The beautiful intersection of simulation and AI

You can check out all the on-demand sessions here.

Simulation can help businesses shorten time to market and lower design costs. Simulation is used for a lot of applications.

• Using a virtual model (also known as a digital twin) to simulate and test their complex systems early and often in the design process.
• Maintaining a digital thread with traceability through requirements, system architecture, component design, code and tests.
• Extending their systems to perform predictive maintenance (PdM) and fault analysis.

Artificial intelligence is being incorporated into model-based designs to improve simulation capabilities. When used together, these two fields create value for engineers and researchers. The strengths and weaknesses of these technologies are perfect for businesses.

Challenge 1: Better training data for more accurate AI models with simulation

Simulation models can synthesise real-world data that is hard to collect into good, clean and cataloged data. Artificial intelligence models are exposed to new data that may not be captured in the training set. Engineers will spend hours trying to figure out why the model isn't working if these models aren't noticed.

Engineers can use simulation to overcome problems. Time spent improving the training data can often yield more extensive improvements in accuracy.

The intelligent security summit is on demand.

Understand the role of artificial intelligence and machine learning in case studies. You can watch sessions on demand.

Engineers can improve a model's performance with an iterative process of simulating data, updating an artificial intelligence model, observing what conditions it cannot predict well, and collecting more simulation data for those conditions.

Challenge 2: AI for new in-product features

Simulation is an important part of the design process for engineers using embedded systems for applications. Virtual sensors are devices that don't measure directly from the available sensors. Engineers are turning to artificial intelligence-based approaches because they have the flexibility to model the complexity of real-world systems. They use data to train an artificial intelligence model that can predict the unobserved state from the observed states, and then integrate that model with the system.

The controls algorithm that ends up on the physical hardware is usually programmed in a lower level language like C/C++. Technical professionals may need to try multiple models and compare trade-offs in accuracy and on-device performance if the machine learning models are restricted by these requirements.

Reinforcement learning takes this approach to a whole new level. Building this type of model requires an accurate model of the environment as well as massive computational power to run a large number of simulations, and this technique has proven effective in some challenging applications.

Challenge 3: Balancing ‘right’ vs. ‘right now’

Time-to-market has been a problem for businesses. The harm to the brand is caused by organizations that push buggy solutions. Also-rans in established markets have difficulty gaining traction. Simulations were an important design innovation when they were first introduced, but their steady improvement and ability to create realistic scenarios can slow the pace of innovation. The risk that the market will have moved on is caused by organizations trying to build simulation models that take too long to build.

Technical professionals need to acknowledge that there will always be environmental nuances that can't be recreated. Even if the models are approximations for high fidelity systems, they should not be trusted blindly.

The future of AI for simulation

For nearly a decade, artificial intelligence and simulation technologies have been building and maintaining their strength. Engineers are starting to see a lot of value at their intersection due to their strengths and weaknesses.

Artificial intelligence and simulation will become more important tools in the engineer's toolkit as models continue to serve more complex applications. The ability to develop, test and validate models in an accurate and affordable way will only continue to grow in use.

At MathWorks, DeLand is the product marketing manager.

The VentureBeat community welcomes you.

Data decision makers can share data related insights and innovation.

Join us at DataDecisionMakers to read about cutting-edge ideas and up-to-date information.

You could possibly contribute an article of your own.

Data decision makers have more to say.