Fault Detection using TinyML, Design and Control of Three-Phase Two-Level Inverter

The depletion of natural sources of mineral oil has led to a search for alternative sources, particularly in the transportation industry where electric vehicles (EVs) are gaining traction. However, current EV technology is still relatively primitive and requires significant improvements to ensure better and prolonged service. Efforts are underway to design and develop more efficient EV models, with a key focus on the powertrain. The powertrain, comprising the electric motor, power electronics, and transmission, is crucial for efficient energy conversion in EVs[1]–[2]. Through literature surveys and iterative processes, various parameters for Permanent Magnet Synchronous Motor (PMSM) based drives have been identified[4]. These parameters include electromagnetic speed, peak and RMS current, inductances (Ld, Lq), flux, inertia (J), resistance (Rs), damping coefficient (B), and pole pairs. The control strategy employed is Field Oriented Control (FOC), which encompasses Field-Weakening Control[7] and Maximum Torque Per Ampere (MTPA) as the operating regions. A three-phase 2-level inverter utilizing Space Vector Pulse Width Modulation (SVPWM)[5] for precise voltage control has been selected as the inverter of choice. This design also incorporates Insulated-Gate Bipolar Transistors (IGBTs) as switching devices, along with TinyML for comprehensive powertrain operation efficiency in EVs. The objective of this comprehensive design approach is to develop a three-phase two-level voltage source inverter with fault detection using TinyML (Edge Impulse) that optimizes the performance of electric vehicles to meet the growing demand for efficient transportation solutions.