Introduction to expert systems, artificial intelligence and decision analysis in mechanical engineering. Fundamentals of logical inference, predicate calculus, multivariate logic, probability theory, diagnostic reasoning, risk assessment, qualitative reasoning, and analytical design. Applications to expert systems in automated manufacturing, mechanical engineering design, real time monitoring and supervisory control, and failure diagnostics. Use of automated influence diagrams/Bayes’ networks to codify expert knowledge, perform probabilistic inference, and to evaluate optimal control or design decisions. All theory will be presented with engineering applications.
Week Lecture Topics
1-17,19
1 Overview of artificial intelligence, knowledge engineering, and expert systems. (Read Sects. 1.1-1.3 of Dym & Levitt.)
1-24,26
2 Historical perspective of computer intelligence. Introduction to expert systems. Summarize applications in mechanical engineering design, monitoring and supervisory control and failure diagnostics.(Read Sects. 1.4-1.5 of Dym & Levitt, Chaps. 1A, 2 from the Reader; selected publications.)
1-31,2-2
3 Discuss advantages and limitations of expert systems. Selection criteria for rule-based systems. Artificial intelligence, problem formulation and heuristic search techniques: backward and forward chaining, Hill-climbing, generate-and-test, depth-first search, breadth-first search. (Read Chap. 2 of Dym & Levitt and Chap. 1B of Reader.)
2-7,2-9,2-14,2-16
4,5 Best-first search, branch-and-bound, A* algorithm. Deduction with formal logic; propositional logic. Introduction to first-order predicate logic: syntax, interpretation, and representation. Meaning, truth, and logical implication. Quantification and concepts of logical inference: unification, instantiation, resolution, and completeness. (Read Chap. 3 and Sects. 5.1-5.2 of Dym & Levitt and Chap. 3-7 of Reader.)
2-21,2-23
6 Languages for symbolic computation. (Read Chap. 4 of Dym & Levitt and selections from Wilensky.)
2-28,3-2,
7 Measures of uncertainty, knowledge representation using other logics: nonmonotonic logic, fuzzy logic, certainty factors and subjective probability. Review basic concepts and theorems of probability theory using event sets and event algebra. (Read Sects. 5.5-5.7 of Dym & Levitt and Chap. 8-9 of Reader.)
3-7,3-9, 3-14,3-16
8,9 Conditional probability and Bayes’ Theorem in event algebra. Dependence and independence and influence diagram/Bayes networks representation. Discrete probability functions. Conditional and joint probability distributions. Definition of moments. Bayes’ Theorem. Expansion and probability trees. (Read Chap. 10-11 of Reader and introductory chapters of Raiffa.)
3-21,3-23
10 Application of probabilistic inference in expert systems for failure diagnostics. Review for midterm. Midterm
3-28,3-30 Semester Break
4-4,4-6
11 Knowledge acquisition, subjective probability and knowledge engineering. Assessment of expert opinion. Influence diagram based expert systems. Applications to real time monitoring, diagnostics and control of mechanical systems. (Read journal publications provided in the Reader).
4-11,4-13
12 Addition of control variables. Decision analysis and decision trees. Deterministic and stochastic dominance. Expected value optimization. (Read Raiffa and selections from the Reader.)
4-18,4-20
13 Value of information and additional testing. Value theory, expected utility, and certain equivalence. (Read from Raiffa and Reader.)
4-25,4-27
14 Application to intelligent real time problem solving and design. (Read journal publications provided in the Reader.)
5-2,5-4
15 Second-generation expert systems. Qualitative reasoning in design, databases/blackboarding, machine learning and misc. topics, time permitting. (Read Chap. 7-8 of Dym & Levitt.)
Final Term Projects: Project serves as a take-home final exam. Due on either last day of class or scheduled day of final exam (exam group 12; May 17), depending on class preference.
PREREQUISITES: Graduate standing, prior courses in probability or logic, or consent of instructor. A previous course in AI or expert systems is not required as this is an introductory survey course.
REQUIRED TEXTS:
Dym, Clive L. and Raymond E. Levitt, Knowledge-Based Systems in Engineering, McGraw-Hill, Inc., 1991.
Purchase of class Reader (Hybrid Expert Systems; selected papers and includes necessary pages from Raiffa); $30. The Reader will be broken into two parts. Part I will be available in class during week 2. Part II will be available in week 7.
RECOMMENDED READING:
Wilensky, Common Lispcraft, W.W. Norton & Company, New York, N.Y. (on reserve in the Engineering Library).
REFERENCE READING:
Raiffa, Howard, Decision Analysis: Introductory Lectures on Choice Under Uncertainty, Addison-Wesley Publishing Company, Menlo Park, California.
Genesereth, Michael R. and Nils J. Nilsson, Logical Foundations of Artificial Intelligence, Morgan Kaufmann Publishers, Inc., 1987.
GUEST LECTURERS: Guest lecturers from research labs and industry will be invited to describe specific AI/expert system applications.
HOMEWORK: One homework set every 3 weeks and one individual project. Late homework will not be accepted. Solution sets will be provided on the due dates.
EXAMS: One midterm; individual project in lieu of final examination.
GRADES:
Homework(4) 40%
Midterm 30%
Project 30%