Research Of Incipient Fault Diagnosis Of Ammunition Supply System Based On Information Entropy And Information Fusion

As one of the core components of the artillery,the ammunition supply system is a complex mechanical,electrical and hydraulic integration of high-speed running system,which needs to complete ammunition transportation,automatic filling and firing tasks in high temperature,high pressure and high impact environment.It is one of the subsystems with a high frequency of faults,and its reliability has always restricted the performance of the artillery.Therefore,it is of great value and significance for its early fault diagnosis.In this paper,the working principle and common faults of the ammunition supply system are analyzed,the modern signal processing method is used to filter.And adopting analysis methods to denoise and extract the weak features of information entropy,and the information fusion technology is used to diagnose the early faults of the ammunition supply system.Firstly,the composition,structure,working principle and common faults of the ammunition supply system are summarized and analyzed.On this basis,the arrangement of sensor points and multi-field information collection are carried out.Through the research and analysis of the measured multi-field information signals,the vibration measuring point 3,the vibration measuring point 5 and the sound pressure measuring point 2 are selected for the normal,deteriorating and faulty working conditions.A large amount of noise is often mixed in the original signal,so the adaptive generalized morphological filtering is used to denoise the signal.Information entropy has excellent characteristics.After analyzing the concept,definition and properties of information entropy,a method based on Approximate Entropy and Sample Entropy weak feature extraction is proposed.Approximate entropy and sample entropy extraction were performed on a total of 194 projectile signals during the shooting.After that,the Elman neural network was used to identify the early faults of the ammunition supply system and the initial diagnosis results were obtained.The correct rate of fault diagnosis for a single measuring point is not very high.In order to fully reflect the operating state of the ammunition supply system and improve the accuracy of diagnosis,a multi-field information fusion method based on D-S evidence theory is proposed.The output of the Elman neural network is normalized as the basic probability distribution of the evidence body.In order to avoid the conflict between the evidence bodies,the weighted average fusion model of the evidence correlation coefficient is used,and then the information fusion of decision-level is carried out,the final decision diagnosis result are obtained.The analysis of the diagnosis results shows that the adaptive generalized morphological filtering,information entropy,Elman neural network and information fusion theory are applied to the early fault diagnosis of the ammunition supply system,which is effective and superior.After the evidence-weighted average information fusion,149 of the 159 test samples were correctly diagnosed and identified with an accuracy rate of 93.71%,which indicates the method proposed in this paper provides a new idea and has certain practicability for the early fault diagnosis of the ammunition supply system.