Development of fitted line and fitted cosine curve for recognition and analysis of unnatural patterns in process control charts

Abstract

Unnatural patterns in process control charts exhibit out-of-control conditions. Therefore, increase in sensitivities in control charts is mandatory to study these situations. Because of the existence of inevitable natural variations, real-time detection and analysis of the significant patterns is a problem, especially when sensitivity level of the process to unnatural patterns formation is high. In the previous studies, most researchers have applied neural networks techniques to monitor significant patterns. Although this approach is effective, but structures of networks are complex and their architectures are difficult. The current paper develops fitted line and fitted cosine curve of samples to recognize and analyze the unnatural patterns. This simpler solution is more efficient and consumes less feedback time. The proposed model alarms occurrence of single and concurrent patterns and estimates their corresponding parameters. These fitted line and curve facilitate recognition and analysis of significant patterns at different levels of sensitivity, while the presented models often face with patterns misclassification error when high level of sensitivity is desired for unnatural patterns discrimination. To implement the proposed model, S2 control chart has been selected as a case study. The accuracy and precision of the proposed tools are evaluated by simulated experiments.