Research On The Fault Diagnosability Evaluation And Design Of Affine Nonlinear System

The actual system mostly are nonlinear and hard to build the criterion of fault diagnosability using linear system theory,not to mention the random effects which noise jamming brings into the systems.There isn't any fault diagnosability evaluation and design method which is universally applicable to affine nonlinear systems.Based on differential geometry theory,this paper studies the fault diagnosability evaluation and design method for affine nonlinear system,affine nonlinear control system and affine nonlinear system under noise.The main contents are arranged as follows:Fault controllability distribution and fault observability dual distribution are defined.For the affine nonlinear system with one fault at the same time,a differential geometry approach based on system decomposition is proposed on fault diagnosability problem of affine nonlinear systems.On this basis the influence of system control on the fault diagnosability is studied..For single output affine nonlinear system under noise,three quantitative indexes,similar degree,missing report rate,and fault detection rate,are defined based the difference of observability subsystems between faulty system and non-faulty system.Comparing the observability subsystems between two fault,repeat rate and fault isolation rate are defined to measure how difficulty to separate the faults.For multiple output system,we can analyze its fault diagnosability by converting it into multiple single output systems.The problem about optimized configuration of minimum number sensors meeting the fault diagnostic is studied.A method is proposed to establish the directional graph of fault propagation.An improved greedy algorithm is offered to achieve the goal that system can still satisfy with fault diagnosabllity when a sensor is broken.An optimization model,with the constraint of maxing fault controllability and observability subsystem,is proposed.And an improved pagerank algorithm is designed to optimize sensor configuration by globally assessing the independence of every measure point.The proposed methods are applied to point mass satellite system.And the effectiveness of these methods is verified again.