Keeping Equipment Healthy by Sensing Things that Go Bump

Remotely monitoring sensor data enables businesses to protect their bottom line, from detecting unauthorized tampering with remotely installed assets to identifying the first signs that manufacturing equipment needs to be adjusted. Vibration sensing can provide valuable intelligence in either role, whether simply to trigger an alarm or to enable detailed analysis of wear in machinery.

A number of techniques are viable for vibration monitoring. A low-cost threshold sensor can be created simply by coiling a spring around a metal pin. When exposed to vibration, the spring contacts the pin and closes the switch. On the other hand, multi-axis MEMS motion sensors have a large bandwidth and wide dynamic range suitable for monitoring machine health.

Basic Theft Prevention

A simple tamper detector can be built using low-cost hardware such as the MikroElektronika Vibra Sense click board, which contains a simple threshold sensor. The board has a digital output, which is capable of indicating the intensity of the vibrations detected and can be set as an interrupt. An on-board potentiometer enables the user to adjust the interrupt threshold. The Vibra Sense board provides a cost-efficient means of implementing an anti-tamper or anti-theft mechanism capable of initiating a response to events such as tilting of a gaming or vending machine, or attempts to remove almost any type of unsupervised, networked equipment. This could include equipment installed outdoors, such as traffic counters, air-quality monitors, weather recorders, or any other appliances used for remote monitoring.

The Vibra Sense sensor connects directly to an embedded processor board via standard MikroBus headers. A wide variety of MikroElektronika computing boards are available, based on well-known microcontrollers from manufacturers such as Atmel, Cypress, Microchip, NXP, STMicroelectronics, and Texas Instruments. Code examples for all MikroElektronika compilers are available from MikroElektronika’s Libstock website.

Monitoring Equipment Performance

More detailed vibration data is needed if the objective is to monitor the health of industrial machinery. Vibration can result from defects such as excessive wear, imbalances or loose fittings, and needs to be corrected before process control becomes unacceptable or – ultimately – the machine becomes unserviceable. To avoid the losses in productivity that can result from either situation, corrective maintenance may be performed at short intervals, but this can be expensive and disruptive. More recently, equipment operators have begun adopting vibration monitoring to tailor maintenance activities to the needs of individual items of equipment. Correlating vibration data with typical wear-out behavior of mechanisms such as bearings, gears, chains, belts, brushes, shafts, coils, or valves enables equipment operators to take quick action when warning conditions are detected.

Until now, vibration monitoring has typically been performed using simple piezoelectric sensors mounted on the machine, or using handheld data collectors. Both approaches have disadvantages: low-noise piezoelectric sensors are expensive and capture only limited information, while handheld equipment introduces the expense of a human operator and the repeatability of tests can be compromised if the probe is not placed in exactly the same position each time. The advent of the Industrial Internet of Things (IIoT), as well as affordable, high-precision, multi-axis MEMS inertial sensors, now allows cost-effective remote monitoring using vibration-sensing equipment that is permanently fixed to the machinery.

MEMS and More on Module

Analog Devices has a vibration-sensing solution that addresses both the local processing and connectivity challenges around implementing vibration monitoring directly on the machine. The ADIS16227 and ADIS16228 are complete vibration-sensing system modules that combine triaxial MEMS acceleration sensing with all the functionality needed (Figure 1) to capture and analyze the MEMS signals, and make the results available via an SPI thereby allowing easy transfer into the host system. Compact dimensions allow the module to be attached at a suitable location on the machine. Using an SPI or timer signal, or other external trigger, the system can be made to wake periodically, record motion data in both axes, generate spectral records, provide access to data and results, and then go back to sleep. Clearly, this type of sensor is more sophisticated than the simple tilt/threshold sensor described earlier, capable of generating detailed, quantified data as well as simple pass/fail threshold results.

Figure 1: Complete spectral vibration analysis is performed in the sensor.

Capable of performing time-domain analysis, including decimation filtering and selective windowing, and subsequent frequency-domain including a 512-point Fast Fourier Transform (FFT) for each axis with FFT averaging to minimize noise effects, the system is able to track changes over time and thereby monitor machine health. The frequency response of the system enables dangerous vibration patterns to be isolated from other vibrations that result from normal operation of the machinery, such as a milling or cutting process. Capturing and storing baseline data from the sensors fitted to any given machine helps to identify significant changes or trends in machine performance.

Because all the signal analysis takes place locally in the device, effects such as transmission noise are unable to contaminate the sensed motion data and thereby compromise accuracy. The module’s embedded capability also reduces the design burden on the system developer, and relieves the computational load on the host system. For most situations, the task is simplified to a process of identifying normal, warning or critical states.

The sensor integrates several user-configurable settings, such as dynamic range, sensitivity and scale, as well as windowing options including Hanning, rectangular and flat-top windows that are commonly used in vibration monitoring. FFT averaging minimizes the vibration noise floor to allow measurement of low-magnitude vibrations. This allows even subtle changes in vibration profiles to be identified. Figure 2 illustrates how the raw data is processed to generate spectral records.

Figure 2: Spectral analysis helps distinguish vibration patterns of interest.

The ADIS16228/PCBZ evaluation board helps accelerate development of machine-health applications using the ADIS16228. The board provides access to the sensor module via a 16-pin dual-row sensor, and has mounting holes that enable the board to be fixed to the frame of the system to be monitored. Evaluation software allows interaction with the sensor-system module via a PC, and gives access to the device for reading register content, setting alarms, and writing and capturing data. Tutorials and video instructions are also available to assist project development.

Figure 3: The evaluation kit for ADIS16228 streamlines development of machine-health monitoring applications.

Conclusion

Highly integrated and easy-to-use smart sensors that combine MEMS devices with signal processing simplify the adoption of vibration monitoring in industry, which has the potential to improve equipment reliability and longevity while also reducing maintenance costs and driving up productivity. In addition, remote access to data from multiple channels, via convenient and license-free RF communications, brings the opportunity to implement vibration monitoring as an IIoT application.

Where simple threshold monitoring is required, such as for tilt or tamper detection, a lower-cost solution like Vibra Sense can be effective, economical, and easy to implement.