Choose the right accelerometer

Comparing vibration sensor technologies for condition monitoring, which are increasingly being used for high volume, smaller machinery.

By Bjorn Ryden, TE Connectivity October 31, 2018

Accelerometers are used for condition monitoring of heavy, high-end machinery such as windmills, industrial pumps, compressors, and HVAC systems. More recently, though, trends toward digital transformation have increased the call to develop accelerometers for high volume, smaller machinery. This includes applications such as machine spindles, conveyor belts, sorting tables, and machine tools. Doing so, however, requires not only better predictive maintenance but an ability to more effectively manage machine downtime to enhance the customer experience and overall profitability. These changing market dynamics have heightened the need to choose the optimal accelerometer for industrial condition monitoring applications: piezoelectric (PE) accelerometers or variable capacitance (VC) sensors. To support the decision-making process, the following is a review of these two different technologies. Select specification parameters are used to compare functionalities—all of which are critical to deliver long-term, reliable, stable, and accurate performance. Piezoelectric vibration sensors PE accelerometers incorporate self-generating piezoelectric crystals and provide a signal when stressed by external excitation caused by vibrating machinery. Most piezoelectric sensors are based on lead zirconate titanate ceramics (PZT), which are poled to align the dipoles and make the crystals piezo-electric. PZT crystals are ideal for condition-monitoring applications since they offer wide temperature range, broad dynamic range, and wide frequency bandwidth (usable to >20kHz). In today’s market, there are primarily two types of PE accelerometer designs: compression mode and shear mode. Compression mode designs are assembled by stressing the PE crystal into compression-typically by loading a mass on top of the crystal and applying a preload force. These designs are outdated and becoming less popular due to their performance limitations, as the constructions are susceptible to mounting base strain and higher thermal drifts. Shear mode designs typically have an annular shear crystal and annular mass that is secured to a support post. This design offers superior performance because the base is isolated and, with less susceptibility to thermal stresses, stability is improved. Most PE accelerometers offered today are shear mode designs as this is clearly the more effective choice. Variable capacitance sensors Variable capacitance (VC) sensors derive the acceleration measurement from a change in capacitance of a seismic mass moving between two parallel capacitor plates. The change in capacitance is directly proportional to the applied acceleration. VC accelerometers require an integrated circuit to be closely coupled to the sensing element to convert the very small capacitance changes into a voltage output. This conversion process often results in a poor signal-to-noise ratio and limited dynamic range. The VC sensors typically are manufactured from silicon wafers and are fabricated into miniature micro-electro-mechanical systems (MEMS) chips. As indicated earlier, a meaningful comparative review of PE accelerometers and wide bandwidth MEMS VC accelerometers requires deploying critical performance specifications. With both accelerometers having a full-scale range of ±50g, these specifications include: wide frequency response; measurement resolution and dynamic range; long term stability with minimum drift; operating temperature range; packaging options and ease of installation; and sensor output options. Wide frequency response The frequency response of the accelerometers was tested on a SPEKTRA GmbH CS18 HF high frequency calibration shaker with a range of 5Hz to 20KHz. The sensors were mounted securely to ensure accurate results over the full test range. Two sensors of each technology, PE and MEMS VC, were tested to ensure the results were consistent. A maximum ±1dB amplitude deviation was assumed as the usable bandwidth although a tighter deviation of ±5% is often used for bandwidth tolerance. The data indicates that the VC MEMS sensor has a usable bandwidth up to 3KHz while the PE sensor has a usable bandwidth >10KHz (this particular PE sensor was within specification up to 14KHz). It is worth noting that the low frequency cut-off for the PE sensor was at 2Hz while the MEMS sensor responded down to 0Hz since it is a DC response device. Measurement resolution and dynamic range To determine the measurement resolution and dynamic range of the piezo-electric and VC MEMS sensors, the samples were tested in a noise-isolated chamber with state-of-the-art measurement equipment capable of micro-g measurement resolution. The units were installed in the same chamber and tested at the same time to eliminate errors from outside environmental interference. The measurements were conducted at four distinct bandwidth settings and the residual noise was measured at each setting. The results are detailed in Table 1. The measurement resolution and dynamic range were calculated based on a 0.03-10KHz bandwidth and are detailed in Table 2. The resolution of the PE sensors is approximately nine times better than the VC MEMS sensors. This results in a significantly better dynamic range, which enables the end user to detect potential problems at a much earlier stage. Long-term stability, minimum