AI-Powered Material Property Prediction for Compounding

Ehrenmüller's solution leverages artificial intelligence to predict material properties in the recycling and compounding of post-consumer recyclates (PCR). By accurately forecasting key properties such as elastic modulus, melt mass-flow rate (MFR), and melt volume-flow rate (MVR), this solution enables producers to precisely control material characteristics, reduce scrap rates, minimize downtime, and optimize resource efficiency.

Key Challenges Addressed:

• Varying composition of post-consumer recyclates across batches
• Unknown material properties after processing and compounding
• Contaminants and foreign substances affecting quality
• Fluctuating processing parameters difficult to define for unknown materials
• Limited suitability of recyclates for high-value applications without reliable property prediction

Solution Architecture: The solution employs a property-based two-stage modeling approach. The first stage predicts blend properties from easily measurable material characteristics and blend ratios. The second stage incorporates additive influences to refine predictions.

AWS Services Integration: The solution is built on robust AWS infrastructure to ensure scalability, reliability, and performance. Amazon SageMaker is used for training, tuning, and deploying the machine learning models that predict material properties. Amazon S3 provides secure and scalable storage for material datasets, processing parameters, and model artifacts. AWS Lambda enables serverless execution of prediction requests and data preprocessing pipelines. Amazon API Gateway exposes prediction endpoints for integration into existing manufacturing execution systems. AWS IoT Core can be leveraged to ingest real-time sensor data from compounding equipment, feeding live processing parameters into the prediction models. Amazon CloudWatch provides monitoring and logging for model performance tracking.

Business Value:

• Targeted and reliable use of recyclates in high-value applications
• Reduced scrap and downtime through accurate property predictions
• Accelerated development of new polymer blends with specified properties
• Consistent product quality despite fluctuating material composition
• A clear step toward greater economic efficiency and a true circular economy

Developed by Ehrenmueller KI-Experten, specialists in custom AI development for the innovative mid-market sector, with proven references across medical technology, food production, mechanical engineering, and manufacturing industries.