MultiphysicsAI has achieved a revolutionary breakthrough to enable biomedical ultrasonic applications. In a notable development from earlier this year, they’ve released a game-changing new recipe for designing piezoelectric micromachined ultrasonic transducers (PMUTs). This transformative platform profoundly changes the design process by integrating cloud-based finite element method (FEM) simulation with neural surrogates. This evolution turns previously heuristic trial-and-error iterations into a data-driven, systematic inverse optimization model. With the help of state-of-the-art artificial intelligence, engineers can more effectively and efficiently optimize PMUT designs.
The example application case study with MultiphysicsAI illustrates the true powerful capabilities of the platform. It provides the lowest weight while meeting optimization criteria that across 10,000 coupled piezoelectric-structural-acoustic simulations. This highly-detailed simulation serves to backup the approach’s demonstrated capabilities. Finally, it surfaces the largest performance wins with the least amount of engineering effort involved. Those manual iterations were labor intensive and time consuming. Today, they take place in a matter of seconds, allowing for the rapid, informed exploration of data on commonly available computational resources.
Enhanced Optimization with Minimal Overhead
With MultiphysicsAI, the process of PMUT circuit and material optimization has forever been altered. It produces AI surrogates that achieve a nominal mean error of just 1%. This degree of precision makes it possible for the designs created to adhere to strict performance standards. As far as inference speed for your KPIs? It’s lightning quick. It boasts sub-millisecond speeds and has metric transmit sensitivity, center frequency, fractional bandwidth and electrical impedance.
The approach’s efficiency is illustrated by its ability to perform Pareto front optimization. This simultaneous antibandwidth process is what increases fractional bandwidth from 65% to 100% while improving sensitivity by 2-3 dB. Most significantly, it adeptly keeps the center frequency at 12 MHz well within the ±0.2% tolerance. These developments mark a big step forward in allowing critical PMUT designs to be promptly optimized, keeping important biomedical applications on track.
Practical Applications and Future Implications
The long-term effects of MultiphysicsAI’s technology go far past just improving performance. This powerful tool delivers an AI-accelerated, real-world ready workflow. It gives engineers and researchers the tools to more efficiently investigate complex design trade-offs between sensitivity and bandwidth. This unprecedented ability allows teams to confidently move forward with decisions grounded in data rather than guessing or months of trial-and-error.
Additionally, training on 10,000 randomized geometries gives MultiphysicsAI a strong foundation to address various design challenges. Leveraging common cloud infrastructure, the platform keeps development accessible and scalable for organizations working to advance the science and craft of ultrasonic transducer development.