A Diverse Neural Network Ensemble Team for Mean Shift Detection in X-Bar and CUSUM Control Charts

In manufacturing processes, maintaining quality is associated with proper process mean parameters and product quality metrics. the early detection of mean changes is important to reduce the number of defectives or non-conformities in the production. in this work, a diverse ensemble of artificial neural networks (anns) with a leader network have been used to achieve this purpose, then a performance comparison was conducted on two types of control charts: x-bar and cusum in addition to comparing it to individual neural network performance. the traditional and individual neural network performances were obtained from published literature. it was found that, the diverse ensemble of anns detects small shift in process mean far earlier (shorter average run length (arl)) than individual ann’s, traditional x-bar and cusum control charts. the postulated reason for this improvement is that the ensemble ann system analyses more than one sample point, rather, it considers the points pattern and overcomes the instabilities in individual ann’s.