Research On On-line Monitoring Method Of Wear State Of No-return Valve On Injection Molding Machine

Predictive maintenance,as an important part of intelligent factory,is attracting more and more attention from injection molding practitioners.Predictive maintenance is the maintenance based on the state of the equipment,and the state information is used to determine the possible future failures.There are many kinds of faults in the long-term production of injection molding machine.The no-return valve is a key component in injection molding,which is prone to wear failure caused by the design process requirements of no-return valve,injection molding process parameter setting,physical properties of polymer raw materials,no-return valve types and other factors.Factories often delay the overall injection molding production plan because of the wear and failure of injection no-return valve,which affects the quality of injection molding products and injection molding production efficiency.With the rapid development of society and The Times,consumers' requirements for the quality of injection products and the society's requirements for injection productivity are increasing day by day,so it is necessary to realize the on-line monitoring of no-return valve wear degree to improve the injection production efficiency and product quality.The main work of this paper is as follows:Firstly,the movement of the no-return valve was simulated,and the simulation results of the injecting stage and the holding stage of the no-return valve with different wear degrees are analyzed in the non-deflection running state.However,in the analysis of its pressure field,it was found that the no-return valve will deflect because of the uneven pressure distribution.Therefore,the no-return valve with different wear degrees after the deflection was simulated again.Finally,the simulation results of the two operating states of the no-return valve are compared.The leakage degree of non-deflected no-return valve increases with the increase of wear degree,while the leakage degree of no-return valve decreases after deflection.Combined with the simulation work,the injection experiments were designed under different process conditions and different wear degrees,and the key signals such as screw position,load element pressure and the pressure of the die cavity near the far gate were collected.The characteristic values of these signals were extracted,including the pressure values of load element and near cavity at the end of injecting stage,the pressure values of far cavity at the end of holding stage and the forward displacement of screw at the holding stage.Through the analysis of the four characteristic values and the simulation results,it could be concluded that the no-return valve is in the deflection state during the injection molding process.The influences of different wear degrees of no-return valves on injecting stage,holding stage and on the quality of injection products were analyzed.The impact is mainly reflected in: With the increase of wear degree,the load element pressure and cavity pressure increased at the end of injecting stage,while the cavity pressure increased at the end of holding stage,while the screw displacement decreased.Product quality will be unstable with the increase of injection speed.Through the process data collected by the above experiment,a total of 57 data features including the above features were extracted and feature engineering was carried out,and 37 effective features were selected.Through the sample features after feature engineering,three classification models were trained,including logical regression,decision tree and k-nearest neighbor.The model was evaluated through the four dimensions including learning curve,confusion matrix,error matrix and accuracy.After the training of the model,it was found that the accuracy of k-nearest neighbors is the best,which was as high as 98.11%,which was much higher than the decision tree 85.28%.