Every maintenance manager would like to predict when a breakdown will occur. But how? The IoT lends a hand.
Using sensors and new technologies that monitor signs of deterioration and changes in the machine’s performance, managers can estimate when a failure will occur.
And so a new type of maintenance is born, predictive maintenance.
Predictive maintenance is a proactive maintenance strategy that aims to prevent breakdowns. Depending on the data collected and predefined predictive algorithms, it tries to estimate when a breakdown will occur. Maintenance activities are then scheduled based on these forecasts.
Now, we need to make a little parenthesis. Predictive maintenance is often confused with condition-based maintenance, but there is an important nuance that distinguishes them. While predictive maintenance schedules tasks based on collected data and formulas, condition-based maintenance only acts when these parameters reach alarming levels.
In theory, it is easy to understand how predictive maintenance works. It gathers data about your assets and, from that, extracts information that allows you to calculate when you need to do maintenance.
In practice, we can divide the process into 3 stages:
As we have already seen, the idea is to predict when a breakdown will occur. As this depends on obtaining quality data, the first step is to install sensors capable of collecting information in real-time on the performance and ‘health’ of the equipment.
The data that these sensors need to measure and collect depends on the techniques you intend to use to monitor the equipment. You can control vibration, temperature, pressure, noise level, or corrosion levels, among others, depending on which is best for your equipment. We will explore several predictive maintenance tools in a moment.
Accumulating data about your assets is of no use if you don’t know how to exploit them. It’s the Internet of Things (IoT) that allows sensors to send all information to a central system or software that allows you to analyse what is going on. Predictive maintenance is much more effective, and much more surgical, in systems where the various assets are integrated.
Some people may believe that predictive maintenance stops here. But if you are only acting when the sensors detect anomalies, you are just performing condition-based maintenance – as we saw above. The most differentiating component of predictive maintenance is to build and apply algorithms that offer a prognosis.
In the beginning, it can be based on equipment history, maintenance logs, and statistics (your CMMS reports are extremely useful in this step). However, as Artificial Intelligence becomes increasingly sophisticated, it is possible to detect anomalies even earlier, find correlations and receive intelligent suggestions to prevent a breakdown. This intelligent maintenance is giving rise to a new type of maintenance, prescriptive maintenance.
The idea is planning maintenance according to the data. That’s why non-destructive tests (NDT) are so important to diagnose failures within the infrastructure. NDTs do not compromise equipment and can be performed while it’s running, just like routine check-ups for humans. Blood tests, ultrasound scans, or eye exams: we do them every year, and none of them hurt us.
👉 Among the predictive maintenance techniques used in the industry are:
- Vibration analysis;
- Oil analysis;
- Infrared or thermal imaging tests (thermography);
- Motor circuit analysis.
- applications: electrical connections and systems, heating systems, fluid analysis, discharge patterns, roof maintenance
Thermographic analysis can also be applied to several types of equipment to detect wear, rusting, delaminations, and disconnections that might go unnoticed to the naked eye.
The advantage of infrared is that it allows you to make temperature maps, even from a distance. This technique is used, for example, in the maintenance of heating systems. Temperature variations are useful for assessing the condition of motors and mechanical and electrical components, the building insulation, and even the condition of solar panels. The major disadvantage is that it cannot be used on reflective surfaces.
🤖 In the future, CPU temperature analysis will be one of the main ways to test the health of robots.
- applications: assess engine degradation, shaft and rotor alignment, insulation, gears, scan for short-circuits
Motor circuit analysis uses a technique known as electric signature analysis (ESA), to find anomalies in electric motors. In addition to analysing the circuit and its components, it evaluates the voltage and current entering the motor. Not least important: it works on both AC and DC motors and can be used while the equipment is running.
- applications: turbines, hydraulic and electro-hydraulic systems, evaluate engines, transmissions, gears, lubricant levels
The purpose of oil analysis is to test the viscosity, the amount of water and the presence of other materials, including metals, to determine the wear of the equipment.
Let’s take into consideration a hydraulic system that consists of two essential components – the rotating parts and their lubricant. As equipment ages, the sample will show byproducts of overheating and erosion. Different particles reveal many problems before you can guess a breakdown is coming.
Different particles reveal different damages to different components. For example, the presence of silicone above 15 ppm can indicate that the insulation is wearing out, while sodium can reveal contamination with salt water. Follow a metal guide (page 71) when using the spectrometer.
- applications: test component alignment, detect imbalances, clearances, resonances, gear failures
Vibration analysis is perfect for rotating equipment and machines, such as compressors, water pumps, and engines. Thus, this predictive maintenance technique is ideal for infrastructures with a complex water supply system, such as hotels, spas, or water parks.
What we do is connect them to a sensor that can detect movement or acceleration, depending on what’s appropriate. The sensor works by detecting sound waves created by movement, which generate electrical impulses and make equipment vibrate.
