FMEA based prescriptive model for equipment repair guidance.

Area of Science:

• Manufacturing Engineering
• Data Science
• Machine Learning

Background:

• Minimizing manufacturing downtime and enhancing production efficiency are critical.
• Accurate prediction of machine repair needs is essential for proactive maintenance.

Purpose of the Study:

• To demonstrate how machine learning can support maintenance teams in selecting optimal repair methods.
• To utilize machine failure data and Failure Mode and Effects Analysis for predictive maintenance.

Main Methods:

• Adoption of the Design Science Research paradigm.
• Integration of CRISP-DM and PDCA methodologies for data science quality assurance.
• Formulation of repair action prediction as a classification problem using multivariate time-series data.

Main Results:

• Evaluation of two modeling approaches: merging time series vs. preserving them as 3D arrays.
• Application of Multi-Layer Perceptron (MLP), Convolutional Neural Networks (CNN), and Fully Convolutional Networks (FCN) models.
• Assessment of model performance based on classification accuracy and the impact of time-series processing and architecture.

Conclusions:

• Highlighting effective strategies for predicting machine repairs.
• Advancing prescriptive maintenance in manufacturing environments through accurate predictions.
• Demonstrating the value of machine learning in optimizing maintenance operations.