Diesel Engine Fault Diagnosis Based On LMD And Neural Network

The diesel engine is the most common mechanical equipment, as the coreportion of cars, ships, aircraft, etc. The normal operation of the entire apparatus isinfluenced by its behavior. It’s great practical significance to found fault in a timelymanner, quickly diagnose the fault, and troubleshoot the fault.LMD as a new time-frequency analysis method can separate signal to a series ofPFs which have physical significance. It’s better in suppressing extreme point effectthan EMD. This paper put forward a new way to replace the Mirror continuationmethod which is attenuation continuation method by studying the affection of slidingspan, the effect of extreme points, etc to LMD. At the same time, this part putforward a concept of Antidumping Signal to solve the influence of LMD’s endeffect.The signal noise reduction is an important step in diesel engine fault diagnosis.A new de-noising method is brought up which is LMD-WD. This new method iscompared to Singular value differential noise reduction, wavelet de-noising andEEMD-WD. This part uses the four noise reduction method to vibration signal of thediesel engine in various working conditions, and compares these four methods withindexes of signal-to-noise ratio and root mean square error.This paper traverses the procedure diesel engine fault diagnosis based on LMDand neural network. Five working conditions’ vibration signal of a V12diesel enginewhich is normal operating conditions, the leakage of left1-cylinder fuel injectionpump, FCO of Right6-cylinder,2.5degrees increasing of fuel supply advanceangle,air filter plugging,is disassembled to a series of PFs by the improved LMD, and first five PFs’energy make up to feature vector for RBF training. Finally usethe trained RBF to test feature vector of various working conditions.