Accelerating EV Design with Neural Concept’s Deep Learning Based Prototyping

Subscribe To Download This Insight

By Michael Larner | 2Q 2021 | IN-6134

The future of industry product development is reliant on effective and extensive simulation modelling.

Registered users can unlock up to five pieces of premium content each month.

Log in or register to unlock this Insight.


Industrial Firms are now Hastening the Product Design Process with Neural Concept


Neural Concept, based in Lausanne, Switzerland, was founded in 2018 by Dr. Pierre Baque. Today, it employs 18 people whose expertise in computer vision and 3D geometry looks to reduce the time required for the likes of Airbus, Bosch, PSA Groupe, and other Tier One automotive suppliers when designing new products and components.

Historically, simulation engineers received the designs and prototypes and were tasked with optimizing them by testing them against criteria. Today, simulation is no longer a separate task but interwoven into the design process. By collecting and assimilating all the relevant data, as well as including all human-designed machine learning models previously produced, Neural Concept’s deep learning capabilities can help customers to create an optimal product quicker than previous product creations.

Designing EVs with Deep Learning


The move away from the internal combustion engine is now irreversible with automotive original equipment manufacturers (OEMs) leaning on their suppliers to help them to ensure they are not left behind. Neural Concept is working with automotive suppliers to provide new types of components for their clients, including cold plate cooling for electric batteries, heat exchangers, and ventilation systems with the company able to treat data related to, for example, model computational fluid dynamics, heat exchangers, as well as test items in crash conditions and repetitive impacts such as potholes. The objective is to reduce the time taken to produce design options from days, to hours, and even down to just minutes.

Many design projects take weeks or months because they require the customer to create models from scratch and then go through lengthy iterations involving physics simulation solvers. Neural Concept produces designs quicker by using the raw computer-aided design (CAD) data to use deep learning to shortcut the simulation directly and to generate new design suggestions. Neural Concept has developed a set of deep-learning models specifically adapted to these applications, which are inspired by 3D computer vision techniques and by generative deep neural networks. Neural Concept also leverages learnings from previous projects and makes them accessible to design engineers, multiplying the number of simulations. Engineers receive predictions in real time for optimizing the designs which can be used by design and simulation engineers, plus manufacturing teams, to validate.

The solution is licensed as NCS/Expert and NCS/Prod with interfaces tailored to users’ simulation expertise. NCS/Expert enables simulation engineers to utilize all the capabilities to build and make models work. NCS/Prod is used by design teams that want to use the models developed by the experts. The solutions are offered both on-premises and via the cloud.

Convince Simulation Experts to Avoid being Squeezed Out


One of the inhibitors to investing in simulation software for designers is that staff don’t entirely trust the findings and often have their own working practices. By focusing on supporting simulation engineers with NCS/Expert, simulation engineers can retain their position of ensuring what a ‘good’ simulation looks like and have confidence that output from CAD designers and engineering teams will work in practice.

The speed and sophistication of the analytics engine, that can evaluate thousands of permutations in a short space of time, brings big improvements to existing workflows. Neural Concept often works with the incumbent vendors at customer sites in order to plug the gap in automating the design process. At the moment, the company lacks the comprehensive simulation capabilities of the likes of Siemens and Dassault Systèmes, who can provide simulation tools to understand the impact of electromagnetics, fluid dynamics, thermal, and acoustic consideration on product performance, as well as simulations of the bill of materials and costings. However, Neural Concept can digest data generated by all these disciplines to build a rapid response that does not require designers to have simulation expertise.

However, while deep learning in this area remains a specialist area for now, Neural Concept needs to continue refining its deep learning capabilities as other larger vendors develop their artificial intelligence capabilities. But by engaging and proving to simulation leaders that it’s deep learning capabilities are ahead of more generic suppliers should offer Neural Concept protection, for the time being.

For more insights on simulations software and an evaluation of leading suppliers, please refer to the Simulation Software Competitor Ranking CA-1317.



Companies Mentioned