United States
Environmental Protection
Agency
Hazardous Waste Engineering
Research Laboratory
Cincinnati OH 45268
Research and Development
EPA/600/S2-86/030 Aug. 1986
c/EPA Project Summary
A Requirements Study of an
Automated Advisory System
for Review of RCRA Permits
J. N. Brown, M. E. Deerhake, R. S. Truesdale, and E. L. Fisher
Under the Resource Conservation
and Recovery Act, hazardous waste
management facilities must apply for
operating permits. Review by the U.S.
Environmental Protection Agency
(EPA) of the present backlog of permit
applications will cost approximately
$40 million. The objectives of this proj-
ect were to develop a conceptual de-
sign and implementation plan for an au-
tomated advisory system to make the
review process more uniform and pro-
ductive, and to aid in training new staff.
The conceptual design of the advisory
system is based on expert system tech-
nology. The system will be able to as-
sist reviewers in making decisions that
normally require expertise, in exercis-
ing analytical models and in generating
correspondence and reports. A three-
year implementation plan is proposed
that incorporates field tests of proto-
type advisory system modules after 6
months time and deployment of par-
tially completed advisory systems after
the first year. Plans for advisory system
maintenance, performance monitoring,
and feedback from permit reviewers
are recommended.
This Project Summary was devel-
oped by EPA's Hazardous Waste Engi-
neering Research Laboratory, Cincin-
nati, OH, to announce key findings of
the research project that is fully docu-
mented in a separate report of the same
title (see Project Report ordering infor-
mation at back).
Introduction
As a result of the Resource Conserva-
tion and Recovery Act (RCRA) of 1976,
all hazardous waste management facili-
ties must apply for operating permits.
As of the end of 1985, the EPA Regional
Offices and states will have an esti-
mated backlog of more than 750 appli-
cations for operating permits. The re-
view of each permit application is
estimated to cost, on average, $52,000
in 1986 dollars, totaling approximately
$40 million over the next 3 to 4 years.
Applications for operating permits
under RCRA are reviewed by EPA Re-
gional Offices and state agencies to en-
sure that they satisfy all regulatory re-
quirements. Many of these
requirements are stated in terms of gen-
eral technical performance criteria. Both
detailed calculations and expert judge-
ment are often required from the permit
reviewer to determine whether the ap-
plication satisfies the criteria. Examples
of calculations that must be performed
include estimates of surface runoff to
check for protection against overtop-
ping of surface impoundments and
slope stability analysis to ensure the
safety of dikes. Expert judgement is re-
quired in such areas as evaluating the
chemical compatibility of synthetic
membrane liners and evaluating the
protectiveness of a proposed facility
site.
Currently, there is no uniform basis
for the review of RCRA permit applica-
tions. Variability is inherent in the inter-
pretation of information in permit appli-
cations and this variability can result in
an application's being approved in one
region and being denied in another. Ad-
ditionally, the complex calculations in-
volved and the necessity of obtaining
outside expert consultation introduce
significant delays in the permit applica-
tion review process.
-------
Measures to enhance the productivity
and uniformity of the RCRA permit ap-
plication review process are being
sought by EPA. An automated advisory
system based on expert system tech-
nology can accomplish this goal. To-
wards this end, the objectives of this
project were:
• to develop a conceptual design for a
computer-based automated advisory
system to assist in the review of
RCRA hazardous waste management
facility permit applications;
• to determine and specify the re-
sources required to develop, deploy,
and maintain the automated advisory
system; and
• to recommend an implementation
plan.
Expert Systems
The proposed advisory system is to
be based on expert system technology.
Expert systems are computer programs
that solve problems normally requiring
the aid of an expert. An expert system
consists of a knowledge base, an infer-
ence engine, and a user interface. The
knowledge base contains information
and experience possessed by the expert
and additionally, the heuristics that the
expert uses in solving problems. The in-
ference engine consists of procedures
that draw inferences from the informa-
tion supplied by the user and from the
information stored in the system's
knowledge base. The user interface al-
lows the user (e.g., the permit applica-
tion reviewer) to communicate with the
expert system.
The operation of an automated advi-
sory system based on expert system
technology is as follows: The reviewer
enters general information and design
details concerning one section of a per-
mit application (e.g., dike design). The
expert system attempts to draw infer-
ences that confirm the acceptability or
identify deficiencies in the design. The
system may request additional specific
information about the design and oper-
ation of the facility being reviewed. The
conclusions of the expert system are
presented to the user in the form of rec-
ommendations and suggestions.
