This research is performed in order to forecast the parameters of a piecewise linear profile to be able to investigate its control status. It introduces a new phase, called phase III, further to former phases: construction (Phase I) and monitoring (phase II) in statistical quality control of a profile. Initially, classical regression techniques including maximum likelihood (ML) and ordinary least squared (OLS) methods are used to estimate the parameters of the profile. Then for evaluating the control limits for future intervals, two different methods are developed based on forecasted profile parameters and the trend of historical control limits. Finally, the performances of these methods are evaluated and compared for simulated data. It reveals that ML method generates more accurate forecasts for slope than OLS method. Additionally, trend analysis for control limit forecasting has a higher type II error than ML or OLS method.

Keywords