabilization time, that conclusion should be reported.
Note the method would still need to conduct the full 24 tests at the longer stabilization period
before claiming to meet the EPA performance standard.
5.3.4 Test For Adequate Temperature Compensation
This section allows one to test whether the method's performance is different for various
temperature conditions. A total of eight tests will have been performed with each of the three
temperature differentials, Ti, T2, and T3 (the nominal values of 0°, -10°, and +10°F will have
been randomly assigned to Ti, T2, and T3). The 24 tests have been ordered by temperature
differential and test number in Table 5 for the example order of sets from Table 2. In general,
group the tests by temperature condition. The test results from the three temperature conditions
are compared to check the method's performance in compensating for temperature differentials.
If the temperature compensation of the method is adequate, the three groups should give
comparable results. If temperature compensation is not adequate, results from the conditions
with a temperature differential will be less reliable than results with no temperature difference.
The following statistical procedure (Bonferroni t-tests) provides a means of testing for
temperature effect on the test results. With three temperature differentials considered in the test
schedule, three comparisons will need to be made: Ti vs. T2, Ti vs. T3, and T2 vs. T3.
36
-------
Table 5. Organization Of Data To Test For Temperature Effects
Test No.
Pair No.
Set No.
Nominal
temperature
differential
(°F)
Absolute leak rate
difference |L - S|
(gal/hr)
Group
No.
5
3
2
Ti
d5
6
3
2
Ti
d6
7
4
2
Ti
d7
8
4
2
Ti
d8
Group 1
dn
17
9
5
Ti
18
9
5
Ti
di8
19
10
5
Ti
di9
20
10
5
Ti
d2o
1
1
1
t2
di
2
1
1
t2
d2
3
2
1
t2
d3
4
2
1
t2
d4
Group 2
13
7
4
t2
di3
14
7
4
t2
di4
15
8
4
t2
dis
16
8
4
t2
di6
9
5
3
t3
d9
10
5
3
t3
dio
11
6
3
t3
dn
12
6
3
t3
di2
Group 3
21
11
6
t3
d2i
22
11
6
t3
d22
23
12
6
t3
d23
24
12
6
t3
d24
If the additional four tests were conducted at the lowest product level, include those in the
appropriate temperature differential group.
37
-------
Step 1. Calculate the average of the absolute differences in each group.
Mi = H5l d-i/ngi where gi denotes the subscripts in Group 1
M2 = d.i/ng2 where g2 denotes the subscripts in Group 2
M3 = Yjq3 d-i/ng3 where g3 denotes the subscripts in Group 3
Step 2. Calculate the variance of the absolute differences in each group.
Varx = ^\dt- M-tf/df
9i
Var2 = ^\di M2)2/df
92
Var3 = di - M3)2/df
9 3
Step 3. Calculate the pooled variance of Van, Van, and Van.
dfVar-t + dfVar2 + dfVar3
Varv =
^total ^
or
Var-i + Var2 + Var3
Varv = i ^i
Step 4. Compute the standard error (SE) of the difference between each pair of the means,
Mi, M2, and M3.
1/2
Varn ( 1 )
ngi ng2
or
SE = -yJVarp
Step 5. Obtain the 95th percentile of the Bonferroni t-statistic with (ntotal - 3) = df and
three comparisons. This statistic is t = 2.60 if 24 tests were conducted.1
1Miller, Rupert G., Jr. 1981.Simultaneous Statistical Inference. Second Edition. Springer-Verlay, New York, New York.
38
-------
Step 6. Compute the critical difference (D) against which each pairwise difference
between group means will be compared.
D = SE xt
Step 7. Compare the absolute difference of the three pairwise differences with D.
Compare \Mt Mz \ with SE x t
Compare \Mt M31 with SE x t
Compare \M2 M3 \ with SE x t
If any difference in group means, in absolute value, exceeds the critical value of SE x t, then
conclude the method's performance is influenced by the temperature conditions.
If the results are statistically significant, the method's performance is affected by the temperature
conditions. If the overall performance evaluation meets the EPA standards, the effect of a 10°F
temperature difference on the method does not degrade performance severely. However, this
does not eliminate the possibility that larger differences could give misleading results. If the
overall performance did not meet the EPA performance standards, and the temperature effect
was significant, then the vendor needs to improve the method's temperature compensation or
stabilization period to meet EPA performance standards. Again, an evaluation testing the
modified ATGS would need to be conducted to document the performance before the ATGS
could claim to meet the performance standards.
5.3.5 Test For Effect Of In-Tank Product Volume
The procedures outlined in Section 4 required the tank be either half full or filled to between 90
percent and 95 percent capacity. As shown in Table 2, 12 tests will have been run with the tank
half full, and 12 tests with the tank full to 90 to 95 percent capacity. The 24 tests have been
ordered by product volume and test number in Table 6 for the example order of tests from
Table 2.
Compare the test results from the two volume levels to check for the effect of product volume on
the method's performance. If the effect is negligible, the two groups of results should be
comparable. If the method's performance is affected by the product level, then this calculation
can identify which product level of the ATGS affects the overall results of meeting or not
meeting EPA performance standards. If it does meet the performance standards at the levels in
the standard 24 tests, it can be used in the test mode at any product level. However, if there is a
significant difference in performance at the two levels, it might be advisable to recommend the
ATGS be used in its test mode only for certain product levels or advisable to perform the
additional four optional tests at the lowest detectable level. Note that this optional part of the test
procedures may only be applicable for magnetostrictive probe technology. If the performance is
not adequate for one of the product levels, the performance of the ATGS is probably marginal.
39
-------
Table 6. Organization Of Data To Test For Product Volume Effect
In-tank
Absolute Leak Rate
Test No.
Pair No.
Set No.
Product
Volume
Difference |L - S|
(gal/hr)
Group No.
1
1
1
90-95% full
di
2
1
1
90-95% full
d 2
5
3
2
90-95% full
d5
6
3
2
90-95% full
d6
9
5
3
90-95% full
d9
10
5
3
90-95% full
dio
Group 1
13
7
4
90-95% full
di3
14
7
4
90-95% full
di4
17
9
5
90-95% full
dn
18
9
5
90-95% full
di8
21
11
6
90-95% full
d2i
22
11
6
90-95% full
d22
3
2
1
50% Ml
d3
4
2
1
50% Ml
d4
7
8
4
4
2
2
50% Ml
50% Ml
d7
d8
11
6
3
50% Ml
dn
12
6
3
50% Ml
di2
Group 2
15
8
4
50% Ml
dis
16
8
4
50% Ml
di6
19
10
5
50% Ml
di9
20
10
5
50% Ml
d20
23
12
6
50% Ml
d23
24
12
6
50% Ml
d24
25
13
1
Lowest % full
d25
26
13
1
Lowest % full
d26
Group 3
27
14
2
Lowest % full
d27
28
14
2
Lowest % full
d28
40
-------
The operation of the test function could be restricted to the product level where the performance
was adequate.
One of the consequences of using an ATGS to test at various levels of product in the tank is the
test can only find leaks below the product level used in the test. The performance standard calls
for detecting a leak from any portion of the tank that normally contains product. Ideally, the test
should be run with the tank as full as it is filled in practice so that leaks can be detected from any
part of the tank. If the test results were restricted to testing when the tank was half full, for
example, the test could not find leaks in the upper half of the tank.
Step 1. Calculate the average of the absolute differences in the two groups.
91
where gi denotes the 12 subscripts in Group 1
92
where g2 denotes the 12 subscripts in Group 2
Step 2. Calculate the variance of the absolute differences in the two groups.
V (di ~ Mi)2
V^=L ii
91
or
V (di - M2)2
Var'=L 11
92
Step 3. Calculate the pooled variance of Van and Van.
llVar-t + 11 Var2
or
Var-i + Var2
Varv = ^
The following statistical procedures (two-sample t-test) provide a means for testing the effect of
product volume on the test results.
41
-------
Step 4.
