Artificial intelligence for aging identification of real applications of exhaust catalysts.

Public joint project funded by BMWK.

Project goals

The goal is the development of an AI method for the analysis and reduction of aging phenomena of exhaust aftertreatment systems

  • Generation of training data for artificial intelligence (AI)
  • Detection of aging mechanisms from measurement data and logic of the AI
  • Prediction of aging mechanisms

Project contents

  • Investigation of real aging behavior
  • Comparison of measurement methods: Test bench vs. OBD
  • Aging detection in the vehicle
  • Aging mechanisms
  • Real driving efficiency
  • Interactions of catalyst combinations
  • Big Data and data evaluation with "artificial intelligence”
  • Vehicle 5.0

Architecture AI model 1:

  • Gradient boosted tree algorithm
  • Supervised machine learning
  • Interpretable to identify variables relevant to aging

Architecture AI model 2:

  • Convolutional Neural Network
  • Supervised machine learning
  • Deep learning
  • Less interpretable

Input of the AI model

  • Aggregated quantities and data of the analytic model and metadata
  • Time series using importance scores of selected channels
  • Training data (input and output)

Output of the AI model

  • Description of the aging state of the catalysts from multi-layered data analyses

Project partners:

Funded by: