When is the right time to service a machine? Until now, this question could only be answered on the basis of estimates and empirical values. The result: if machines are serviced too late, this can lead to expensive production downtime. If, on the other hand, machine parts are replaced too early, unnecessary costs are incurred. Today, reliable data is used to answer this question, on the basis of which maintenance can be controlled with foresight. Predictive maintenance is the buzzword and is based on advanced analytics. With big data technology, future scenarios can be derived using historical data and special prediction models. The advantage is that even complex correlations can be analyzed quickly in this way. Based on this, well-founded decisions can be made.

But how exactly does the technology behind application scenarios such as intelligent maintenance work?

The basis for advanced analytics and therefore also for predictive maintenance is the collection of data - on a production plant or machine, for example. In order to derive forecasts for the future, this data must be analyzed and recorded in mathematical models. There are two different variants: The time-based prediction provides an outlook on how a value will behave in the future. The value-based prediction shows how a value will behave if another value is changed. For example, this model can determine how much energy a plant will consume if the production volume changes. The basis for calculating such forecasts are mathematical models, known as predictive models, which either work with statistical regression analyses or rely on machine learning using neural networks.

Maintenance as required

Every machine needs to be serviced. But the question is: when is the right time? There is no general answer to this question. This is because it depends on various parameters, such as how old the machine is, how heavily it is used or for what purpose it is being used. If you always maintain your machines at the same intervals according to a fixed cycle, you are not meeting these individual criteria. This can lead to rising costs. For example, if a component is replaced more frequently than necessary. Or if it is replaced too late, resulting in machine damage and production downtime.

The right time for the next maintenance date, which previously could only be determined on the basis of estimates and empirical values, can now be calculated precisely - based on historical data and statistical prediction models. This requires data from ongoing operations, which is analyzed. Based on these data sets, the system is constantly learning and enables live data to be interpreted using a model. This allows individual maintenance schedules to be defined for each machine and each component. As a result, this makes production much smarter and reduces costs and the risk of breakdowns. This is because maintenance is then based on the actual load on the machine: if the load is higher, the maintenance period is reduced, thus preventing machine damage. If the load is low, the maintenance time is postponed accordingly. This avoids unnecessary costs and downtime, thereby increasing efficiency. Spare parts can be ordered in the right quantity at the right time and the technicians' working hours can be planned optimally.

Detect quality deviations in advance

A typical example of the use of predictive maintenance is the repair of filling and packaging systems in the food industry. Here, certain process parameters of machines and systems are monitored. This allows any quality deviations to be detected automatically in advance. This is controlled via the zenon software platform from Copa-Data, which can be used to visually process and analyze the recorded data. This provides valuable information, such as the wear of a bottle coating station. If signs of wear are detected at an early stage, maintenance appointments can be planned accordingly in advance. It is also possible to classify events according to severity: is it just a warning - or do we really need to assume a fault?

In addition to predictive maintenance, advanced analytics is also used in resource management and in production planning and control . The data from a production plant can be used to create a precise picture of resource consumption. And this is not only possible for the past, but also as an extrapolation up to the end of the accounting period. This enables companies to plan and control their resources precisely.

Another area of application is process optimization: if it is known how individual parameters affect the manufacturing process, the production result of a machine can be improved. Advanced analytics applications are able to determine correlations between individual values and data points. In this way, the process can be accelerated enormously.

According to documents from Copa Data.

This article first appeared on our sister portal www.scope-online.de