Conceptual Design of the Advi-
sory System
The functions performed by the per-
mit application reviewers include: find-
ing specific data and information in the
permit application; interpreting infor-
mation presented in the application;
performing calculations and exercising
analytical models; comparing data and
information in permit applications with
information found in regulations and
guidance documents; and making deci-
sions that range from the very simple to
complex ones requiring expertise.
The conceptual design of the permit
review advisory system is presented in
block diagram form in Figure 1. The es-
sential components of the system are
the supervisory module/inference en-
gine, the user interface, and the expert
system library. These three major com-
ponents of the advisory system form a
single expert system and are the mini-
mum set of components for a working
system. The expert system library may
be thought of initially as a single expert
system knowledge base. It contains the
facts, heuristics, and system messages
associated with the review of one part
of a permit application. The working
memory associated with this knowl-
edge base stores case-specific data en-
tered by the user and the results of in-
ferences made by the expert system.
The supervisory module/inference en-
gine performs all system control func-
tions and inference making. The user in-
terface provides for simple information
exchanges between the system and the
user.
The analytical model library is func-
tionally the same as the expert system
library. Each knowledge base in the ana-
lytical model library contains knowl-
edge about how to exercise a particular
model such as, for example, the EPA's
HELP (Hydrologic Evaluation of Landfill
Performance) model. Among other
forms of assistance in using analytical
tools, the system can assist the reviewer
in assembling input data sets and inter-
preting output data.
The query/answer and explanation
subsystems are components of the user
interface. The query/answer subsystem
allows the advisory system to request
specific information that it needs and
allows the user to respond by entering
that information. The explanation sub-
system allows the user to request expla-
nations of the system's queries, infer-
ences, and recommendations.
The report generator is a software
printer interface that formats various re-
ports based on the review of an applica-
tion. For example, if a deficiency is
found in the completeness check of an
application, the system can generate
correspondence to the applicant noting
the deficiency and requesting the
specific information needed for the re-
view to continue.
The DBMS interface to an external
data base is primarily for the storage of
data generated in permit applications
reviews for later analysis and for incor-
poration in a data base of waste man-
agement facilities.
The advisory system can be imple-
mented in one of two configurations: a
separate module configuration and an
integrated system configuration. In the
separate module configuration, the ad-
visory system is a collection of individ-
ual expert systems each of which pro-
vides assistance to the reviewer in one
specific part of the review process.
There is no communication between
the individual knowledge bases that
make up the expert system library.
In the integrated version of the advi-
sory system, the supervisory knowl-
edge base contains the knowledge re-
quired to coordinate the entire permit
application review and to provide to
each expert system access to knowl-
edge entered and inferred during previ-
ously executed expert systems. The
knowledge bases that form the expert
system library in the integrated system
are called task knowledge bases. The
supervisory expert system knowledge
base also is referred to as the job knowl-
edge base.
In the integrated advisory system,
two expert systems are operating in a
coordinated manner at the same time.
The supervisory or job expert system
controls the overall review of a specific
permit application. It controls the se-
quencing of task expert systems in the
detailed review of specific sections of
the application. The job expert system
provides communications between the
individual task expert systems such that
once specific information is entered by
the reviewer, it is available for use by all
task expert systems.
Implementation Plan
The recommended implementation
plan schedule is shown in Figure 2. This
schedule emphasizes the need to pro-
vide assistance to permit reviewers as
soon as possible and to provide the as-
sistance where it is most needed.
In the first year of the implementation
phase, the basic modules of the advi-
sory system are to be developed. These
include the supervisory module and in-
ference engine, the user interface, and
four knowledge bases. The develop-
ment software will incorporate basic
control (supervisory) functions and an
inference engine as well as user inter-
face software modules. The specific
-------
knowledge bases recommended to be
developed first are: synthetic liner se-
lection, waste analysis plan, the HELP
model, and permit application com-
pleteness check.
It is recommended that EPA make the
decision to deploy the separate module
or the integrated configuration of the
advisory system at the end of the first
year of the implementation program. At
the end of the first six months, proto-
type modules of the advisory system
can be demonstrated in a Regional Of-
fice; specifically, the completeness
check and waste analysis plan modules.
Early demonstration and test of the ad-
visory system by the end user, the per-
mit application reviewer, will allow their
recommendations to be incorporated
into the system as early as possible.
