Research On Quality Parameter Online Inspection,Modeling And Optimization Of Injection Molding Product

Due to the continuously improvements of the injection molding technology, people are more familiar with the injection molding process and require higher quality of injection molding product. On the basis of the current technology and equipment, the laggard inspection technology is the main resistance to improve the quality of injection molding product. Based on a great deal of studies, it can be overcome by using fast and accurate quality inspection methods to obtain the quality information of injection molding product online. During the injection molding process, the product quality is affected by many injection molding parameters. Base on the model between injection molding parameters and product quality, optimized injection molding parameters can be obtained to produce products whose quality satisfy customers’needs. Therefore, the dissertation focuses on researching quality parameter inspection online, modeling and optimization of injection molding product.(1) On the research of quality parameter inspection online of injection molding product, shape defect and size index of injection molding product are detected by machine vision inspection system. An area based inspection method and a difference image algorithm based inspection method are proposed for satisfying different requirements of users. An improved active contour model algorithm is proposed to extract contour of injection molding product from its image. An improved virtual triangle method is proposed to improve the accuracy of feature matching in the process of image registration. Pin-hole model and lens distortion model are applied to construct the relation between space coordinates and image coordinates. Robust metric calibration method is applied to calibrate the camera models above. A black and white X-corner detection method is proposed to detect X-corner in chess board image automatically.(2) Data based modeling method is studied to construct the relation between injection molding parameters and product quality. A neural network (NN) ensemble approach is proposed to improve the generalization ability of ensemble neural network by combining BAGGING and negative correlation learning algorithm with a selection strategy. Computer aided engineering (CAE) software are applied to generate data for modeling, since high experiment costs are need if the data for modeling are generated by experiments on actual production process. However, CAE software cannot substitute actual production process. In this thesis, a new modeling method is proposed to improve the accuracy of NN ensemble model by assisting only few experiments on actual production process.(3) The model obtained by using the proposed modeling method is applied for optimizing injection molding parameters. However, the optimized parameters obtained by using the above model cannot satisfy the requirement of users. In this thesis, a space mapping algorithm is applied to optimize injection molding parameters. An improved space mapping algorithm is proposed to save experiment costs. Meanwhile, since injection molding process is often affected by disturbances, robust injection molding parameters optimization method is studied to guarantee the quality of injection molding product. Since high experiment costs should be paid by using the existing robust parameters optimization method, a two-stage approach for optimizing the robust injection molding parameters is proposed to save experiment costs.At the end of the dissertation, the potential further research direction in the area of quality parameter inspection, modeling and optimization of injection molding product is discussed after summarizing the whole work.