Expert system of cold forging defects using risk analysis tree network with fuzzy language

The authors have developed an expert system for detecting the risk of forging defects and their causes in cold processes. The system employs risk analysis for a computer-aided process planning system. In addition, the authors investigated topics that affect typical forging defects. Based on the investigation, the authors developed risk analysis tree networks to evaluate the risk potential and describe the risk potential by fuzzy language [G. J. Schmucker, Translated to Japanese by T. Onizawa, Fuzzy Set, Natural Language and Risk Analysis, Keigaku-Shuppan (in Japanese)]. The risk analysis tree networks are developed for each kind of typical defect and the location on the forging part at which they occur. The risk analysis network is a hierarchic network, and its nodes represent the topics that affect forging defects. The value of the node represents the “risk” (e.g., the “certainty the topic comes true”). First, the computer-aided process planning system uses rules in its knowledge base and the result of FE analysis for the estimation. Second, the risk rate of the upper node is determined by the average rates of its daughter nodes with weight values. Thus, the risk that a certain defect occurs at a certain location is estimated through executing the above procedure from the bottom to the top. Finally, the expert system displays the nodes that have higher risk values. They are considered the main causes of defects. The system can rapidly indicate rough but appropriate causes and the location of forging defects.