Computer-aided engineering (CAE) is taking an increasing role in design, extending beyond its roots in structural validation to predict every aspect of physics, including the behaviour of constituent materials and the suitability of manufacturing processes. This is contributing to engineers having more data at their fingertips than at any other time in history, says John Janevic, COO of MSC Software.
Digital Twins are an example of that data being harnessed effectively. By combining all data pertinent to the challenge at hand, they can boost innovation by making insights about details of the product’s performance available as early as possible, as well as monitor that performance by feeding post-production data back into the simulated design and testing process, and significantly increasing the technological leaps between generations. You can, for example, explore how different powertrain and control system configurations affect an electric vehicle’s effective range, or understand how swapping steel for composites affect handling, and incorporate those learnings into the next design.
Most engineering disciplines currently lack the means to effectively feed these insights into their design and engineering process, however. Physics-based simulation alone lacks robust mechanisms to assimilate long-term data from physical testing or products deployed in the field, and unless the computational efficiency of simulation is improved by several orders of magnitude, the potential of Digital Twins will remain under-exploited throughout the product development cycle.
Artificial Intelligence (AI) is the key. An AI model which is trained by simulation can be used to deliver insights much faster and more efficiently, making it possible to iterate through more scenarios at a more rapid pace.
It also helps us simulate more: performing bigger, more complex tests interactively so that they can be used for design and optimisation and not just validation. A “what if?” use case like this requires extremely fast and efficient simulations. AI applied well can help achieve this without huge computational cost. Think of completing an automotive crash analysis in seconds rather than hours, enabling more thorough safety testing while simultaneously investigating how changes in material or production affect outcomes.
Combining AI and Machine Learning (ML) with pertinent IIoT sources and measurements allows tools like Digital Twins to capitalise on the massive amounts of data we are generating every day, and makes it possible to bring new innovations to market with ‘Smart Manufacturing’ approaches.
An AI model which is trained by simulation can be used to deliver insights much faster
With the help of AI, we can use virtual and physical test data together to improve design for manufacturing, or to improve quality in serial production. Applying real measurements from production means we have, for example, examined how the properties of a machine part’s material transform through the manufacturing cycle, decreasing its development time and the amount of material wasted.
Much of the data and physics-based simulation required to achieve these Digital Twin goals exists today, as do the data management tools and processes for simulation, materials and IIoT. So why are these not techniques we all use every day – why is so much of our data still not being used to its full potential?
It may be closer than you think. For Digital Twins to demonstrate value and guide product development, we need to have confidence that the behaviour – not merely the appearance – of a product is being taken into account, and that requires open platforms that connect with CAE and physical data being applied throughout the product lifecycle.
Forrester reported late last year that the COVID-19 industry slow-down has forced manufacturers to recognise the value AI presents in optimising the computational resources engineers need to design and test their products. Now, those scalable and open platforms that combine cloud and ML technologies with simulation and physical measurement at each part of the product lifecycle are entering the market.
The 2020s had a rather inauspicious start for most of the manufacturing industry, but as a result will be the first decade to see widespread AI usage within engineering, improving the product design process through new data-driven insights and saving cost and time – a crucial consideration for many engineers under intense pressure to deliver innovation faster than ever.
John Janevic, COO of MSC Software, part of Hexagon’s Manufacturing Intelligence division