Innovation project: Up to 75 percent of operator time for data analysis saved on blade testing

Based on artificial intelligence and image recognition, leading researchers and companies from Denmark have optimized the work with testing the wear of wind turbine blades.

When the wind really takes hold, the wind turbine blades spin at top speeds of 320 km/h.

At these speeds, raindrops and dust particles turn into projectiles, which is why the leading edges of the wind turbine blades often need to be serviced.

With a Rain Erosion Tester, you can predict the lifetime of wind turbine blades, but large parts of the data collected by the tester are never used, as it is a very time-consuming process to analyze data. The innovation project Requim does something about this.

“As a rule of thumb, we say that ten hours of testing requires an hour of analysis afterwards,” says Jesper Dal Hasager, Technical Product Manager at R&D Test Systems A/S, who is the problem owner in the innovation project and together with DTU Wind and Energy Systems and Wind Power Lab solves the challenge.

The EU Regional Fund supports financially, and Energy Cluster Denmark facilitates.

AI saves time

Through an innovative AI solution, the engineers will be able to reduce the amount of time spent manually analyzing the test results.

“The primary purpose of the project has been the AI model, which means that those who work with erosion tests can save 50-75 percent of the time. This allows you to move brainpower to other tasks,” he says.

Charlotte Bay Hasager from DTU Wind and Energy Systems sees value in establishing an artificial test environment that resembles the actual conditions offshore.

“Usually you use your eyes: here’s an injury. It’s a bit of a tedious process to sit and look at the process manually. What is new is that you use image recognition and an algorithm in a computer to find the damage. When do the injuries happen? When is there a hole in the coating? It’s a more cunning way to do it,” she says.

At the level of expert

The new AI tool can point out where the damage has been done. The machine is trained to recognize errors, and is in tests at the level of specialists:

“We have looked at an expert who is used to finding errors manually and measured his troubleshooting with the algorithm’s results. They are quite identical,” says Charlotte Bay Hasager.

Anders Røpke, CEO of Wind Power LAB, is also excited about a successful outcome of the innovation project:

“It has been a really good learning experience for all of us, and I expect that AI analyses will be the foundation for good data structures in the future. It takes away a lot of wasted time,” he says.

For Anders Røpke, there is still a lot of development potential to investigate:

“We can also use the product in aviation, and the scaling possibilities are therefore quite large. We hope that it will become an EUDP project with the same partners in the future,” he says.