Real-Time AI and IoT Integrated Quality Assessment for Advanced Automotive Manufacturing Environments

Abstract

Traditionally, quality assurance (QA) in the automotive manufacturing industry has relied on a set of expensive tests and visual inspections on a sample of produced vehicles. The growing presence of Internet of Things (IoT) sensors in production lines gives manufacturers new opportunities to move from QA to Predictive Quality Control. Predictive Quality Control uses Machine Learning/Artificial Intelligence (ML/AI)-driven predictive analytics to assess quality and risk in real-time, enabling root cause analysis, yield optimization, and improved process control during production. A system that uses the streaming data of connected sensors to quantify the performance of these techniques has been deployed in a major automotive manufacturer.

The first case study describes the application of predictive modeling to two stamping and assembly processes. A spike in process risk led to increased costs and testing; however, better risk assessment and detection of sensor anomalies improved the management of sensor data quality. A predictive quality model for the most affected body components was also developed but is not yet fully operational. In the second case study, another ML-driven yield prediction model has been implemented to reduce paint shop downtime and costs arising from coarse and fines segregation in weld joint areas. During the last months of operation, the segmentation of fine joints has shown a predictive accuracy above 94% and increasingly stable precision and recall ratios.