1. Catalog Description

Course Title: IME 499, Artificial Intelligence in Engineering (3 units)

Material to be covered: Basic concepts in Database Systems, Expert Systems, and Machine Learning techniques. (For more information, please refer to the course outline)
 

2.   Required Background or Experience

This course is designed for students from various engineering fields with some limited experience in programming.  It is required (Prerequisite) to have upper division standing in engineering.
 

3.  Expected Outcomes

At the satisfactory completion of this course the student will achieve the following outcomes:
C Become familiar with the basic concepts in areas of Databases, Expert Systems, and Machine Learning.
C Understand the limitations of Artificial Intelligence (AI).
C Recognize and appreciate the application of AI as a tool to solve ill-structured problems in engineering.
C Be capable of applying tools, covered in this course, in other courses or in industry.
C Improve communications skills such as team-work, writing professional reports, and presentations.
 

4.  Text and Readings

Texts:

[1]  Supplementary instructor notes and transparencies

[2]  Access Notes on Dr. F. Janger's web page, Civil Engineering Department at Cal Poly Pomona  http://www.engineering.csupomona.edu/civil/faculty/janger/janger.htm

[3]  J. W. Satzinger and T. U. Orvik, Object-Oriented Approach: Concepts, Modeling, and System Development, International Thomson Publishing Company, 1996.  ISBN 0-7895-0110-4

[4]  Peter Jackson, Introduction to Expert Systems, Third Edition, Addison-Wesley Pub Co., 1999. ISBN 0-201-87686-8

[5]  Carol E. Brown and Daniel E. O'Leary, Introduction to Artificial Intelligence and Expert Systems, web page at   http://www.bus.orst.edu/faculty/brownc/es_tutor/es_tutor.htm#1-AI

[6]  Ram D.Sriram, Intelligent Systems for Engineering: A Knowledge-Based Approach, Springer-Verlag London Limited, 1997.  ISBN 3-540-76128-4

[7]  Adedeji B. Badiru, Expert Systems Applications in Engineering and Manufacturing, Prentice Hall, Inc., 1992.  ISBN 0-13-278219-7

[8]   Patrick Henry Winston, Artificial Intelligence, 4th Edition, Addison Wesley Publication Co., will be published 1999   ISBN 0-201-53377-4, 1992

[9]  Elaine Rich, Artificial Intelligence, ISBN 0-07-052261-8, 1983

[10] Elmasri and Navathe, Fundamentals of Database Systems, Benjamin/Cummings Publishing Company, Inc., 1989.  ISBN 0-8053-0145-3

[11] John Kelly, Artificial Intelligence, A Modern Myth, ISBN 0-13-338559-0, 1993

[12]  Video on Expert Systems by Dr. Sriram

[13]  CLIPS, A Tool for Building Expert Systems web page at  http://www.ghgcorp.com/clips/CLIPS.html
Also  http://www.ghgcorp.com/clips/CLIPS-FAQ ,     http://www.ghg.net/clips/CLIPS.html ,     http://www.ghg.net/clips/download/documentation/ ,    http://www.ghg.net/clips/download/executables/examples/ ,   http://www.ghgcorp.com/clips/download/documentation/ ,    http://www.ghgcorp.com/clips/download/documentation/usrguide.pdf
Question and comment can be sent to Gary Riley at     clips@ghg.net
 
[14]  Video on AI and ES in Manufacturing by SME

[15]  T. J. O'Leary and L. I. O'Leary, Microsoft Office 97, Professional, McGraw-Hill Companies Inc., 1998.   ISBN 0-07-115475-2

[16]  Tom Mitchell, Machine Learning, McGraw-Hill Co. Inc., 1997.  ISBN 0-07-042807-7

[17]  Sima Parisay, The Application of Inductive Learning in Simulation of Queuing Systems, Ph.D. Dissertation, 1996.

[18] Toshinori Munakata, Fundamentals of the New Artificial Intelligence, Springer, 1998.  ISBN 0-387-98302-3,   Located at the Library with call number Q 335 M86
 

• Richard Miller, Artificial Intelligence Applications for Manufacturing, Madison, Georgia: SEAI Technical Publications, 1984.  Located at the Library with call number Q335.M58 1984.
• R.A. Adey, Artificial Intelligence in Engineering Design, Southampton, Boston: Computational Mechanics Publications, 1993. Located at the Library with call number TA174 .A77 1993.
• Graham Winstanley, Artificial Intelligence in Engineering, J. Wiley, 1991. Located at the Library with call number TA345 .A78 1991.
• Craig Eddy; Paul Cassel, Sam's Teach Yourself Access 97 in 21 Days, ISBN: 0672312980, Find at URL:   http://www.mcp.com/help/super_results.cfm
• Thomas M. Mitchell, Machine Learning, McGraw-Hill Series in Computer Science, 1997.   ISBN: 0-07-042807-7

The class format is a dynamic combination of lecture, discussion, and demonstration.  Video presentation will be utilized to present the application cases. There will be assignments that use software.  A group research term paper or case study with presentation will be required during the course of the term.
 

