Improving condition monitoring and diagnostics and enabling overall system optimization are some of the core challenges people face today when using mechanical facilities and technical systems. This topic is becoming increasingly important not only in the industrial sector, but wherever mechanical systems are used. In the past, machines were maintained according to the plan, and delayed maintenance could be at risk of production downtime. Nowadays, people process a machine's data to predict its remaining useful life. In particular, critical parameters such as temperature, noise and vibration can be used to determine the optimal operating conditions and even the number of maintenance required. This avoids unnecessary wear and tear and enables the early detection of potential problems and causes. With this condition monitoring, the availability and effectiveness of the facility can be tapped into considerable room for optimization, resulting in a decisive advantage. For example, after implementing this monitoring, ABB1 has been shown to reduce downtime by 70 percent, extend the service life of electric motors by 30 percent, and reduce the energy consumption of the facility by 10 percent in one year.
An important component of preventive maintenance is condition-based monitoring (CBM), which typically monitors rotating machines such as turbines, fans, pumps, motors, etc. CbM can be used to record the operating status information in real time. However, failure or wear predictions are not provided. These can only be provided through preventive maintenance, thus bringing about a turning point: with smarter sensors, more powerful communication networks and computing platforms, people can create models, detect changes, and calculate service life in detail.
To build effective models, vibrations, temperatures, currents, and magnetic fields need to be analyzed. Today's wired and wireless communication methods enable facility monitoring across the plant or company. Cloud-based systems open up more analytical possibilities for operators and service technicians to obtain data about machine status information in a simple way. However, the machine must have a local intelligent sensor and communication infrastructure, which is a prerequisite for additional analytical capabilities. What these sensors look like, what requirements need to be met, and what are the key features—this article explores these and other issues.
Life cycle presentation of the machine
Regarding condition monitoring, the following basic questions may need to be considered: How long will the equipment be able to operate before the necessary maintenance is implemented?
In general, logically, the shorter the interval between the discovery of a problem and the start of maintenance, the better. However, in order to optimize operating and maintenance costs, or to fully utilize the maximum efficiency of the facility, professionals familiar with the characteristics of the machine need to be judged by knowledge and experience. These professionals are mainly from the field of bearing/lubrication and have little experience in motor analysis, which is the weakest link. The professional will ultimately decide whether repairs or even replacements should be made depending on the actual life cycle (Figure 1) and the deviation from the normal state of the actual state.
Figure 1: The life cycle of the machine.
Unused machines are initially under what is known as a warranty period. This is an early stage of the life cycle and does not rule out failures at this stage, but the odds are relatively small and are generally related to production failures. Only during the following regular maintenance phases will the appropriately trained service personnel begin targeted interventions. Regardless of the actual state of the machine, they perform routine maintenance on the machine at a specified time or after the specified period of use has been reached, for example, changing the oil for the machine. In this case, the chance of failure during maintenance intervals is still very low. As the machine's service time increases, it gradually reaches the condition monitoring phase. After that, you should be prepared for failure. Figure 1 shows the following six variations, starting with changes in the ultrasonic range (1) followed by vibration changes (2). By analyzing the lubricating oil (3) or by slightly increasing the temperature (4), the first signs of impending failure can be detected by perceptible noise (5) or heating condition (6) before the actual failure occurs. Vibration is often used to confirm aging. Figure 2 shows the vibration pattern of three identical devices over their lifetime. All three machines are in the normal range at the initial stage. However, from the mid-stage onwards, the vibrations increase more or less rapidly depending on the specific load situation; by the later stages, they increase exponentially to the critical range. Once the device reaches the critical range, immediate action is required.
Figure 2. Vibration parameters change over time.
Implement condition monitoring with vibration analysis
Parameters such as output speed, gear ratio and number of bearing assemblies are closely related to the vibration mode analysis of the machine. In general, the vibration caused by the gearbox is reflected in the frequency domain as a multiple of the shaft speed, and the characteristic frequency of the bearing usually does not represent the harmonic component. In addition, vibrations caused by turbulence and cavitation are often detected. They are often associated with airflow and/or flow in fans and pumps and, as such, are generally considered random vibrations. They are usually stationary, and statistically there is no difference. However, random vibrations also have cyclic stationarity and therefore also have statistical properties. They are produced by the machine and change periodically, similar to the situation where an internal combustion engine is ignited once per cylinder per cycle.
