Skip to content

Efficient Process Parameter Optimization Using Machine Learning

  • today
  • access_time - ET
  • location_onMcCormick Place
  • gamesConference
  • monetization_onPaid Upgrade
  • schoolIntermediate

BMW has been using additive manufacturing for more than 30 years to make end use parts in a wide variety of applications (both metal and polymer). This presentation is a case study of how BMW has successfully used machine learning to help the Additive Manufacturing Center’s metals group drastically improve the process of finding optimal process parameters for an aluminum alloy produced on a quad-laser powder bed fusion system. The process parameters that were developed using the machine learning approach exceeded the performance of the conventionally optimized parameters – even though taking less time. This was validated by making real end-use parts.  Furthermore, by using the new process parameters BMW succeeded in building prototyping-parts that had failed in previous builds. Finally, BMW was able to achieve such optimal process parameters while also reducing the number of data points needed. This resulted in fewer builds, less testing, less time and less cost.  This presentation will describe two different projects that were undertaken. The first project was focused on finding process parameters for good surface finish and high tensile strength, while the second was focused on minimizing distortion and high tensile strength. Furthermore, the presentation explains to the user how the machine learning approach is different from conventional approaches and walks the audience through the steps that BMW took to rapidly find optimal parameters while minimizing data generation time and costs.

Learning Objectives:

  • Understand how the machine learning approach can drastically cut down on time and cost for parameter optimization and how the machine learning approach differs from a conventional approach
  • Understand a framework with which to analyze the relationships between process parameters, material properties and mechanical performance of metallic components
  • Evaluate whether or not the machine learning approach can be used by the participant to efficiently develop process parameters for their own applications