Research On Fault Diagnosis Of Complex AGV System

The advent of the era of industrial intelligence puts forward higher requirements for Automated Guided Vehicles(AGV).AGV have independent decision making capabilities,which enable them to work safely and autonomously in unknown and complex environments.However,the failure of AGV during operation can have disastrous effects,which can lead to mission failure.Therefore,faults need to be detected and isolated correctly and quickly in order to avoid catastrophes.In recent years,the model-based fault diagnosis has received more and more attention.The key idea is to convert the process of the system operation into a model,and compare the estimated output of the model with the output of the real system to diagnose the fault.The fault diagnosis based on the linear model is not applicable because of the non-linear non-Gaussian characteristics of the AGV system.A tool for solving the fault diagnosis problem of AGV nonlinear complex models,namely the particle filter,has proved to be a very powerful method to deal with nonlinear nonGaussian system problemIn this paper,the fault diagnosis based on improved particle filter is studied combining the characteristics of complex AGV system and related fault diagnosis theory.Finally,the fault diagnosis is effectively applied to complex AGV system.The major work and the results of research are presented as follows:(1)The kinematics model of AGV is analyzed in detail.The common faults of AGV and the difficulties in AGV fault diagnosis are summarized.(2)An improved particle filter using butterfly algorithm is proposed.The key idea is to optimize sampling process of the particle filter through the improved butterfly algorithm,driving the invalid particles to the regions where they have larger values of posterior density function and enhancing the role of particles.Simulation results show that the improved particle filter has better performance of state estimation than generic particle filter.(3)A new method based on improved particle filter and residual is proposed for fault detection in nonlinear complex systems.Firstly,the estimated measurement of nonlinear system is obtained through the improved particle filter,and then the residual is formed between the model's estimated measurement and the system's actual measurement.Finally,the fault detection is performed by evaluating the residual.Simulation results show that the proposed fault detection method is superior to other methods when the miss alarm rate and false alarm rate are comprehensively measured.(4)In combination with the AGV system,fault isolation steps are perfected to form a complete AGV fault diagnosis.Finally,the fault diagnosis is applied to the AGV system in the simulation environment.The simulation results show that the proposed AGV fault diagnosis can effectively detect and isolate faults.