It all starts with selecting the right industrial IoT vibration sensor and developing the fitting data-processing algorithms; each sensor type has its own characteristics. In this blog, we will discuss available sensor types, implementation challenges, and data processing trade-offs.
Vibrations in industrial machinery and assets include the back-and-forth movement or oscillation of components or moving parts. We can identify vibrations in different types of industrial equipment, including heavy-duty vehicles or machinery, and assets with moving parts, such as rotary equipment (pumps, etc.) or machines with linear slides or stages. In extreme cases, you can feel these vibrations, for example when a housing is shaking, or even hear them, as when something is rattling. These vibrations are often signs of wear or tear in the industrial machinery or asset, and can therefore be used as indications of the need for maintenance in a predictive approach.
IoT vibration sensors can measure vibrations in different ways: as a displacement, a velocity, or an acceleration. When acceleration is measured, it can be integrated to velocity and doubly integrated to (periodic) displacement, which is what a vibration effectively is. The spectral content can be derived from the measurements: what frequencies are present in the vibration signal? This is important information, especially the changes that occur in the frequencies due to external (potentially maintenance-related) influences.
Vibration sensors can be used in an industrial IoT (Internet of Things) setting for collecting vibration data. This data can subsequently be analyzed either locally or in the cloud, for predictive maintenance or other purposes.
There are a large number of different measurement principles on which vibration sensors can be based, and they appear in a wide variety of applications. Let’s dig a little deeper into the different types of IoT vibration sensors.
Physical measurement principles
Each vibration sensor is based on a specific physical measurement principle and comes with its own pros & cons. Here are just a few examples:
- Accelerometers measure accelerations and, in some cases, are combined with a gyroscope in what is known as an inertial measuring unit. They are the most popular sensor type for vibration measurement in IoT applications, covering a wide vibration spectrum from low to high frequencies.
- Strain gauges measure displacements, but their mounting is not straightforward. They require electronics for converting the measured change in electric resistance into a mechanical displacement value.
- Eddy-current sensors measure the electric current that is caused by a changing magnetic field (or in the case of vibrations, a displacement in a static, non-uniform field). Their drawbacks include a high power consumption and the necessity of a clean (lab) environment. On the other hand, their contactless operation is a plus.
- Laser displacement sensors enable the non-contact measurement of a target’s height, position, or distance, and any (vibratory) changes therein. They are very accurate – and equally expensive.
As a subcategory of vibration sensors, accelerometers also come in different types. For example, the piezoelectric effect, i.e. a force-induced displacement being translated into an electric signal (or vice versa), is applied in piezoelectric and piezoresistive sensors. Overall, accelerometers offer good performance, but they come with a hefty price tag, which can be a setback for IoT solutions employing large numbers of sensors.
An inexpensive alternative is offered by MEMS (micro-electro-mechanical systems) technology. In its simplest form, a MEMS vibration sensor contains a cantilever beam with a seismic mass that starts to move under the influence of external accelerations. Its deflection can be determined by measuring the capacitance variation between the mass and a reference position, providing a reliable measurement. When the MEMS sensor has a digital interface, no custom front-end analog electronics have to be developed.
MEMS vibration sensors also come in a lot of flavors. The criteria for choosing the right MEMS sensor include:
- The specific type of signal to be measured.
- The intensity and frequency content of the vibrations.
- The conditions under which the sensors must operate.
- Their power consumption and the available communication network protocol.
It requires a careful study of the various data sheets to select the best option in terms of resolution, price, availability, power consumption, etc. Luckily, availability is usually not much of an issue, even in the current global supply chain crisis.
Measuring vibrations is one of the most common ways to monitor asset condition and establish machine health. As there are many different types of vibration sensors and many different applications, there are many factors that determine whether a specific IoT vibration sensor is suitable for the job. These include ease of installation, power supply and (wireless) connectivity, as well as sensor data integrity, data acquisition and processing, edge and cloud computing, and cybersecurity. Let’s take a look at some of the factors to consider when choosing an IoT vibration sensor.
The battery life of industrial IoT vibrations sensors is among the first factors to consider. After all, a dead battery is a dead sensor. Vibration sensors are usually battery powered, so power consumption has to be extremely low to achieve a typical battery lifetime of 2-5 years. Therefore, the rate at which data is sampled has to be as low as possible, while still being high enough not to lose any information. Novices in IoT often opt for the highest sampling frequency, “so that we have all the data,” but every bit that is collected, processed and dispatched, takes a bite out of the battery energy budget.