Compute the SE of the difference between Mi and Ma.
SE =
Vrp(-\ + ±)
SE =
Var
Step 5.
Step 6.
Calculate the t-statistic.
t =
SE
From the t-table in Appendix A, obtain the critical value corresponding to a t with
(12 + 12 - 2) = 22 df and a two-sided 5 percent significance level. This value is
2.074.
Step 7. Compare the absolute value of t, abs(t), to 2.074. If abs(t) is less than 2.074,
conclude the average difference between measured and induced leak rates
obtained with a tank half full is not significantly different (at the 5 percent
significance level) from the average difference between measured and induced
leak rates obtained with a tank filled to 90 to 95 percent capacity. In other words,
the amount of product, in this given range, has no significant impact on the leak
rate results. Otherwise, conclude the difference is statistically significant, that is,
the method's performance depends on the amount of product in the tank.
5.3.6 Option To Test The Evaluator-Determined Minimum Fill Height Level
Four is the minimal number of additional tests, including the third fill level, that are needed to
analyze of the effect of product volume. These tests include at least one test at each nominal
leak rate and one at the zero-leak rate. Choose the order of these four additional tests comprising
Group 3 (in Table 6) at random. The modification to the above described test procedures is then
as follows:
Step 1. Calculate the average of the absolute differences in each group.
d JYl where gi denotes the 12 subscripts in Group 1
M2 = Y,g2 dJYl where g2 denotes the 12 subscripts in Group 2
M3 = di/4 where g3 denotes the 4 subscripts in Group 3
42
-------
Step 2. Calculate the variance of the absolute differences in each group.
Vari = ^idi - MJ2/!!
91
Var2 = di - M2)2/11
9 2
Var3 =^(di-Ms)2/3
9 3
Step 3. Calculate the pooled variance of Van, Van, and Van.
llVar-t + 11 Var2 + 3 Var3
Var =
v 28-3
Step 4.
Compute the SEi of the difference between Mi and M2
SE, =
1 1
Var( + )
pV12 12
SEt =
Var
and the SE2 of the differences between Mi and M3, and M2 and M3.
SE7 =
1 1
Varp( + ~)
Step 5. Obtain the 95th percentile of the Bonferroni t-statistic with (28-3) = 25 df and
three comparisons. This statistic is t = 2.582.2
Step 6. Compute the critical differences, D, against which each pairwise difference
between group means will be compared.
= SE-l X t = SE-l X 2.582
D2 = SE2 x t = SE2 x 2.582
2 Miller, Rupert G., Jr. 1981. Simultaneous Statistical Inference. Second Edition. Springer-Verlay, New York, New York.
43
-------
Step 7. Compare the absolute difference of the three pairwise differences with D.
Compare \M1 M2\ withSE1 x 2.582
Compare \Mt M3 \ with SEZ x 2.582
Compare \M2 M3\ withSE2 x 2.582
If any difference in group means, or in absolute value, exceed the critical value, then conclude
that the method's performance is influenced by the product fill height conditions.
Note that if the operator would like to increase the power of the above procedure to detect
differences among the fill heights involving the evaluator-determined minimum, then additional
tests at the nominal leak rates can be added. If there are a total r3, tests conducted in Group 3,
then make the following modifications to the above procedure:
M3 = Yjq., di/r3 where g3 denotes the r3 subscripts in Group 3
Var3 = ^33
laMi~M3)2
r3-l
Varp =
11 Var-t + 11 Var2 + (r3 1 )Var3
24 + r-, 3
SE2 =
Var'(j2+^>
And lastly the relevant Bonferroni t-statistic for three comparisons and 24 + r3 3 df can be
found from the following table:
t-statistic
2.566
2.552
2.541
5.4 Outline Of Calculations For Alternative Approach
This section describes the data analysis required for the alternative approach described in Section
4.5.
The water sensor data will be identical to that obtained with the standard test procedure outlined
in Section 4.4. Consequently, the same data analysis will be used. Refer to Section 5.2 for the
details.
44
-------
5.4.1 Calculation of P(fa) and P(d)
Using the leak rate reported by the ATGS and the actual leak rate (zero for tight tank tests,
measured for the induced leak rate tests), calculate the differences between the measured and
actual leak rates. Calculate the mean and SD of these differences as in Section 5.1.1. Perform
the test for significant bias and estimate the P(fa) and the P(d) as described in that section.
Calculate the variances of the differences separately for the data from the tests on the tight tanks
and those from the tests on tanks with induced leak rates. This calculation can be done as in
Section 5.3.3, except the two groups are now defined by the leak status of the tanks and the
sample sizes will not be equal. Let the subscript 1 denote the tight tank data set and 2 denote the
data from the tests with induced leaks.
Let % be the number of test results from tight tank tests and m be the number of test results from
induced leak rate tests. Denote by dji the difference between measured and induced leak rates for
each test, where j=l or 2, and i=l, ..., nx or n2. Then calculate
where the summations are taken over the appropriate groups of data, and where dj denotes the
mean of the data in group j, and is given by
form the ratio
and compare this statistic to the F statistic with (m-l) and (ni-1) df for the ratio at the 5%
significance level. The larger S value should be in the numerator and the smaller S value in the
denominator. If the calculated F statistic is larger than the F value in an F-Table (from a
statistical reference book), conclude the data from the induced leak rate tests are significantly
more variable than those from the tight tanks. If this is the case, it might impair the ability of the
ATGS to detect leaks. Recompute the P(d) (see Section 5.1.3) using the SD calculated from just
the induced leak rate tests, S2, to verify that P(d) is still at least 95 percent.
and
712 -
n2 V (d2j ~ d2Y
2 L (n2 - 1)
1 = 1
45
-------
5.4.2 Limitations On The Results
The limitations on the results must be calculated from the actual test conditions. Since the
conditions were not controlled but rather observed, take the following approach to determine the
applicable conditions.
Size Of Tank
Due to the variety in geometry and materials of tanks currently on the market, using one scaling
factor does not consider all factors. The test method may be used for tanks with less volume and
a smaller diameter than the one used during testing; however, scaling has historically been
acceptable to tanks that are 50 percent larger than the test tank used in the evaluation. This 1.5
scaling factor will continue to be used as a simple approach.
Maximum Allowable Temperature Difference
Calculate the temperature difference between the product in the tank and that of newly added
product for each delivery in the data set. Note the temperature of the delivered product can be
calculated from the temperature of the product in the tank immediately before delivery, the
temperature of the product in the tank immediately after delivery, and the volumes of product by
the following formula:
TaVa - TbVb
D~ V
VD
The subscript A denotes product in tank after delivery, B denotes product in tank before delivery,
D denotes product delivered, T denotes product temperature, and V denotes volume.
Calculate the SD of the temperature differentials and multiply this by 1.5. Report this result as
the maximum temperature differential for which the ATGS evaluation is valid.
When the calculations are complete, enter the results on the standard results reporting form in
Appendix B. Also check the box on that form to indicate the evaluation was done using the
alternative approach.
Average Waiting Time After Filling
Use the time interval between the most recent fill or product delivery and each following test as a
stabilization period. Order these times from least to greatest and determine the 20th percentile.
Report this result as the minimum and maximum stabilization period.
Average Data Collection Time Per Test
The tests often have a constant or nearly constant duration prescribed by the ATGS. If so,
simply report this duration as the test data collection time. If the ATGS software determines a
test time from the data, report the average test time taken by the test and note the ATGS software
determines the applicable test time.
