Integrated use of artificial neural networks and Shewhart, CUSUM and EWMA control charts in statistical process control: A case study in forest industry enterprise

In this research, Statistical Quality Control (SQC) charts, together with Artificial Neural Networks (ANN), were used to improve the overall quality and minimize the costs in a particleboard industry. The data regarding the some mechanical properties (internal bond strength, modulus of elasticity, surface soundness, screw withdrawal strength) of particleboards, regularly received from the company for a period of six months, were grouped into two quarterly terms in accordance with the working plan and the applied methods. In the first stage of the research, basic concepts of quality and quality control, Shewhart Control Charts, Cumulative Sum (CUSUM) Control Charts, Exponentially Weighted Moving Average (EWMA) control charts, ANN subjects and particleboard industry were addressed after a literature survey. In the application stage, the control charts were prepared using the data received between February/April 2016 to determine the factors that impair quality, and the most suitable control charts for the company were specified accordingly. Control charts Shewhart, CUSUM and EWMA were used for varying quantities to achieve the targeted quality level. In the last stage of the research, estimations were made using ANN to predict the future state of the process and minimize the evaluation costs. It was concluded at the end of the research that, inter-dependent and independent evaluation of the data, observed during particleboard manufacturing, was required, thus, collective use of CUSUM and Shewhart control charts was proposed. Also, the low MSE, MAPE and MAD performance values obtained with ANN estimations indicated that, some of the quality characteristics could be estimated without the need for measurements.