MODELLING AND SIMULATION
effectively learns to perform a task through repeated trial-and-error interactions in a dynamic environment, allowing it to autonomously make decisions. But reinforcement learning requires a lot of training in order to reach acceptable performance. ‘Even for relatively simple applications, training time can take anywhere from minutes to hours or days,’ Martin added. That won’t stop reinforcement learning
Virtual commissioning, seen here using MapleSim and B&R Automation studio, allows engineers to see real-time 3-D visualisations of their machine operation, which can serve as a test platform for control strategy testing
damage to their machine – a cost they wanted to avoid. By using results from virtual
commissioning, the company could identify the precise loading requirements for its new motors and motion profile, eliminating the added costs of oversized motors. ‘Furthermore, the machine’s control strategy was thoroughly tested against the dynamic model, preventing the risks of damaging the actual machine during testing. With our help, the new machine could be offered at a lower price than before, while still offering the same high standard of reliability required in the injection molding industry,’ Harduwar added.
Next-generation trends What’s next for design tools? We can expect many of these trends to continue growing and expanding, as Dagg explained: ‘We anticipate greater cooperation between departments and specialised domains. To support this, we will be continuing to extend our toolset to drive increased early simulation at more organisations, regardless of who performs it. Our goal is to enable greater product innovation through better collaborative decision making across product teams of designers, analysts, and manufacturing engineers.’ ‘The ‘democratisation of simulation’
isn’t just about putting some simulation functionality into tools for designers. It must address the larger goal of making simulation an effective tool to perform rapid ‘what if’ studies at the speed of concept design,’ he added.
www.scientific-computing.com | @scwmagazine Simulation and modelling will continue
to lower a primary barrier to successful AI adoption – lack of data quality – according to Martin. ‘We know training accurate AI models
requires lots of data. While you often have lots of data for normal system operation, what you really need is data from anomalies or critical failure conditions. This is especially true for predictive maintenance applications, such as accurately predicting remaining useful life for a pump on an industrial site,’ he explained. ‘Since creating failure data from
”With virtual commissioning, the engineers developing control code can use machine simulation models as virtual test platforms for validating their control code”
physical equipment would be destructive and expensive, the best approach is to generate data from simulations representing failure behaviour and use the synthesised data to train an accurate AI model. Simulation will quickly become a key enabler for AI-driven systems.’ Another emerging form of machine learning – reinforcement learning – will also move into the mainstream. Using reinforcement learning, a computer
from extending its reach in the design tools space. Martin explained: ‘This year and beyond, reinforcement learning will go from playing games to enabling real-world industrial applications particularly for automated driving, autonomous systems, control design, and robotics. We’ll see successes where reinforcement learning is used as a component to improve a larger system.’ ‘Key enablers are easier tools
for engineers to build and train reinforcement learning policies, generate lots of simulation data for training, easy integration of reinforcement learning agents into system simulation tools and code generation for embedded hardware,’ he added. An example is improving driver
performance in an autonomous driving system. AI can enhance the controller in this system by adding a reinforcement learning agent to improve and optimise performance – such as faster speed, minimal fuel consumption, or response time. ‘This can be incorporated in a fully autonomous driving system model that includes a vehicle dynamics model, an environment model, camera sensor models, and image processing algorithms,’ Martin added. Remote working is another trend, which will transform not only how we work but also our simulation and modelling tools. Brown added: ‘We need to not only provide take-home licenses but also ensure the security of our software solutions as people increasingly work remotely. To protect our customers, vendors have to ensure the software is free from any trapdoors or entry points for malicious players. At Siemens, every business line has a security officer to provide this diligence.’ As the years pass, simulation will
continue to provide cost savings, according to Harduwar, who said: ‘Fixing issues with mouse clicks is always going to be cheaper than fixing physical machines. For that reason, we’re expecting that virtual design techniques will continue to grow, and adoption will increase across industries.’
Summer 2020 Scientific Computing World 31
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38
