Titre :
Automated detection of manufacturing geometric deviations using knowledge-driven supervised learning
Journal, revue, page ... :
International Journal of Computer Integrated Manufacturing
Type de publication :
Article de Journal
Abstract :
Quality control in manufacturing relies on clear communication and continuous feedback among
production teams. However, limited expertise or incomplete information often leads to errors that
compromise process reliability. During inspection, one of the critical aspects of verifying mechanical part
conformity lies in assessing geometric deviations from the ideal geometry defined by the reference
model. This analysis not only reveals significant variations in mechanical parts but also helps trace
defects back to their underlying sources, improving quality. This paper introduces a supervised learningbased methodology for detecting the sources of geometric errors. The proposed approach leverages
a technological knowledge base of typical defect patterns to automatically classify measurement data,
identify shape defects, and infer their most likely origins through an integrated learning subprocess. To
support this methodology, a decision-support framework, referred to as MCMoL, was developed within
the MATLAB®. The method improves quality control performance in manufacturing and inspection
processes. The model is trained on generated datasets and evaluated using representative industrial
radial geometric defects defined according to ISO 1101, demonstrating the effectiveness of the
proposed methodology. Finally, the advantages and limitations of the proposed approach are
discussed.
