Product Development Why AI and Multiphysics Are a Dream Team

Related Vendors

Altair Engineering GmbH

FAULHABER Drive Systems

Wuxi InfiMotion Technology Co., Ltd.

In product development, increasingly complex relationships and interactions of various physical phenomena must be considered. This article demonstrates how multiphysics simulation and artificial intelligence can become a decisive competitive advantage.

Product development today faces increasing complexity: performance improvements and efficiency gains can only be achieved through a deep understanding of the physical interactions between mechanics, electricity, thermodynamics, control engineering, electronics, and software. From plastics to household appliances to electric motors, various physical phenomena must therefore be considered and incorporated during the development of a single product.

Multiphysics simulation enables a realistic, holistic consideration of these disciplines. At the same time, the use of artificial intelligence opens up new ways to accelerate development processes, automatically identify optimization potentials, and make data-driven design decisions. Together, they form the foundation for a new generation of intelligent and innovative products, paving the way for fully digitized and interconnected product development.

What Role Does Multiphysics Play in the Industry?

In the development of construction and agricultural machinery, not only the close integration of hydraulics, mechanics, and electricity is crucial, but also the interaction with the ground plays a central role. For example, agricultural equipment must handle granular materials such as soil or seeds, and snow groomers must interact with snow and the ground. In the development of these devices, particle and flow simulation (CFD—Computational Fluid Dynamics), sometimes coupled with each other, play an important role. To provide meaningful results for development engineers, an efficient interplay of various simulation disciplines is essential.

The situation is similar in the pharmaceutical sector: an integrated multiphysics simulation can drive pharmaceutical development and production. In pharmaceutical manufacturing processes such as form filling, tablet coating, or powder mixing, the Discrete Element Method (DEM) is used to simulate discrete matter, which can represent the dual nature of the liquid and solid properties of granular material flow. Using virtual tests, engineers can gain crucial insights into the interaction of materials and equipment under a range of operating and process conditions.

What Requirements Must Be Met for Realistic Simulations?

The prerequisite for a realistic representation of particles for applications in agriculture, batteries, and the pharmaceutical industry is the provision of physics-based models that can purposefully improve workflows. DEM software such as Altair EDEM precisely simulates the behavior of granular materials and provides crucial insights into processes that are difficult to capture experimentally.

Why Does Multiphysics Play A Significant Role in Drive Technology?

The strong dependence of various subsystems and components, and thus physical domains, on each other is particularly evident in the design of electric drives—electromagnetics in the stator & rotor, mechanics in the drive shaft, and electronics & control in the inverter.

An excessive simplification or neglect of a component can significantly distort the evaluation and potentially lead to suboptimal or even incorrect design decisions. The drive serves as a means to perform a mechanical task, often resulting in a design that is heavily mechanically oriented. Designers frequently reach their limits in the realm of electricity, voltage, and high-frequency switching processes. However, it is precisely here that the optimization potentials lie, which can only be fully exploited through a perfect interplay of electricity and mechanics.

Electric Drive Technology

Why Multiphysics is Becoming Increasingly Important

To ensure efficiency and reliability throughout the entire development process and operation, drive systems must be designed with both usage and manufacturing in mind. Many manufacturing steps can lead to deviations from ideal performance and, in the worst case, to system failure. Typical examples include poor solder joints in electronics, magneto-mechanical stresses on rotor and stator sheets, as well as damage to the insulation or enamel coating of windings. A holistic consideration of all disciplines is crucial for the success of product development.

How Does Multiphysics Optimize Complex Interactions?

This is precisely where multiphysics simulation comes in, as the interaction of the various domains can be easily evaluated and optimized within the overall system using the appropriate tools. With Altair solutions, the modeling depth can be adjusted according to the task, ensuring that the required accuracy of the results and the computational effort needed are always in the right balance.

• For certain applications or effects, such as disruptive sound radiation or increased heat losses due to higher frequency components in the current from the inverter, the direct coupling of a dedicated solver for power electronics (e.g., Altair PSIM) with an FEM code for the electromagnetics of the electric motor is, for example, useful.
• For other tasks and influences, such as non-linear motor inductances and harmonics resulting from motor geometry, reduced order models (ROMs) based on high-fidelity models are derived and combined into an overall system model. Various technologies can be used to create ROMs, including equivalent circuits, static characteristic maps, response surfaces, state-space matrices, or even neural networks (AI models).

For all these approaches, suitable interfaces are crucial—ideally, ones that work independently of the tools being used. Only in this way can a seamless data flow between different simulation environments be ensured, and the full potential of multiphysical modeling be leveraged.

Which Trends Are Currently Shaping Multiphysics?

Two central trends are currently setting the course:

1. On the one hand, improving the interaction of various simulation disciplines (e.g., DEM with CFD or CFD with MKS). This requires optimizing the exchange of models from different manufacturers based on standardized model exchange formats.
2. On the other hand, the focus is increasingly shifting to the acceleration of simulation. This can be achieved on the hardware side through the use of high-performance computing, or through AI technology such as Altair romAI or Altair physicsAI to speed up physical simulations and make simulation accessible to non-experts.

How Do AI And Multiphysics Work Together?

Although various tools are already being combined for holistic system analysis, comprehensive optimizations are often still associated with high computational effort. This process becomes significantly more efficient through the use of AI methods in combination with optimization tools: detailed simulation models can be linked with automation tools and design-of-experiment approaches to create a data foundation for training machine learning models with just a few simulations. These models deliver results many times faster than the original simulations and can be integrated into optimization loops—enabling comparisons of hundreds to thousands of variants in the shortest possible time.

In short: AI plays a very important role in connection with multiphysics.

In short, AI plays a very important role in the context of multiphysics. It will not replace physical simulation, but it will help make better decisions earlier and faster by allowing larger parameter spaces and significantly more variants to be considered with less effort. It is important that the results can be reviewed by a domain expert, meaning AI approaches must be easy to use and must enable data from various simulators to be utilized for training neural networks.

Furthermore, AI can be used to generate behavioral models for physical phenomena based on measurement data, which can then be used within a multiphysical overall system model. This applies, for example, to aging processes in batteries, which are captured in (long-term) measurements but are difficult to describe analytically.

Only those who understand the complex physical relationships and simultaneously leverage the potential of intelligent algorithms can purposefully optimize products and shorten development times.