46
-------
Section 6: Interpretation
Each function of the ATGS is evaluated separately based on data analysis of experimental test
results. This section covers the release detection function, water detection function, and
measurement of maximum water level change (MLC). The entire evaluation process results in
performance estimates for the release detection modes of ATGS. The results reported are valid
for the experimental conditions during the evaluation, which have been chosen to represent the
most common situations encountered in the field. These should be typical of most tank testing
conditions, but extreme weather conditions can occur and might adversely affect the
performance of the ATGS. The performance of the release detection function should be at least
as good for tanks smaller than the test tank. However, the performance evaluation results should
only be scaled up to tanks of 50 percent greater capacity than the test tank. The performance of
the water sensor in terms of minimum detectable level and MLC are independent of the tank
size. However, the volume that corresponds to these heights of water phase depends on tank
size. It should be emphasized the performance estimates are based on average results obtained
during the evaluation. Vendors are encouraged to provide a measure of the precision of a test,
such as a SE for their calculated leak rate at that site, along with the leak rate and test results.
6.1 Release Test Function Evaluation
The relevant performance measures for proving that an ATGS meets EPA standards are the P(fa)
and P(d) for a leak rate of 0.20 gal/hr. The estimated P(fa) can be compared with the EPA
standard of P(fa) not to exceed 5 percent. In general, a lower P(fa) is preferable, since it implies
the chance of mistakenly indicating a leak on a tight tank is less. However, reducing the false
alarm rate may also reduce the chance of detecting a leak. The P(d) generally increases with the
size of the leak. The EPA standard specifies that P(d) be at least 95 percent for a leak of 0.20
gal/hr. A higher estimated P(d) means there is less chance of missing a small leak.
If the estimated performance of the ATGS did not meet the EPA performance requirements, the
vendor may want to investigate the conditions that affected the performance as described in
Section 5.3. If the stabilization period, temperature condition, or the product level can be shown
to affect the performance of the ATGS, these results can be used to improve the ATGS. It may
be possible to improve the performance simply by changing the ATGS method procedure (e.g.,
waiting longer for the tank to stabilize) or it may be necessary to redesign the hardware. In
either case, a new evaluation with the modified method is necessary to document that the ATGS
meets the performance standards.
The relationship of performance to test conditions is primarily of interest when the ATGS did not
meet the EPA performance standards. Developing these relationships is part of the
supplementary data analysis that may be useful to the vendor but is not of primary interest to
many tank owners or operators.
6.2 Water Level Detection Function
The minimum water level detected by the ATGS is estimated from the average threshold of
detection, and the variability of the water level threshold is estimated by the SD of the test data.
47
-------
The minimum water level that will be detected at least 95 percent of the time is the level to be
reported. Statistically, this is a one-sided tolerance limit.
The tolerance limit calculated in Section 5.2.1 estimates the minimum water level the ATGS can
detect above the bottom of the probe. If the installation of the ATGS leaves the probe at a
specified distance above the bottom of the tank (for example, 1 inch), then this minimum
distance needs to be added to the reported minimum detectable water level.
6.3 Minimum Water Level Change Measurement
Since ATGSs operate with the product at all levels of normal tank operation, the water sensor
can be used to test for leaks in the event of a high groundwater level. If the groundwater level is
above the bottom of the tank, there will be an inward pressure when the product level is
sufficiently low, and if there is a hole in the tank, water will flow into the tank under these
conditions. Based on the ability of the water sensor to detect a change in the level of water phase
in the product and using the independently measured level, the evaluator can determine how
much water must enter the tank for an increase in the water level to be detected. From this
information the evaluator can conservatively calculate the size of a leak of water into the tank the
ATGS can detect at a given time. It should also be noted that this water phase increase at the
bottom of the tank is affected by the level of mixing and water miscibility of the product.
Therefore, the increase in water phase at the bottom of the tank during testing is a better estimate
than assuming no miscibility in the calculation; however, there may still be a large amount of
variability.
The SD of the differences between the change in water phase level measured by the water sensor
and the change independently measured during the tests is used to determine the ability of the
water sensor to detect changes in the water level. A two-sided 95 percent tolerance interval is
then calculated for this detection ability (Section 5.2.2).
The MLC that can be detected is used to compute a minimum change in water volume in the
tank. This conversion is specific to the tank size. Using the minimum change in water phase
volume the water sensor can detect, the time needed for the ATGS to detect an incursion of water
at the rate of 0.20 gal/hr can be determined (Section 5.2.3). This calculation indicates the time
needed for the water sensor to identify an inflow of water at the minimum leak rate and to alert
the operator the water level has increased. If the ATGS has a water alarm, and if the conditions
for activating the water alarm are specified, the length of time for that alarm to be activated can
be calculated. This calculation assumes that the ATG probe is mounted at the midpoint of the
tank; otherwise tank tilt becomes a factor that must be addressed.
6.4 Limitations
The limitations on the results of the evaluation are to be reported on the Results of U.S.EPA
Standard Evaluation Form (Appendix B). The intent is to document that the results are valid
under conditions represented by the test conditions. Section 5.1.4 describes the summary of the
test conditions that should be reported as limitations on the results form. These items are also
discussed below. The test conditions have been chosen to represent the majority of testing
situations, but do not include the most extreme conditions under which testing could be done.
48
-------
The test conditions were also selected to be practical and not impose an undue burden for
evaluation on the vendors.
One practical limitation of the results is the size of the tank. Tests based on volumetric changes
generally perform less well as the size of the tank increases. Consequently, the results of the
evaluation may be applied to tanks smaller than the test tank. The results may also be extended
to tanks of 50 percent larger capacity than the test tank. Thus, if testing is done in a 10,000-
gallon tank, the results may be extended to tanks up to 15,000 gallons in size.
A second limitation on the results is the temperature differential between the product added to
the tank and that of the product already in the tank. Often a leak test must be performed shortly
after a tank has been filled. The reported results are applicable provided the temperature
differential is no more than that used in the evaluation. During the EPA national survey,3 EPA
found that temperature differentials were no more than 5°F for at least 60 percent of the tests.
However, larger differences could exist. These test procedures are designed to use 10°F
temperature differentials, reporting those actually used. The results cannot be guaranteed for
temperature differentials larger than those used in the evaluation.
A third limitation on the results is the stabilization period needed by the ATGS. The Individual
Test Logs call for recording the actual stabilization period used during the testing. The mean of
these stabilization periods is reported, as are the shortest and longest periods. The results are
valid for stabilization periods at least as long as those used in the evaluation.
The duration of the data collecting phase of the test is another limitation of the ATGS. If the
collection time and amount of data collected is shortened during a test, the method's performance
may be adversely affected and the results will be invalid. This is primarily of concern when
documenting that a tank is tight. Results that clearly indicate a leak can sometimes be
determined in less time than needed to document a tight tank, particularly if the leak rate is large.
Thus, while the false alarm rate may be larger if the test time is shortened, this is not usually a
problem if test results indicate a leak and efforts are made to identify and correct the source of
the leak.
There is potentially an additional limitation on the results regarding the ability of the water
sensor to function sufficiently with ethanol-blended fuels. The minimum depth of water phase
the water sensor can detect and the minimum change in water level that the water sensor can
detect is reported. Note that the calculations in Section 5.2 do not consider a water phase with
ethanol and is therefore a conservative estimate of the time to detect water ingress at this rate.
Depending on the test procedure applied in accordance with Section 4.4, any limitations or
expected effect on performance regarding use of the water sensor with ethanol-blended fuels
must be noted in applicable forms.
3 Flora, J.D., Jr., and J.E. Pelkey, "Typical Tank Testing Conditions," EPA Contract No.68-01-7383, Work Assignment 22, Task
13, Final Report, December 1988.
49
-------
Finally, the same reporting forms provided in Appendix B can be used to document limitations
for the alternative evaluation described in Section 4.5. The data analysis for the alternative
approach is described in Section 5.4. This analysis will result in reporting observed average
conditions during the evaluation. The limitations are based on the observed conditions instead of
experimentally controlled conditions, but the results are reported on the same form. The
Individual Test Log form should be applicable to the induced leak rate tests under the alternative
evaluation procedure. However, the evaluator may find it more efficient to design a different
data collection form for recording the data from the many tight tank tests.
