Start-up monitoring for intermittent manufacturing based on hierarchical stationarity analysis

Abstract

Intermittent manufacturing is becoming increasingly popular due to its capability of coping with dynamic changes in market demands. The operation of the intermittent manufacturing equipment is subject to frequent restarts as the type of product being produced is switched, and it is challenging to achieve consistency in production during the start-up phase when restarts take place. Timely identification and monitoring of this phase is critical for avoiding the waste of materials and improving the product quality for intermittent manufacturing. In this work, a hierarchical stationarity analysis is proposed for the monitoring of the start-up phase process operation of intermittent manufacturing to extract two types of stationary information. First, consistently invariant process variations among batches are separated using a Kullback-Leibler divergence-based feature extraction method. In this way, process variations in each batch are divided into two subspaces – the stationary subspace and the remaining non-stationary subspace. Cointegration analysis is performed on the non-stationary subspace to capture the stationary information to find the long-term equilibrium during the start-up phase. Based on the divided subspaces, the start-up phase is identified, and process abnormalities can be online monitored. The efficacy of the proposed method is illustrated through a plastic molding process.