drift With more than 30 years of field installations, the long-term stability of PE sensors has been well demonstrated. PE crystals are known to be inherently stable with long-term drift parameters dependent on the crystal formulation used. While quartz offers the best long-term stability in any PE accelerometer, it is rarely used in condition monitoring applications due to limited output and cost. Instead, PZT crystals are most commonly used in PE accelerometers and are increasingly becoming the choice for crystal in most applications. VC MEMS accelerometers also have wide specification limits for long-term drift depending on the MEMS design structure. A bulk micromachined MEMS sensor will have the best long-term drift but also will be significantly more expensive. As a result, they are typically used only in inertial applications. As a result, MEMS vendors offer surface micromachined VC MEMS sensors for condition monitoring, which are much less expensive but sacrifice measurement resolution and long-term stability. Operating temperature range The operating temperature ranges for PE and VC MEMS accelerometers are comparable. Both sensors are equipped to work within condition monitoring applications where the typical environment spans from -40 to 125°C. In certain extreme installations where a higher temperature range sensor may be required, a charge mode PE sensor would be the recommended choice. Charge mode PE accelerometers that do not include an onboard charge converter circuit can be used for temperatures exceeding 700°C. Packaging options and ease of installation For condition monitoring installations in smaller machinery, the size and mounting options can become an important factor in choosing an accelerometer. Larger machinery typically uses an external TO-5 stud mounted accelerometer but for machinery with smaller bearings and rotating shafts, an embedded or miniature accelerometer is necessary. Most VC MEMS accelerometers are offered in a surface mount technology package, which is ideal for high-volume printed circuit board assembly. The VC MEMS sensors are also offered in small sizes to create more packaging options. PE accelerometers are offered in various configurations. The SMT mount version is available, similar to VC MEMS, but the size of the SMT package typically is larger than VC MEMS designs. PE accelerometers also are offered in rugged TO-5 can packages with a stainless-steel housing. These designs allow for direct mounting to bearing housings or embedded installations. Sensor output options Depending on installation and application, a choice of sensor output signal options may be necessary. Most current predictive maintenance installations require an analog signal from the sensor, allowing the end user to decide which parameters to monitor for a particular piece of machinery. Typically, the signal output is driven by the data acquisition or PLC interface and an analog output (±2 or ±5V) is the most common choice. For installations requiring long cable lengths, however, loop-powered 4-20mA sensors are also common. As Industrie 4.0 focuses on automation and data exchange to reshape manufacturing technologies, digital factories are the future. Digital output signals will become more common, as will smart sensors with onboard microprocessors, to give end users the information they need to make immediate maintenance decisions. These output signal options will be available in both PE and VC MEMS sensor technologies. The clear choice Based on this analysis, PE sensors are the obvious choice for condition monitoring installations, given their proven technologies, reliability, and long-term stability. What’s more, the embedded PE accelerometers are ideal for both low- to high-speed machinery, given their wide frequency response and ability to offer a better signal resolution for earlier failure detection. Terms and definitions Wide frequency response: to detect all possible failure modes in the machinery, for bearing monitoring the frequency response of the accelerometer should be 40 to 50 times the shaft rpm (revolutions per minute). For fans and gearboxes, the minimum upper limit of the accelerometer should be 4 to 5 times the blade passing frequency. Measurement resolution and dynamic range: The measurement resolution of a vibration sensor is a function of the amplitude of the output signal in relation to the broadband noise of the onboard electronics. An accelerometer with superior signal output allows for the measurement of smaller vibration levels in the machinery, enabling the end user to predict a fault much earlier than a sensor with a lower dynamic range. As a rule, the output signal should be 10x higher than the noise level of the sensor for the output to be a reliable measurement. Long term stability: Long-term drift is a shift in the sensitivity or zero output measurement (zero output drift applies to MEMS sensors only). A shift in the sensitivity of the accelerometer could trigger a false alarm over time in the monitoring application, while a shift in the zero-output measurement can have the same effect. Since Piezoelectric sensors do not have a DC response, they are not susceptible to zero drift, only sensitivity drift. MEMS VC sensor can have both zero and sensitivity drifts over time. Bjorn Ryden is a global product manager for vibration sensors at TE Connectivity. This article appears in the IIoT for Engineers supplement for Control Engineering and Plant Engineering. See other articles from the supplement below.