6.   Evaluation of Outcomes

The course grade is based on two exams (midterm and final), homework assignments, participation, and a term paper or case study.  The term paper or case study will be a group assignment with presentation.

The major topics covered in this course are explained below.

1.  Database Systems:

Databases are discussed as a collection of related data.  The process to develop a database system will be covered as: specifying (defining) the types of data to be stored, storing the data (constructing the database), and manipulating the database.  Database manipulation refers to retrieving specific data, updating the database, and generating reports from data.  Relational databases and object-oriented databases will also be discussed.

a) Relational databases

The Access software will be utilized for this illustration. This is not a course on the Access software, however, this software will be taught to the extent that it helps in developing our discussions on database topics.  Also this software is used for a homework assignment.  You can also select to use Access for your project.  Access is considered to be a relational database.  Topics to be covered with Access for beginning and intermediary users are mentioned below.

Access 97:
 
RObject (table, form, report, query, macro, and modules)
RTable, record, field (data type, field properties, field size, format, input mask, default, and validity), search, sort, primary key, and find
RForm
RReport (mailing address)
RQuery (filtering, AND and OR operators, and joins and relationships)
RPrint, save, and delete
 

b) Object-oriented (OO) databases

The Object-oriented approach has a considerable impact on the way engineering data and knowledge will be handled in the future. Topics to be covered for beginners are listed below.
 
RObject-oriented approach (thinking)
RConcepts in objected-oriented area (object, class, attribute, method, relationship, encapsulation, inheritance, classification hierarchies, and message sending)
RObject-oriented database management
RBenefits and problems with the object-oriented approach
 

2.  Expert Systems:

Engineers have to deal with ill-structured problems for which there are no straightforward algorithmic solutions.  Engineer's heuristic knowledge, which is based on experience, plays a role in solving such problems. Artificial Intelligence (AI) finds tools, for example Expert Systems (ES), to imitate human behavior.  However, Artificial Intelligence has its limitations as compared to human intelligence.

a) Knowledge Representation

Expert Systems present the knowledge-base by production rules.  The form of production rules is IF-THEN rules with conditions in the IF part and conclusions or actions in THEN part.  Rules will be graphically presented by means of an inference or goal network.

Semantic networks will be discussed as another means for knowledge representation with nodes as objects of real world and arcs as links of these objects.  We will cover different classes of nodes (individual and generic) and different classes of arcs (generalization, classification, and aggregation).

b) Knowledge-Centered Problem Solving Strategies

Several problem-solving strategies will be discussed that are used for knowledge-centered problems.  Forward chaining will be covered as a widely applied strategy that moves from an initial state of known facts to a goal state. Backward chaining will be discussed as another strategy that tries to support a goal state by checking the known facts. Inexact inference methods will be covered that deal with erroneous, inexact, and incomplete information. Certainty and Fuzzy Logic are examples of the later approach.

c)   Expert Systems Project Management

Various issues involved in developing Knowledge-Based Systems (KBS) will be discussed such as problem identification, problem conceptualization, mapping of concepts into more formal representations, implementation, verification, and validation.  The knowledge acquisition process will be explained with its different modes.   This process is where the knowledge engineers acquire knowledge from the domain experts and translate it into some formalism.

d) The CLIPS software will be demonstrated as an expert system shell.  There will be presentation of several examples on application of expert systems in engineering area.
 

3.   Machine Learning:

As mentioned before the knowledge acquisition is the bottle-neck of the knowledge engineering process.  There are several attempts to automate this process.  Many of these attempts try to learn from examples using a machine learning paradigm such as neural networks and inductive learning.

 Neural network (or connectionist system) is constructed from a set of simple processing units (neurons) connected to many others in a network.  Each processing unit is capable of a few computations and keeps one piece of information. Processing units communicate with each other through connections with associated weights.  The network is capable of learning from the presented examples of a concept.  Neural networks are suitable for applications where there is little available structured knowledge.

 Inductive learning programs generate a concept by using the positive and negative examples of the concept.  Quinlan's ID3 algorithm is one of the several programs in this area.  ID3 generates a simple decision tree that generalizes the information in the examples.  The generated decision tree will also highlight the important features of these examples.  It is believed that when the problem is acquiring rules for expert systems, and there exits no strong domain theory, inductive learning is a suitable paradigm to select.  Contrary to inductive learning, it is difficult to obtain an explanation for the behavior of neural networks.
 

Please email me your comments and suggestions at sparisay@csupomona.edu

Last updated: September 1999