Interval Filtering Based Fault Diagnosis Under Noise Uncertainty

For the case where the noise of a system to be diagnosed is uncertain,noise is assumed to meet certain probability distribution conditions in the traditional research.However,as the complexity of the system increases,it is assumed that the noise in the bounded interval is more suitable for practical application requirements.In order to reduce the amount of calculation and improve the accuracy of interval estimation and fault diagnosis results,the research on the actuator fault diagnosis method based on interval filtering is carried out in this paper.It has significant value for theoretical research and practical application in the field of fault diagnosis.The specific work of this paper includes the following three aspects:1.For a linear time-invariant system with unknown but bounded noise,the aug-mented state vector is composed of the state vector and the actuator fault firstly,and the fault estimation problem is transformed into an augmented state estimation problem.The H_∞technique is used to realize the augmented state interval estimation based on in-terval filtering.In order to reduce the wrapping effect of interval calculation,the interval set inversion filtering algorithm is used to further contract the fault interval.The batch operation of the interval is realized by the Boolean operation of the vector,which effec-tively reduces the computational complexity.And the multi-time measurement output data makes the fault tracking effect better.The simulations of two-tank system and DC motor system have verified the effectiveness of the vector interval set inversion filtering based fault estimation algorithm.2.For the inherent high conservatism of interval algorithms,zonotope is selected as the feasible set of state,and the minimal conservative interval observer is designed by using the properties of the zonotope and the state error minimization idea.In order to further contract the fault interval,dimensional vector set inversion interval contraction algorithm is proposed.Interval bisection in a fixed dimension can reduce the number of bisections in the contraction process effectively,and reduce the number of intervals for batch processing at the same time.The proposed algorithm is compared and analyzed from the perspectives of computational complexity,memory requirements and solution set relations.The simulation of three-dimensional DC motor verifies the effectiveness of the interval observer and contract algorithm.3.For the dual uncertainty linear discrete-time system with mixed Gaussian noise and unknown but bounded noise,a novel interval observer filtering based fault diagnosis algorithm is proposed.The weight ratio of the two kinds of noise is determined according to the signal average power principle.And the interval filter observer is designed by the optimization problem based on state error,so that the obtained real-time state estimation results can contain the real state of each time instant more accurately.Considering the property of Gaussian noise,the interval indicator for fault detection is designed according to the residual probability ratio to realize dynamic detection.And the accuracy of fault estimation has been improved by means of sliding window data.The effectiveness of the algorithm is verified by simulation of numerical system and DC motor system.