For some years now, new digital technologies have been in active use. Choosing the right technology and especially the right provider is not always an easy task. The spectrum of solutions is broad. The most important question is not which provider to choose, but what actual use case do I have and what expectations such a solution must cover. Christopher Dreke and Marco Noerenberg from Berliner Wasserbetriebe draw their first conclusion after the pilot project with Industrial Analytics at the company's third largest wastewater treatment plant. For the pilot project, Industrial Analytics monitors a compressor in Schönerlinde (Berlin).

Mr Dreke, you are responsible for strategic execution at Berliner Wasserbetriebe, and that includes investing in new technologies. With what challenges are you presented?

Christopher Dreke: The market offers a wide range of new technologies and for Berliner Wasserbetriebe the question is, which solutions do we choose? Currently, there is an oversupply of automation solutions. We came across Industrial Analytics through a start-up scouting with our partners Hamburg Wasser and Gelsenwasser. This enabled us to select start-ups from a shortlist on the topic of predictive maintenance. For us, the main questions in the selection process are: What makes sense solution-wise and business-wise? What is economically efficient for us? What is the best choice for us? Superficially, many companies have the same problems, but the devil is in the details. We have special challenges at Berliner Wasserbetriebe.

Berliner Wasserbetriebe operates system-critical infrastructure. How can AI solutions help to ensure security of supply? And how do you get your personnel to use new solutions?

Christopher Dreke: System failures are the worst case for us. Especially at weekends or on holidays, because there is less staff at the plant. AI can help to avoid these breakdowns. Currently, we have not yet reached the point where we leave the decision to an AI solution. We see it as a tool that facilitates our decision-making-processes. The experience and decisions of the employees must be taken into account. 

Marco Noerenberg: A good example is the pilot project in Schönerlinde, where we’re monitoring a compressor with the help of Industrial Analytics. We are making the same diagnoses with the help of the AI algorithm, we would normally do with our staff. But without AI algorithms, it takes much more time and requires a lot more manpower. To successfully implement new solutions, we need a practical example that is well-developed and convinces our plant managers.

Christopher Dreke: When implementing such a technology, it is important to involve the employees and establish a close relationship to them. We have had negative experiences with simply rolling things out. If the plant personnel is not convinced, it doesn't work.

What are the motives to apply such a technology? On what do you base your decision on?

Marco Noerenberg: Climate change, profitability and demographic change within the workforce are motives to test the performance of such technologies in a real environment and see the potential of future reliefs.  

Christopher Dreke: We keep all these factors in mind, especially the environmental ones, and compare different processes and methods when making a decision. We don't use AI technology, just because it is an AI. What’s important is that it suits us. If AI is the tool of choice, then that's what we’ll use.

You use Industrial Analytics to monitor one of your compressors with retrofitted vibration technology. What exactly was done there, and what strategic goals and expectations were associated with it?

Marco Noerenberg: Compressors are essential for wastewater treatment, so they are particularly critical and their maintenance is expensive. We therefore wanted to prove to our operator that the saving potentials in maintenance and the stretching potential of maintenance cycles are big, and in some cases the maintenance cycles can even be doubled. With Industrial Analytics we built a simple and lightweight pilot that basically acts like an ECG on the machine. The system is giving us deeper insights of the machine and tells us what is happening inside the machine, like detecting changes in condition. What we’ve learned is that the calibration of the analysis took much longer than originally expected. 

And what are your expectations for a future cooperation with Industrial Analytics? 

Christopher Dreke: What we have learned is that the technology has its raison d'être. But it can't be a plug-and-play solution, that's just because of the different types of machinery. We need to work together for a longer period and compare whether the ratio between the benefits of the system and its economic efficiency is in the right balance. 

Marco Noerenberg: The knowledge growth for Berliner Wasserbetriebe through the cooperation with Industrial Analytics was the greatest success for me. And, of course, showing that sensor-based condition monitoring that is AI-supported brings added value compared to time-consuming manual analysis. In the dashboard, we see if the machine is in good condition. We get active information when measured values change to a critical range. And the whole thing is not based on fixed maximum or minimum values, but dynamically depending on the current machine condition. These indicators for changes in condition are important for detecting real faults and avoiding total failures later on.

Interview originated on Ruhrhub blog: https://ruhrhub.de/blog-single/wir-nehmen-keine-ki-weil-es-eine-ki-ist-wichtig-ist-dass-es-zu-uns-passt