Statistical Modeling of Industrial Process Parameters

Abstract Identification of models of process parameters provides a way to clarify some hitherto unexplained patterns of deviation from design values, leading to enhanced opportunities of quality improvement. While most standard procedures are based upon normal distribution hypothesis, the latter sometimes is liable to fail to accommodate actual data even to a first approximation. Skew, bounded, multimodal data sets call for reasonably close description if meaningful inferences are to be drawn. Graphic representation may pose challenges, the aspect of grouped data being materially affected by a more or less arbitrary choice among several options. Issues in modeling are discussed in the light of an actual case, concerning a critical bore realization on an automotive component.