Fault Diagnosis of Tractor Start System Based on Structural Analysis Method

Abstract

In order to achieve accurate positioning of the electrical failure of the tractor starting system, this paper designs a fault diagnosis system based on structural analysis. Firstly, through the analysis of the traction battery and the starting motor, this paper selects four key faults: battery resistance fault, motor armature resistance fault, motor armature fault and motor joint fault, and establishes the electrical fault model of the tractor starting system. Secondly, this paper divides the variables in the fault model into known variables, unknown variables and fault variables. Then the MATLAB dmperm command is used to analyze the observability and isolation of the unknown variables. We find that we can detecte and isolate four faults by adding a voltage sensor based on the original engine speed sensor. Then select three equation sets as the test set in all the structural minimum set of over-determined equations (MSO sets), and design the residual value of each test set as the observer of the system. The faults detected by different observers are different, but 3 observers can detect 4 faults. Finally, we design the diagnostic strategy of the system: we select [-1,1] as the feasible domain of the observation. If the value of the observer exceeds the feasible domain, the fault characteristic parameter is 1, otherwise it is 0. The fault feature vector is composed of three fault characteristic parameters and used as system fault judgment. In order to verify the correctness of the diagnostic system design, we use MATLAB to build the system simulation platform, which generated 4 faults in 3s-4s, 6s-7s, 9s-10s, 12s-13s respectively. The simulation results show that the results of the three observers are in line with the pre-test. It is determined that the fault feature vector can accurately locate the critical fault. This also shows the advantage of structural analysis for the electrical fault diagnosis of tractor starting systems, which can accurately locate system faults with only a small number of sensors.