Enhancement of a rule-based diagnosis system with BBN

• Main Author: Wang, Fei
• Other Authors: Chen Lihui
• Format: Final Year Project
• Language: English
• Published: 2010
• Subjects: DRNTU::Engineering
• Online Access: http://hdl.handle.net/10356/40820
• Institution: Nanyang Technological University

The author’s approach generates diagnosis model in the face of uncertainty in the relationship among device components and status, observations as well as the effect of actions on device status. A series of quantitative and qualitative approximations for problematic diagnosis under uncertainty are described. Included in our approach is a graphical probabilistic model for rule-based reasoning in diagnostics under uncertainty. The model utilizes Bayes’ Theorem and a special fishbone-structure Bayesian Belief Network (BBN) to correlate diagnosis cases, with “fish head” representing failure symptoms, “sub-bones” representing root causes and categories. Particular considerations are given to the design of the BBN model structure, determination of prior and conditional probabilities, and diagnostic procedures for both single and multiple symptoms. The proposed model is capable of guiding the diagnosis with a probability assignment and suggesting possible recovery actions. The model has been constructed to a software assistant tool for diagnosing manufacturing device and the results show that the model can support decision making promptly.