WHAT IS NEWTWEN?

NEWTWEN is a company that provides virtual sensing solutions. We have developed a new engineering software platform that takes high fidelity models, used for design purposes, to create real-time digital replicas of these components and systems. From these digital replicas you can extract physics-based information like temperatures, vibration, fluid dynamics, that can then be used for real-time control purposes, including thermal management applications.

The technology has been applied to different components and systems, from electric motors to power modules, inverters, chargers, electric heaters and so on. Thanks to virtual thermal sensors, that can replicate in real-time the thermal behaviour of systems and components on granular levels, you can effectively measure the temperatures where real sensors can’t be placed. It becomes a key enabling technology to increase the performance of the thermal management control.

SO YOU REUSE THE SAME MODELS THAT ARE USED IN DESIGN SOFTWARE?

No. That is one ingredient, the starting point, but the model will always be very different from models used in design that can't represent the reality, because you have manufacturing errors, nonlinearities that are not taken into account in your high fidelity models, physics behaviour that can't be modelled or is unknown or uncertain. Our technology takes these high fidelity models as a first principle, but they are then compressed, parameterised and calibrated against test bench experiments. At the end they are transformed into firmware that is implemented into third party platforms, inside the microcontroller units where the thermal management controls are executed.

YOU ARE ONE OF THE FOUNDERS OF NEWTWEN. HOW DID THIS GET STARTED?

It started in early 2020. I was doing my last year of PhD and together with two colleagues, one who was at that time working at Robert Bosch in Stuttgart, and my supervisor from electric engineering at the University of Padua, we wanted to establish a company to bring to market this engineering software platform solution that can create virtual sensors. We started with a focus on virtual temperature sensors and thermal models executed in real-time. Now we are also scaling the technology to different physics, for different applications, like vibrations, electrochemistry, electromagnetics and so on. There are a lot of barriers, but those barriers will also be there for our future competitors. What is a challenge today, will be our competitive advantage in the future.

ONE OF THE BARRIERS, I IMAGINE, WOULD BE TO GET SENSORS INTO SYSTEMS THAT YOU NEED TO EXTRAPOLATE OTHER DATA FROM. OR NOT?

Sensors are always there for functional safety reasons. To obtain very good and calibrated virtual sensors does require some effort, also during the R&D phase of the product itself, but the good news is that this kind of functional testing, endurance testing, all the data required to calibrate good virtual sensors, they are already there. The main audience and Tier 1s in electric powertain components and systems have all the necessary ingredients to work with our recipe, with our technology to enhance their product with virtual thermal sensors to achieve the next generation of thermal management on top of that.

DOES ARTIFICIAL INTELLIGENCE (AI) PLAY A ROLE IN NEWTWEN'S SOLUTION?

Artificial intelligence plays a role in transforming a model as designed to a model as manufactured, considering uncertainties and all kinds of variables that are more statistical based and need to be treated with a more stochastic approach. In our engineering software platform Twin Fabrica there is an API to create these discrepancy models based on AI, but it has to be executed on the edge, so Edge AI. In our solution this discrepancy model is coupled with the reduced order, high fidelity ones from the customer to create a hybrid solution that can gain competitive advantage in terms of accuracy, in terms of real-time feasibility for control purposes and so on. So yes, AI plays a role, but it is not the fundamental core of our technology.

IF AI IS NOT THE FUNDAMENTAL CORE, WHAT MAKES THIS THE RIGHT TIME FOR PREDICTIVE THERMAL MANAGEMENT? WHY HAS THIS NOT BEEN DONE BEFORE? WHY NOW?

Electrification is still a huge trend. If you ask the main players who are providing electric power train solutions and involved in thermal management challenges, they would like to have a real-time running model that can predict the temperature in the most critical points, like the junction temperature in semiconductor models, rotor magnets inside electric machines, the state of the discrete semiconductors inside the onboard chargers, the coolant fluid and the flow rate of the coolant inside the cooling systems, all these variables that are not possible to be measured in production, but should be available if you want to enhance the efficiency, performance and safety of the controls. Today these people are competing on design and material technology, but the main margin is at the control level. The challenge is that control is still blind, since the state-of-art sensing solutions cannot provide accurate insights in the most critical spots at the component level.