Find out when to use condition monitoring, what technologies work best to eliminate unplanned downtime and reduce risk, and what factors to consider when selecting a condition monitoring system.
From food and water to fuels, paper and medicines, countless machines power critical industrial processes that nourish our lives. These mechanical assets are often subjected to harsh conditions, process changes and simple wear and tear. For that reason, it’s important to keep an eye on the actual condition of machines in order to plan and schedule proper maintenance actions before they break down. This is where various condition monitoring technologies support maintenance teams in preventing unplanned downtime, optimizing maintenance scheduling, saving on maintenance costs and increasing asset reliability.
As the name suggests, condition monitoring is the name for a class of techniques that measure one or more physical parameters of equipment’s condition—current, voltage, lubricant, temperature, vibration, stress waves, etc.—to identify changes in a machine’s behavior that signal a developing fault.
These techniques are utilized to enable predictive maintenance (also known as condition-based maintenance), where attempts are made to estimate the actual, or even future, condition of machines to schedule timely maintenance interventions before their functional failure.
Today maintenance teams across industries make use of various maintenance strategies within their arsenal to ensure uptime, increase reliability and improve employee safety. Some of the most well-known maintenance strategies include: reactive, corrective, preventive, time-based, predictive and condition-based. Some of them are different ways to say the same thing, and some are meta-strategies that bundle several others. Whatever names are used, every strategy has its place—even run-to-failure. (Consider a facility’s light bulbs, batteries and ceiling tiles.)
So then when should you use condition monitoring?
Strategies based on condition monitoring are useful for the most critical machines whose failure would be very costly. That cost might be financial, such as lost revenue or labor, or it might be reputational, as with a sewage spill. Condition-based monitoring provides a means to prevent these losses.
Alongside condition monitoring, there are other traditional methods that are still widely used by maintenance teams that help to avoid failures in the most critical assets. One of them is time-based, also known as preventive maintenance, and another is machine redundancy, coupled with a solid parts inventory management. However, with the rise of artificial intelligence (AI) and the industrial internet of things (IIoT), condition monitoring solutions have become serious contenders. Their ability to gather condition monitoring data round-the-clock and perform real-time analysis on the asset health data can sometimes provide better results at a much lower cost.
As discussed above, condition monitoring technologies based on AI and the IIoT have become serious contenders to more traditional maintenance methods when it comes to avoiding catastrophic and costly failures in plants’ most critical machines. Below are several advantages brought by the marriage between artificial intelligence and the industrial internet of things that set such condition monitoring systems apart:
Instead of relying on manual data collection, where a maintenance engineer goes around the plant to take readings at a set interval (time-based maintenance), tried-and-true condition monitoring techniques can now be used to collect asset health data points not just once a month or every six weeks, but thousands of times per second, all day every day. This machine health data, as is common for any smart condition monitoring program today, is collected via permanent sensors (wireless in most cases) that continuously capture high-frequency signals.
As soon as the asset health data gets captured and stored on the cloud or edge, machine learning algorithms start processing that data in real time to automatically detect developing faults and pinpoint where in the machine they’re arising. They are at work 24/7/365, processing terabytes of data without cease. Apart from continuous and rapid analysis, machine learning algorithms are able to detect tiny changes the human observer would usually miss. All this means that condition monitoring systems based on AI consistently close in on perfection in terms of catching developing faults.
Another advantage of AI-based condition monitoring technologies is that they are constantly improving their predictions based on new data that is coming in every second, and their conclusions are always based on fact, never on intuition. All this means that they get better and better at what they do over time. As their libraries of known failure patterns (what we call “fingerprints of failure”) grow, they ultimately learn to associate these patterns with their underlying cause, enabling proactive and not just predictive maintenance.
The benefits mentioned above are what makes 21st-century condition monitoring systems so scalable. The AI software does all the heavy lifting, only alerting when there’s an actual impending failure and pinpointing where in the machine it’s developing. By using online condition monitoring systems based on AI and the IIoT, maintenance teams receive fact-based insights in real time about their most critical equipment that can help them drive better decisions and outcomes.
As we have been discussing so far, condition-based monitoring strategies are used today for critical machines whose failure would be costly. The rise of AI and the IIoT made modern condition monitoring systems a serious contender to more traditional maintenance strategies such as time-based maintenance as they can provide better and faster results at a lower cost. This continuous condition monitoring insight provides important benefits to maintenance and operations teams such as:
The ability to plan downtime in an industrial environment is very beneficial because the true cost of unplanned downtime due to a failed machine is often wildly underestimated. Below are several cost factors which are routinely ignored:
- lost production output at a specified quality level
- emergency costs to replace the asset (delivery, installation, etc)
- depending on the severity and type of machine break, other machines may be damaged because of the equipment fault
- the cost of needing to store and manage large numbers of spare parts in case of a breakdown
- reputational damage (as with a sewage or chemical spill)
By accurately signaling when a machine will break, condition monitoring not only enables teams to schedule inspections, repairs or replacements well before functional failure, but it also provides the means to prevent these crippling, unnecessary maintenance costs.
