Monitoring production processes with robust and efficient adaptive techniques when model assumption is unknown

Traditional techniques for monitoring process parameters require certain assumptions for the efficient detection of process anomalies. This study focuses on addressing the limitations of traditional control charts in monitoring process parameters, particularly when the normality assumption of quality characteristics is violated in the industrial processes. This study introduces two novel adaptive control charts, namely the nonparametric (NP) unbiased function-based exponentially weighted moving average signed rank (NPUAEWMA-SR) chart and the NP unbiased-function based cumulative sum signed rank (NPUACUSUM-SR) chart. The study highlights the enhanced performance of the proposed charts in terms of run-length characteristics for specific and over the range of shifts. The NPUAEWMA-SR and NPUACUSUM-SR charts exhibit superiority for in-control resistance to non-normality and outperform the existing methods. The statistical metrics like average run-length (ARL), expected ARL, extra quadratic loss and relative mean index have been used for relative assessment. The proposals are applied to real-world cases, including a piston ring manufacturing process, and two artificial datasets. The findings suggest that the proposed charts are robust and adaptable tools for improving process monitoring in industrial settings.