Why use STM32H7 instead of ESP32 for predictive maintenance?

The STM32H7's 480 MHz clock and 1 MB SRAM enable high-frequency vibration analysis (up to 10 kHz) that the ESP32 cannot match. This captures bearing defect frequencies (BPFO, BPFI) that appear above 1 kHz. For standard rotating machinery monitoring below 1 kHz, the ESP32 is sufficient and cheaper.

What vibration sensor works best with STM32H7?

ST's IIS3DWB accelerometer is purpose-built for vibration monitoring with 6.3 kHz bandwidth and low noise (70 µg/√Hz per datasheet). Connect via SPI for maximum throughput. For lower-cost setups, the ADXL355 (4 kHz bandwidth) via SPI is widely used in industrial monitoring.

Can Edge Impulse handle multi-sensor predictive maintenance on STM32H7?

Yes. Edge Impulse supports multi-axis accelerometer input natively. You can combine vibration (3-axis accel), temperature, and current data into a single inference pipeline. The STM32H7's 1 MB SRAM accommodates the larger feature vectors from multi-sensor input without memory pressure.

Hardware Guide

STM32H7 for Predictive Maintenance with Edge Impulse

The STM32H7 paired with Edge Impulse delivers industrial-grade predictive maintenance. The 1 MB SRAM and 480 MHz Cortex-M7 handle multi-sensor vibration analysis at sample rates up to 10 kHz, while Edge Impulse's spectral analysis pipeline simplifies the DSP and ML workflow.

Hardware Specs

Spec	STM32H7
Processor	ARM Cortex-M7 @ 480 MHz
SRAM	1024 KB
Flash	2 MB
Key Features	Double-precision FPU, L1 cache (16 KB I + 16 KB D), JPEG codec, Chrom-ART Accelerator (DMA2D)
Connectivity	Ethernet, USB OTG HS/FS
Price Range	$8 - $20 (chip), $30 - $80 (dev board)

Compatibility: Excellent

The STM32H7 is overkill for basic vibration monitoring — and that is exactly the point for industrial deployments. The 1 MB SRAM enables complex feature pipelines: high-resolution FFT (2048+ bins), multiple concurrent sensor channels, and large inference buffers. The 480 MHz Cortex-M7 with L1 cache processes 10 kHz vibration data in real-time, capturing high-frequency bearing defects that slower MCUs miss. Edge Impulse has official STM32 support with optimized CMSIS-NN deployment. Their spectral analysis block handles windowing, FFT, and feature extraction automatically — you configure parameters in the web UI rather than writing DSP code. The STM32H7's lack of Wi-Fi is less of an issue in industrial settings where Ethernet or RS-485 fieldbus connections are standard. The STM32H7B3-DK and NUCLEO-H743ZI2 boards are both officially supported by Edge Impulse.

Getting Started

1

Set up Edge Impulse with STM32H7

Flash the Edge Impulse firmware to your STM32H7 board (NUCLEO-H743ZI2 or H7B3-DK). Connect via USB and verify the device appears in the Edge Impulse Studio's Devices tab.
2

Connect high-frequency vibration sensors

Use an ADXL355 (4 kHz bandwidth) or IIS3DWB (6.3 kHz bandwidth) accelerometer via SPI for high-frequency vibration capture. ST's IIS3DWB is designed specifically for industrial vibration monitoring with ultra-low noise.
3

Configure spectral analysis and train the model

In Edge Impulse Studio, set up a Spectral Analysis block with appropriate FFT size (512-2048) and frequency range. Collect data during normal and degraded operation. The classifier learns to distinguish healthy from faulty vibration signatures.
4

Deploy as STM32CubeIDE library

Export from Edge Impulse as a CMSIS-PACK or C++ library. Integrate into your STM32CubeIDE project. The deployment includes optimized CMSIS-NN kernels for the Cortex-M7 automatically.

Alternatives

ESP32 with Edge Impulse

Built-in Wi-Fi for wireless reporting. 520 KB SRAM handles standard vibration models. Lower cost but limited to ~1 kHz vibration analysis — sufficient for most rotating machinery.

STM32F4 with TFLite Micro

Lower cost Cortex-M4 for standard vibration monitoring. 192 KB SRAM and 168 MHz. Sufficient for sub-1 kHz analysis. Better cost-performance ratio for less demanding applications.

Explore More

More STM32H7 guides More Predictive Maintenance guides All resources Find the right MCU

FAQ

Why use STM32H7 instead of ESP32 for predictive maintenance?: The STM32H7's 480 MHz clock and 1 MB SRAM enable high-frequency vibration analysis (up to 10 kHz) that the ESP32 cannot match. This captures bearing defect frequencies (BPFO, BPFI) that appear above 1 kHz. For standard rotating machinery monitoring below 1 kHz, the ESP32 is sufficient and cheaper.
What vibration sensor works best with STM32H7?: ST's IIS3DWB accelerometer is purpose-built for vibration monitoring with 6.3 kHz bandwidth and low noise (70 µg/√Hz per datasheet). Connect via SPI for maximum throughput. For lower-cost setups, the ADXL355 (4 kHz bandwidth) via SPI is widely used in industrial monitoring.
Can Edge Impulse handle multi-sensor predictive maintenance on STM32H7?: Yes. Edge Impulse supports multi-axis accelerometer input natively. You can combine vibration (3-axis accel), temperature, and current data into a single inference pipeline. The STM32H7's 1 MB SRAM accommodates the larger feature vectors from multi-sensor input without memory pressure.

Orchestrate Predictive Maintenance with ForestHub

Devices score condition on-device; ForestHub on the Linux edge gateway aggregates over MQTT/Modbus, reasons across the line, and acts — an inspectable, auditable graph.

Get Started Free