Why Industrial Analytics in the Cloud

Predictive maintenance using industrial analytics is not a vision of the future for a time beyond the horizon, but a practical reality. The technologies and algorithms required for this, as well as data from the systems and machines, are often readily available for use and "only" need to be meaningfully evaluated and assessed.

Although modern machines and systems consist of established, mostly standardized and interchangeable components, some of which are produced by competitors, we rarely use generally known analyses, best practices or the like in the field of industrial analytics, but instead keep reinventing the wheel.

Of course, this is only because our machines and systems are completely different from those of our competitors. But are they really, when the components, control systems etc. are similar and often even substitutable?

Yes, this is certainly possible if the basic population of all systems really enables such best practices. Large corporations may be able to do this, but German mechanical and plant engineering in particular is a highly diversified market. This may have made Germany the global market leader, but it could be counterproductive for future machine and plant services. This is because the individual development of an industrial analytics infrastructure and the development of meaningful analyses usually involve considerable investment, which not every small and medium-sized machine and plant manufacturer is willing or able to make.

It can also be observed that many companies implement their own analyses via newly established digitalization departments in order to detect anomalies, for example, but then fail to scale and cover the end-to-end process (in particular the process of actual troubleshooting).

For example, red flags can often be raised for an anomaly in a component that signals an anomaly, but it remains unclear what this means.

What was the probable cause? What troubleshooting measures were helpful in similar cases? How long can the component still be used? Will it last until the planned maintenance? Is it necessary to react immediately? Can the problem be repaired, is a replacement part necessary or is a different parameterization simply sufficient for the system?

All these questions often remain unanswered. But aren't these questions at least as important as the anomaly itself? In addition, the anomaly detections described above may work for one component or assembly, but not for all the other hundreds of components in the system and their interaction. Scaling in traditional mechanical and plant engineering cannot usually be taken into account here due to the scope of the analyses.

We have found that the analyses based on components or assemblies are quite comparable and "only" the interaction of the components or their analyses actually need to be considered on a company-specific basis. This means that it makes perfect sense to jointly develop certain analyses for individual components or to draw on analyses already developed by other companies.

Following the installation of established technical components from suppliers in your own machines and systems, the use of established analyses from industrial analytics providers can now open up new ratio potential.

How does it work?

However, the further standardization of such analyses requires a certain degree of standardization in terms of methods and IT infrastructure. This is where modern cloud solutions come into their own. Customers can largely avoid major investments in their own infrastructures and IT and, above all, OT (Operational Technology) technical expertise and benefit from the shared infrastructure, but above all from its "specifications". This is because such standards can also help to avoid getting bogged down in IT technical details that are far removed from the core business of a machine and plant manufacturer and instead focus on the individual analysis of the component analyses provided out of the box with the cloud solution. This is where the real added value of the specialist supplier of a system lies, not in the umpteenth development of a time series analysis for a sensor to detect pump problems. It therefore does not make sense for every company to develop its own industrial analytics solutions based on an analytics platform, such as Microsoft, Siemens, Bosch or IBM, but to obtain ready-made business or component analyses from an industrial analytics cloud for precisely these platforms and simply add them on. This is precisely the advantage of modern cloud infrastructures.

In contrast to generic IT platforms, the use of cloud solutions offers the advantage that some analyses are already available for the above pump problem, for example, which can be easily accessed without having to allocate your own capacities. Such solutions are currently being developed, but still need to be more widely accepted in industry. There are still too many fears of relinquishing key expertise and therefore essential competitive advantages by using such a cloud platform. But are analysis processes for a pump really of such high competitive relevance or is it not the interaction of all analyses, corresponding to the interaction of the mechatronic components, that provides the real competitive advantage?

Reducing return on investment and time to market

Cloud solutions for industrial analytics can therefore ensure a significantly faster return on investment and time to market. In today's dynamic, globalized economy, these factors are perhaps much more important than the time-consuming conversion of a company's existing detailed knowledge of a vibration sensor into its own analyses with its own IP. Cloud solutions based on German legal systems in particular also offer the advantage that each participant in such a system can safely define the extent of their willingness to cooperate with other companies on this platform. Do they only want to participate in the component analyses provided or also make their own analyses available to the community in order to further develop the platform? The individual orchestration of the component analyses provided by the cloud solution, analogous to the mechatronic components, to create a unique selling proposition for the system generally remains accessible only to the company and (if desired) its customers. The same naturally applies to the original sensor data.

"Progress thrives on the exchange of knowledge."

Albert Einstein propagated this idea of sharing knowledge for personal benefit many years ago. Applied to industrial analytics cloud platforms, this can be interpreted to mean that providing your own component analyses (regardless of whether they were developed in-house or by a service provider) on such a solution will lead to others sharing their content. A crowd-sourcing approach that could be particularly interesting for small and medium-sized companies whose capacity is not sufficient to build up a parallel innovation process for industrial analytics analyses in addition to innovations in their core business. For these companies in particular, it may therefore make sense to set up their offerings for predictive maintenance services relating to their products on an analytics cloud platform with other companies in the industry rather than on a purely IT platform.