50
-------
Section 7: Reporting of Results
Appendix B is designed to be the framework for a standard evaluation report. There are five
parts to Appendix B, each with instructions for completion:
Part 1: Results of U.S. EPA Standard Evaluation Form. This form, completed by
the evaluator, is an executive summary of the findings and intended for the tank
owner or operator that uses this method of release detection. The report should be
succinct so the results form can be widely distributed.
Part 2: Description of the ATGS. This form should be completed by the evaluator
with help from the vendor.
Part 3: Reporting Form for Leak Rate Data. This form summarizes the test results
and contains the information on starting dates and times, test duration, leak rate
results, etc.
Part 4: Individual Test Log. This log should be used to record data in the field.
While the completed daily test logs are optional in the standard report, the evaluator
should keep copies for three years in case questions arise. These logs serve as the
backup data to document the performance estimates reported.
Part 5: Reporting Form for Water Sensor Evaluation Data. This form contains
the minimum detectable water level data and records of the minimum water level
changes. A separate form is filled out for each test. See Section 4.4.
If the optional calculations described in Section 5.3 are performed, they should be reported to the
vendor. These results should be reported to the vendor in a supplemental report, distinct from
the standard report. The vendor would still have the supplemental information available if
needed and has the option to share the results.
51
-------
Appendix A
Definitions And Notational Conventions
A-l
-------
Definitions of terms used throughout the test procedures and the Student's t distribution table
(Table A-l) are presented here. For more information on the statistical approach and
relationships between the statistics calculated in these test procedures see the General Guidance
For Using EPA's Standard Test Procedures For Evaluating Release Detection Methods.
Accuracy: The degree to which the calculated leak rate agrees with
the induced leak rate on the average. If a method is
accurate, it has a very small or zero bias.
Calculated Leak Rate, R: A positive number, in gallons per hour (gal/hr), estimated
by the ATGS method and indicating the amount of
product leaking out of the tank. A negative leak rate
could result from water leaking into the tank,
miscalibration, or other causes.
False Alarm: Declaring that a tank is leaking when in fact it is tight.
Induced Leak Rate, S: The actual leak rate, in gal/hr, introduced in the
evaluation data sets, against which the results from a
given method will be compared.
Mean Squared Error, An estimate of the overall performance of a test method.
MSE:
Method Bias, B: The average difference between calculated and induced
leak rates. It is an indication of whether the ATGS
method consistently overestimates (positive bias) or
underestimates (negative bias) the actual leak rate.
Precision: A measure of the test method's ability in producing
similar results (that is, in close agreement) under identical
conditions. Statistically, the precision is expressed as the
standard deviation of these measurements.
Probability of Detection,
P(d):
The probability of detecting a leak rate of a given size, R
gal/hr. In statistical terms, it is the power of the test
method and is calculated as one minus beta (P), where
beta is the probability of not detecting (missing) a leak
rate R. Typically, the power of a test is expressed in
percent, as 95 percent
Probability of False
Alarm, P(fa):
The probability of declaring a tank leaking when it is
tight. In statistical terms, this is also called the Type I
error, and is denoted by alpha (a). It is usually expressed
in percent, as 5 percent.
A-2
-------
Root Mean Squared The positive square root of the mean squared error.
Error, RMSE:
Threshold, Th: The leak rate above which a method declares a leak. It is
also called the threshold of the method.
Variance: A measure of the variability of measurements. It is the
square of the standard deviation.
Table A-l. Percentage Points Of Student's t Distribution
o t t
df
a = .10
a = .05
a = .025
a = .010
a = .005
1
3.078
6.314
12.706
31.821
63.657
2
1.886
2.920
4.303
6.965
9.925
3
1.638
2.353
3.182
4.541
5.841
4
1.333
2.132
2.776
3.747
4.604
5
1.476
2.015
2.571
3.365
4.032
6
1.440
1.943
2.447
3.143
3.707
7
1.415
1.895
2.365
2.998
3.499
8
1.397
1.860
2.306
2.896
3.355
9
1.383
1.833
2.262
2.821
3.250
10
1.372
1.812
2.228
2.764
3.169
11
1.363
1.796
2.201
2.718
3.106
12
1.356
1.782
2.179
2.681
3.055
13
1.350
1.771
2.160
2.650
3.012
14
1.345
1.761
2.145
2.624
2.977
15
1.341
1.753
2.131
2.602
2.947
16
1.337
1.746
2.120
2.583
2.921
17
1.333
1.740
2.110
2.567
2.898
18
1.330
1.734
2.101
2.552
2.878
19
1.328
1.729
2.093
2.539
2.861
20
1.325
1.725
2.086
2.528
2.845
21
1.323
1.721
2.080
2.518
2.831
22
1.321
1.717
2.074
2.508
2.819
23
1.319
1.714
2.069
2.500
2.807
24
1.318
1.711
2.064
2.492
2.797
25
1.316
1.708
2.060
2.485
2.787
26
1.315
1.706
2.056
2.479
2.779
27
1.314
1.703
2.052
2.473
2.771
28
1.313
1.701
2.048
2.467
2.763
A-3
-------
df
a = .10
a = .05
a = .025
a = .010
a = .005
29
1.311
1.699
2.045
2.462
2.756
30
1.310
1.697
2.042
2.457
2.750
40
1.303
1.684
2.021
2.423
2.704
60
1.296
1.671
2.000
2.390
2.660
120
1.289
1.658
1.980
2.358
2.617
inf.
1.282
1.645
1.960
2.326
2.576
Table A-2. One Sided Normal Tolerance Limits, Confidence Interval, K
100-7=95%
100(l-a)
df
90%
95%
99%
2
20.58
26.26
37.09
3
6.156
7.656
10.55
4
4.162
5.144
7.042
5
3.407
4.203
5.741
6
3.006
3.708
5.062
7
2.756
3.4
4.642
8
2.582
3.187
4.354
9
2.454
3.031
4.143
10
2.355
2.911
3.981
11
2.275
2.815
3.852
12
2.21
2.736
3.747
13
2.155
2.671
3.659
14
2.109
2.615
3.585
15
2.068
2.566
3.52
16
2.033
2.524
3.464
17
2.002
2.486
3.414
18
1.974
2.453
3.37
19
1.949
2.423
3.331
20
1.926
2.396
3.295
21
1.905
2.371
3.262
22
1.887
2.35
3.233
23
1.869
2.329
3.206
24
1.853
2.309
3.181
25
1.838
2.292
3.158
30
1.777
2.22
3.064
35
1.732
2.167
2.995
40
1.697
2.126
2.941
50
1.646
2.065
2.863
60
1.609
2.022
2.807
80
1.559
1.965
2.733
100
1.527
1.927
2.684
200
1.45
1.837
2.57
lOOy is the confidence level in percent
lOO(l-a) is the percentage of population below (or above) tolerance limits
A-4
-------
Table A-3. Two-Sided Normal Tolerance Limits, Confidence Interval, K
100-7=95%
lOOfl-g)
df
90%
95%
99%
2
32.02
37.67
48.43
3
8.38
9.916
12.86
4
5.369
6.37
8.299
5
4.275
5.079
6.634
6
3.712
4.414
5.775
7
3.369
4.007
5.248
8
3.136
3.732
4.891
9
2.967
3.532
4.631
10
2.829
3.379
4.433
11
2.737
3.259
4.277
12
2.655
3.162
4.15
13
2.587
3.081
4.044
14
2.529
3.012
3.955
15
2.48
2.954
3.878
16
2.437
2.903
3.812
17
2.4
2.858
3.754
18
2.366
2.819
3.702
19
2.337
2.784
3.656
20
2.31
2.752
3.615
21
2.286
2.723
3.577
22
2.264
2.697
3.543
23
2.244
2.673
3.512
24
2.225
2.651
3.483
25
2.208
2.631
3.457
26
2.193
2.612
3.432
27
2.178
2.595
3.409
28
2.164
2.579
3.388
29
2.152
2.554
3.368
30
2.14
2.549
3.35
35
2.09
2.49
3.272
40
2.052
2.445
3.213
50
1.996
2.379
3.126
60
1.958
2.333
3.066
80
1.907
2.272
2.986
100
1.874
2.233
2.934
200
1.798
2.143
2.816
500
1.737
2.07
2.721
100
1.709
2.036
2.676
GO
1.645
1.96
2.576
1 OOy is the confidence level in percent
100(1-a) is the percentage of population included between tolerance limits
A-5
-------
Appendix B
Reporting Forms
B-l
-------
Appendix B provides five sets of blank forms. Once filled out, these forms will provide the
framework for the standard report. The forms consist of the following:
1. Results of U.S. EPA Standard Evaluation - Automatic Tank Gauging System
2. Description - Automatic Tank Gauging System
3. Reporting Form for Leak Rate Data - Automatic Tank Gauging System
4. Individual Test log - Automatic Tank Gauging System
5. Reporting Form for Water Sensor Evaluation Data - Automatic Tank Gauging
System
Each set of forms is preceded by instructions on how the forms are to be filled out and by whom.
