Fault diagnosis of PTA oxidation process based on decision tree and ant colony optimization

Data mining technique can extract desired knowledge from existing databases and ease knowledge acquisition bottleneck of fault diagnosis. A new knowledge acquisition method combined of decision tree and rough set theory is thus proposed for fault diagnosis in this paper. Based on the reduction by rough set theory, decision tree extract diagnostic knowledge from the reduced decision tables in the form of symbolic trees. Trend correlation degree, as the heuristic knowledge, is proposed to evaluate the significance of condition attributes for the construction of decision tree model. Also a modified ant colony algorithm is introduced to determine the optimal fault test sequence of the decision tree model. The proposed method is applied to the fault diagnosis of purified terephthalic acid (PTA) oxidation reactor, which is the key unit in AMOCO PTA production technique. The results show the method is satisfactory and suitable for industrial application.