Here’s a short collection of the most asked questions.
Contact us to learn more about our solutions.

Frequently Asked Questions

What is the typical experience the customer should expect?

We typically engage in a co-development relationship with our clients; in fact, we enter the early development phase and we accompany the client to series production with the goal to empower the client with a new powerful software tool to integrate into the in-house development process and finally in the embedded system. We see the initial collaboration on three levels:  

1) The engineering phase, where we collaborate closely with the client’s design engineering team to develop, calibrate, and deliver the digital twin model of the target component.  

2) The validation phase, where we generate and deliver the embedded C application and support the client during the experimental validation.  

3) The early production phase, where we ensure together with the client that the embedded firmware solution functions properly in volume production.  

Once the customer has perceived the impact of the technology in terms of increasing product value (the objective of the co-pilot) and is willing to not only industrialize the solution that is the subject of this collaboration proposal but also extend it to other products, we offer a license of our software platform on an annual basis. 

What is the client's typical resource investment during a pilot?

Typically, we need to interface with one simulation engineer who can help us collect all the required data for migrating the 3D high-fidelity model into our platform during the initial phase of the project (expected 2-4 hours of effort) and an application engineer for providing us information regarding the software architecture and the hardware platform as soon as we enter in the validation phase (expected 4 -6 hours of effort). Furthermore, the client should account for at least 2 days of test-bench measurements for collecting experimental measurements and verification. 

What is the duration of a pilot project?

The duration of a pilot project typically spans around three months. This time frame allows for sufficient planning, execution, and assessment of the project’s objectives and outcomes. However, it is important to note that the exact duration may vary depending on the specific requirements and complexity of the project at hand. 

What direct use can the customer make of the digital twin?

Customers can directly utilize the embedded digital twin to enhance the safety and reliability of the device. This is achieved through virtual sensor redundancy, which complements the physical temperature sensors to enable early fault detection and prevent severe damage in the event of a malfunction. Additionally, integrating the digital twin into the control architecture allows for optimized thermal management, focusing on the system’s most critical components. 

How does NEWTEWEN cope with functional safety requirements in the source code?

Typically, we need to interface with one simulation engineer who can help us collect all the

NEWTEWEN is committed to delivering high-quality, safe, and reliable source code. We achieve this through our well-defined and automated twin-generation library, which minimizes human intervention and reduces the risk of errors. Our digital twin is guaranteed to be a stable system, with zero as the equilibrium point. Key features of our safety measures include input validation logic, error checking, and virtual sensor redundancy to continuously monitor deviations between measurements and estimations. We establish acceptable deviation thresholds based on the system’s requirements and desired accuracy levels. 

Each digital twin application is provided as a standalone C-code package. The digital twin function is coded in plain C code, adhering to MISRA-C 2012 standards. To ensure proper operation within the design operating range, our production code is thoroughly tested and validated, including tests against abnormal situations and noisy measurements. Additionally, our code is well documented to facilitate understanding and maintenance. 

Is it necessary to upgrade the hardware to install the digital twin?

Upgrading your hardware is not necessary when installing the digital twin. Our production code is designed to be completely hardware-agnostic and is written in plain C-code, using either floating-point or fixed-point precision. This flexibility allows for seamless integration with any hardware platform. Moreover, we have full control over the model complexity and accuracy during the digital twin generation process. This enables us to optimize the final code size to ensure it fits within the available hardware resources, eliminating the need for hardware upgrades to accommodate the digital twin. 

Can the digital twin exclusively monitor temperatures?

No, the digital twin is not limited to monitoring temperatures exclusively. Our software library is designed to handle a wide range of physical phenomena simultaneously, such as thermodynamics, electromagnetics, and electrochemistry, allowing the simulation of a broad spectrum of quantities. In addition, it is also capable of managing the typical non-linearities that arise in these types of multi-physics problems. This versatility enables our digital twin solution to provide comprehensive monitoring and control for a variety of applications and use cases. 

Is there the need of large amount of training data for calibrating the digital twin?

NEWTWEN’s methodology involves two steps for training the digital twin. Firstly, the digital twin is calibrated against the high-fidelity model, which is a quick and straightforward process due to our deterministic approach. Secondly, the digital twin is compared to experimental data to calibrate its parameters. The amount of data required and the number of test cases needed to achieve an effective calibration depend on several factors, such as the reliability of the starting high-fidelity model, the quality of the experimental data, and errors due to manufacturing and external conditions. However, since the model is mainly physics-based and includes adaptivity features, the training process requires significantly less effort and data than equivalent electrical circuits and fully data-driven black box models.