What happens when you combine the knowledge of Wärtsilä experts and artificial intelligence? The outcome is a unique, cutting-edge predictive maintenance service.
Wärtsilä is launching its new Expert Insight predictive maintenance service as a part of its Lifecycle Solutions portfolio. Expert Insight is a condition monitoring system powered by artificial intelligence (AI) specially designed to detect
anomalies and potential equipment failures before they occur.
“It makes us far more predictive and proactive. We can now start to give advice to customers based on forward-looking predictions rather than looking backward at what’s already happened,” says Paul Kohle, Director, Performance Services, Wärtsilä Marine.
The system benefits from AI that learns as it accumulates data. Unlike other products, which focus on troubleshooting after an issue has taken place, Expert Insight flags issues before they cause actual problems.
“What is different today is that we tend to work by measuring the selected parameter with set alarm values, and only when these values are exceeded does an alarm register” says Kohle. “With Expert Insight, we predict the value itself. It can recognize a deviation at the earliest possible moment, long before we get to an alarm state.”
The solution incorporates a neural network, a set of algorithms based loosely on the architecture of the human brain designed to recognise patterns in numerical data. It was built by Wärtsilä data scientists using Google TensorFlow, which powers the calculations behind the programme and allows Expert Insight to crunch a year’s worth of engine data in only a few minutes. Thanks to this speed, Expert Insight can identify anomalies in near-real time.
“Around 15 minutes after an anomaly happens, it will be flagged in Wärtsilä Expert Insight. This timeframe will continue to narrow in the future as the hardware becomes more powerful and connectivity is faster,” says Frank Velthuis, Wärtsilä’s Director of Digital Product Development.
For more than two years, Wärtsilä engineers and data scientists worked on improving the AI to identify anomalies in data from the regular operations of engines, and to use both its successes and failures to improve itself. Despite the superior
speed of AI in processing data, Expert Insight still requires humans to interpret the findings. When it flags an anomaly, experts at Wärtsilä Expertise Centres worldwide step in to take a look.
“Unfortunately, AI today doesn’t yet have the capability to interpret the anomalies they discover. A computer can’t determine whether an anomaly is a precursor of a severe failure or a less problematic defect, such as a broken sensor, which has no effect on machinery performance,” explains Velthuis.
Nevertheless, by eliminating the need to pour over reams of data looking for problems, Expert Insight allows equipment experts to focus on supporting the customer proactively, preventing issues before they even occur, improving the overall performance of the systems in use.
An important part of the process has been collaborating with customer experts, such as chief engineers or fleet managers, to identify and confirm underlying problems causing anomalies in data.
Because Expert Insight is developed wholly by Wärtsilä, it is easy to integrate into the company’s product portfolio, says Erik Ristiluoma, General Manager of Maintenance & Operations Management. Additionally, having the product in-house also avoids expensive licensing fees.
“We’ll be able to bring this to our customers much more easily,” Ristiluoma says.
Expert Insight also allows Wärtsilä to be more precise about the life span and maintenance requirements of its products.
“From a long-term service agreement perspective, the application can give us an opportunity to give better guarantees that equipment will function as they should and without interruptions, as we are better able to catch potential problems at an earlier stage,” explains Ristiluoma.
In the marine sector, extending the lifecycle solution offering with Expert Insight means less downtime and fewer days in port. Expert Insight’s real-time monitoring of shipboard performance allows for Remote Services to step in and determine if
a port call is needed. Identifying potential anomalies in real time also means that ships can make the most of scheduled port calls and better predict maintenance costs.
“It’s really about a reduction of unplanned downtime and maintenance, which are costly for marine customers,” says Paul Kohle.
There are two paths being explored for expanding Expert Insight in the future. One path is to further deepen the AI’s knowledge of four-stroke engine systems, incorporating even more data to increase understanding of the root causes of problems.
“There is a large potential population of engines that we can install the system to, and this will allow for gathering even more data to teach the system. The experts can then have even more time to dive deeper into helping customers solve problems,” says John Kop, Wärtsilä Chief Digital Product Owner.
The other path is to expand the AI’s capability to work with other equipment, such as 2-stroke engines, scrubbers and gas systems on LNG carriers.
Both options open up numerous possibilities for Wärtsilä customers.
“Expert insight gives us a very valuable building block for creating value, adding lifecycle solutions for our clients,” says Kohle.