Abstract
Although process control has been extensively studied in injection molding, the process still relies on experienced engineers to tune the processing parameters on the shop floor today. Hence, due to the inconsistent product quality, excessive manpower is still needed for meeting production requirements in the industry. In this paper, an optimization algorithm based on two-linked Radial Based Function Networks (RBFN) models with multi-losses criteria capable of controlling the part quality has been studied. The objective is to automatically control the process with little operator intervention. First, processing parameters, geometry of the mold, and properties of resins are entered into a Computer Aided Engineering program for finding the processing window. Design of Experiment procedures are conducted with the CAE results for using inputs and outputs on training of networks in order to establish both a process controller and a quality predictor. Then, the controller and predictor are employed for on-line regulating part qualities. Final experimental results have indicated that the controller can automatically adjust the machine settings and reduce the fluctuations of part qualities from shot to shot. Conclusively, the product qualities are satisfactorily controlled under no operator intervention.