- applications: pipes and plumbing, condensers, vacuum systems, fans, air compressors
We’ve already established that sound waves are our allies. Acoustic analysis is a technique used to detect problems in the material’s technical performance, pinpoint the source of the problem, and perform “check-ups” on the equipment’s overall health. How? By detecting changes in sound frequencies.
Every working machine makes some sort of noise, but its frequency and range change whenever there are leaks or pressure changes, for example. This predictive maintenance technique is especially useful for pipelines carrying liquids or gas. Some more modern tools incorporate thermometers and cameras to deepen the analysis at a distance.
Predictive maintenance was born to avoid breakdowns, but let us not be under any illusions. There will always be random failures that are impossible to predict or prevent.
In addition, we must not forget that predictive maintenance requires a large infrastructure. Therefore, predictive maintenance is only recommended for critical assets and with predictable failure modes.
- The main advantage of predictive maintenance is to act in a timely manner, which reduces downtime and increases asset availability.
- As maintenance is scheduled according to needs, it avoids wasting stock and labor in unnecessary maintenance.
- By reducing the emergency repairs and the wastes we mentioned above, it helps to better control your maintenance budget.
- Downtime is planned in advance, which allows for better maintenance and normal company activity.
- Optimal use of the equipment throughout its life cycle.
⚙️ Predictive maintenance is also a cornerstone of lean maintenance and just-in-time production.
- The need to invest in specific monitoring equipment, as well as to train personnel to use and interpret the data collected.
- For assets with low criticality, predictive maintenance may not offer great savings over the alternatives.
- It is not suitable for assets with random failure modes or without initial data to predict malfunctions (in these cases, it is preferable to start with condition-based maintenance and gradually make the transition).
We already mentioned that the cost of implementation is high, but even so, predictive maintenance has a high ROI, which can be up to 10 times the investment. A study by the United States Department of Energy in 2010, when installing sensors was even more expensive than now, points to 25-30% reductions in maintenance costs, 35-45% less downtime, and 70-75% fewer breakdowns.
Compared to reactive maintenance, it resulted in savings of 30-40%. Compared to preventive maintenance, savings of 8-12% were achieved. According to the RCM methodology (reliability-centered maintenance), the ideal is that 45-55% of maintenance is predictive, 25-35% preventive and only 10% reactive or corrective.
Deloitte, in a 2017 study, is less optimistic. Still, predictive maintenance is very promising. This study suggests that uptime will increase by 10-20%, and maintenance costs will decrease by 5-10% in Industry 4.0. Maintenance planning can take 20-50% less time.
Taking into account the advantages and disadvantages mentioned above, we can generalise that predictive maintenance pays off the investment for:
- companies with high operating expenses and a lot of invested capital;
- companies where downtime implies great losses;
- companies with assets whose breakdowns are a security risk.
Therefore, it is not surprising that aviation was a pioneer in predictive maintenance. Failure prediction is used both during flight – with engine temperature and vibrations monitoring to prevent accidents – and on land, to reduce delays and cancellations.
Manufacturing is another sector that adhered greatly to predictive maintenance. It is easy to see why: a halt in production can cause huge losses. In the coming years, it is likely that it will also gain greater expression in fleet management, healthcare, mining industry, energy extraction and production.
If you have already decided to invest in predictive maintenance in your company, recap the 4 steps you need to implement it:
The first step is to identify the priority assets to include in the predictive maintenance strategy. Prioritise critical assets for your operations and assets with high repair costs.
The second step is to gather the necessary information to be able to transform the data you collected into actions. If you already have a CMMS, it is easier to organise the history and develop the first algorithms. At the same time, it is convenient to establish failure modes and the probability of occurrence.
Now that you have defined priorities and failure modes, you can start implementing the sensors. In new equipment, it is usually a simple process, but it can be more complicated in older machines. Investigate with the manufacturer the best way to integrate old models with new technologies.
Test the operation of the sensors and the accuracy of the algorithms only on some machines, following a PDCA cycle. When you are able to schedule maintenance and meet your goals, expand this strategy to other machines!
Predictive maintenance is already a strategy that bears fruits and returns for many companies. In the future, as the IoT and digital transformation take hold, it will be even more dominant, accessible, and effective in predicting breakdowns.
However, one thing is clear: predictive maintenance benefits from intelligent and integrated systems, such as those that Intelligent Maintenance Management Platforms provide.
📌 Talk to one of our experts to understand how Infraspeak supports your operations, today and in the future!
💡Did you know?
An Intelligent Maintenance Management Platform can help you create, manage and evaluate your maintenance strategy. Infraspeak provides you with a multitude of apps, integrations and IoT hardware that will supercharge your team, business and maintenance operations.
📌 To know more, schedule a free demonstration with one of our specialists.