Siemens Worldwide

Pictures of the Future



Mr. Sebastian Webel
Mr. Sebastian Webel


Mobile: +49 172-7169762

Werner-von-Siemens-Straße 1
80333 Munich

Pictures of the Future
The Magazine for Research and Innovation

The Future of Manufacturing

Toward Models that Accurately Predict Product Behavior

CT researcher Vinay Ramanath in the simulation lab in Bangalore. The aim of his team: a better digital twin.

How accurate is a simulation in terms of representing a future product? Siemens scientists in India are using advanced machine learning and probabilistic techniques to develop a simulation-based algorithm that enable accurate quantification of model parameters with reduced uncertainties in their estimates. Their model calibration technology is set to substantially improve the robustness of simulations and precisely predict behaviors of complex systems.

Today’s most complex products and processes are designed, tested and calibrated in the virtual world before being manufactured in the real world. Model calibration involves creating and simulating software models of future products, eventually leading to the creation of “digital twins”. The software models are validated with test data to ascertain, how their real world counterparts will behave in actual environments. Product designers can then observe critical product behaviors such as what parts can fail at what stages, what ranges of temperatures and loads a product can withstand, etc., so that the insights derived can be employed to design the product for robustness. Further, the models are calibrated to detect anomalies in various parameters including heat transfer coefficients, material properties, loads, boundary conditions and geometric tolerances. Once a model performs in the desired manner, its design is rendered into a real product, which is then manufactured and installed in its real-world environment.

Simulation experts at Corporate Technology in India are researching into new ways to radically improve the predictability of model simulations for future product development, maintenance and optimization.

Heading for Automated Calibration

Digital twins may be blurring the lines between the virtual and physical worlds, but they still need to be systematically calibrated once their physical counterparts begin providing valuable field data. Such data often captures uncertainties that were previously unaccounted for in a computer model’s test data. In the Simulation Solutions Research Group, at Corporate Technology in Bengaluru, India, Vinay Ramanath who is a Senior Key Expert Engineer for advanced probabilistics, says, “Real world uncertainties spare no one, not even digital twins. The differences between simulated and field data need to be eliminated in order to automate the whole model calibration process and increase the predictability of digital twins for real-time maintenance and monitoring.”

“The algorithm will help us to characterize a larger and wider range of uncertainties from field data received by virtual models."

At their Bengaluru lab, Ramanath and his team are using advanced machine learning and probabilistic techniques to develop a disruptive algorithm that will help improve the predictability of digital twins. “The algorithm will help us to characterize a larger and wider range of known and unknown uncertainties from field data received by virtual models. The uncertainties can then be updated in virtual model libraries to implement real-time detection of every kind of problem that a product could encounter during its lifetime.” The algorithm has already demonstrated its abilities in a Siemens business.

Brainstorming before the next development steps: Vidyabhushana Hande (left), responsible for the simulation solutions at CT India, talks with Vinay Ramanath.

Predictions from the Virtual World

The new calibration technique is being developed for applications across all areas of future automation and cyber-physical systems, where simulation models are expected to be essential. At the design, development and manufacturing levels, digital twins based on the new technique will become increasingly predictable to those who calibrate them. As a result, product designers will be able to predict the outcome of product and component behaviors accurately at the design stage and build in robust design margins based on key quality and performance metrics. Dependence on product testing and prototyping will be reduced, as estimates of uncertainty parameters increasingly migrate to the design level itself. This will lead to increased productivity and optimized manufacturing practices. As for long-term product maintenance is concerned, the new calibration technique is expected to significantly offer an “as-is” view of a physical product’s working condition, thereby accelerating the decision-making process for maintenance and monitoring tasks.

Looking ahead, the new calibration technique will allow businesses to exploit the full potential of the digital twin concept, which can only be realized when the virtual and physical worlds become indistinguishable from each other in terms of their actions and reactions.

Tomonica Chandran and Vinay Ramanath