In this Forming Technology Forum the focus will be on the application of hybrid models to forming processes.
Hybrid models merge classical physics-based models with purely data-based modeling techniques with the aim to combine the strength of both approaches and cancel their weaknesses. Typically, high accuracy, high speed and high transparency or explainability are favoured. In forming processes, applications of hybrid models include enhanced performance in classical off-line simulations as well as fully integrated data-acquisition and control of ultimate product properties.
In the first category, e.g. fast advanced material models are developed using machine learning, but trained by time-consuming offline simulations and constrained by fundamental physical laws, such as physics informed neural networks (PINNs) and thermodynamics based artificial neural networks (TANNs). Another example in this category is the correction of relatively simple and fast analytical process models by large scale measurement data and machine learning to improve the overall accuracy.
In the second category, inline process measurements and fast simulation models estimate the status of relevant process parameters that cannot be measured directly, like e.g. internal stresses in a formed part. Based on models, so-called soft sensors can be developed. In many cases, this requires a statistical approach. The connection between a physical process and the virtual counterpart creates a pure form of a Digital Twin and can ultimately be used to control product properties.
Topics
- Data-based model enhancement
- Digital Twins of forming processes and formed parts
- Physics informed neural networks (PINNs)
- Thermodynamics based artificial neural networks (TANNs)
- In-line sensors for forming processes (e.g. draw-in, vision, eddy-current, …)
- Soft sensors and smart tools
- Actuators specific for forming (e.g. adaptable beads/spacers, blank holders, …)
- Hybrid approaches (AI combined with classical methods, e.g. FEM and ML)
- Using real-world data and synthetic data for training
- Material scatter, process uncertainty and state estimation
About Forming technology forum
Forming Technology Forum is a 2-day conference with a limited number of carefully selected presentations and sufficient time for in-depth discussion. Each year another theme is selected.
Forming Technology Forum sets itself apart from other similarly focused conferences with its highly interactive, supportive and collaborative environment that is achieved by having only plenary sessions, encouraging lively discussions after each talk. It is common that the discussions continue at the conference dinner and strong international bonds are formed between the participants.
It is the intention of the three chairs, Prof.Dr. Ton van den Boogaard, University of Twente, Enschede, Prof.Dr. Dirk Mohr, Institute for Mechanical Systems, ETH, Zurich, Prof.Dr. Wolfram Volk, Technische Universität München, Munich, that the above mentioned traits continue to define the FTF conference series.
The conference brings together researchers and practitioners in production technology, materials, modelling and process control to share and benefit from each other’s experience by high quality presentations and lively discussions.