Presenting a Novel Motion Capture-based Approach for Walk Path Segmentation and Drift Analysis in Manual Assembly

Automotive industry is currently facing the challenge to cope with the market demand for mass-customization whilst remaining competitive. In production planning, this trend towards product-diversification leads to a rising complexity, since growing numbers of variants are hitting mixed-model assembly lines. Due to these changing preconditions, traditional planning models and respective simulations tend to decreasingly reflect reality. Actual manual assembly processes can deviate significantly from their corresponding plans due to simplified assumptions of simulation models, methods and tools. In order to contribute to a better prediction quality of planning models, this paper investigates walk paths in real assembly situations with regard to their deviation from corresponding plans. A novel algorithm set for walk path reconstruction and neural network based classification of work tasks is introduced. Therewith, data gathered by a mobile tracking setup can be automatically segmented and subsequently assigned to the process plans. This novel approach enables an assessment of predetermined assembly times by comparing reference to real walk paths. The method's technical performance is verified in laboratory evaluation scenarios and its applicability is proven in a productive automotive final assembly line during operation.