| Summary: | 博士 === 中華大學 === 科技管理學系(所) === 99 === The plastic injection molding (PIM) has been widely researched and routinely applied in many high-tech industries; the accuracy and precision of this technology are being severely scrutinized in the wake of the evolutionary trend of making products diverse and creative. In this regard, how to effectively grasp the best and suitable quality of products is always the crucial issue associated with the throughput and the yield. By virtue of the underlying plastic injection molding system causing the process parameter settings in injection molding machines to be complicated, the nonlinear control model for the injection molding system is hard to be obtained with exponentially changed complexity while more unknown parameters are added.
In the past, PIM product quality was usually measured by one single quality
characteristic or by multiple quality characteristics with independent parameters one another. With the increasing complexity of product, this dissertation proposes a three-stage integrated optimization system to generate the optimal process parameter settings of multiple-quality characteristics. In the first stage, the significant PIM process parameters can be determined by DOE screening experiments. In the second stage, the optimal process parameter settings are obtained via computer aided engineering (CAE) simulation integrated with response surface methodology (RSM) and genetic algorithm (GA) , which are taken as practically initial settings of process-related parameters. As for the previous two stages, the mold-flow analysis software (i.e., Moldex3D and Rhinoceros 4.0) is used as a working platform to develop mesh with a sufficient resolution, and predict possible defects observed on one product by way of a computer’s simulated results for reduced cycle time, increased reliability, improved efficiency in die sinking, and decreased cost from CAE simulation instead of trial-and-error process prior to die sinking.
In the final stage, Taguchi method and back-propagation neural network (BPNN) are utilized for developing a signal-to-noise (S/N) ratio predictor and executing analysis of variance (ANOVA) to analyze the factors’ significance effects of initial process-related parameter settings, then the BPNN S/N ratio predictor conducting with simulated annealing (SA) to optimize the process-related parameter settings for S/N ratios characteristic and minimize/stabilize variations of one process; moreover, the BPNN quality predictor also works in with particle swarm optimization (PSO) for multi-quality/ multi-objective characteristics (i.e., length and warpage) optimization of process-related process parameters.
In addition, the above-mentioned CAE simulations and intelligent integrated process parameter optimization system are employed to search the optimal PIM process parameters,and can further obtain the optimal die design and effectively reduce the costs.
|
|---|