Sensor orientation is also critical. If a single-axis sensor is used to measure the main linear vibration, the sensor must be adjusted in the direction of the vibration. Multi-axis sensors can also be used to record vibrations in all directions, but based on their physical properties, single-axis sensors have lower noise, a wider measurement range, and a larger bandwidth.
The need for vibration sensors
In order to make extensive use of vibration sensors for condition monitoring, two important factors must be considered: low cost and small size. Piezoelectric sensors used to be used, but mems-based accelerometers are increasingly used. They offer higher resolution, excellent drift characteristics and sensitivity, and a higher signal-to-noise ratio, in addition to detecting very low frequency vibrations that are almost close to the DC range. It is also very energy efficient, making it ideal for battery-powered wireless surveillance systems. There is another advantage over piezoelectric sensors: the entire system can be integrated into a single housing (system-in-package). These so-called SiP solutions are continuously integrated into other important functions that work together to build intelligent systems: analog-to-digital converters, microcontrollers with embedded firmware (for dedicated preprocessing), communication protocols and common interfaces, in addition to various protection functions.
Integrated protection is important because excessive stress on the sensor element can cause damage. The integrated overrange detection function protects the sensor element from damage by issuing a warning or disabling the sensor assembly in the gyroscope by turning off the internal clock. The SiP solution is shown in Figure 3.
Figure 3: MEMS-based system-in-package (left).
As the demand in the CBM field increases, so does the demand for sensors. For an effective CBM, the sensor measurement range (full scale, i.e. FSR) is typically ±50 g.
Since the acceleration is proportional to the square of the frequency, these high acceleration forces can be reached relatively quickly. Equation 1 proves this:
The variables represent acceleration, f represent frequency, and d represents vibration amplitude. Thus, for example, at a vibration of 1 kHz, an amplitude of 1 μm produces an acceleration of 39.5 g.
As for noise performance, this value should be very low over the widest possible frequency range (from near dc to the middle range of tens of kHz), so that bearing noise can be detected at very low speeds, among other factors. However, it can also be seen that manufacturers of vibration sensors are facing a major challenge, especially for multi-axis sensors. Only a few manufacturers are able to offer multi-axis sensors with bandwidths greater than 2 kHz and low enough noise. Analog Devices, Inc. (ADI) has developed the ADXL356/ADXL357 family of three-axis sensors for CBM applications. This series of products offers excellent noise performance and temperature stability. In addition to the limited bandwidth of 1.5 kHz (resonant frequency = 5.5 kHz), these accelerometers are still capable of providing important condition monitoring readings for low-speed equipment such as wind turbines.
The single-axis sensors in the ADXL100x family are suitable for higher bandwidths. They offer bandwidths up to 24 kHz (resonant frequency = 45 kHz) and up to ± 100 g g range at very low noise levels. Thanks to their high bandwidth, the sensor series can detect most failures in rotating machinery (plain bearing damage, imbalance, friction, looseness, tooth defects, bearing wear and cavitation).
An analytical approach that can be employed for condition-based monitoring
Machine condition analysis in CBM can be done in a variety of ways. The most common methods are time domain analysis, frequency domain analysis, and sharing both.
1. Time-based analysis
In a time-domain vibration analysis, valid values (rms), peak-to-peak, and vibration amplitude are considered (see Figure 4).
Figure 4. Amplitude, rms value, and peak-to-peak of harmonic vibration signals.
Peak-to-peak reflects the maximum skew of the motor shaft and therefore the maximum load can be derived. The amplitude value indicates the amplitude of the vibration and identifies abnormal vibration phenomena. However, the duration of the vibration or the energy during the vibration is not taken into account, as well as the destructive force of the vibration. Therefore, the rms value is generally the most meaningful value because it takes into account not only the duration of the vibration, but also the value of the vibration amplitude. By analyzing the dependence of all these parameters on motor speed, the correlation of statistical thresholds for rms vibration can be obtained.
This type of analysis turns out to be very simple because it requires neither basic system knowledge nor any type of spectral analysis.
2. Frequency-based analysis
Frequency-based analysis allows time-varying vibration signals to be decomposed into frequency components by means of fast Fourier transforms (FTTs). The resulting amplitude and frequency-dependent spectrogram helps to monitor specific frequency components and their harmonics and sidebands (see Figure 5).
Figure 5: Spectral diagram of vibration and frequency relationship.
