Beyond Pareto Analysis: A Decision Support Model for the Prioritization of Deviations with Natural Language Processing

Abstract

In the manufacturing domain, the systematic problem-solving (SPS) process is essential to eliminate the root causes of deviations from expected performance. The major goal of SPS is to prevent the recurrence of known deviations. However, due to time and resource limitations, the deviations that occur on the shop floor should be prioritized before applying SPS. Therefore, a method to support the decision-making process for prioritization of deviations is required. Traditional methods, such as the Pareto analysis, are widely accepted and applied for easy use. But their performance is no more sufficient for the production environment with large fluctuations nowadays. Therefore, this paper proposes a decision support model - the error score - to prioritize deviations on the shop floor. The error score is calculated based on the process data as well as textual data found in the deviation documentation. As the quality of textual data in the deviation documentation has great effects on the performance of the model, Natural Language Processing (NLP) methods are developed to pre-process the unstructured text. To validate the model, it is applied to a real-world use case in the automotive industry to demonstrate and evaluate the performance. The study shows that the proposed model can effectively support the decision-making process on the shop floor and is superior to traditional methods.