Veryst Engineering has used COMSOL Multiphysics and computational fluid dynamics (CFD) to study blood damage in medical devices such as pacemakers, heart valves, and catheters. Using the U.S. FDA’s benchmark nozzle, the team developed two deep neural network surrogate models: one to predict blood velocity and another to estimate hemolysis, where red blood cells rupture and release hemoglobin.
The models use flow rate, converging length, neck diameter, and diverging length as inputs and were trained using finite element simulation data. A physics-informed model also incorporates the no-slip condition at the nozzle walls. The study found that changes in nozzle geometry had little effect on hemolysis, with the approach aiming to assess blood damage across different device designs and operating conditions.
Image generated by: ChatGPT
