Research Of Fault Diagnosis Of Fans On Neural Network

Fans play an important role in industrial production.It not only hasa direct impact on the entire production line,but also can cause significant economic lossesand even disastrous accident when it’s fault occurs.In order to ensure these equipmentsoperate safely,reduce maintenance costs and improve the equipment utilization rate havebecome an important research area to develop fault diagnosis method that canautomatically obtain knowledge and reason in a high speed in the study of the Fan.This paper select fan fault diagnosis and condition monitoring for research. ChoosePDM2000Date Collector/Analyzer to acquire the vibration signal of the fan and thecharacteristic frequency.According to the spectral analysis method of equipment vibrationdiagnosis technology,we analyze the fan fault symptoms and obtain the results of the faultdiagnosis.In the same time, select the analysis method of BP Neural Network to do furtheranalysis on the fault diagnosis of fan.According to the structure and algorithm of BP neural network,Use three methods toimprove the BP neural network algorithm.Through the actual measurement of theoperation and compare three kinds of improved algorithm,and select the operation speedis faster,more accurate judgment of Levenberg-Marquardt algorithm for training the BPneural network analysis.Use the Matlab software to train the characteristic whichmeasured from the factory.Use the BP neural network which has been trained to test it,andcome to the conclusion that the fan also exsist the fault of rotor imbalance,rotorrub-impact and mild rotor misalign.The diagnosis in conformity with the analysis resultsof the actual measurement data.In this paper,through the comparation of the actual measurement of the fan vibrationdata analysis results and the theoretical calculation results, the results show that thediagnosis method which is based on the inproved algorithm BP Neural Network of the fanhas certain practicability and feasibility.