Understanding whole shape variability of stamped sheet metal parts for root cause analysis

In automotive and aerospace industries most of the manufacturing processes deal with very flexible parts which makes hard to fully take under control the final shape. The quality control should be capable to capture size variations in some key points and local shape defects caused by manufacturing steps. For this reason surfacebased metrology systems can be used but when dealing with big data sets new challenging issues arise related to interpretation and classification of faults as well as the link to the faulty causes. The present on-going research is a step towards a root cause analysis of faults related to stamped sheet metal components acquired with an optical scanner. Against the common costly and time consuming trial-and-error approaches, the aim of the research is to set a methodology to forecast geometric errors for given set of process parameters. The methodology is summarized as follows: (i) to capture data (as dense Cloud-of-Points, CoP) and extract the deviation field by mapping CoP with nominal CAD model; (ii) to decompose the high-dimensionality of the deviation field into orthogonal (i.e. independent) significant error modes; (iii) to build analytical model linking design parameters to the decomposed deviation field. The proposed analytical-based model can be used to forecast geometric errors and narrow down root causes of failure, usually unforeseen if only based on heuristic approaches. An industrial case study is used to validate the methodology.