Orca-RBFNN: A New Machine Learning Method for Control Chart Pattern Recognition

Abstract

Supervising the production process in different factories and industries is one of the important and basic measures for the production of high quality goods and is of special importance. This is accomplished by monitoring the behavior of a system. Control chart is one of the most widely used and accurate statistical quality control tools that has been used in recent years in various industries to monitor the production process. In this study, a new method for detecting control chart patterns (CCPs) with the aim of online monitoring of the production process is proposed. In the proposed method, the radial basis function neural network (RBFNN) is used as a classifier of CCPs and a combination of shape and statistical features is used as input. In the proposed method, unlike the conventional methods in the literature, which use a set of shape or statistical features as input, the features are used intelligently and at different steps. In the RBFNN, center of clusters, number of clusters and their spread has a high impact on the network performance. Therefore, their optimal value must be determined correctly. In the proposed method, Orca optimization algorithm (OOA) is used to determine the value of these parameters. The proposed method was tested on a data set containing 800 samples and the simulation results showed that the proposed method is able to identify eight CCPs with 99.41% accuracy.