Research On Fault Diagnosis Of One-way Valve Of Diaphragm Pump In Slurry Pipeline Transportation System

The diaphragm pump is the core power equipment of slurry pipeline transportation,and its operation status affects the transportation efficiency of mineral raw materials and the production efficiency of enterprises directly.The check valve is one of the core components of the diaphragm pump,which has good tightness and pressure resistance,and its safe and stable operation guarantees the efficiency and safety of the diaphragm pump.The harsh operating environment and frequent reciprocating motion lead to the check valve easily damaged,and its failure is related to factors such as structure,material,slurry characteristics and the working conditions of the diaphragm pump.The fault diagnosis method and maintenance and replacement strategy adopted by metallurgical enterprises depend on subjective experience,and the reliability is not high.Besides,the collected vibration signal is a nonlinear signal composed of fault signal,multi-part vibration signal and noise,and the signal has non-stationary characteristics due to changes in the characteristics of the slurry and operating conditions.The complex characteristics of the signal presents a challenge to the fault diagnosis of the check valve.Information entropy can measure the complexity of nonlinear signal effectively,and twin support vector machine?TSVM?perform well in nonlinear classification,so this thesis studies the feature extraction and fault diagnosis methods of check valves based on entropy and TSVM.The main work and contributions of the thesis are summarized as follows:?1?It is difficult to determine the failure status and maintenance and replacement strategy of the check valve,so a new check valve fault detection method based on sliding dispersion entropy?SDE?and adaptive variational modal decomposition?VMD?was proposed in this thesis.Firstly,the introduction of sliding window downsampling and mapping function improved the trend and characterization capabilities of SDE features of check valve.Then,the fault warning point can be determined preliminarily by continuously updating SDE features and the status warning line.Finally,the adaptive VMD model based on the energy and correlation of the check valve vibration signal can further detect the fault state of the check valve near the warning point.The SDE can track the evolution of the fault state of the check valve in real-time,and can detect the fault warning point early,adaptive VMD can determine the fault status of the check valve at the warning point effectively.?2?In the fault feature extraction of check valve,the multiscale permutation entropy?MPE?has the defects of information loss and poor anti-noise,and the classification performance of TSVM is poor.Therefore,A feature extraction method of check valve based on improved multiscale weighted permutation entropy?IMWPE?was proposed,and a fault diagnosis model of check valve based on least squares twin support vector machine?LSTSVM?was constructed.Firstly,the introduction of composite coarse-grained and permutation mode weighting overcomes the shortcoming of information loss in check valve vibration signals.Then,the pre-filter based on improved VMD enhanced the anti-noise performance of the fault feature.Finally,the IMWPE fault features of the check valve were extracted and the fault identification was completed by LSTSVM model.The simulation results and fault diagnosis results of check valve show that IMWPE not only solved the problems of information loss and poor anti-noise performance of MPE,but also improved the stability and anti-noise performance of the check valve fault features.In addition,the LSTSVM model improved the accuracy of fault diagnosis.?3?The IMWPE has the shortcomings of equal amplitude and low efficiency,so we replaced the permutation pattern with the dispersion pattern and extracted the multiscale dispersion entropy?MDE?features of the check valve.In order to improve the stability and accuracy of MDE features,a feature extraction method of check valve based on improved multiscale fluctuation Rényi dispersion entropy?IMFRDE?was proposed,and a diagnosis model based on the optimal binary tree LSTSVM?OBT LSTSVM?was constructed.Firstly,the improved coarse-grained method enhanced the stability of the entropy feature of the check valve,and the introduction of Rényi entropy improved the accuracy of the entropy feature.Then,the IMFRDE feature of the check valve was extracted and the fault diagnosis accuracy of the check valve was improved by OBT LSTSVM model.The simulation results and fault diagnosis results of the check valve show that IMFRDE not only overcame the shortcomings of IMWPE but also improved the stability and accuracy of MDE feature,and OBT LSTSVM further improved the fault diagnosis accuracy of the check valve.?4?The environmental noise and fuzzy transition of fault states may lead to outliers in the feature samples of check valve,and the generalization performance of the model is not high.Therefore,a fuzzy regularization LSTSVM?FRLSTSVM?fault diagnosis model was proposed and combined with IMFRDE feature to improve the reliability of the fault diagnosis of check valve.Firstly,the IMFRDE feature of the check valve was extracted.Secondly,the L₂ norm regularization term was introduced into the objective function of LSTSVM,which solved the problem that the generalization performance of the model is not high.Then,the membership function S₃ was constructed based on the support vector domain description?SVDD?,and the problem related to the outlier was solved.Compared with the LSTSVM model,the average accuracy of the FRLSTSVM diagnostic model was increased,and the standard deviation of accuracy was reduced.In addition,the minimum accuracy was significantly improved.The results show that the FRLSTSVM fault diagnosis model based on SVDD outlier detection method and membership S₃ has better generalization performance and anti-outlier capability,lower sensitivity to parameters,and has a higher reliability.In this thesis,the diaphragm pump check valve for the slurry pipeline transportation is the main object of study.This thesis has completed the study of fault feature extraction methods and fault diagnosis methods,and provided new methods for fault diagnosis of mechanical parts in the metallurgical industry.