Secondmind K.K.
AI-powered, data-driven MBD - Are all those calculations and tests really necessary?
Secondmind is an engineering AI company based in Cambridge, UK. Using proprietary machine learning technology based on Bayesian optimisation and Gaussian processes, it derives optimal solutions and design spaces with fewer simulations and experiments, thereby reducing engineering complexity and shortening development times. Its flagship products are Secondmind for DSE, which supports conceptual system design, and Secondmind for Calibration, which supports system tuning. A key feature is its ability to assign quantitative probabilities to each point in the model, thereby enabling decisions to be explained using data.
Secondmind’s engineering AI eliminates the ‘endless, trial-and-error exploration’ faced by many engineers, dramatically streamlining development. It reduces the number of simulations and measurements on physical equipment—traditionally required to determine optimal design values and control maps—by up to 80 per cent, thereby shortening development times.
Products & Services
Secondmind for DSE
-- Can you explain the rationale behind that design approach? Design decisions backed by data-driven AI --
The cost of development rework increases the longer it takes to detect design issues. With conventional methods, design approaches had to be finalised before sufficient evidence was available, making it difficult to provide a logical explanation for ‘why that particular design was chosen’. Secondmind for DSE visualises the design space that meets all requirements with minimal simulation, thereby demonstrating the impact of each parameter on performance through data and providing solid justification for the optimal design approach. Changing the requirements immediately reveals their impact, enabling confident reassessment without further simulation.
Secondmind for Calibration
-- Can you explain the rationale behind that suitability? Suitability backed by data-driven AI --
Secondmind’s AI technology builds models from small amounts of measurement data and selects only high-value experiments. As it assigns a quantitative confidence score to each point on the model, areas of uncertainty can be identified at a glance. Even when engineers with limited experience are in charge, they can explain the rationale behind the fitting results—including which constraints were taken into account and what optimisation was performed—enabling rapid decision-making. It also reduces the number of experiments by up to 80 per cent, thereby shortening the fitting period.