Risk Analysis Based on Time Series Models for Process Cotrol Data

摘要

In this paper, we propose an application method for risk evaluation of autocorrelated process control data based on time series analysis handling ARMA and EWMA models. Parameters of the models are estimated by maximum likelihood methods using algorithms of the Kalman filter. The information criterions are applied to the models for choice one fitted data. Risks of process control are evaluated by difference between predictions and observations based on the chosen model for a set of data. It is shown by cases of the real data analysis that the time series analysis is useful for detection of the risks.