This marks the first significant step forward in biomedical ultrasonic applications since the early 1990s. Meet MultiphysicsAI—an advanced, cloud-based, finite element method (FEM) simulation technology powered by neural surrogate models. This cutting-edge technology is changing the way piezoelectric micromachined ultrasonic transducers (PMUTs) are designed. Eliminating the previous trial-and-error approach, it integrates a new systematic inverse optimization method, increasing efficiency and precision in design.
MultiphysicsAI is specifically tuned to optimize PMUTs relying on training database constituted of 10,000 randomized geometries. As a result, it generates such AI surrogates with an incredible mean error of just 1%. The platform reaches sub-millisecond inference for most performance indicators including transmit sensitivity, center frequency, fractional bandwidth and electrical impedance. This achievement is a major step toward using AI to accelerate the design and performance of medical ultrasound devices.
Enhanced Design Capabilities
With MultiphysicsAI, engineers can make informed design decisions across competing design dimensions. The tool provides a high level of flexibility and can quickly evaluate multiple configurations, especially in the tradeoff between sensitivity and bandwidth. The novel technology has effectively doubled fractional bandwidth from 65% up to a trailblazing 100%. It introduced 2-3 dB increased sensitivity.
The platform consistently achieves a center frequency of 12 MHz. It functions at a very narrow tolerance of ±0.2%, underscoring its potential to generate superior quality ultrasonic devices. Such precision of this order of magnitude is important, especially for medical applications where the capability of the device affects the health of patients.
Accelerated Performance Improvements
Perhaps the greatest strength of MultiphysicsAI is its computational speed. Further contrasting traditional approaches, which may take days, it provides real-time validated performance improvements in minutes. This quick turnaround speeds up the research and development pipeline. Beyond saving taxpayer dollars, it provides tremendous engineering efficiencies reducing the amount of time engineers spend on nonproductive design iterations, so they can focus on innovation.
By building on top of proven, standard cloud infrastructure, MultiphysicsAI provides extensive scalability and accessibility to users in academia, government, and industry biomedical engineering. By democratizing the use of sophisticated simulation tools, organizations of all sizes can leverage this groundbreaking technology. They can finally do so without huge hardware investments.
Future Implications
The implications of MultiphysicsAI go well beyond efficiency gains. The technique’s speed would create new possibilities for research and development of new PMUT-based ultrasonic technologies. They can immediately investigate a larger variety of design alternatives. This provides them with the unique ability to customize devices in ways that better address particular medical conditions.
With the launch of MultiphysicsAI, the design process of PMUTs is revolutionized. It has the potential to revolutionize biomedical ultrasonic applications by improving robustness and decreasing development cycles.
