Accelerometers can effectively detect the vibrational state of machines. If it is possible to integrate artificial intelligence and machine learning to conduct vibration analysis, they can grasp the abnormal vibration of machines more accurately and conduct preventive maintenance for machines when needed. This text will introduce ADXL1002 accelerometer launched by ADI to help you realize the key role of accelerometers played in IIoT.
Artificial intelligence makes vibration analysis accurate
Using IIoT to monitor machines’ health status is beneficial for personnel to achieve predictive maintenance, conduct failure prediction and thus substantially save operating cost. After using sensors, especially accelerometers, personnel can analyze machines’ operating state instead of replacing worn components at intervals.
Manufacturers often use multiple parameters to confirm the time of starting maintenance operation, which include vibration, noise and temperature measurement etc. Among the measurable physical quantities, vibration spectrum measurement can provide more information aiming at the root of problems in rotating machines (engines and generators etc.) Abnormal vibration may be caused by ball bearing fault, axis misalignment, imbalance, excessive looseness and other problems. Each problem has its own symptom, such as vibration source of rotating machines.
There are multiple vibration analysis techniques at present. Except for digital filtering used for overcoming the parasitic vibration of the process itself or other components of machines, it is also able to use mathematical tools to assist such as calculation of the average, standard deviation, crest factors, kurtosis and other instruments. The analysis can be conducted in the time domain, but only frequency analysis can provide more information about the abnormal phenomenon and reasons. Frequency analysis can even be used to calculate the cepstrum of the frequency spectrum which is assimilated into signal frequency spectrum (inverse Fourier transform is used for calculating the logarithm of signal Fourier transform). However, no matter which analytical method is adopted, the difficulty lies in confirming the optimum alert threshold value in order to prevent the maintenance operation from being too early or too late.
At present, it is able to adopt one method to replace the traditional alarm threshold value configuration, which refers to introducing artificial intelligence in the failure identification process. In the stage of machine learning, the cloud resource is used as the representative machine model created by the data from vibrating sensors. After finishing the model creation, it can be downloaded to the local processor. Using embedded software can not only conduct real-time identification of what is happening, but also identify the transient events in order to detect abnormal phenomenon.
Except for creating model for predictive maintenance, artificial intelligent and cloud visit also opens the door leading to multiple possibilities. By associating vibration measurement data with the data from other sensors (pressure, temperature, rotation and power etc.), it is able to infer lots of information about system state which is far more than the data size required by maintenance. Basic data merging can further optimize the device model, which can be used not only in detecting mechanical failure, but also in handling problems (such as empty conveyer belts, pumps without fluid and mixers without being pasty etc.).
The ADXL1002 accelerometer launched by ADI is a good partner of IIoT, which can not only provide ultra-low noise density in the extended frequency range with two full-scale range options, but also provide optimized industrial condition monitoring ability. The ADXL1002 (±50 g) has the typical noise density of 25 μg/√Hz, stable and repeatable sensitivity and the ability of bearing external shocks up to 10,000 g.
With integrated All-electrostatic Self Test (ST) function and Over Range (OR) indicator, the ADXL1002 can provide advanced system-level functions and can be used in embedded applications. Depending on low-power dissipation and 3.3V-5.25V single power supply, the ADXL1002 can also conduct wireless sensing product design. It also provides 5 mm * 5 mm * 1.80 mm LFCSP packaging and can be operated in the temperature range from -40℃ to +125℃.
Being a single-in-plane axis accelerometer with analog output, the ADXL1002 has the direct current to 11 kHz linear frequency response range (3 dB point) with 21 kHz resonant frequency. With the characteristics of Over Range sensing and DC coupling for achieving fast recovery time, comprehensive electromechanical self test and sensitivity performance etc., it has the temperature sensitivity stability of 5%, linearity of ±0.1% of the full scale range, cross axis sensitivity up to ±1% (ZX), ±1% (YX). Adopting single power supply, it uses the output voltage in ratiometric to supply and has low-power consumption of 1.0 mA as well as power saving standby operation mode in order to provide fast recovery and achieve RoHS compatibility. The main application fields of the ADXL1002 include condition monitoring, predictive maintenance, asset health, test and measurement, Health Usage Monitoring System (HUMS).
ADI also launches the EVAL-ADXL1002Z evaluation board which allows users to conduct quick performance evaluation for the ADXL1002 vibrating sensor. The EVAL-ADXL1002Z is specially designed to be mounted onto the mechanical shaker, which is composed of extra thick printed circuit boards (PCB) with the size of 0.8 inch square. Threaded holes are provided for rigid mounting to the shaker block. This design allows users to evaluate the complete performance range of the ADXL1002 vibrating sensor without soldering the device on a separate test board. The output end provides simple RC low-pass filter and the −3 dB bandwidth is about 20 kHz. The component can be replaced so that users can have the low-pass filter suitable for their own applications at the output of the device.
The EVAL-ADXL1002Z evaluation board has 2 sets of spaced via holes that are used for installing 6 pin headers and can be easily attached to prototyping board or PCB, of which the small size and board stiffness have minimal impact on user system and acceleration measurement.