Implementation of a real-time multivariate data analysis methodology in injection molding and high frequency welding process

In this research two methodologies for data collection using a multivariate data analysis technique are proposed. The Continuous Methodology and Combination Methodology for data collection are used and evaluated in two distinct processes, injection molding and welding by radio frequency. These processes, which are of a multivariate nature, contain variables that have an effect on the quality of a product. The Continuous Modeling Methodology proposed the use of process data at three levels to create sub-matrices, which provided the most information of all the available batch characteristics. The Combination Methodology utilized gradual changes to critical variables of the process or product defect. A signal monitoring system is utilized for the validation of the methodologies proposed for each process. For the evaluation of the methods proposed, type I and type II errors are used along with process scrap rates. It has been shown that with the continuous and combination methodology, the type I error or the scrap rate are decreased. The variables that are significant for the identification of product defect have been identified.