Smart and Resilient Manufacturing

Towards a resilient manufacturing using digital twins and cognitive assistant systems, human-focused and data-based approaches are essential. Work skill gaps and differences among employees still belong to the most common challenges that have a high impact in manual manufacturing processes. For measuring and inspecting of parts manually in between manufacturing process to secure perfect quality, it is essential that those measurements are performed in the correct way. Especially in manual processes, lack of data collection and usage for analysis purposes result in general in loss of competitiveness, whereas extracting data knowledge and reacting to skill level gaps immediately via algorithm predictions allows continuous improvement of measuring and inspection processes and employee skills. Hence, the main purpose of this study is to empirically explore and predict skill related patterns that affect process performance and the quality of 3D hand-based measurements for quality inspection tasks in the automotive industry context using machine learning approaches. The setting is a manual measuring and inspection station in the ARENA2036 in Stuttgart, Germany using a battery case as an example part for the performed measurements. The algorithms are selected following the human-in-the-loop-approach, where understanding of the statistical data and the involvement of the process experts to validate results should be addressed. In this sense, principal component analysis (PCA) and clustering algorithms are used in the first exploration stage of linear separability. Subsequently, predictions are done using multiple classification algorithms as K-Nearest Neighbours (KNN), Random Forest, Support Vector Machines (SVM) and Naive-Bayes to evaluate accuracy. Thus, the novelty of this paper lies in validating predictions for qualification levels within the manual inspection context and promptly developing modules and guidelines for the use of data to enhance cognitive digital assistance systems for resilient factories.