Font Size: a A A

Study On Plunger Pump Fault Diagnosis Method Based On Deep Learning

Posted on:2021-02-03Degree:MasterType:Thesis
Country:ChinaCandidate:M Z DuFull Text:PDF
GTID:2392330614455165Subject:Mechanical engineering
Abstract/Summary:PDF Full Text Request
Hydraulic pump is energy source for hydraulic system for its compact structure,less leakage,high volume efficiency,high pressure and easy control of flow,which is unfortunately a fault source as well.Because of its concealing,diversity and complex,it is difficult to discover the loss of primary function.Therefore,it is necessary to extract and diagnose the typical fault characteristics of axial piston pump.Firstly,the structure and working principle of plunger pump were analyzed to understand the location and characteristics of common faults on plunger pump,the signals of those faults were obtained by the acquisition system.Then,one-dimensional and two-dimensional convolutional neural networks were utilized to diagnose piston pump,respectively.The results stated that both can diagnose plunger pump faults effectively,but the latter can reduce convolutional layers,and the training time is shorter,which encourages the application of two-dimensional convolutional neural network,so the fault diagnosis performance is better than onedimensional convolutional neural networks.The average fault diagnosis accuracy rate was 98.26%,and the generalization and robustness of the CNN-2D model was verified by the rolling bearing fault data set.Following that,EMD and EWT algorithms were used to decompose fault signals into limited intrinsic mode function,which would appear in the grid of two-dimensional convolutional neural network latter.The results indicated that EWT combining with twodimensional convolutional neural networks outperforms EMD with less layers and the accuracy of 100%.The last but not least,the LSTM model in the circulating neural network was used for the fault diagnosis of plunger pump,whose parameters were researched as well.Based on the investigation,a conclusion could be made that LSTM also works for fault diagnosis.Diagnosing and classifying plunger pump faults with deep learning methods are of importance in the working status monitoring,theoretical and application research on plunger pump.Figure 46;Table 11;Reference 63.
Keywords/Search Tags:fault diagnosis, deep learning, CNN, LSTM, plunger pump
PDF Full Text Request
Related items