Artificial Neural Network‐Based Adaptive Cumulative Sum Control Chart for Process Monitoring

ABSTRACT The key requirement in industrial process monitoring is the capability to rapidly and accurately detect changes, that is, those that occur as outliers, since ensuring high quality levels and the equity of functioning are paramount priorities. Control charts give a systematic and statistically valid foundation to the identification and management of such deviations, which are usually brought about by assignable causes, particular and identifiable sources of fluctuation, but not by the randomness of the process. In this study, the proposed research suggests an innovative adaptive Cumulative Sum (CUSUM) chart enriched with Artificial Neural Network (ANN) methodology, which is specifically aimed at increasing sensitivity in tracking the behavioral changes of a process, small or medium‐scale. Using the capacity to see trends and change according to learning, this scheme dynamically varies the reference parameter of the traditional CUSUM chart according to the real‐time data characteristics. Such an adaptation process facilitates more adaptive and precise monitoring that is crucial when the dynamics of processes are changing gradually or suddenly. In order to illustrate the practicality of the given model, the suggested adaptive CUSUM chart is applied to a real engineering data set consisting of measurements of piston diameters, and minor deviations in the dimensional accuracy may adversely affect the engine operation to a considerable degree. Simulations using Monte Carlo prove the higher efficiency of this control chart to detect anomalies at an early stage, which is an effective and powerful means of quality control in contemporary production facilities.