APPLICATIONS OF ROUGH SETS THEORY IN CONTROL OF FOUNDRY PROCESSES

摘要

In contemporary manufacturing industry many varied methods are used for controlling process parameters, ranging from paper SPC charts to automated closed loop systems. In recent years a remarkably increasing interest in application of supporting systems based on computational intelligence methods, utilizing company recorded data, is observed. Most frequently, methods of automated knowledge extraction from data in the form of logic rules are utilized, based on classification learning systems. A growing interest of industrial applications of rough sets theory (RST) is observed in that area, which can provide with various types of important information concerning complex manufacturing processes. In the present paper the main elements and characteristics of RST are expounded and compared to another learning system – widely used decision trees (DTs). Current applications of the both methods in manufacturing industry are briefly reviewed. Results of the paper authors' research concerning two important issues of RST and DTs applications in foundry production are presented. They include assessment of correctness of relative significances of process parameters of arbitrary nature (e.g. physical, human, organizational etc.) and evaluation of reliability of engineering knowledge in the form of logic rules. The numerical experiments, carried out on simulated and real data, related to foundry technology, have shown that RST can be a valuable tool for control of complex manufacturing processes and it performs remarkably better then DTs.