This early field test will facilitate the de-
ployment of partially complete advisory
systems in one or more Regional Of-
fices by the end of the first year.
In the second year of the implementa-
tion phase, additional high priority task
knowledge bases will be developed. Ex-
amples of appropriate knowledge bases
include: cover systems, soil liners, and
facility location information. Knowledge
bases for assistance in exercising addi-
tional analytical models should be in-
cluded. The final selection of specific
knowledge bases to be developed
should be made following input from
EPA Headquarters and, especially, Re-
gional Office staff. The development of
all remaining knowledge bases is to be
completed in the third year.
It is recommended that the advisory
system be developed on symbolic pro-
cessors using sophisticated software
development tools and be delivered
using microcomputers. This approach
allows EPA to take advantage of the fea-
tures of symbolic processors during the
development phase. The use of micro-
computers as the delivery vehicle
avoids the necessity of having to pur-
chase additional hardware in order to
deploy the system. The cost of transfer-
ring the knowledge bases from the for-
mat appropriate for the development
software to one appropriate for the de-
livery software will be small compared
to the cost of knowledge base develop-
ment.
Maintenance of the advisory system
will be required for the following rea-
sons: amendments to RCRA, user sug-
gested changes and enhancements to
the system, and system enhancements
resulting from new knowledge coming
from EPA research and development
projects. It is recommended that EPA
establish a capability for advisory sys-
tem maintenance within a contractor or-
ganization. This capability should be es-
tablished in the second year of the
development phase and should be, at a
minimum, one knowledge engineer
with development hardware and soft-
Query/Answer
Subsystem
V 1 '
Explanation
Subsystem
User
Interface
Supervisory Module/
Inference Engine
Job
Knowledge
Base
r
Expert System
i
Task I
Knowledge
Base
Task
Working
Memory
Expert
System
Library
Job
Working
Memory
i
Report
Generator
DBMS
Interface
Analytical
Model
Library
External
Data
Bases
Figure 1. Automated advisory system.
-------
ware; additional capability should be
available on an as-needed basis.
It is estimated that the development
of the four knowledge bases beginning
in the first year of implementation will
require 44 person-months. Develop-
ment of the explanation module and re-
port generator will require an additional
6 person-months as will modification of
the supervisory/inference engine. Field
tests and demonstration of prototype
systems in Regional Offices will require
5 person-months; development of soft-
ware to demonstrate and test the inte-
grated advisory system configuration
will require approximately 11 person-
months. Cost estimates for years two
and three are 8 to 12 person-years of
effort each.
Recommendations
Based on the results of the project
summarized herein, it is recommended
that the EPA proceed with specific im-
plementation plans and a program to
develop the automated advisory system
to assist in the review of RCRA permit
applications as soon as possible.
Specifically, it is recommended that the
implementation incorporate the follow-
ing:
• The integrated version of the advi-
sory system should be evaluated in
the first year of the implementation
program and a decision on system
configuration should be made at the
beginning of the second year.
• Knowledge bases for assistance in re-
viewing waste analysis plans and
synthetic membrane liner selection,
in checking the completeness of ap-
plications, and in using the HELP
model should be developed in the
first year of the program.
• A program to obtain feedback on sys-
tem utility and system enhancements
from permit applications reviewers
should be established early.
• Field tests of prototype expert sys-
tems in EPA Regional Offices should
be initiated six months after develop-
ment is started.
• Arrangements for system mainte-
nance and system performance mon-
itoring should be completed in the
second year of the program.
The full report was submitted in fulfill-
ment of EPA Contract 68-02-3992 by the
Research Triangle Institute under the
sponsorship of the U.S. Environmental
Protection Agency.
Advisory System Modules
Supervisory Module and
Inference Engine
Explanation Module and
Report Generator
y \
S/S/J1 }
"1 /
Expert System Library
Synthetic Liner Design''
Waste Analysis Plan'
Dike Design and Construction
Cover Systems2
Soil Liners
Leachate Collection Systems
Facility Location Information
Remaining (SelectedI Knowledge
Bases
Job Know/edge Base
Completeness Check
Supervisory Functions
Analytical Model Library
HELP
Remaining (Selected) Models
Field Test
Deployment
Maintenance
'Knowledge acquisition partially complete.
2Part of Closure Plans.
Figure 2. Tentative implementation schedule.
^HI^H Separate Module Implementation
Y///////A Integrated Advisory System
------- |