An unexpected machine failure can pose a major threat to employees’ safety, especially the ones who are working around it. By accurately signaling an impending failure, plant managers can ensure timely repairs and replacements, and help facilitate a safe working environment.
In the industrial world, there are two types of downtime: planned and unplanned. We have discussed the unplanned downtime case above. On the other hand, planned downtime occurs when maintenance employees perform recurring scheduled maintenance, usually at a fixed time interval. With condition monitoring, maintenance managers can reduce both types of downtime and move away from doing maintenance “just in case.”
By monitoring the machine’s condition, a maintenance team can know which equipment is healthy and which is failing, and service only the machines that need it—which is essential when it comes to moving away from time-based maintenance. And by accurately detecting an impending failure, maintenance teams can schedule repairs well in advance of asset breakdown, reducing not only unplanned downtime events but also the need for asset redundancy. This also enables maintenance teams to optimize their maintenance scheduling by deploying their key resources on equipment that actually needs help.
In order to accurately detect developing failures, condition monitoring requires sophisticated analysis on a reliable source of information-rich data. There are five major technologies, categorized by data source: lubricants, vibrations, stress waves, temperature and electrical signals.
Lubricants are indispensable for the proper working of bearings, gearboxes and hydraulic systems. There are several analytical techniques that are routinely used to analyze the quality and the composition of lubricants to reveal any presence of contamination or aging, in turn providing a diagnosis of the machine’s condition. Historically, these techniques have been performed offline, but recent advancements in sensor technology now make it possible to analyze lubricants in real time. Below are some examples of oil analysis techniques:
- Atomic emission spectroscopy
- Fourier transform infrared spectroscopy
- Karl Fischer titration
- Online oil condition monitoring (sensor technology)
Article continues below the form
Download this condition monitoring comparison guide in PDF form to read later or share with others
As with the everyday use of the word “vibration,” all industrial equipment produces some vibrations. In condition monitoring, vibration analysis is a process for collecting, measuring and analyzing a machine’s vibration patterns for any signs of unusual changes that can point to a potentially developing failure. Vibration measurements can be collected either through a handheld device or various types of vibration sensors installed directly on the asset. Here, too, there are several analytical techniques that measure different vibration characteristics:
- Shock pulse monitoring
- Kurtosis measurement
- Cepstrum analysis
- Discrete frequency monitoring
Some faults, such as cracks, leaks and fiber breakage, produce elastic waves (or stress waves) within the material that’s cracking, leaking or breaking. This is the basis for acoustic emission analysis. Acoustic stress waves are a particularly good data source to detect surface and near-surface cracks and pitting, delamination, and corrosion fatigue in concrete, metal and fiberglass. As with the other technologies above, the measurements are collected via sensor technology, typically piezoelectric transducers.
Common techniques include:
- Airborne ultrasound analysis
- Structure-borne ultrasound analysis
Failing machines, corroded electrical cables and other parts can lead to unusual temperature changes. In condition monitoring, infrared thermography is utilized to measure and analyze the radiation (heat) emitting from a machine or a whole area in order to determine its condition. For measurement, temperature sensors range from simple thermocouples that measure the temperature at one location to infrared cameras that can capture the heat map for a whole area.
Here, too, there are several techniques, including:
- Comparative thermography
- Absolute thermography
- Lock-in thermography
There’s an array of electrical condition monitoring methods out there, which fall into two basic categories: electrical signature analysis (ESA), which is performed while the machine is operating normally (or “online”), and motor circuit analysis (MCA), which is performed while the machine is deenergized (“offline”). Where other types of condition monitoring techniques analyze vibrations, lubricants or temperature, ESA analyzes current and voltage to determine an equipment’s condition.
Common techniques include:
- Motor current signature analysis
- Motor voltage signature analysis
- Power quality analysis
Each of the 5 data sources that we mentioned in the previous section carries different information about the condition of a machine, which means there’s no “best” choice across the board. In ideal conditions, maintenance teams would use combinations of various condition monitoring techniques for their critical machines to ensure maximum insight. In practice, the constraints of cost and accessibility will force them to choose. To find the best fit for each machine and application, it’s important to consider the following factors:
As we mentioned before, condition monitoring is used for critical machines whose failure would be very costly in terms of financial and reputational impact. Each plant and industrial process has a select list of “bad actors,” or in other words, critical equipment within key systems that are most prone to breakage and whose failure would result in serious consequences.
In order to pick the right condition monitoring system or systems, it’s important to know and prioritize the most critical assets for condition-based monitoring. A common approach is to perform a criticality analysis, a process used by maintenance and reliability teams to assign an asset a criticality ranking based on potential risks its failure could have on operations. If your organization has done such an analysis, use this to inform your choice of condition monitoring technology. (If not, check out this useful article on how to start this process.)
