A revolutionary new technology in biomedical engineering has come… Deployment of MultiphysicsAI brings together cloud-based finite element method (FEM) simulation powered by neural surrogates. This exciting new methodology completely changes the way piezoelectric micromachined ultrasonic transducers (PMUTs) are designed. This approach radically simplifies a complex, cumbersome trial-and-error process into a systematic, efficient inverse optimization strategy. By predicting and optimizing dozens of geometric parameters, MultiphysicsAI tangentially but dramatically speeds and improves the design process, providing a higher-order benefit of improved performance and efficiency.
The Power of MultiphysicsAI
Now, MultiphysicsAI makes it possible to find the optimum value of four key geometric parameters over a phenomenal 10,000 coupled piezoelectric-structural-acoustic simulations. That massive ground truth training is what lets the AI get validated step-function performance gains with very little engineering effort. Today, engineers can accomplish in seconds what once took days of a human’s manual back-and-forth. They accomplish this by leveraging off-the-shelf computing power.
The system’s AI surrogates show incredible precision, with a mean error of just 1%. MultiphysicsAI provides fast insights to fundamental metrics such as transmit sensitivity, center frequency, fractional bandwidth and electrical impedance. It does this at sub-millisecond inference latency for critical high-performance computing (HPC) indicators. This increased efficiency leads to a faster overall design process while improving the quality of PMUTs that are developed.
Enhanced Performance Metrics
The implications of using MultiphysicsAI are profound. An ultimate goal of the new system is to enable Pareto front optimization. This improvement would increase the fractional bandwidth of PMUTs, which is already an impressive 65%, to a full 100%. At the same time, it increases sensitivity by 2 to 3 decibels with center frequency of 12 MHz held constant to ±0.2%. These improvements provide more flexibility and increase performance across biomedical applications. Consequently, PMUTs are better suited for applications such as imaging and therapeutic ultrasound.
The investigation of intricate design trade-offs between sensitivity and bandwidth becomes possible with MultiphysicsAI. This capability empowers engineers to make informed decisions about design choices that best meet the demands of specific biomedical applications.
A Practical Workflow Transformation
Integrating MultiphysicsAI into the design workflow provides a feasible solution to optimize PMUTs. The AI-accelerated workflow goes beyond automating the processes of the past to deliver a more insightful view of design efficiencies. This change of paradigm is a huge step not only in engineering but as well in technologists facilitating engineers to design ultrasonic devices for medical applications.
Engineers might utilize this groundbreaking technology to give them a superpower of efficiency. This frees them up to focus on overall design principles rather than hair pulling iterations. Not only does this automated process save time, expediting research and development timelines, it increases accuracy. In doing so, we provide more complex biomedical technologies in a timelier manner.