Data-driven distributed collaborative fault detection and isolation for large-scale dynamic processes in simultaneous-fault cases

Abstract

This paper proposes a data-driven distributed collaborative fault detection and isolation scheme for large-scale dynamic processes, interconnected by heterogeneous subsystems with both actuator and sensor faults. The residual generator in each subsystem is directly constructed with both local and neighboring process data based on subspace identification. Utilizing the rank deficiency property, the healthy mode residual covariance matrix is estimated with noisy data, and thus the test statistic and detection threshold are obtained. Then, a collaborative simultaneous fault isolation method is developed, where the local residual is decoupled with neighboring actuator fault’s effects and a series of robust residual generators are constructed by solving corresponding optimization problems. In each subsystem, both local fault isolation and neighboring sensor fault isolation are realized. Thus, sensor fault detection and isolation for the whole process can be achieved with a reduced number of monitoring subsystems. Finally, case studies on the Tennessee Eastman (TE) process and a real hot strip mill process (HSMP) verify the feasibility and good performance of the proposed scheme.