Overview

Predictive maintenance – make or buy?

When the concept of Industry 4.0 is to be illustrated using a specific use case, predictive maintenance is often described. There are three main reasons why this use case is so popular: Firstly, the topic is relevant for manufacturing companies in all industries. After all, anyone using machines and systems wants them to perform as well as possible – and to fail as rarely as possible. Predictive maintenance promises exactly that. Secondly, predictive maintenance is not a completely new concept that companies need to get used to. Predictive maintenance is merely the continuation of condition-based maintenance or condition monitoring. And thirdly, the necessary technologies are reasonably well established. Operating data and machine data have been digitally recorded for years. Databases designed to handle large volumes of data in real time are available. Algorithms that can derive forecasts from the large amount of data are also available.

The importance of predictive maintenance in the context of Industry 4.0 was emphasized by the study “Predictive Maintenance – Service of the future – and where it really stands”, which was presented at this year’s Hannover Messe and initiated by the VDMA and Deutsche Messe and conducted by Roland Berger. One of the key findings of the study is that predictive maintenance is a key topic for the manufacturing industry. However, only machine and plant manufacturers from different segments were surveyed. Users, on the other hand, were not.

Predictive maintenance – relevant, but rare

One observation is remarkable: although predictive maintenance is considered to be so highly relevant and although the best conditions exist for implementing the concept, it has rarely become a reality. Of the machine and plant manufacturers surveyed for the study, only 11 percent stated that they already had a comprehensive offering on the market, while 30 percent had at least a basic offering. 40 percent are in the development phase. And 19 percent have not even started yet. Comparable figures are not available for the situation among users. However, there is much to suggest that predictive maintenance is still a real rarity here.

Strategic question: Make or buy?

On the one hand, this is due to the fact that the range of services offered by machine and plant manufacturers is not yet particularly plentiful. On the other hand, manufacturing companies are faced with the strategic question of how to organize predictive maintenance. There are basically two options:

Buy: Companies rely on machine and system manufacturers and buy in predictive maintenance as an end-to-end service. The advantage of this is that they have little or no need to worry about data acquisition and evaluation. However, as there are no machines from just one manufacturer in any factory, companies then have to deal with a large number of services. This also means that the responsible production managers have to keep an eye on several service dashboards. What’s more, this manufacturer-oriented view makes it difficult to take interdependencies between machines into account. However, this is essential for maintenance planning in particular. Another disadvantage is that such end-to-end services are possible for models of tomorrow or the day after tomorrow. However, machines and systems that have been in use for some time and are expected to run for a few more years are left out.

Make: The manufacturing companies themselves ensure a predictive analytics approach in which they integrate all machines and systems – regardless of the supplier and how old they are. This would give them a complete overview and allow them to take all dependencies into account. However, the technological effort required in this case is much greater. And this is particularly true because many manufacturing companies – despite Industry 4.0 being an ongoing topic – are not yet particularly advanced when it comes to digitalization. This is the case, for example, when it comes to the basic recording of operating and machine data. Paper and Excel lists are still widely used. Although the use of manufacturing execution systems has increased in recent years, they are still far from standard.

He who hesitates, loses

In order for companies to make a well-founded decision as to which path is the right one for them, they should first analyze the current situation on the shop floor in detail. Answering a series of questions is helpful here:

• Which machines and systems are available and what data do they provide?
• Which machines and systems will need to be replaced in the foreseeable future?
• Can machines and systems that are already older but are still to be used be retrofitted with sensor technology – in order to generate the necessary data?
• Which manufacturers already offer which predictive analytics services, what do they look like and how much do they cost?
• What manufacturer-independent forecasting software is available and how can it be integrated into the existing infrastructure?

If this inventory shows that a significant proportion of machines and systems can be maintained in future via an end-to-end service – and in the best-case scenario by just two or three manufacturers – a buy strategy is the obvious choice. Today, however, this is likely to be an absolute exception. However, doing nothing for the time being and hoping for increasing standardization is definitely not the right decision. After all, this can too quickly lead to competitive disadvantages that are difficult to compensate for later.

It makes more sense to start by consistently recording the existing data and processing it using a manufacturing execution system. On the one hand, this allows meaningful key figures on the current situation to be obtained – which is a considerable benefit. On the other hand, maintenance can be successively developed and expanded on the basis of such a solid database. Because one thing is certain: Many companies today have not even implemented condition-based maintenance or condition monitoring. Implementing these concepts is therefore a necessary first step. Forecasting procedures can then be tested for individual machines – perhaps even for entire production lines – in lighthouse projects. Experience will be gathered in order to determine the appropriate strategy more and more precisely. Hybrid approaches are also conceivable: Some machines are then integrated into the end-to-end service of a manufacturer. However, the manufacturer only analyzes the data and sends the results back to the company. The company can then combine these with the results from other sources. This creates the desired overall view.