Knowledge modeling for root cause analysis of complex systems based on dynamic fuzzy cognitive maps

This paper proposes a knowledge model for root cause analysis (RCA) of complex systems based on fuzzy cognitive maps (FCMs) and particle swarm optimization algorithm (PSO). The process knowledge and experience of technicians can be captured by FCMs that are characterized by briefness of knowledge modeling and execution. The traditional methods for RCA based on FCMs are restricted to fixed incidence matrix. However, the individualized features are there existing in each system of the same kind, therefore fixed weights are unreasonable. PSO is introduced to detect the weight that can reveal the individualized features of systems among concepts of FCMs. And then a dynamic knowledge model for RCA is obtained, including predictive, diagnostic, and hybrid RCA. The three types RCA can be used for forecasting future event of output, identifying root cause and presenting measures of abnormal event. The effectiveness of proposed method is validated in aluminum reduction process, and the experiments results show the proposed method is effective and application potential.