Once the criticality analysis is performed to determine the most critical assets and components within systems, the next useful step would be to perform a Failure Modes, Effects and Criticality Analysis (FMECA) on the top 20% of the most critical assets. Each failure mode has a distinct pattern in terms of data source (vibration, stress waves, current, etc), and some of these patterns are so pronounced that a sensor can pick them up as soon as they start to develop, while others might not even reach a measurable level until the system breaks down. That’s why determining how useful each condition monitoring data source will be depends on what failure modes are essential to monitor in terms of criticality.
Another important factor to consider when selecting a condition monitoring system is the environment where a selected critical machine operates. As with any smart condition monitoring program today, most of the time the data collection is performed via wireless sensors. These condition monitoring sensors are delicate pieces of equipment, meaning that they must be shielded from environmental extremes like very high temperatures, corrosive substances and so on. Plus, it can be difficult to mount sensors directly on hard-to-reach equipment like submersible borehole pumps. The same goes for equipment located in ATEX zones and other restricted locations.
After selecting critical machines and failure modes as well as considering the environment they operate in, the next step would be to learn more about various suppliers of the selected condition monitoring solution. This is where it’s important to understand how each system collects and measures data, what the installation process is like and whether the selected technology meets all connectivity and regulatory requirements.
If you are planning to implement a predictive maintenance program and don’t know where to start, we wrote a guide that is designed to walk you through the concrete steps from idea to implementation. Plus, there are several supporting worksheets to help you at each step. Be sure to check it out here.
A tank oil storage facility has multiple centrifugal pumps that transfer liquids at predetermined flow rates. The team determined that these pumps are prone to cavitation, a critical failure mode to monitor and detect. Though cavitation will not immediately result in a functional failure of the pump, it will lead over time to seal and bearing wear, erosion damage, and possibly sudden impeller breakage. Plus, prolonged cavitation shortens a pump’s equipment lifetime and wastes energy. To determine which condition monitoring technology might be the best choice, the maintenance team considered all 5 major technologies. They found out that vibration, stress wave and electrical sensors are the top choices, because all 3 data sources can pick up the patterns of cavitation in advance. Plus, they learn that through the pump affinity laws, electrical data can also be used to track a pump’s real-time pressure and flow, to additionally identify real-time operation of the pump in relation to its best efficiency point (this might pinpoint where cavitation is likely occurring in real time). They proceed with doing further research into selected vendors to determine if the technology meets all regulatory and connectivity requirements.
As we’ve been discussing, there are top 5 major condition monitoring technologies that are categorized by data sources. One of them is electrical signature analysis (ESA), which is an electrical monitoring method to determine a machine’s condition. It’s performed while the machine is operating online, enabling real-time condition monitoring insight into the machine’s health state. Samotics’ ESA system called SAM4 uses current transformers and voltage taps that install in the motor control cabinet, where they capture all three phases of the current and the voltage at a high frequency around the clock. But the system goes further by implementing advanced machine learning algorithms to analyze data in real time and pinpoint early on when damage starts to develop. These capabilities in sum provide several unique benefits which we haven’t discussed so far:
Because SAM4’s sensors install within the safety of the motor control cabinet, it enables maintenance teams to capture reliable data about the asset’s condition at a distance. This is especially useful when it becomes difficult and impractical to mount sensors directly on hard-to-reach equipment such as submerged pumps or motors encased in larger machines. ATEX zones and other restricted locations also pose a problem. In addition, the motor control cabinet is a cheap, safe and convenient place for sensor installation, which usually takes no longer than 30 minutes per machine.
It’s a common misconception that ESA only excels at detecting electrical faults. ESA-based systems can detect and localize electrical as well as mechanical faults throughout the drive train. For example, mechanical failures like misaligned coupling will cause vibrations that will influence the air gap between the motor’s stator and rotor, causing magnetic field disturbances, and in turn, affecting the supply voltage and operating current. To date, Samotics’ SAM4 technology detects over 90% of failures—both mechanical and electrical, up to 5 months in advance.
As we noted above, ESA systems can track a host of additional metrics that help you raise efficiency, lower costs and shrink your company’s environmental footprint. All these metrics require current and voltage information; they can’t be calculated from vibration, thermal, acoustic or oil-based data. SAM4 has developed the following advanced tools to provide you with performance and energy insights. (Not every ESA system will provide these, so be sure to compare features.)
Three of SAM4’s advanced energy and performance features:
- a real-time pump monitor to help you steer a pump back to its best efficiency point, reducing cavitation and raising bearing and seal life. Over time, the data can tell you where redesign or replacement would score major cost and efficiency gains.
- an energy monitor to track an asset’s operational efficiency. Over time, the data can tell you where redesign or replacement would score major cost and efficiency gains. (Read more in our sustainable industry white paper.)
- a power quality monitor to identify and solve supply-side issues such as voltage unbalance and harmonic distortion.
If electrical signature analysis is among your selection of condition monitoring technologies, we’d be happy to tell you more about our solution. Please contact us to book a no-obligation demo at your convenience.