News information

The development trend of MEMS sensing monitoring will no longer be limited to condition monitoring

Release on : Apr 18, 2022

The development trend of MEMS sensing monitoring will no longer be limited to condition monitoring
MEMS sensor monitor
When talking about the combination of sensors and machine learning, it was mentioned before that in order to monitor a certain device, sensors began to be combined with machine learning to perform predictable state monitoring of the device. Of course, our focus here is not on machine learning, but on sensors.

When monitoring the operation and health status of equipment, it is necessary to select the most suitable sensor to ensure that the device can accurately obtain equipment information, and detect, diagnose and even predict failures. Let's take an industrial motor as an example. Bearing damage is a fault that can often be encountered during use. Vibration and sound pressure sensing devices are the most used to detect such faults. The faults in the rotor, winding, etc. are mostly measured by the current transformer when the motor is powered.

Vibration sensor detection range

Vibration sensing can generally be used in motor detection to detect the following faults, bearing condition, gear meshing, pump cavitation, motor misalignment, motor unbalance and motor load conditions. For such faults as unbalanced and misaligned, the noise performance requirements of the sensor devices are not strict, and the bandwidth requirements only need to reach 5× to 10× fundamental frequency. Simultaneous detection of multiple axes; faults such as bearing defects and gear defects have extremely high noise and bandwidth requirements. The noise range must be controlled at <100 µg/√Hz, while the bandwidth requirement is >5kHz.

Faults such as motor unbalance/misalignment can be detected during motor vibration; bearing or gear defects are less obvious, especially in the early stages, and cannot be identified or predicted by increasing the vibration frequency alone. Addressing these failures typically requires pairing vibration sensors with low noise <100 µg/√Hz and wide bandwidth >5 kHz with high-performance signal chains, processing, transceivers, and post-processors.

Comparison of MEMS vibration and piezoelectric vibration

Accelerometers are the most commonly used vibration sensors, and piezoelectric accelerometers have low noise and frequencies up to 30kHz, which is their advantage. The frequency of MEMS accelerometers is generally around 20kHz, which is more advantageous in cost, power and size. Condition detection applications have grown rapidly in recent years thanks to the push for wireless installations, which must consider sensor size, integration, and power consumption. Therefore, although the bandwidth and noise performance of piezoelectric accelerometers are significantly stronger than MEMS accelerometers, everyone prefers MEMS accelerometers under comprehensive consideration.
Sensing Type Bandwidth Noise DC Response
Piezo Accelerometer 2.5kHz-30kHz 1 µg/√Hz-50 µg/√Hz None
MEMS accelerometer 3kHz-20kHz 25µg/√Hz-100 µg/√Hz Yes
(Comparison of MEMS and Piezoelectric)

In detection that requires high bandwidth and low noise, both sensors actually have satisfactory bandwidth and low noise, but MEMS accelerometers can provide DC response, which is not available in piezoelectric accelerometers. DC response can be found in Detects motor unbalance and tilt at very low RPM.

(MEMS accelerometer, ST)

Another point is that the MEMS accelerometer has a self-test function to verify the usability of the sensor itself. It should be said that the smaller size and higher integration of MEMS accelerometers are more in line with the current development trend of condition monitoring.

MEMS accelerometers monitor other more prominent capabilities

In terms of noise and bandwidth comparison, MEMS accelerometers do not show an overwhelming advantage over piezoelectric sensing, but from another perspective, MEMS-based monitoring capabilities are outstanding. In addition to the DC response we mentioned above that can detect very low frequency vibrations in the near DC range, higher resolution, excellent drift characteristics and sensitivity are also more prominent than piezoelectric sensing capabilities.

(MEMS Accelerometer Monitoring Module, ADI)

High-frequency MEMS accelerometers can provide output signals well beyond the sensor's resonant frequency range to monitor frequencies beyond the 3dB bandwidth. This performance relies on the output amplifier, which needs to support a signal bandwidth of 70kHz to support the accelerometer to complete the monitoring of the overclocking range. Aliasing noise is inevitable with amplifiers, so filters are also essential.

The combination of condition monitoring and machine learning is still a big trend

There are many technologies used for vibration monitoring and analysis, such as digital filtering, frequency analysis, etc. No matter which analysis method, the most critical point is how to determine the most suitable alarm point under condition monitoring. When sensing and machine learning are combined, machine learning AI can be used in the fault identification process to create a representative machine model based on data from vibration sensors. After the model is created, it is downloaded to the local processor, and then embedded software is used. Not only the ongoing events can be identified in real time, but also transient events can be identified, enabling the detection of anomalies.

In addition, AI-introduced condition monitoring can correlate vibration monitoring data with other sensor data, and the inferred condition information should be more than the amount of data required for maintenance. Further use of the acquired data can complete more dimensions of equipment analysis, not just a single state monitoring.