Research On Optimization And Control Of Injection Molding Manufacturing Process

Injection molding is a major plastic forming process in which plastic pellets are molten and injected into a mold of desired shape where it cools and hardens to the final product. It is ideal for producing plastic parts with complex shape. Some advantages of injection molding are high production rates, minimal scrap losses, repeatable high tolerances and low labour cost. It’s one of the most important problem that how to set the injeciton molding prcess parameters for optimizing the part quality index that users care. While these index are influenced by combined effect of process pareameters of each stage of injection molding. Mold trial is the procedure mostly used in practice to set processing parametes and few of the parameters are under closed-loop control, and even if those are, the steady-state accuracy and response time of them can hardly meet the needs. Therefore, the optimization and control of process parameters are mainly discussed to develop effective and practical control strategies and optimization methods for injection production with high quality and repeat accuracy in this thesis. The main topics in this thesis are listed as follows.Process parameters control for injection molding processing:(1) According to the control requirements of barrel temperature, open-loop control based on Time Optimal Control (TOC) is adopted during initial heating-up period to minimize the start-up time from initial temperature to the setting temperature, while adaptive Multivariable Generalized Predictive Control (MGPC) is used during steady-state period around the operating temperature to improve steady-state precision, and these two methods complement each other to obtain excellent control effect.(2) According to the characteristics and control requirements of injection velocity, a feedforward and feedback control method is designed, in which ILC is used as feedforward part and proportional control as feedback part. In order to overcome bad learning transients in ILC, a zero-phase filter has been employed to filter the feedforward part of control value. Finally, monotonic decay of learning transients has been achieved and so has the accurate velocity profile tracking.(3) As for the filling-to-packing switch point setting, a detection method of filling-to-packing switch point based on wavelet transform modulus maxima has been presented, according to the jump feature of cavity pressure at this point. This method can bring about a realtime and accurate detection and it also has the characteristics of high degree of automation, little influence from measument noise and variation in the operating conditions.Process parameters optimization for injection molding processing:(1) As for the optimization of injection velocity profile, firstly, a criterion that cavity entrance pressure following a ramp with time has been proposed as base of optimization to guarantee uniform filling of thin-wall mold cavity. And then a neural network model has been developed for modeling the dynamic relationship from injection velocity to cavity pressure, based on which the profile problem is transformed into an optimization problem for the predicted cavity pressure to track a given ramp. Subsequently an optimization strategy with the properties of automatic division of velocity profile and iterative optimizaiton of decision variables has also been presented to execute offline optimization. Finally, aimming to avoiding the effect on optmization result from model mismatch, on-line profile correction has been adopted, with which actual cavity pressure can be corrected to approach the given profile from batch to batch.(2) As for the optimization of packing profile, and the single-and multi-objective optimization problems encoutered in practical production, a step-by-step optimization method based on the kriging surrogate model and statistical improvement criterion has been employed. In single optimization, a modified expected improvement criterion has been proposed to improve optimizaitoin precision and global searching capability at the expense of mildly increased infill data, and then it is applied to the optimization of packing profile of injection molding process for obtaining best shrinkage evenness of molded part. While in multiobjective optimization, expected improvement criterion of Pareto front has been used to provide direction in which additional training samples could be added to find trade-off solutions between optimal shrinkage average and evenness, and meanwhile a data preprocessing method has been put forward to avoid the negative effects brought about by the singular points. Simulation results of the above both applications show that a much better shrinkage quality of molded part can be achieved using the proposed method with smaller number of trials.