Using machine learning prediction models for quality control: a case study from the automotive industry.

Area of Science:

• Industrial Engineering
• Manufacturing Process Optimization
• Machine Learning Applications

Background:

• Automotive manufacturing relies on precise component dimensions, such as milled hole locations in bumper beams.
• Strict tolerance limits are critical for ensuring the quality and safety of automotive parts.
• Current quality control methods may not detect deviations early enough to prevent production disruptions.

Purpose of the Study:

• To investigate the use of machine learning for predicting milled hole locations in automotive bumper beams.
• To assess the effectiveness of different machine learning algorithms in early tolerance violation detection.
• To enhance the quality control process in automotive manufacturing through predictive analytics.

Main Methods:

• Time series data analysis of historical product measurements.
• Implementation and training of standard neural networks, Long Short-Term Memory (LSTM) networks, and Random Forest algorithms.
• Comparative evaluation of predictive capabilities across different machine learning models.

Main Results:

• All tested machine learning models demonstrated predictive capabilities for milled hole locations.
• Long Short-Term Memory (LSTM) and Random Forest algorithms showed a slight advantage in prediction accuracy.
• Successful prediction for some holes indicates potential for improved quality control, while others highlight data limitations.

Conclusions:

• Machine learning, particularly LSTM and Random Forest, can effectively predict milled hole locations to improve automotive part quality control.
• Early detection of tolerance violations is achievable, leading to a more efficient production flow.
• The accuracy of predictions is contingent on the quality and relevance of the available historical data, indicating challenges in real-world data problems.