Predictive maintenance is a way of performing maintenance tasks based on predictions about the ideal time to carry them out. This process seeks to optimize the incident solution process and minimize unplanned downtime of a specific asset or industrial machine.
To better understand what predictive maintenance consists of, it is convenient to explain what different maintenance tasks generally occur in the industrial world. Industrial maintenance usually has two approaches:
it is aimed at solving faults and incidents once they have been detected. They can be the result of a routine inspection or even an emergency, if they cause a serious unscheduled incident. This type of maintenance is part of the traditional maintenance that any industry carries with it.
This pursues the objective of preventing serious failures or unscheduled incidents from occurring. It is an investment to avoid having to incur higher expenses (those that occur when a failure occurs).
Preventive maintenance relies on different factors to decide when it should be done. The most common and widespread factor is time: scheduled maintenance tasks are carried out periodically, in which a review of the components of industrial assets or equipment is made and it is checked whether any require repair or replacement.
If instead of using factors such as time, we rely on data that gives information about the intensity of use, performance and how the asset is behaving, and based on that data we can create models to develop maintenance tasks, that is when we can start talking about predictive maintenance.
To perform predictive maintenance, we first need to be able to monitor certain parameters and extract data. To capture these parameters and see their evolution, technologies such as IoT come into play. IoT technologies make it possible to collect data from sensors, actuators and other equipment, in real time, and have this information available anywhere.
The parameters to be monitored will depend a lot on the type of asset we want to review and its function. They can be such as:
Noise: The sound pattern, or simply the volume of the noise generated by certain parts of a machine can give information about its possible wear or foreseeable breakdown.
Vibration: Changes in the vibration of machinery can be a sign of a possible breakdown or wear of a component.
Temperature (through sensors or even thermographic images): Overheating points can be an indicator of a possible failure.
Pressure: Changes in hydraulic pressure in certain elements can be used to detect obstructions or leaks.
The evolution of the measurements of these parameters, alone or in combination, are what provide information that is used to develop the prediction models.
Over time, in fact, the models have become more complex as they use more data sources to make their predictions. This also includes technologies oriented to data intelligence, such as machine learning, edge computing or artificial intelligence, which automate the development of the prediction models themselves, and generate the necessary alerts.
Predictive maintenance is something that is spreading throughout many industrial sectors. It is especially attractive for sectors whose assets have a high intensity of use and whose involuntary or unscheduled shutdown can have a very negative impact on the company (either financially or otherwise).
A good example is the sectors more oriented to manufacturing, such as the automotive, ceramic or plastic industry. In these types of industries, stops in the manufacturing chain have a direct impact on the volume of production and therefore on the turnover, so any improvement that helps to minimize these situations has a high positive impact.
Another example is that of the electricity distribution sector. The transformation centers are points of intense activity and high impact. In them, the electricity that arrives in high or medium voltage is transformed into low voltage so that citizens can consume it.
When a breakdown occurs in a transformation center (for example due to overload due to excess demand, or due to a physical problem in one of the equipment in the center), the impact for the company is not only economic for the company, but also has important consequences for the population, and also generates considerable reputational damage. That is why this is one of the sectors where more progress is being made in equipping its assets with intelligence and working on different improvement mechanisms such as predictive maintenance.
Similar to the previous case, the water sector is a sector whose performance has a direct impact on the population. Its problem is similar: any failure in any equipment that impacts on the drinking water supply chain to citizens, is damaging to both the company and its customers, and that is why this is another of the sectors that most seeks optimization and digitization of its processes, including predictive maintenance among its modernization priorities.
For a long time, maintenance has been considered more of a cost factor than a profitability factor in companies. However, with the incorporation of new technologies such as Artificial Intelligence, companies are seeing that they can achieve significant competitive advantages.
The decision on whether to invest in predictive maintenance techniques, logically, is usually based on the return on investment that companies can achieve. Although, there are sectors that can be affected from a reputational point of view and not just an economic one.
In 2016, most companies that were considering implementing predictive maintenance processes did not see clearly what ROI they would be able to obtain. However, in 2021, more than 80% of the companies that are already implementing it claim to have a positive ROI, and half of them also say they have recovered the investment before even a year.
The forecasts for the coming years, with the improvement both in IoT technologies and in machine learning and artificial intelligence models, is that predictive maintenance will become an indispensable tool in any industrial company.