摘要：AbstractIndustrial manufacturers are facing increased demand for custom products, which comes with an increase of process variability. For example, in the case study presented in this paper, the company put business rules in place to mitigate the non-conformity of color of the textiles they produce. We decided to use historical data to build predictive models to help the manufacturer design a more efficient quality control strategy. The performance of these models was compared to the business rule previously used by the company. It was found that the random forest model outperformed their business rule by 12% (reduction of false negative ratio) for the same number of quality tests performed. Also, the proposed method allows the business to better choose the tests to perform based on their budget and the number of non-compliant products that are deemed acceptable for them.
关键词：KeywordsProbabilistic & statistical models in industrial plant controlIndustrialapplied mathematics for productionIndustry 4.0Data analyticsData miningmultivariate statistics
