Condition Monitoring

Early Fault Detection and Predictive Maintenance

This case study demonstrates motor fault diagnosis and predictive maintenance using a MEMS accelerometer equipped with the latest MEMS sensing technology. The MEMS accelerometer mounted on the motor can be connected directly to an industrial Raspberry Pi with a built-in STM32 microcontroller, or connected via a wireless sensor box. In this project, measurement data is transmitted from the wireless sensor box (sensor node) to a wireless gateway installed on a PiLink PL-R4 industrial Raspberry Pi via the 920 MHz sub-GHz band. The system acquires nearly 20,000 data points per measurement at a sampling interval of 50 μs and transfers the data at approximately 300 kbps. Measurements are performed about five times per day. At this frequency, the wireless sensor box can typically operate for 3 to 5 years without battery replacement.
The raw data received by the industrial Raspberry Pi is processed locally using open-source software such as SciPy. Mathematical processing such as Fast Fourier Transform (FFT) and envelope analysis is performed as needed on the Raspberry Pi, followed by graph generation, fault diagnosis, and data file transmission to upper-level systems such as the customer’s server.

Although this method does not require a database of machine-specific parameters for each motor, it enables highly accurate early fault detection by continuously measuring and comparing fault evaluation indices from the initial healthy operating condition.

The condition monitoring application is provided free of charge, including source code. Users are free to customize the screen layout, diagnostic algorithms, and add machine-learning logic as needed.

Data collected from sensors and equipment controllers can also be processed using various AI libraries. This enables the development of AI algorithms optimized and precisely tailored to the customer’s equipment without relying on expensive commercial AI software.

PiLink works closely with industrial machinery manufacturers, sensor manufacturers, end-user maintenance teams, and other partners. In addition to custom expansion board development and prototyping, and contract software development for measurement and AI-based diagnosis, we also support system integration with customer equipment controllers, PLCs, major cloud services, on-premises file servers, AI servers, and other systems. PiLink provides added value beyond hardware supply.

Measurement Innovation Enabled by MEMS Sensors

MEMS (Micro Electro Mechanical Systems) refers to IC chips that integrate miniature mechanical sensing elements for sensing functions. This technology has greatly contributed to the miniaturization and cost reduction of sensors. While common MEMS sensors such as ToF distance sensors and temperature/humidity sensors are already well known, innovative MEMS sensors suitable for industrial applications have recently emerged, including Doppler sensors for precise velocity measurement and gas sensors for detecting hazardous substances in the air. MEMS sensors typically use low-voltage analog outputs, SPI, I²C, and other standard interfaces, allowing them to be driven directly by microcontrollers such as the STM32 mentioned above. For example, a MEMS accelerometer sensor equipped with a MEMS acceleration sensor chip capable of measuring up to 20 kHz can be connected directly to a sensor-interface model of PiLink’s industrial Raspberry Pi without additional converters. By combining MEMS sensors, high-performance STM32 microcontrollers, and industrial Raspberry Pi platforms, it is now possible to build advanced IoT measurement systems at dramatically lower cost.

Motor Fault Diagnosis and Predictive Maintenance Using High-Performance MEMS Vibration Sensors

MEMS accelerometers and gyroscopes are already widely used in smartphones and other devices. However, general-purpose MEMS accelerometers typically offered measurement bandwidths of only 5 to 10 kHz and relatively high noise levels, making them insufficient for motor and rotating machinery diagnostics. Harting’s MEMS-based vibration sensor achieves high-frequency bandwidth, low noise, and high sensitivity beyond the conventional limits of MEMS accelerometers by combining a next-generation industrial MEMS accelerometer capable of measuring frequencies beyond 20 kHz with proprietary high-order filter circuitry, while maintaining a price range in the 20,000-yen class. This makes it possible to detect high-frequency abnormal vibrations caused by bearing seizure and other mechanical faults. By connecting this sensor directly to a PL-R4 sensor-interface model equipped with a microcontroller, or to a wireless sensor box, high-speed sampling at 20 kHz (50 μs intervals) can be achieved.