Capturing cause and effect relationships based on distance correlation

Capturing cause and effect relationships has diverse applications in industrial systems; however, it is a challenging task due to complicated dynamics as well as the presence of many variables in industrial plants. This paper, presents a causality detection method based on distance correlation. The proposed method utilizes distance correlation to capture cause-effect relationships. Moreover, a causality measure index is proposed for detecting direct/indirect cause-effect relationships. Furthermore, in order to reduce computational burden and required memory size for the prevalent empirical distance correlation, the recursive empirical distance covariance (REDC) is presented. Compared to conventional causality detection methods, the proposed method provides acceptable performance for causality detection in linear and nonlinear systems. Simulation results confirm that the REDC algorithm significantly reduces computational time compared to the direct method. In addition, the effectiveness of the proposed method in causality detection is verified by simulation results in experimental and industrial benchmarks.

Automated summary

This paper proposes a causality detection method based on distance correlation, utilizing causality measure index and REDC to reduce computational burden. The method provides acceptable performance for causality detection in linear and nonlinear systems, significantly reducing computational time.