Load Identification Research Of Mine Hoist Based On SVM

Mine hoist is an important part of mining machinery,which plays an significance throat-role in the mining production.Because of the complexity of the mine production environment,mine hoist is usually worked under complex loads,which include the normal loads of loading work and the dangerous loads of malfunction.These are closely related with the safe operation of the unit.However,limited by the working environment,the loads usually can't be measured by direct method.So the load identification method is proposed,the load is identified by the dynamic response under load,which is relatively easy measured.In order to put forward effective load identification method,it is necessary to accurately understand the dynamic response of the mine hoist.During loading,the characteristics of mine hoist running is first influenced,then the running current is affected,as well as the vibration of the fixed support.Considering the complexity of the electric current signals and vibration signals,as well as the error of transitivity,the load identification based on running characteristics is put forward.Based on transfer matrix method assumes,a reasonable mine hoist physical model is simplified.With the principle of conservation of energy and power,the rotational inertia,stiffness and torque is done equivalent conversion.With the principle of the conservation of magnetic field and power,the asynchronous motor electromagnetic model is decoupled.And then the MATLAB/Simulink electromechanical coupling model and vector control simulation model is established.Accordingly,the normal operation and failure typical load(steady state,impact,transient,linear,sine load)operating characteristic curve of mine hoist can get from the simulation models.Then it is found that speed characteristics and load characteristics have a consistent relationship,the greater the load,the greater the speed reduce in numerical value.In sinusoidal load,speed frequency is the same as the load frequency,and the greater the load amplitude,the smaller the speed amplitude.These features,can be used to bypass system characteristic parameters.By taking advantage of speed change features as independent variables and load characteristics as the dependent variable,the load identification of mine hoist can be realized.In order to accurately identify the load,compared traditional regression forecast method and the new regression forecast method,the SVM load identification method is introduced.The SVM is theoretical perfection,has good generalization ability;need to adjust less parameters,has strong robustness;only need a simple optimization technology,the calculation is simple;always belongs to one of the best ways in the solution of practical problems.Therefore,SVM model is used to identified load in this article,and combined with the typical representative of global optimization algorithm-Grid search,directly heuristic algorithm-GA search and based on groups heuristic algorithm-PSO search.Identification results show that the good recognition effect is often get from the SVM model Grid and PSO optimized,and the Grid-SVM identification effect is superior to the PSO-SVM generally,which is related to the distribution of data in this article.The largest absolute error of the identification result is always occurred on the predictions of the minimum load,and presenting a convergence trend from maximum absolute error to the minimum.In data processing,no matter no-feedback or feedback simulation model in SVM load identification,the better handling of the independent variable and dependent variable ways are all not normalized and independent variables normalized,dependent variables not normalized.But the independent variable normalization or not normalized has little effects to the recognition,the most optimal way can be chose from the testing of the two method,and the PCA dimension also has little effect.During the SVM load identification in feedback simulation model,it is found that the size of the dependent variable has larger effect on the final recognition performance,and through the proper proportion of change,can greatly reduce the absolute value of the maximum relative error,which has little effect on the overall recognition,even can optimize the overall recognition effect.Finally,designing the test-bed,collecting the data under different loads.By the method of test,the effectiveness of the SVM load identification method in the engineering practice is also proved.Above all,the research can provide a theoretical basis on load identification and fault diagnosis of mine hoist.