When machines whisper...

...our experts listen! How LANXESS uses AI-based maintenance to prevent plant downtime, save millions —and make its operations even more digital.

Sometimes it's just a slight drop in pressure in the lubricating oil system, a gentle increase in fan speed, or an inconspicuous rise in temperature at a pump bearing. Deviations from normal signal behavior that are barely noticeable – but can mean a lot. Bastian Höfer, Head of Predictive Maintenance, and his team in the Group Function PTSE have learned to hear precisely these “whispers.” Using AI-supported software, modern sensor technology, and deep insights into machine and process data, they discover problems before they even arise. After about a year of rollout, it is clear that this early listening is paying off.

Mr. Höfer, LANXESS is currently rolling out a tool for predictive maintenance. Where do we stand at the moment?

Bastian Höfer: About one-third of all particularly critical plant components are now connected to our monitoring system – around 100 objects in all business units and almost all regions. We are currently focusing on large rotating machines such as centrifuges, compressors, and agitators, because damage to these can quickly become very expensive. We are currently detecting an average of one emerging problem per week – and the trend is rising.

What have you been able to achieve so far?

Bastian Höfer: A lot: we have already detected around 40 potential problems in good time. Without our analyses, many of these would only have been noticed once they had already had serious consequences. Everything from bearing damage and leaks to blocked pipes and defective sensors was included. And it is precisely this diversity that demonstrates the strength of the system: it not only detects individual fault patterns, but also acts as a fine-meshed monitoring network. Overall, we are now talking about savings in the seven-digit range – not to mention avoided delivery failures and angry customers.

How exactly does the monitoring work?

Bastian Höfer: First of all, nothing works without a robust software environment and a reliable data infrastructure. Both are provided by our colleagues from GF IT, with whom we work closely. Our tool then analyzes the interaction of values for e.g., current, pressure, temperature, and vibrations from our central data archive (AVEVA PI) and thus learns the “normal behavior” of a machine, so to speak. If values deviate from this, the system sounds the alarm and shows us precisely which signals triggered the anomaly. Then the real teamwork begins: we check the data, discuss it with our experts in the plants, and decide together whether and what measures are necessary. It is important to note that AI does not make decisions. It provides information. The expertise of our employees remains the benchmark.

Once the rollout is complete, do you see further potential?

Bastian Höfer: Absolutely! We are currently investigating several use cases. One focus is on monitoring control valves. These are found in large numbers in our plants, and the software designed for this purpose can not only detect types of faults, but even calculate the expected remaining service life. This will enable us to avoid unnecessary valve revisions and unplanned downtime in the future. Monitoring steam traps is another highly scalable use case. If a trap is defective, steam may escape unnoticed for weeks, leading to significant energy losses. Continuous monitoring allows us to detect such defects at an early stage. The investment usually pays for itself within a few months. At the same time, we are making a noticeable contribution to energy efficiency and thus to greater sustainability.

Are our systems already digital enough for this?

Bastian Höfer: In many cases, yes. We now have access to data from around 600,000 sensors in our central archive – an excellent basis. For some special applications, however, we need to install additional sensors to become even more transparent. To do this, we use modern, wireless technology that can be easily integrated – without major interventions and high costs. Each additional sensor expands and strengthens the network at the same time: Not only do we gain access to the data relevant for monitoring, but we also give digitalization in production and maintenance a powerful boost.