Another important factor to be considered when choosing an IoT vibration sensor is data processing. To save energy by reducing data processing in the vibration sensor (i.e., edge computing), all the data can be sent to a gateway (mains powered or with a bigger battery) or directly to the cloud. However, to reduce costly data communication, it is preferable to do as much pre-processing and data compression in the edge as possible . Taking into account the specific application, this trade-off has to be made. One way of saving energy is to design a sensor system that is “asleep” for most of the time, receiving a kind of wake-up trigger to do the relevant measurements in time.
Incorporating the cloud into an IoT solution means that data communication issues have to be resolved. Basically, there are two questions to be answered:
- What are the (wireless) connectivity options available?
- What is the coverage of wide-area networks for traveling assets or machines?
A common type of long-range connectivity is provided by LTE (Long-term evolution), a standard for wireless broadband communication for mobile devices and data terminals. Customized for machine-type communication, it has evolved into LTE-M (LTE for Machines), a telecommunications standard used in IoT networks that can connect battery-powered devices directly to the 4G network, without gateways.
In the Netherlands, we are spoiled with an extensive network infrastructure that provides coverage everywhere, but across the border it can be a different situation. So “local” coverage has to be checked and the IoT solution has to be adapted to that, as part of the edge-cloud trade-off. For example, if coverage is poor, more data processing and storage has to be performed on the edge, while data and information can be uploaded when coverage is present, for example at a base station that has a local WiFi network.
When making the edge-cloud trade-off, cybersecurity has also to be taken into account. The more communication with the cloud required, the more often there are “open lines.” This offers cyber criminals more opportunities to hack the system, and this in turn will demand that more provisions are taken for securing the communication. Ultimately, this will mean that even more data communication costs are incurred.
Calibration and mounting
Once an adequate vibration sensor has been selected, implementation can start. The first step is calibration. Sensors may come with factory calibration, but for these mass-produced commodity items, manual calibration is usually required. Sometimes an offset correction is advised for improving the data accuracy and reliability. Mounting the sensor then also requires careful attention. After all, vibrations are of a mechanical nature, so the mechanical mounting of the sensor has to be correct, otherwise things in or around the sensor can start to rattle, giving rise to artifacts in the sensor signal.
Data processing and analysis
The next and perhaps biggest challenge is data processing and analysis. This is where the real magic starts: cleaning up and filtering the data (removing noise and artifacts), and averaging, while taking a deep dive into the data.
The main approach to processing the data may be the same across different projects, but the devil is in the details. The signal that is derived from the vibration sensor not only depends on the type of sensor and the type of asset or machine it is mounted on, but also on the exact manner of mounting, including the housing or casing that is involved. In fact, the complete mechanical set-up of an IoT vibration sensor system can make or break the IoT solution, a lesson that is usually learned in practice, either the gentle or the hard way.
Then, the ultimate question is what can be learned from these signals:
- Which characteristic of the data coming out of that sensor can be attributed to which mechanical effect?
- Does the data convey something about the condition of (a specific part of) the asset or is it just an artifact?
Learn more about the 8 most common predictive maintenance challenges
Finally, it’s time to run the application; i.e. use the IoT vibration sensors for real-time monitoring of the health of machines and assets. In order to detect machine failures before they happen, it is important to understand what is causing the vibrations. Here are some examples:
- Wear and tear through overuse, or poor installation.
- Unbalancing of parts through overuse, or poor installation.
- Looseness of parts or bearings.
- Misalignment of parts.
These types of malfunctioning, each with their specific cause and effect, are typically observed in equipment such as motors, pumps, compressors, rail tracks, automatic doors, elevators and escalators, and many other types of industrial machinery.
Regardless of the cause of the vibrations of the machine or asset, what really matters is anomaly detection; finding anomalies in measurements that can indicate (imminent) faulty behavior. Depending on the application and the nature of the data, vibration sensor readings can be used for straightforward detection of an anomaly, for example when the vibration intensity exceeds a predefined limit.
Alternatively, the readings are used to analyze, for example, the vibration frequencies, using either standard statistics or a machine-learning algorithm to detect any changes. If changes in the vibration spectrum exceed a certain quantitative or qualitative threshold value, then anomalous behavior is detected. Expert judgment may still be required to assess whether a detected anomaly is real, or just a (sensor) artifact.
It will have become clear by now that deploying IoT vibration sensors in an IoT solution for predictive maintenance is far from plug & play. The hardware itself may be a commodity, but the selection of the proper sensor and its implementation requires due attention. Then, the real challenge lies in data processing and analysis. Data on its own does not tell the story: it takes a multidisciplinary effort, including domain experts, to convert vibration data into useful information.
If you want to use this information to create business opportunities for improving machine diagnostics, we can help you with selecting and implementing the best IoT vibration sensor solution. Get in touch with our experts to learn more about the opportunities that IoT vibrations sensors offer.