Enhancing injection molding barrel temperature control performance using a data-guided simplex search method

Abstract

The barrel temperature control system is one of the key components for process control of the injection molding machine, with its performance heavily influenced by the control parameter settings. However, the tuning process for these parameters is often both costly and cumbersome. Currently, data-driven techniques for parameter tuning are increasingly widespread, but the existing methods generally fail to exploit the information embedded in the previous iterations or datasets. To improve optimization efficiency through more effective data utilization, a Data-Guided Simplex Search method based on adjacent historical Centroid information (CDG-SS) is proposed. By reformulating the simplex iteration mechanism to establish the concept of quasi-gradient estimation, this method uncovers the intrinsic similarity between the gradient-free simplex search algorithm and conventional gradient-based methods in terms of their shared approximate gradient search properties. Building upon the concept of quasi-gradient estimation, this method utilizes historical centroid data from adjacent simplices to identify the current trend states of the optimization progress. Based on these states, a dynamic compensation mechanism is then designed according to distinct trend states, enabling adaptive adjustment of the optimization step sizes. This approach thereby improves the efficiency of the barrel temperature parameter tuning for injection molding machines. The simulation results demonstrate that the CDG-SS method significantly improves the efficiency of optimization for control of the barrel temperature. Compared to the original simplex search method, CDG-SS reduces the average number of iterations required for the Integral of Time multiplied by Absolute Error (ITAE) by 16.6% and for steady-state error by 12.1%, while maintaining comparable accuracy.