Highly parallel multi-physics simulation of muscular activation and EMG

Abstract

Simulation of skeletal muscle activation can help to interpret electromyographic measurements and infer the behavior of the muscle fibers. Existing models consider simplified geometries or a low number of muscle fibers to reduce the computation time. We demonstrate how to simulate a finely-resolved model of biceps brachii with a typical number of 270.000 fibers. We have used domain decomposition to run simulations on 27.000 cores of the supercomputer HazelHen at HLRS in Stuttgart, Germany. We present details on opendihu, our software framework. Its configurability, efficient data structures and modular software architecture target usability, performance and extensibility for future models. We present good parallel weak scaling of the simulations.