Based On Vibration Signal Neural Network Fault Diagnosis Technology

Fault diagnosis of diesel engines is difficult because of its complicated structure,higher occurrence rate of malfunctions and its severe working conditions. It's important to diagnose the faults in time so some measures could be taken to prevent the engine from unexpected accidents.This could greatly reduce cost spending on maintenance and unexpected losses accompanied with faults.In this way,we can insure that the engine could work under more secure and reliable situations and it's also of great significance in both research and social ecnomic values.Feature extraction is an important part of fault diagnosis, and determines whether or not the diagnosis would success. The vibration signal of diesel engine includes rich working status information and fault feature information, extracting characteristics parameters from the surface of the vibration signal can be effectively identify diesel engine working status and the fault. For the past few years,fundamental theories about fault diagnosis have been summarized.In general,the fault diagnosis is still in the stage of sudy and lacks actual practice.In this paper,on the basis of summing up previous research achievements and according to the real circumstances and the subject item requirement,a fault diagnosis method based on vibration signal and artificial neural network has been proposed.This paper gives a fault diagnosis method of diesel engine based on neural network analysis of vibration signals considering the current research on it home and aboard and then executes empiric test analysis on it. After analyzed the force of the vibration, a sound test system is designed, and the vibration signals are measured on the type WP7under four fault states of working conditions, which including valve clearance fault, the injection advanced angle fault,the injection quantity fault and the injection pressure fault. Making analysis on time domain and frequency domain of vibration signal of diesel engine in order to extract parameters of time domain, then the diesel engine state will be judged by contrasting the characteristics parameters of the different fault conditions vibration signal. Propose a method of judging fault diagnosis by the energy characteristics, which based on wavelet packet frequency band energy analysis technology, using this method to decompose and reconstruct the signal and then extract the vibration signal of fault characteristic. The BP neural nets are established used the software of MAILAB, extracting feature parameters and putting the samples into BP neural network for training and testing to achieve the desired effect. One swatch is used to prove in each states, point lays and working conditions, and the neural nets could draw right judgments. So it is said that the method is feasible. The method of extracting fault symptom vectors from the vibration signal of diesel engine is simple, so it is easy to realize the fault diagnosis, even for diagnosis without disassemble. The method of using artificial neural network in diesel fault diagnosis had good reliability of identification to fault condition.lt will make more effects to the development of the fault diagnosis of the marine diesel engine as the development and improvement of the neural network theory.