Fault Diagnosability Evaluation for Nonlinear Networked Control Systems Under Weak Transmission Conditions

This article is devoted to providing a method for quantifying fault diagnosability of nonlinear networked control systems under weak transmission conditions. The receiving signals of fault detection and isolation systems are characterized under packet dropouts and communication constraints with the help of Bernoulli random variables and a round-robin protocol. Then, the statistical characteristics of system dynamics under the impact of different faults are characterized by defined statistical functions combining probability density functions. Furthermore, a quantitative fault diagnosability evaluation framework for nonlinear networked control systems is established based on Takagi–Sugeno fuzzy models and Kullback–Leibler divergence. Moreover, an evaluation metric is proposed, which can be applied to system dynamics following both Gaussian distribution and mixed distribution consisting of Bernoulli distribution and Gaussian distribution, eliminating the constraint of assuming purely Gaussian distribution. In addition, a necessary and sufficient condition is given to reveal the relations between the qualitative fault diagnosability evaluation results and the proposed metric.