• AI start-up Secondmind has developed an application-agnostic technology platform for model-based engineering that uses a machine learning approach to solve complex optimization problems.

• Secondmind has focused on solving the automotive industry's time-to-market challenge, promising efficiency gains resulting in an 80% reduction in the number of simulations and 50% faster time to optimize feasible design solutions.

• Powertrain design and development was the first application the company applied its solution to, but the technology platform can also be applied to other segments in the automotive development chain, including software-defined vehicles, ADAS and autonomous vehicles, digital cockpit, infotainment and connectivity.

Secondmind, an AI start-up based in Cambridge, UK, has developed a technology platform for model-based engineering. The platform uses a unique machine-learning approach and cloud-native software to accelerate automotive design and development by solving complex optimization problems through Active Learning. This solution can be applied across the whole product lifecycle, from concept to customer. It is an application-agnostic tool that applies probabilistic modeling to develop automated Design of Experiments (DoE). In the automotive context, it can be applied to the vehicle, system or component design.

There are two products that Secondmind currently provides for the automotive sector:

1.         Secondmind for System Design enables engineers to implement set-based design techniques, where multiple design options are front-loaded, assessed for trade-offs for requirement and design parameters, and then optimized progressively to ensure the final solution meets all requirements.

2.         Secondmind for Calibration develops model-based calibration using machine learning techniques to intelligently automate the Design of Experiments by streamlining data acquisition. It minimizes the amount of data required to generate calibration maps and reduces the time for full-system calibration.

Secondmind has been working with automotive OEM Mazda, which has benefitted from cost and time savings in its design and development process for powertrains. Mazda has used a Secondmind solution to calibrate ECUs for its SKYACTIV engine technology. Secondmind machine learning helps Mazda navigate complexity resulting from tighter emission regulations, increasing customer demands and improvements in its development process and sustainability. The Secondmind solution has helped Mazda cut engine calibration time by up to 50%, reduce data acquisition and processing costs by up to 80% and prototype materials use by up to 40%.

Accelerating automotive development cycles

Initially, Secondmind has focused on solving the automotive industry’s time-to-market challenge where Chinese automakers and suppliers have been raising the bar, shifting vehicle production timelines, from concept to production, to under two years with 40% lower labor costs. The company’s two tools can assist engineering teams in the automotive space to speed up the development cycle through their application in the concept, design and development phases.

In the automotive sector, many legacy OEMs are used to a point-based approach to development where an OEM chooses a single concept for a system and then develops that design sequentially. This involves designing based on engineering judgments. However, in the design review stage, engineers can miss performance targets, resulting in a costly and lengthy design rework loop.

In the lean development practice of set-based design, multiple design concept options are developed in parallel within the design boundary conditions. Engineers then eliminate concepts throughout the design phase. Set-based design, through extensive use of brute-force DoE and simulations, provides flexibility by reducing incidents of late-requirement changes in the product development cycle. However, as the complexity and the number of data points increase, it can become very resource heavy. Large automotive organizations like Toyota follow a set-based design approach, but for smaller OEMs and start-ups, it is highly resource- and time-intensive. This is where Secondmind Active Learning can enable organizations to reduce their development time and cost within the V development cycle, enabling design testing and validation earlier in the cycle.