Dynamic artificial immune system with variable selection based on causal inference

Abstract

Abstract Large numbers of variables are measured in typical chemical processes. In most of the fault detection and diagnosis methods, all the measured variables are selected. However, the incorporating measured variables that do not provide any additional information about faults degrades monitoring performance. Dynamic artificial immune system (DAIS) is a new artificial intelligence methodology shows strong ability of self-learning and self-adaptability. Much work has been done on its application in process monitoring and fault diagnosis in chemical processes. However, there is little research in the variable selection in DAIS. In this paper, we propose an approach to select variable used for DAIS. The approach is based on the causal inference among the measured variables. Through causal inference, the cause-effect relations among variables can be found and then variables that do not reflect the main trends of changes of the processes can be picked out. Then the variable selection of the DAIS can be optimized and the performance of the fault diagnosis of DAIS can be improved. Case studies based on the Tennessee Eastman process are performed to illustrate the effectiveness of our approach.