The following is an overview on various responsibilities.
1. Results of U.S. EPA Standard Evaluation. The evaluator is responsible for
completing this form at the end of the evaluation.
2. Description of Automatic Tank Gauging System. The evaluator assisted by the
vendor will complete this form by the end of the evaluation.
3. Reporting Form for Leak Rate Data. The evaluator completes this form. In
general, the statistician analyzing the data will complete this form. A blank form can
be developed on a personal computer, the database for a given evaluation generated,
and the two merged on the computer. The form can also be filled out manually. The
input for that form will consist of the field test results recorded by the evaluator's field
crew on the Individual Test logs (below) and the ATGS test results.
4. Individual Test Logs. The evaluator completes and uses these forms. These forms
need to be kept blind to the vendor during testing. The evaluator should reproduce a
sufficient number (at least 24 copies) of the blank form provided in this appendix and
produce a bound notebook for the complete test period. The individual log sheets are
optional in the evaluation report and should be archived with the evaluator if not
included.
5. Reporting Form for Water Ingress Sensor Evaluation Data. These forms provide
a template for the evaluation data. They are to be used and completed by the
evaluator. The evaluator should reproduce a sufficient number (at least 20 copies) of
the blank form provided in this appendix and produce a bound notebook to be used in
the field.
At the completion of the evaluation, the evaluator will collate all the forms into a single standard
report in the order listed above except for the test logs (that may be archived and not part of the
final report). If the evaluator performed additional, optional calculations (see Section 5.3 of the
test procedures), those results can be attached to the standard report; however, there is no
reporting requirement for these optional calculations.
B-2
-------
If the alternative EPA test procedures described in Section 4.5 was followed, then the reporting
is essentially the same as that for the standard evaluation procedure. The major difference is
that the Results of U.S.EPA Standard Evaluation form will be completed using the results of the
calculations described in Section 5.4. A box is provided to indicate which evaluation procedure
was used. Archive individual test logs should questions arise during review of the report.
Summarize the tank test results (no-leak and induced leak rate conditions) on the Reporting Form
for Leak Rate Data. There will be no changes in the reporting of the water sensor performance
since only one testing procedure is presented.
Distribution Of The Evaluation Test Results
The organization performing the evaluator will prepare a report to the vendor describing the
results of the evaluation. This report consists primarily of the forms in Appendix B. The first
form reports the results of the evaluation and is designed to be distributed widely. A copy of this
form will be supplied to each tank owner and operator who uses this system of release detection.
The owner and operator must retain a copy of this form as part of their record keeping
requirements. The owner and operator must also retain copies of each tank test performed at the
facility to document that the tank(s) passed the tightness test. This form will also be distributed
to regulatory authorities who must approve release detection methods for use in their jurisdiction.
The evaluator submits the complete report, consisting of all the forms in Appendix B, to the
ATGS vendor. The vendor may distribute the complete report to regulatory authorities who wish
to see the data collected during the evaluation. It may also be distributed to customers of the
release detection method who want to see the additional information before deciding to select a
particular release detection method.
The evaluator provides the optional part of the calculations (Section 5.3), to the ATGS vendor.
This is intended primarily for the vendor's use in understanding the details of the performance
and perhaps suggesting how to improve the method. It is left to the vendor whether to distribute
this form.
The evaluator provides the report to the vendor. Distribution of the report to tank owners,
operators, and implementing agencies is the responsibility of the ATGS vendor.
B-3
-------
Results Of U.S. EPA Standard Evaluation
Automatic Tank Gauging System (ATGS)
Instructions For Completing The Form
The evaluator fills out this form upon completion of the evaluation of the ATGS. This form will
contain the most important information relative to the ATGS evaluation. All items are to be
filled out and the appropriate boxes checked. If a question is not applicable to the ATGS, write
'NA' in the appropriate space.
This form consists of five main parts.:
ATGS description
Evaluation results
Test conditions during evaluation
Limitations on the results
Certification of results
ATGS Description
Indicate the commercial name of the ATGS, the version, and the name, address, and telephone
number of the vendor. Some vendors use different versions of their ATGS when using it with
different products or tank sizes. If so, indicate the version used in the evaluation. If the vendor
is not the party responsible for the development and use of the ATGS, then indicate the home
office name and address of the responsible party.
Evaluation Results
The ATGS's threshold (Th) is supplied by the vendor. This is the criterion for declaring a tank to
be leaking. Typically, a method declares a tank to be leaking if the measured leak rate exceeds
C.
P(fa) is the probability of false alarm calculated in Section 5.1.2. Report P(fa) in percent. P(fa)
may be rounded to the nearest whole percent.
P(d) is the probability of detecting a leak rate of 0.20 gal/hr and is calculated in Section 5.1.3.
Report P(d) in percent. P(d) may be rounded to the nearest whole percent.
The minimum detectable water level and the minimum detectable water level change (MLC) that
the water sensor can detect will have been obtained from the calculations in Sections 5.2.1 and
5.2.2.
If the P(fa) calculated in Section 5.1.2 is 5 percent or less and if the P(d) calculated in Section
5.1.3 is 95 percent or more, then check the first 'does' box. Otherwise, check the first 'does not'
B-4
-------
box. If the MLC calculated in Section 5.2.2 is less than or equal to 1/8 inch, then check the
second 'does' box. If the MLC exceeds 1/8 inch, then check the second does not box.
Test Conditions During Evaluation
Insert the information in the blanks provided. Fill in the nominal volume of the tank in gallons
as well as the tank material, steel or fiberglass. Also, indicate the tank diameter and length in
inches. Report the product used during the testing. Give the range of temperature differences
measured as well as the standard deviation of the observed temperature differences. Note, if
more than one tank, product, or level was used in the testing, indicate this and refer to the data
summary form where these should be documented.
Limitations On The Results
The size (in gallons) of the largest tank to which these results can be applied is calculated as 1.5
times the size (in gallons) of the test tank.
The temperature differential, the stabilization period after adding the product until testing, and
the total data collection time should be completed using the results from calculations in Section
5.1.4.
If the alternative evaluation procedures described in Section 4.5 have been followed, then report
the results obtained from the calculations in Section 5.4.
Certification Of Results
The evaluator certifies which test procedures were followed and provides his or her name and
signature, and the name, address, and telephone number of the evaluator's organization.
B-5
-------
Results Of U.S. EPA Standard Evaluation
Automatic Tank Gauging System (ATGS)+
This form presents whether the automatic tank gauging system (ATGS) described below
complies with the performance requirements of the federal underground storage tank (UST)
regulation. The evaluation was conducted by the vendor according to EPA's Standard Test
Procedures for Evaluating Release Detection Systems: Automatic Tank Gauging Systems. The
full evaluation report also includes a form describing the ATGS method and a form summarizing
the test data.
UST owners and operators using this release detection method should keep this form on file to
prove compliance with the federal UST regulation. UST owners and operators should check
with the implementing agencies to make sure this form satisfies their requirements.
