A positive effect of Internet of Things and Industry 4.0 results from the fact that the use of sensors generates a large amount of production data and measured values. For example, the energy consumption of entire machines and plants as well as individual devices and components. Companies that professionally record and analyze their energy consumption not only benefit from cost savings through efficient energy management. The data obtained also makes it possible to implement predictive maintenance with the aim of optimising the time required for maintenance work on production plants and machinery. This is achieved by generating reliable information about the expected remaining service life of a component on the basis of the collected, analyzed and evaluated data. The sensor-based predictive maintenance avoids repairs, anticipates failures and enables the user to take preventive measures.
This predictive procedure is known as the heart of industry 4.0 under the name Predictive Maintenance. With the promise of early anticipation of failures and malfunctions, this proactive approach is increasingly spreading to small and medium-sized industrial enterprises. This is because where production capacities do not fail in the first place, companies save significant costs. In addition, the knowledge gained makes it possible to increase productivity through optimized plant performance. Today, the solution is used for the maintenance of engines in turbines, aircraft or wind turbines, among other things.
Up to today, maintenance in medium-sized industrial companies often looks the same: A malfunction occurs during operation, production comes to a standstill or has to be discontinued completely. Only now do the technicians search for and analyze the problem and take measures to eliminate the malfunction. It often takes days for a required spare part to be delivered and installed. High downtimes, in the worst case even delivery delays and financial losses are the result. Preventive maintenance is one step further: based on experience, it tries to determine which components are statistically susceptible to failure at which point in time. Replacement takes place at fixed intervals. This has the advantage that more failures are prevented than with reactive maintenance – but not all of them, because machines and systems do not adhere to statistics. A financial disadvantage also arises from the fact that the rigid process also replaces functional wear parts as a precaution, which in many cases would have continued to work.
Predictive Maintenance radically pursues the approach of preventive maintenance and uses the current technological developments around IoT, such as Big Data, to perfect it. The aim is to find the optimum time for replacement of all components individually as early as possible. The approach uses data collected live from machines. Not only the result of an individual sensor is included in the evaluation, but the entirety of all data collected in the plant is taken into account. This combination of measured values enables intelligent and predictive evaluation. Algorithms draw the right conclusions for the future from the data.
The result: companies can prepare and initiate the necessary maintenance measures long before the effects are felt.
A deviation from the predicted consumption serves as a characteristic for defect recognition.
The process of predictive maintenance is complex and only successful if sufficient production data and measured values are permanently recorded, stored, analyzed and evaluated over long periods of time. Finally, it is critical that the algorithms reliably calculate how likely certain events are to occur. Predictive Maintenance relies on techniques and databases that are also used for big data solutions in order to be able to process the large amounts of data.
The energy data of a company obtained with the ENIT agent can be used with the help of predictive maintenance for fault condition detection. The comparison of real data with a statistical prediction provides valuable insights: If a deviation from the predicted consumption is detected, this serves as a characteristic for fault detection. The predictive maintenance system from ENIT Systems uses anomaly detection algorithms to detect fault conditions. An essential feature of predictive maintenance is that the algorithms learn continuously. The larger the data basis, the greater the advantage of complex self-learning algorithms. In concrete terms, this means that algorithms learn from the past and automatically recognize relevant features.