Transfer Learning Method For Crankshaft Clearance Fault Diagnosis Based On Virtual Prototyping

Reciprocating compressor is an important equipment in petrochemical industry,and crankshaft connecting rod mechanism is the most critical component.The connecting rod big end bearing shell and the crankshaft journal will lead to excessive clearance due to long-term wear and tear,which will cause crankshaft clearance fault and seriously affect the safe operation of equipment.Aiming at the problems of crankshaft clearance fault label data shortage,insufficient study on fault response characteristics and probability distribution vary between data,a transfer learning method for crankshaft clearance diagnosis of reciprocating compressor based on virtual prototype is proposed,which is studied from three aspects: motion pair dynamics modeling,fault response characteristics analysis and fault transfer diagnosis.The main contents are as follows:(1)Aiming at the lack of fault label data,the method of obtaining fault label data by dynamic simulation of virtual prototyping machine is proposed.The reciprocating compressor test-bed is used as a physical prototype to simplify the structure of the motion pair.Then a series of three-dimensional solid models of the motion pair with different crankshaft clearance values are established by using Solid Works 2012 software.The constraint conditions,piston load,solver,integrator and contact parameters of the motion pair dynamic model are determined by analysis.The dynamic simulation of crankshaft clearance fault under different working conditions is carried out by using ADAMS 2013 simulation software,and the simulation data of the fault are obtained.(2)Aiming at the insufficient study on fault response characteristics,the method of fault response characteristics analysis for crankshaft clearance based on simulation and experiment is proposed.Firstly,the crankshaft clearance fault experiment is carried outon the reciprocating compressor test-bed,and then the simulation results are compared with the experiment results to verify the correctness of the simulation results.At the same time,the time domain and the frequency domain analysis of the simulation signal show that the increase of the clearance will aggravate the contact friction between the big end bearing shell and crankshaft pin,and the peak value in time domain and the energy in frequency domain of 1k Hz-2k Hz and 3k Hz-5k Hz will raise with the increase of clearance value,which indicates that the response characteristics in time domain and frequency domain can reflect the generation and development of crankshaft clearance fault.(3)Aiming at the problem of differences between data,a fault diagnosis model based on the combination of transfer component analysis(TCA)and support vector machine(SVM)is proposed.TCA-SVM diagnosis model is trained with simulation data,and the trained model is tested with experimental data.The results show that the proposed method can better eliminate the differences of data distribution between different data sources and different working conditions,and the accuracy of crankshaft clearance fault diagnosis is improved.The accuracy reaches 87.19%,which is obviously superior to those of four contrast methods such as SVM direct classification.It realizes the transfer diagnosis of crankshaft clearance fault from simulation data to experimental data and under different working conditions.