Predictive Maintenance: Successful as an individual project

node-divider
5 minutes Reading time
5 minutes Reading time

Predictive maintenance is a vision of many companies, but not an off-the-shelf solution. It is always a customer-specific project based on specific requirements and target images.

The challenge: Avoiding machine downtime

Machines need regular maintenance and replacement of wear parts at certain intervals. In this respect, machine tools, electrical and electronic systems from automation and production technology, wind turbines, charging stations and many other systems are no different. They are checked and maintained at regular intervals to minimize malfunctions and system failures. This avoids or reduces expensive downtime.

The smaller the maintenance intervals, the greater the likelihood that unplanned downtime can be avoided. At the same time, however, the effort and personnel requirements are also increasing. That’s why many companies are looking for alternatives. Digital technologies around the Industrial IoT (Internet of Things) offer a solution approach here: the intelligent monitoring of machine conditions in order to prevent unplanned downtimes as completely as possible in an efficient manner.

This concept is referred to as “predictive maintenance.” This is a system that uses data analysis to predict a machine or product failure with a certain probability. In this way, necessary countermeasures can be initiated in time to avoid an unplanned shutdown.

Implementation partner for solutions of this kind is the All For One Group. The company has extensive experience in developing digitization solutions throughout the GSA region. It focuses its work on the underlying practical problem, its solution and the resulting added value.

The solution: Predictive maintenance based on data analysis

Predictive maintenance exists in two different variants. 

  •  In the first (external) variant, the manufacturer processes the data from its devices in use by customers and makes the analyses available to customers in a cloud application. So the predictive maintenance in this case is the service of the manufacturer.

  • In the second (internal) variant, a manufacturing company collects data from its machines and systems running in its own production. It uses them in its own application to avoid unplanned downtime and to determine the optimum maintenance time from a business point of view. This variant often refers to existing machines from different manufacturers and requires integration into the ERP system.

  • Data acquisition and analysis
A predictive maintenance system requires data for analysis. They are often collected with specific sensors and sent to a cloud application via IoT gateways. With internal predictive maintenance, it is also often possible to use data from existing industrial controls. There are two approaches to collecting the data: 
 
  • Generic data collection involves the collection of as much data as possible In a second step, they are examined for patterns and anomalies that can be used to determine the current status of the plant. This use case is easier to implement with the external variant of predictive maintenance. When all of an operator’s assets are networked, it creates a large database that is better suited for comprehensive analyses.
  • Case-based data collection focuses on a specific problem, in many cases the failure of specific components. The manufacturer takes precautions in his machines to determine wear of important components, for example. Since the data collection is focused on a specific problem, smaller amounts of data are sufficient.

Various methods from data analytics are used to analyze the collected data, such as machine learning. The All for One white paper on Predictive Maintenance provides deeper insights.

The approach of All For One

The implementation of predictive maintenance is customer-specific: every project is different, and the spectrum of individual requirements is very broad. Three examples illustrate this: A company has built up a data lake and now wants to do additional analyses with the data. Another company wants to monitor its assets, but has only a rough idea of what data it even has. A third company wants to network its products to collect data from real-world customer operations in the first place. It does not want to decide on a predictive maintenance scenario until later.

In order to meet these different maturity levels and requirements, All For One often offers its customers a workshop as a first step. There, the current status is clarified and then a target picture is developed. The result does not always have to be predictive maintenance in its purest form. The first thing many companies need is insight into machine status or an accurate analysis of the available data. Often, important data is invisible to the company because it is neither collected nor evaluated.

Only after clarification of the project goals does the technical realization begin. This can actually be predictive maintenance, but also a preliminary stage such as anomaly detection or condition monitoring. In the process, All For One builds either a “proof of concept” or a functional prototype, which is then expanded into a comprehensive solution.

An important aspect of this phase is the digital maturity level of the customer company according to the acatech model. Some companies have all the prerequisites for rapid implementation of the prototype. Others, however, have to create the technological basis in the first place in order to be able to analyze data on a larger scale. In this way, All For One can determine how the project is ideally adapted to the company.

The result: Cost reduction and optimization of production

Predictive maintenance can save costs on a large scale. This concerns downtimes in production, failures at the customer’s site, unplanned “fire extinguishing service calls” or unnecessary work steps, such as too frequent tool changes. All of this can be avoided or at least drastically reduced.

In addition, predictive maintenance can improve the competitive situation. For one thing, products do not fail unannounced at the customer’s site, so quality and customer satisfaction increase. For another, delivery reliability is increased in the manufacturing sector, as unplanned interruptions in production are avoided.

Finally, predictive maintenance supports optimization both in eliminating specific problems in the operation of a machine or product and in terms of general performance.

NEWSLETTER

Get our IoT Use Case Update now

node-divider

Get the free monthly update on exciting use cases, latest podcast episodes and news from our IoT community.