Factory Physics Factory maintenance is an activity to ensure the essential condition of the factory. In order to prevent the machine from breaking down, it will be fixed quickly when it breaks down. No matter how difficult the breakdown is, it is a tough activity because it requires a quick response.
However, this activity is so inconspicuous that even people inside the factory may not know about it. Also, it is not long before outsiders know.
It is one of the specific activities of Risk Management.
Maintenance methods can be broadly divided into reactive maintenance and preventive maintenance.
Preventive maintenance can be further divided into two parts.
"PM" is an abbreviation for both preventive and predictive maintenance, so it is important to be careful. For example, when we are talking about whether to use BM or PM at a factory, we are talking about preventive maintenance. When we talk about machine learning, we talk about predictive maintenance.
It's confusing because BM isn't subdivided into TBM and CBM. The original word for "B" is different.
Conservation data analysis has traditionally been Weibull analysis, but machine learning is also being used. In order to be eligible for data analysis, the data must be prepared, which is usually a particularly important part of the factory.
On the other hand, there are a huge number of things that are subject to maintenance in factories. Maintenance personnel have to deal with everything when something happens.
For a huge number of maintenance objects, the perspective of data management and digital transformation (DX) is required rather than data analysis.
In the case of factory maintenance, there is an option that "reactive maintenance is also possible", so this is the story. Not so much in different fields. For example, in an airplane, it is not possible to break down during flight. It seems that a huge amount of time is being spent on conservation activities for a huge number of conservation targets.
In principle, Weibull analysis in Reliability Engineering can be used as a method of analyzing not only the life of the products produced in the factory, but also the life of the machines and components used in the factory.
However, in the case of factory machines and parts, it is easy to say, "I've only ever broken once" or "This is the only place where I'm using this method," so it's easy to get into a situation where "there is no usable data."
For the same reason, determining the duration of periodic maintenance is quite a difficult problem.
Basically, it is the monitoring of the sensor value itself and the value converted to a Feature by Control Chart. "Feature quantity" is not always difficult, and it is also effective to extract only data when the power is turned on. In fact, it is carried out at the beginning of the workday.
If you can't tell by looking at each sensor, you need to make a decision from the data of multiple sensors, but research is underway to promote predictive maintenance using Machine Learning anomaly detection technology.
As for the fact that there is little data at the time of failure, the One-Class Model is one of the countermeasures.
TPM is an abbreviation for Total Productive Maintenance, and it is a movement that seeks to improve management and corporate culture from a facility-centric perspective. TPM methodologies include equipment maintenance, production efficiency, and quality maintenance.
At TPM, we promote "production efficiency" and "quality maintenance" from the perspective of equipment rather than products.
TPM is the name of a movement. Some TPMs are talking about introducing and enlightening the company's facility-oriented methodology.
NEXT Factory Physics