ATGS Description
Name
Version number
Vendor Phone
(street address) (city) (state) (zip)
Evaluation Results
This ATGS, which declares a tank to be leaking when the measured leak rate exceeds the
threshold of gallon per hour (gal/hr), has a probability of false alarms (P(fa)) of
percent.
The corresponding probability of detection (P(d)) of a ~ 0.20 or ~ o .10 gal/hr leak is
percent. The minimum water level in the tank that the ATGS can detect is
inches.
The minimum change in water level that can be detected by the ATGS is inches
(provided that the water level is above the minimum water level).
Therefore, this ATGS ~ does ~ does not meet the federal performance standards established
by the U.S. Environmental Protection Agency (0.20 gal/hr at P(d) of 95% and P(fa) of 5%), and
this ATGS O does O does not meet the federal performance standard of detecting water in the
bottom of the tank to the nearest 1/8 inch.
ATGS - Evaluation Form
Page 1 of 3
-------
Optional Testing Results
The lowest product level the ATGS can detect is inches. The test results of these
tests are included in the above (P(d) and P(fa) results.
The stabilization period from when product was added to the beginning of the test ranged from_
minutes to minutes, with an average of minutes.
The liquid level sensor detected water ingress percent of the tests.
Water was detected by the pattern of water detection, simulated fuel delivery, then no water
detection after average number of cycles.
Test Conditions during Evaluation
The evaluation testing was conducted in a gallon ~ steel Ofiberglass tank that was
inches in diameter and inches long.
The temperature difference between product added to fill the UST and product already in the
tank ranged from °F to °F, with a standard deviation of °F.
The tests were conducted with the tank product levels and percent full.
If the option of evaluating the ATGS at the third product level was conducted, that level was
percent full.
Limitations On The Results
The performance estimates above are only valid when:
The method has not been substantially changed.
The vendor's instructions for installing and operating the ATGS are followed.
The tank contains a product identified on the method description form.
The tank is no larger than gallons.
The tank is at least percent full.
The stabilization period after adding any substantial amount of product to the tank is
hours.
The temperature of the added product does not differ more than °F from that
already in the tank.
The total data collection time for the test is at least hours.
Other limitations specified by the vendor or determined during testing:
ATGS - Evaluation Form
Page 2 of 3
-------
Certification of Results
I certify that the ATGS was installed and operated according to the vendor's instructions and that
the results presented on this form are those obtained during the evaluation. I also certify that the
evaluation was performed according to one of the following:
~ Standard EPA test procedures for ATGS ~ Alternative EPA test procedures for
ATGS
Printed name Organization performing evaluation
Signature City, state, zip
Date Phone number
ATGS - Evaluation Form
Page 3 of 3
-------
Description Of Automatic Tank Gauging System
Instructions For Completing The Form
The evaluator completes this form with assistance from the vendor, as part of the evaluation of
the ATGS. This form provides supporting information on the principles behind the method or on
how the method works.
To minimize the time to complete this form, the most frequently expected answers to the
questions have been provided. For those answers that are dependent on-site conditions, please
give answers that apply in typical conditions. Write in any additional information about the
testing method that may be important.
There are seven parts to this form. These are:
1. ATGS name and version
2. Product
> Product type
> Product level
3. Level measurement
4. Temperature measurement
5. Data acquisition
6. Procedure information
> Stabilization times
> Test duration
> Total time
> Identifying and correcting for interfering factors
> Interpreting test results
7. Exceptions
Indicate the commercial name and the version of the ATGS in the first part.
NOTE: The version is provided for ATGS that use different versions of the method for different
products or tank sizes.
For the six remaining parts, check all appropriate boxes for each question. Check more than one
box per question if it applies. If a box Other is checked, please complete the space provided to
specify or briefly describe the matter. If necessary, use all the white space next to a question for
a description.
B-6
-------
Description
ATGS Background Information
This section describes briefly the important aspects of the automatic tank gauging system
(ATGS). It is not intended to provide a thorough description of the principles behind the system
or how the equipment works.
1. ATGS Information
Commercial Name:
ATGS Version:
2. Product Information
> Product Type-ATGS Compatibility? (check all
that apply)
I~1 Gasoline ~ Fuel Oil #6
~ Diesel ~ Solvents
~ Aviation fuel ~ Waste oil
I~1 Fuel oil #4 EH Other
> Product Level
Test Product Level:
~> 90% ~> 50% DOther:
Ml full
Does the ATGS measure inflow
of water as well as loss of ~Yes QNo
product (gal/hr)?
Does the ATGS detect the
presence of water in the bottom ~Yes QNo
of the tank?
3. Level Measurement
> Technique used to measure changes in product
volume
l~~l Directly measure the volume of the product
change
l~~l Changes in head pressure
l~~l Changes in buoyancy of a probe
~ Mechanical level measure (e.g. ruler,
dipstick)
l~~l Changes in capacitance
~ Ultrasonic
l~~l Change in level of float. Specify operating
principle (capacitance, magnetostrictive,
load cell, etc.):
~ Other:
4. Temperature Measurement
> Product Temperature Sensor (check those that
apply).
Quantity of Sensors:
Type of Sensors:
~
Single sensor, without
circulation
~
Resistance temperature
detector (RTD)
~
Single sensor, with
circulation
~
Bimetallic strip
~
2-4 sensors
~
Quartz crystal
~
5 or more sensors
~
Thermistor
~
Temperature-
averaging probe
~
Other:
If product temperature is not measured during
test, why not?
~ The factor measured for change in level/volume is
independent of temperature (mass)
~ The factor measured for change in level/volume self-
compensates for changes in temperature
~ Other:
5. Data Acquisitions
> Method of Data Acquisition and Record:
~ manually ~ strip chart ~ computer
6. Procedure Information
> Minimum Waiting Period between adding
product and the beginning of a test
~ No wait period ~ 7-12 hours
~ < 3 hours ~ >12 hours
~ 3-6 hours ~ Variable (explain):
> Test Duration
ATGS - Description Form
Page 1 of 3
-------
Description
ATGS Background Information
I~1 5-10 hours
~ >10 hours
~ Variable (explain):
I I < 1 hour
I I 1 hour
I I 2 hours
I I 3 hours
I I 4 hours
> Total Time
What is the total time needed to test with this
ATGS after delivery?
hours minutes
Sampling frequency for level and temperature
measurements:
~
> once per
second
~
Every 1-15
minutes
~
Every 16-30
minutes
I I Every 31-60 minutes
I I < once per hour
I I Variable (explain):
> Identifying and correcting for interfering
factors
a.
Method of determining the presence and level
of the groundwater:
~
Observation well
~
Presence of water in
near tank
the tank
~
Information from
~
Level of ground
USGS, etc.
water above bottom
~
Information for
of the tank not
personnel on-site
determined
~
Other (explain):
b. Methods of correction for inferences caused
by the presence of groundwater about the
bottom of the tank:
~ Method tests for water
incursion
~ Different product levels
tested and leak rates
compared
~ Other (explain):
~ No Action
~ Wait a specified
stabilization period
before beginning test
~ No procedure
~ Watch the data
trends and begin
test when decrease
in product level
has stopped
~ Other (explain):
d. Are the temperature and level sensors
calibrated before each test?
~ Yes ~ No
e.
If not, how frequently are the sensors
calibrated?
