In one of the 16 sessions at the 33rd International Colloquium on Plastics Engineering, IKV will demonstrate how AI methods can be used to enhance established analytical and numerical models, enabling more informed decisions in development and design. A particular focus will be on the behavior of short-fiber-reinforced thermoplastics.
Models that can predict material and process properties with high accuracy and efficiency are an important prerequisite for the design of modern plastic components. Institute for Plastics Processing (IKV) in Industry and Craft at RWTH Aachen University is conducting research to show how AI methods can expand established analytical and numerical models, thereby enabling more informed decisions in development and design. A particular focus is on the behaviour of short fibre-reinforced thermoplastics.
When manufacturing short fibre-reinforced thermoplastic components, the spatially varying fibre orientation generated in the injection moulding process as a result of local flow mechanics, shear rates and solidification conditions must be taken into account. This microstructure not only influences anisotropic mechanical properties such as longitudinal and transverse stiffness, shear modulus and transverse contraction, but also affects shrinkage, thermal stresses and ultimately the warpage of the component. The ability to map this process-related microstructure is therefore central to the design of fibre-reinforced components and the optimisation of processes.
Scientists at IKV are pursuing various approaches to describe the process-related microstructure of short fibre-reinforced thermoplastics and their influence on mechanical properties as well as shrinkage and warpage behaviour. Examples are the development of machine learning-based stiffness modelling for fibre-reinforced thermoplastic components and AI-adapted material maps that improve the prediction of shrinkage and warpage. This creates tools that enhance classic calculation models and thus enable more precise, efficient and consistent decisions in component development. The approaches will be presented at the 33rd International Colloquium Plastics Technology in Session 10, Artificial intelligence in product development and simulation, and demonstrate how AI-based modelling in plastics technology can make a measurable contribution to the quality and cost-effectiveness of development processes.
Machine learning-based stiffness modelling for the development of fibre-reinforced thermoplastic components
The mechanical properties of short fibre-reinforced thermoplastics are significantly influenced by fibre content, fibre length, orientation tensors and the rheological properties of the matrix. Analytical mean-field models, such as Mori–Tanaka homogenisation, provide quick and physically motivated initial estimates of these anisotropic stiffnesses, but capture more complex effects such as fibre-fibre interactions, non-linearities or local microstructure characteristics only to a limited extent. High-fidelity methods such as FEM-based microstructure simulations can map these relationships much more accurately, but they involve considerable computational effort.
© IKVAt the IKV a multi-fidelity artificial neural network was developed that combines analytical mean-field predictions (low fidelity) with FEM reference data (high fidelity) in order to combine these two levels (Figure 1). The neural network learns systematic deviations of the analytical model and calibrates its trend behaviour using the detailed numerical data. This creates a model that maintains the physical consistency of analytical homogenisation while achieving the accuracy of FEM-based microstructure description. The result is a more accurate description across the entire design space, especially for areas with high stiffness contrast and increased fibre volume content. The method is suitable for efficient material screening and robust design steps in the early development phase.
Use of AI to improve deformation simulations – AI-based adaptation of material cards
Shrinkage and warpage arise from the interaction of thermal, rheological and microstructural effects during cooling in the injection moulding process. Simulation-based predictions require material cards, whose parameters often have to be calibrated manually. This iterative process is time-consuming, highly dependent on experience and only reproducible to a limited extent in industrial practice. At the same time, the accuracy of warpage simulation depends directly on the consistency of these material cards.
© IKVIKV has developed a data-based approach for this purpose, in which an AI model evaluates experimental deformation measurements and numerical simulation results to identify deviations between actual and calculated component behaviour (Figure 2). The neural network then specifically adjusts the relevant parameters of the material card, thereby generating an automated, consistent calibration. The method improves the quality of predictions for different geometries, reduces the need for iterative tool corrections and enhances the validity of simulation-based development processes. This approach is particularly relevant for components where low tolerances are crucial or where the microstructure varies greatly depending on the process.
FRP, methods of digitalisation and AI at the 33rd International Colloquium Plastics Technology
The topics of lightweight construction and fibre composites will be addressed in
- Session 3: Plastics as a key to scaling the hydrogen economy
- Session 6: Mechanical recycling of CFRP: Vitrimers as enablers
- Session 10: Artificial intelligence in product development and simulation
The application of digital methods and AI in plastics processing will be the subject of
- Session 5: Robust injection moulding through adaptive process control
- Session 7: Analysis and improvement of the quality of polyolefin recyclates
- Session 8: Prediction of process-dependent material behaviour in product development
- Session 13: Simulation technologies for precise injection-moulded components
At ‘IKV 360° – Research Live’, IKV scientists bring the topic to life at various stations in the IKV technical centre.