Automotive part manufacturer: Quality prediction and driver analysis for aluminum castings

Automotive part manufacturer: Quality prediction and driver analysis for aluminum castings

Azure AI/ML, Customer Stories, Manufacturing, MLOps
Share

 

Challenge

  • The manufacturer has a population of parts which pass a final inspection but are actually defective
  • The defective parts are used to build cars, which results in costly repairs or recalls
  • Manufacturer desires to understand how to better control process manufacturing in order to better detect defective parts in the factory

Solution-V12

Solution

  • Integrated external data sources that were hypothesized to have an effect on manufacturing quality (ex. weather data where the plant is located)
  • Analyzed manufacturing data to uncover relationships between key manufacturing variables and final product quality
  • Developed machine learning models to predict the probability that a part is defective

Result-V12

Result

  • Utilized machine learning to identify key drivers of part failures which can be adjusted to improve product quality
  • Provided model-driven strategies to improve defect detection
  • Found that parts tend to fail in sequence, i.e. a part produced right before or after a bad part is more likely to fail

You May Also Like

Tips for starting your MLops Journey
2022-05-26
3 tips for starting your MLOps journey
doctor in blue scrubs
2021-01-13
Optimized data delivery to members and providers at a health benefits administrator
2020-12-29
Improving promotion targeting for a credit card company