I~1 Weekly
~ Monthly
I I Yearly or less
I I Never
> Interpreting Test Results
a. Method of converting level changes to
volumes changes
~ Actual level changes observed when known
volume is added/removed
~ Theoretical ratio calculated from tank
geometry
~ Interpolation from tank vendor's chart
~ Not applicable; volume measure directly
~ Other (explain):
b. Method of determining coefficient of thermal
expansion (Ce)
~ Actual sample taken for each test and Ce
determined from specific gravity
~ Value supplied by vendor of product
~ Average value for type of product
~ Other (explain):
c. Method of determining tank deformation after
delivery of product:
ATGS - Description Form
Page 2 of 3
-------
Description
ATGS Background Information
c. Method to calculate leak rate (gal/hr)
~ Average of subsets of all data collected
~ Difference between first and last data
collected
~ From data from last
hours of test period
~ From data determined to be valid by statistical
analysis
~ Other (explain):
d. Threshold value used to declare a leaking tank
I~1 0.05 gal/ hr Q 0.20 gal/hr
~ 0 .10 gal/hr HH Other (explain):
~ Product level when test is conducted
I I When to conduct test
I I Waiting period between filling tank and
beginning test
I I Determination of "outlier" date that may be
discarded
I I None
I I Other:
Additional Explanations or Comments:
e. Under what conditions are test results consider
inconclusive
~ Too much variability in the data
~ Unexplained product volume increase
~ Other (explain):
7. Exceptions
a. Are there any condition under which a test
should not be conducted?
I I Water in the excavation zone
~ Large difference between ground temperature
and delivered product temperature
~ Extremely high or low ambient temperature
~ Invalid for some products (explain):
~ Other (explain):
b. What are acceptable deviations for the standard
test procedures?
I I None
~ Lengthen the duration of the test
~ Other (explain):
c. What elements of the test procedures
determined by personnel on-site?
are
ATGS - Description Form
Page 3 of 3
-------
Reporting Form For Leak Rate Data
Automatic Tank Gauging System (ATGS)
Instructions For Completing The Form
The evaluator fills out this form upon completion of the evaluation of the ATGS in its release
detection mode. A single sheet provides for 28 test results, the minimum number of tests
required in the test procedures, plus the optional 4 tests at the lowest fill height. Use as many
pages as necessary to summarize all the tests attempted.
Indicate the commercial name and the version of the ATGS and the period of evaluation above
the table. The version is provided for ATGS that use different versions of the method for
different products or tank sizes.
A blank form can be developed on a personal computer, the database for a given evaluation
generated, and the two merged on the computer. The form can also be filled out manually. The
input for that form will consist of the field test results recorded by the evaluator's field crew on
the Individual Test Logs and the ATGS test results.
The table consists of 11 columns. One line is provided for each test performed during evaluation
of the ATGS. If a test was invalid or was aborted, the test should be listed with the appropriate
notation (e.g., invalid) on the line.
The Test Number in the first column refers to the test number from the randomization design
determined according to the instructions in Section 4.1 of the test procedures. Since some
changes to the design might occur during the field testing, the test numbers might not always be
in sequential order.
Note that the results from the trial run need to be reported here as well.
The following list matches the column input required with its source, for each column in the
table.
Column No.
Input
Source
1
Test number or trial run
Randomization design
2
Date at completion of last fill
Individual Test Log
3
Time at completion of last fill
Individual Test Log
4
Date test began
Individual Test Log
5
Time test began
Individual Test Log
6
Time test ended
Individual Test Log
7
Product temperature differential
Individual Test Log
8
Nominal leak rate
Randomization design
9
Induced leak rate
Individual Test Log
10
Measured leak rate
ATGS records
11
Measured minus induced leak rate
By subtraction
B-7
-------
The product temperature differential (column 7) is the difference between the temperature of the
product added and that of the product in the tank each time the tank is filled from 50 percent full
to between 90 to 95 percent full. This temperature differential is the actual differential achieved
in the field and not the nominal temperature differential. The difference can be calculated by one
of two methods. If the evaluator measured the temperature of the product added and that of the
product in the tank just prior to filling, then take the difference between these two temperatures.
If the evaluator measured the temperature of the product in the tank before and after filling and
recorded the amount of product added, then calculate the temperature differential based on
volumes and temperatures according to the formula in Section 5.4. The data necessary for these
calculations should be provided on the Individual Test Log.
B-8
-------
Reporting Form For Leak Rate Data
Automatic Tank Gauging System (ATGS)
ATGS Name and Version:
Evaluation Period: from to (Dates)
Date at
Completion
Of Last Fill
(m/d/y)
Time At
Completion
Of Last Fill
(military)
Date
Test
Began
(m/d/y)
Time Test
Began
(military)
Time Test
Ended
(military)
Product
Temperature
Differential
(F°)
Nominal
Leak
Rate
(gal/hr)
Induced
Leak
Rate
(gal/hr)
Measure
d Leak
Rate
(gal/hr)
Meas.
1 rid. Leak
Rate
(gal/hr)
Test
No.
Trial
Run
0
0
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
ATGS-Data Reporting Form
Page 1 of2
-------
Date at
Completion
Of Last Fill
(m/d/y)
Time At
Completion
Of Last Fill
(military)
Date
Test
Began
(m/d/y)
Time Test
Began
(military)
Time Test
Ended
(military)
Product
Temperature
Differential
(F°)
Nominal
Leak
Rate
(gal/hr)
Induced
Leak
Rate
(gal/hr)
Measure
d Leak
Rate
(gal/hr)
Meas.
1 rid. Leak
Rate
(gal/hr)
Test
No.
21
22
23
24
25
26
27
28
ATGS-Data Reporting Form
Page 2 of 2
-------
Individual Test Log
Automatic Tank Gauging System (ATGS)
Instructions For Completing The Form
The evaluator completes the test log form. A separate form is to be filled out for each individual
test including the trial run (at least 25). The information on these forms is to be kept blind to the
vendor during the period of evaluation of the ATGS. These raw data forms are not needed when
submitting the evaluation report; however, they must be retained and archived for a minimum of
three years should any questions arise.
All items are to be filled out and the appropriate boxes checked. If a question is not applicable,
then indicate so as NA. The following provides guidance on the use of this form.
1. Header Information
The header information is to be repeated on all five pages, if used. If a page is not used, cross it
out and initial it. The evaluator needs to print, sign, and date the top of each sheet. The test
number is the number obtained from the randomization design. It is not the sequential running
test number. If a test needs to be rerun, indicate the test number of the test being rerun and
indicate that on the test log (e.g., Test No.5 repeat).
2. General Background Information
Indicate the commercial name of the ATGS. Include a version identification number if the
ATGS uses different versions for different products or tank sizes. The vendor's recommended
stabilization period (if applicable) must be obtained from the vendor prior to testing. This is
important since it will impact on the scheduling of the evaluation. All other items in this section
refer to the test tank and product. Indicate the groundwater level at the time of the test.
Theoretically, this information would remain unchanged for the whole evaluation period.
However, weather conditions could change and affect the ground-water level. Also, the
evaluator could change the test tank.
3. Conditions Before Testing
Fill in all the blanks. If the information is obtained by calculation (for example the amount of
water in the tank is obtained from the stick reading and then converted to volume), this can be
done after the test is completed. Indicate the unit of all temperature measurements by checking
the appropriate box.
Note that the term conditioning refers to all activities undertaken by the evaluating field crew to
prepare for a test. This includes emptying or filling the tank, heating or cooling product, and
changing the leak rate. In some cases, all the above is performed, in others, only one parameter
might be changed.
B-9
-------
Special Case Reporting
Use the Individual Test Log form to record all data pertaining to the trial run. Next, when
emptying the tank to half full and then filling to 90 to 95 percent capacity before performing the
first test, note on the form that this has been done. Indicate on page 1 the dates, time, and
volumes when product was removed and then added. This is the only case where emptying and
filling are performed in sequence without a test being performed in between. Record all other
information (e.g., temperature of product added) as applicable.
4. Conditions At Beginning Of Test
The evaluator's field crew starts inducing the leak rate and records the time. All leak simulation
data are to be recorded using the form.
Once the evaluator is ready with the induced leak rate simulation, and the testing begins, record
the date and time that the ATGS test data collection starts. Also, indicate the product
temperature at that time. Fill out the weather condition section of the form. Indicate the nominal
leak rate which is obtained from the randomization design.
5. Conditions At Completion Of Testing
Indicate date and time when the test was completed.
Again, take manual stick readings and record these readings and the amount of water in the tank.
