Improvement of Shewhart Control Chart for Autocorrelated Data in Continuous Production Process

Abstract

Shewhart Control Chart is widely used to monitor, control and improve quality in many industrial processes. Control chart is based on the assumption that the resulting data is distributed independently. But in the process of continuous production most data are autocorrelated. Autocorrelation is a state in which between sequential observations have a relationship. In order to use the control chart effectively, the autocorrelation in the data must be eliminated. Autocorrelation can be eliminated by mapping residual modeling results using the time series method because of the residuals of the modeling following a normal and independent distribution. In this study Genetic Algorithm is integrated with support vector regression for optimization of support vector regression model parameters for more accurate prediction result. The more accurate the model used, the predicted results will be close to the actual value so that the residual value obtained will be closer to zero. The more residual values close to zero, the average will be zero and the data will spread around the average value. After the calculation it was found that the proposed modeling resulted in a RMSE of 46 % smaller than other modeling and the residual control chart generated from the modeling of Genetic algorithm support vector regression of all data within the control limits.