Physics informed machine learning model for blow molding process optimization

Overview

The proposed solution is a physics-informed machine learning model designed to optimize the stretch blow molding process by utilizing synthetic data generated through physical simulations. This approach addresses the challenge of insufficient training data, enabling the creation of a surrogate model that formulates an optimization problem. The model aids in fine-tuning processes for specific outcomes, such as minimizing bottle weight while meeting load targets, optimizing thickness, and determining ideal container designs.

Technical specifications

Key features:

• Utilizes synthetic data from physical simulations to train the machine learning model, eliminating the need for extensive experimental data collection.
• Capable of integrating synthetic and experimentally measured data to enhance model accuracy and reliability.
• Facilitates the optimization of blow molding processes by addressing parameter variations such as geometry, material, and process conditions.
• Provides a framework for generating supplementary data to improve model performance in targeted areas.

Technology readiness level

This technology is at TRL 6, indicating that it has been demonstrated in a relevant environment and is ready for further validation and optimization in operational settings. Future validation steps include assessing data representativity, defining key features and KPIs, and further developing the machine learning model architecture.