Record all weather conditions as requested.
6. Leak Rate Data
The evaluator's statistician or analyst who performed the calculations should complete this
section. The nominal leak rate is obtained from page 2 (Conditions at Beginning of Test). It
should be checked against the nominal leak rate in the randomization design by matching test
numbers. The induced leak rate is obtained by calculation from the data reported by the
evaluating field crew on page 4 (and 5, if needed) of this form. The measured leak rate is that
recorded by the ATGS for that test. The difference is calculated by subtracting the induced from
the measured leak rate.
7. Additional Comments (if needed)
Use this page for any comments (e.g., adverse weather conditions, equipment failure, reason for
invalid test, etc.) pertaining to that test.
8. Induced Leak Rate Data
The evaluator should complete this form. From the randomization design, the crew will know
the nominal leak rate to be targeted. The induced leak rate will be known accurately at the end
B-10
-------
of the test. However, the test procedures require that the induced leak rate be within 30 percent
of the nominal leak rate.
B-ll
-------
Individual Test Log
ATGS
Name of Evaluator, or Designee
Signature of Evaluator or Field Crew
Instructions:
Use on log for each test. Fill in the blanks and
check the boxes, as appropriate. Keep test log
even if test is inconclusive.
1. General Background Information
a. ATGS name and version:
b.
c.
Product type:
Type of tank:
d. Tank dimensions:
e. Groundwater level
inches above bottom of tank
f. If applicable, recommended stabilization
period before test (per vendor):
2. Conditions Before Testing
a. Start of conditioning test tank
Date: Military Time:
b. Stick reading before conditioning test tank
Product inches gallons
Water
inches
gallons
c. Temperature of product in test tank before
conditioning
~ °F or ~ °C
d. Stick reading after conditioning tank
Product inches gallons
Test Number_
Date of Test
Diameter
(inches)
Length
(inches)
Volume
(gallons)
e. Amount of product (check one only):
~ no change in product
level
~ removed from tank
(by subtraction):
gallons
~ added to tank (by
subtraction):
gallons
3. Conditions At Beginning of Test
a. Test conditions
Date of Test Data Collection:
Start Time of Test Data Collection:
(military)
Temperature of product at start of test:
n°F or n°c
Nominal Leak Rate:
gallon per hour
b. Weather conditions at beginning
Ambient temperature: Barometric pressure:
~ °F or ~ °C
Wind:
~ None
~ Light
I I Moderate
~ Strong
Sky condition:
I I Sunny
I I Partly Cloudy
~ mmHg
~ inches Hg
Precipitation:
I~1 None
~ Light
~ Moderate
~ Heavy
I~1 Cloudy
~ Dark
c. Complete the induced leak rate data sheet
ATGS - Test Log
Page 1 of 3
-------
Individual Test Log
ATGS
Name of Evaluator, or Designee
Signature of Evaluator or Field Crew
a. Test conditions
Date of Test Data Collection:
Time Test Data Collection was Completed:
(military)
Stick Reading at Completion:
Product inches
inches
Water
gallo
gallo
b. Weather conditions at end
Ambient temperature: Barometric pressure:
~ °F
~ °C
Wind:
~ None
~ Light
I I Moderate
~ Strong
Sky Condition:
I I Sunny
I I Partly Cloudy
~ mmHg
~ inches Hg
Precipitation:
I~1 None
~ Light
~ Moderate
~ Heavy
I~1 Cloudy
~ Dark
Test Number_
Date of Test
4. Conditions at Completion of Test
5. Leak Rate Data
> Filled out by the statistician or analyst
who performed the calculation.
a. Nominal leak rate
gallons per hour
(gal/hr)
b. Induced leak rate
gal/hr
c. Leak rate measured by ATGS
gal/hr
d. Difference (ATGS measured rate minus
induced rate)
gal/hr
Additional Explanations or Comments:
ATGS - Test Log
Page 2 of 3
-------
Test Number
Date of Test
Time at
product
collection
(military)
Amount of
product
collected
(mL)
Comments (if applicable)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Individual Test Log
ATGS
Name of Evaluator, or Designee
Signature of Evaluator or Field Crew
Induced Leak Rate Data Sheet
ATGS - Test Log
Page 3 of 3
-------
Reporting Form For Water Sensor Evaluation Data
Automatic Tank Gauging System
The evaluator's field crew completes this form when evaluating the performance of the ATGS
water sensor. A separate form is to be filled out for each individual test (at least 20 replicates).
The form provides a template to record the data and consists of three parts. These are:
1. Header information
2. Template for recording the data obtained to determine the minimum water level that the
sensor can detect in each replicate
3. Template for recording the data obtained when determining the minimum water level
change (MLC) that the sensor can detect in each replicate.
Header Information
The header information is to be repeated on all four pages, if used. If a page is not used, cross it
out and initial it.
Indicate the commercial name of the ATGS. Include the version identification if the ATGS uses
different versions for different products or tank sizes. Complete the date of test and product type
information. The alcohol content of the product will be reported on the first test form filled out
and referenced on the subsequent forms. Indicate the test (replicate) number on each sheet for
each test.
The evaluator, or designee, collecting the raw data needs to print, sign and date of the test on top
of each sheet.
Minimum Detectable Water Level Data
Follow the test procedures described in Section 4.4 and record all data on page 1 of the form.
When the water sensor first detects the water, stop testing for this replicate. The minimum
detected water level is calculated from the total amount of water added until the first sensor
response and the geometry of the probe and the standpipe. This calculation can be done after all
testing is completed and is generally performed by the statistician or other person responsible for
data analysis.
Minimum Water Level Change (MLC)
After the first water sensor response, continue with the test procedures as described in Section
4.4. Record all amounts of water added and the sensor readings at each increment using the table
as necessary. The data to be entered will be calculated once all testing is completed. Again, the
evaluator, or designee, will compute these data and enter the calculated minimum water level
detected in that replicate run.
B-12
-------
Reporting Form For Water Sensor Evaluation Data
Automatic Tank Gauging System
ATGS Name and Version:_
Date of Test:
Product Type:
Product Alcohol-content
on bulk product for all the tests.)
Name of Evaluator:
Signature of Evaluator:
. as measured by method.
(May be determined
Increment
No.
Water Phase
Height (inch)
Sensor Reading
(inch)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Total
Height
(inch)
known from the start, the length of the report
form will vary from test to test.
Calculated Minimum
Detectable Water Level (inches)
Test No.
NOTE: This form provides a template for data
reporting. Since the number of increments is not
ATGS - Water Sensor Form
Page 1 of 3
-------
Reporting Form For Water Sensor Evaluation Data
Automatic Tank Gauging System
ATGS Name and Version:
Date of Test:
Product Type:
Name of Evaluator:
Signature of Evaluator:
Test No.
Increment No.
Water Phase Height Increment, h (in) A
Sensor
Reading
(in)
B
Measured
Sensor
Increment
(in)
C
Increment
Difference
(in)
B-C
Minimum water level detected, X: inches
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
vfOTE: This form provides a template for data reporting.
Use as many pages as necessary.
ATGS - Results Form
Page 2 of 3
-------
Reporting Form For Water Sensor Evaluation Data
Using Liquid Level Measurements And Delivery Simulation
Automatic Tank Gauging System
ATGS Name and Version:_
Date of Test:
Name of Evaluator:
Product Type:
Signature of Evaluator:
Test No.
Increment
No.
Liquid Level
Height Increment,
h (in)
A
Liquid Level
Sensor
Reading (in)
B
Measured Liquid
Level Sensor
Increment (in)
C
Increment
Difference (in)
B-C
Minimum water level detected, X: inches
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
NOTE: This form provides a template for data reporting.
Description and Observations of Fuel Delivery Simulation:
ATGS - Results Form
Page 3 of 3
-------
United States Land And EPA 510-B-19-002
Environmental Emergency Management May 2019
Protection Agency 5401R www.epa.gov/ust
------- |