In-situ workpiece perception: A key to zero-defect manufacturing in Industry 4.0 compliant job shops

• Published in: Computers in industry 2023-06, Vol.148, p.103891, Article 103891
• Main Authors: Babalola, Simeon A., Mishra, Debasish, Dutta, Samik, Murmu, Naresh C.
• Format: Article
• Language: English
• Subjects: Friction Stir Welding; Industrial recommender system; Machine learning; Workpiece perception; Zero-defect manufacturing
• ISSN: 0166-3615; 1872-6194
• DOI: 10.1016/j.compind.2023.103891

Job shop manufacturing is characterized by excellent job flexibility and highly customizable products. The dynamic nature of jobs in this manufacturing segment poses a relatively high challenge to actualizing zero-defect manufacturing (ZDM). The continuous emergence of new engineering materials and improved consumer’s contributions to product development driven by IIoT further complicates the uphill task. This study develops a methodological framework that positions in-situ workpiece perception and industrial recommender system as tools to trivialize the challenge of ZDM in job shops. In-situ workpiece perception was experimented using a case study of friction stir welding (FSW), a thermomechanical manufacturing process where existing algorithms for online quality assessment are material-specific. The novelty in this study is deciphering unique applications for the hitherto jettisoned sensor data acquired prior to the FSW tool traverse (the actual welding stage), ensuring that the material perception process does not induce defects in the products. This study presents two major contributions towards attaining first-time-right ZDM in job shops. First, a novel interlink for IIoT-driven production planning to aid the selection of optimal process parameters; second, the selection of appropriate quality assessment algorithms during the welding process. •Posits in-situ workpiece perception and industrial recommender systems as BoML ZDM KETs for MTO job shops.•A low latency in-situ workpiece perception workflow with 100 % multi-classification accuracy in FSW use case.•Novel use of pre-manufacturing sensor data for ZDM ensures the developed KET does not induce defects in the product.•The industrial recommender system serves as a decision support system, thereby facilitating human-in-loop ZDM solution.