Design of a Linear Permanent Magnet Transverse Flux Motor for Needle-free Jet Injection

This paper proposes the work of analyzing and optimizing three types of linear transverse flux motors in order to build a compact handheld device for needle-free jet injection. The use of finite-element analysis, deep neural network models, and convex non-linear optimization in the study will be explored. The high performance computing infrastructure allowed for rapid creation of 534435 motor design data points via parallel 3-D finite-element analysis jobs. With the available data, a deep regression neural network was trained to make highly accurate predictions (96% accuracy) about any motor with design factors within the given ranges. The optimization process found a number motor configurations which satisfy the given specification but not yet found a practical design for the application. Future effort will focus on widening the search range, and more importantly tuning the optimizer, prototyping and experimental validation.