STMicroelectronics has launched a new machine-learning–enabled software package that helps developers more easily add AI functions to motor-control systems. The FP-IND-MCAI1 function pack provides a ready framework for implementing condition monitoring and predictive maintenance in applications such as industrial drives, robotics, and household appliances powered by electric motors.
The software runs on the EVLSPIN32G4-ACT evaluation board and includes a complete demonstration project for controlling a low-voltage three-phase brushless motor using field-oriented control (FOC). It also comes with hardware abstraction layers, board-specific drivers, and a pre-trained machine-learning model that can identify different motor states. The model classifies motor behavior into three categories: normal operation, high vibration, and unstable running conditions. This allows developers to experiment with AI-based monitoring without having to build a full machine-learning workflow from scratch.
The evaluation platform supports three-phase brushless DC motors with power ratings of up to about 250 W. It also allows connection of external vibration sensors, including the STEVAL-C34KAT1 module or the STWIN.box, giving developers additional data for condition analysis. Motor parameters and control settings can be configured through the STM32 Motor-Control Software Development Kit, while the machine-learning model can be further trained or adapted using NanoEdge AI Studio to support new classifications or application-specific behavior.
At the hardware level, the evaluation board is built around the STSPIN32G4, a compact device that integrates an Arm Cortex-M4 microcontroller, gate drivers, bootstrap diodes, and protection circuitry in a 9 mm × 9 mm package. In the reference design, this device works alongside a MOSFET power stage, current-sensing amplifiers, and temperature-monitoring components. The platform supports both field-oriented control and six-step commutation, along with three-shunt or single-shunt current sensing. It can also interface with digital Hall sensors or incremental quadrature encoders for speed and position feedback.
The FP-IND-MCAI1 function pack is available for download at no charge, and the EVLSPIN32G4-ACT evaluation board can be obtained through distributors or through the company’s online store. The solution will be demonstrated at the embedded world in Nuremberg.