FFT is a widely used method in vibration analysis, especially for detecting bearing damage. In this way, the corresponding component can be assigned to each frequency component. With the FFT, the main frequencies that generate repetitive pulses when the contact of the rolling part with the defective area causes certain faults can be filtered out. Because of their different frequency components, different types of bearing damage (outer ring, inner ring or ball bearing damage) can be distinguished. However, this requires accurate information about the bearings, the motor and the entire system.
In addition, the FFT process needs to provide discrete time blocks that repeatedly record and process vibrations in the microcontroller. Although this analysis requires more computing power than time domain analysis, it is capable of performing more detailed damage analysis.
3. Combination of time domain and frequency domain analysis
This type of analysis is the most comprehensive because it combines the advantages of both approaches. Statistical analysis in the time domain provides information on the vibration intensity of a system over time and whether they are within the permissible range. Frequency domain analysis monitors speed in the form of a fundamental frequency, as well as harmonic components needed to accurately identify fault characteristics.
The tracking of the base frequency is particularly decisive because the rms values and other statistical parameters vary with speed. If the statistical parameters change significantly compared to the last measurement, the base frequency must be checked to avoid false positives.
For all three of these analytical methods, the measured values vary over time. Monitoring systems may first need to record health, or generate so-called fingerprints. It is then compared with the data that is constantly recorded. In cases where the deviation is too large, or the corresponding threshold is exceeded, a response is required. As shown in Figure 6, the possible reactions can be warnings (2) or alarms (4). Depending on the severity, maintenance personnel may need to work immediately on correcting these deviations.
Figure 6. Threshold and response to FFT.
CBM is implemented through magnetic field analysis
Due to the rapid development of integrated magnetometers, measuring stray magnetic fields around motors is another promising method for condition monitoring of rotating machines. The measurement is non-contact; that is, there is no need for a direct connection between the machine and the sensor. Like vibration sensors, magnetic field sensors are available in single- and multi-axis versions.
For fault detection, stray magnetic fields should be measured axially (parallel to the motor shaft) and radial (at right angles to the motor shaft). Radial magnetic fields are usually weakened by stator cores and motor housings. At the same time, it is significantly affected by the magnetic flux of the air gap. The axial magnetic field is generated by the current of the squirrel cage rotor and the end winding of the stator. The position and orientation of the magnetometer are decisive for the measurement of two magnetic fields. Therefore, it is recommended to choose a suitable position close to the shaft or motor housing. At the same time, it is necessary to measure the temperature, which is absolutely necessary, because the magnetic field strength is directly related to the temperature. Therefore, in most cases, today's magnetic field sensors contain integrated temperature sensors. In addition, the sensor should be calibrated to perform temperature drift compensation correction.
FFTs are used to implement magnetic field-based condition monitoring of motors, just like vibration measurements. However, for motor condition evaluation, even low frequencies in the range of a few hertz to about 120 hertz are sufficient. The line frequency is prominent, and in the event of a fault, the low-frequency component spectrum is dominant.
In the case of a rupture of the lever of a squirrel cage rotor, the sliding value also plays a decisive role. It is load-dependent, ideally 0% when there is no load. When the rated load is used, the value is between 1% and 5% for a normally operating machine, and it increases accordingly in the event of a failure. For CBM, measurements should be made under the same load conditions to eliminate the effects of different loads.
The status of preventive maintenance
Regardless of the type of condition monitoring, even with the most intelligent monitoring scheme, there is no 100% guarantee that there will be no unplanned downtime, failure or safety risks. These risks can only be mitigated. However, preventive maintenance is gaining traction and is becoming an important topic in the industry. It is considered a clear prerequisite for the sustainable success of production facilities in the future. However, to do so, unique technologies are needed, and innovation must be continued to accelerate development. The profit and loss deficit is reflected in the comparison of customer benefits and costs.
Nonetheless, many industrial companies have recognized the importance of preventive maintenance, which is an important factor in determining success and therefore an opportunity to start a future business – and not just in the field of repair services. Despite the enormous challenges, especially in the field of data analysis, preventive maintenance is now highly technically feasible. However, the current preventive maintenance has a strong opportunistic character. It is expected that the future business model will mainly depend on software components, and the share of value added brought by hardware will continue to decline. In short, because of the longer machine run times and the higher value generated, the current investment in hardware and software for preventive maintenance is already worth the money.