Research On PLS Quality Prediction For Injection Molding Process

The injection molding production process is a typical multi-stage batch process. In the batch process, the quality of the product is difficult to be measured online. Many of the quality indicators can only be obtained through a variety of tests in the laboratory after the process, and these tests’results are often serious lag in the production process itself. However, the vast majority of process variables measured online in the injection molding production process, not only contain a wealth of information reflecting the state of the process runs, but also contain the information related with the final product quality of the batch process. Therefore, we can pursue the specific relationship between the measured value of the process variables and the value of products quality from the historical data of the injection molding production process, and achieve the prediction of the quality online which is unmeasured online.The multiway partial least squares model is widely used in the batch process product quality prediction. However, for the multiphase batch processe, each phase has different characteristics of the process. If we establish a MPLS prediction model using all phases, it will be difficult to reveal the effect of each phase to the quality of final product. If we can divide phases correctly for the multiphase batch processe, and establish the model for each phase, then it can increase the degree of accuracy of the model.For the multiphase characteristics of the batch process and the disadvantages of the MPLS method, this paper proposes an improved MPPCA to divide phases of procedure. On the basis of in-depth analysis of production and operating mechanism, combined with production and operational data to determine the critical periods and key variables related to product quality, I establish MPLS model used in the injection molding process based on sub phase and achieve the quality prediction online, and the simulation results prove the validity of the proposed method.Finally, the paper takes the production process of injection molding machines as the background, using the C#, database and Matlab as the main development tools, and achieves the design and development of the injection molding process quality prediction experimental platform. The design and development of this platform laid a solid foundation for experiment authentication of product quality prediction method based on multivariate statistical.