Have you ever heard the words “preventive maintenance” or “predictive maintenance”? Regarding the maintenance of equipment in production lines, the Internet of Things (IoT) has recently attracted attention. Also, these two words seem to be similar and have different meanings. This section explains the overall definition of conservation activities such as preventive maintenance and predictive maintenance, the difference between preventive maintenance and predictive maintenance, the benefits of predictive maintenance, and the relationship between predictive maintenance and edge computing.
What are Conservation activities?
What is a conservation activity? JIS (Japanese Industrial Standard) defines maintenance activities as “a general term for activities that eliminate failures and keep equipment in normal and good condition, including planning, inspection, adjustment, repair, replacement, etc.”.
In other words, it can be thought of as a human influence on the production line to maintain the performance of the production line. As defined in the JIS, conservation activities are divided into maintenance activities and improvement activities.
Maintenance activities are activities to maintain the quality of products and the performance of production equipment, that is, activities to maintain the perfect condition of production facilities. This includes preventive and reactive maintenance. On the other hand, improvement activities refer to activities such as “improvement maintenance” that reviews machinery to prevent recurrence when it breaks down, and “maintenance prevention” that replaces machinery and equipment to prevent breakdowns and mistakes.
Preventive maintenance is the prevention of failure by daily inspections and replacement of deteriorated parts before they occur. It includes predictive and periodic maintenance. Follow-up maintenance refers to restoring the function of equipment when a failure is discovered in the equipment due to a malfunction or the like. In other words, it is assumed that it will be “repaired” when the equipment is broken.
Periodic maintenance is the act of determining the cycle based on fault records and equipment characteristics, and replacing and inspecting parts for each cycle. It can also be rephrased as maintenance performed based on elapsed time. Generally, “preventive maintenance” refers to periodic maintenance.
On the other hand, predictive maintenance is to detect or predict deterioration from the state of equipment measured continuously, and to take the best measures at the optimal time before a failure occurs. It is based on the condition of the device.
The main difference between periodic maintenance and predictive maintenance is that preventive maintenance is performed at a certain time cycle regardless of the condition of the equipment, whereas predictive maintenance constantly monitors the condition of the equipment and responds when signs of failure are detected.
Conservation Activities to Date
Conventional conservation activities mainly include periodic maintenance, predictive maintenance, and post-mortem maintenance. Improvement activities (such as modifications and upgrades) may be carried out to extend the life of the equipment, but the response is limited, such as in the case of expensive equipment.
Periodic maintenance, as already mentioned, involves the replacement of parts on a regular basis, regardless of the condition of the equipment. In addition, predictive maintenance was carried out by on-site workers and engineers relying on the intuition cultivated through many years of experience that “it is about time to replace that part.”
Periodic maintenance time intervals are tailored to the most important and shortest-lived parts, but other parts are often replaced during this replacement. The reason is that the life of the parts varies depending on the type, so if the equipment is stopped and replaced each time, the operation rate of the equipment will deteriorate. For this reason, parts that have not yet reached the end of their useful life will also be replaced, and there is a problem that there is a lot of waste in periodic maintenance.
Predictive maintenance conducts is based on the signs between anticipation of a failure and the actual failure. This means that you can think of a maintenance plan at the expected stage and act, such as replacing parts before a failure occurs. This minimizes the condition that the machine is in an emergency stop due to failure.
In addition, if predictive maintenance is automated using IoT, an alarm is issued when there is a sign of failure. Since it is only possible to think about response when an alarm is issued, it does not take time and effort, and it leads to a reduction in labor costs in this aspect.
Specific examples of Predictive maintenance
In manufacturing equipment, motors are frequently used. There is a component called “bearing” that supports the “shaft” (drive shaft) that transmits power from this motor. If the bearing fails, the axis may not turn or the load on the shaft cannot be distributed, which can lead to serious accidents. Therefore, this bearing is very important.
Bearings themselves are inherently highly reliable, but when higher reliability is required, for example, in the process of stretching heated iron at a steel mill (rolling process), bearing monitoring is often performed.
The bearing monitoring system installs a vibration sensor on the bearing to detect the waveform, frequency, and amplitude conditions of the vibration. Then, an alarm is issued when the state of vibration indicates a sign of failure.
For example, if a bearing is damaged in one place, the load applied to the bearing changes. As the load changes, the distance between the vibration sensor and the bearing changes, but the vibration sensor detects a minute change in distance. Then, since the rotating part of the bearing rotates at a constant period, the change in the distance will occur every rotation. This causes vibration, and it is possible to detect this vibration and make it a sign of failure. Skilled engineers took this vibration from changes in machine sound and used it as a sign of damage.
This example is simplified for clarity but predicting failures from sound requires experience because there is no change in vibration even if multiple places are damaged or damaged. For this reason, it was not possible to make such a judgment unless from a skilled engineer. Unfortunately, the number of skilled technicians is decreasing. Artificial intelligence (AI) is considered as a solution. Artificial intelligence, which has been attracting attention in recent years, enables “deep learning” to learn by itself by a structure modeled on human actions. Of course, although it is necessary to have learning first, artificial intelligence can learn and execute like humans who have captured signs of failure from experience. With the number of skilled engineers decreasing, it can be said that it is one effective means to respond to the shortage of human resources.
Predictive maintenance and edge computing
In this way, automation of predictive maintenance is an effective means of eliminating the shortage of human resources and optimizing maintenance activities. However, there are some things to be aware of when automating predictive maintenance.
The structure that imitates human actions of artificial intelligence described earlier is called “neural network” and contributes greatly to the realization of deep learning. However, neural networks are a very complex mechanism. Therefore, it is a practical issue to realize predictive maintenance while maintaining the speed necessary for predictive maintenance.
There is also a method called “neurochip” that realizes neural networks with hardware, but it is not a very common method. The problem here is speed, so cloud computing, which always sends and receives from servers on the Internet, is impractical. So, let’s think about deploying edge computing and putting artificial intelligence on edge servers. This is called edge AI. As a result, it can be said that the best answer at this time is to automate predictive maintenance while maintaining the speed as much as possible.
Possibility of predictive maintenance and artificial intelligence
This paper mainly describes the difference between predictive maintenance and preventive maintenance, and the merits of predictive maintenance. With the development of artificial intelligence, the possibilities for predictive maintenance are greatly expanding. And enabling predictive maintenance using artificial intelligence with edge computing is the most ideal figure at present.