Drone Data and Machine Learning Optimize Wind Turbine InspectionsPosted for Measure
Measure UAS, Inc.
- June 8, 2019
- 602 views
Drones have already proven themselves in the wind energy industry with many efficiency and safety benefits. Drones capture high-quality data while avoiding hazardous man-hours and minimizing downtime.
The focus is now shifting to advancements in data analysis, primarily in automation and machine learning (ML). Measure’s data collection and analysis processes were developed with machine learning in mind, allowing us to build datasets with consistent labeling in a similar data type. Through our work with partners like AES, a global energy company, we have inspected thousands of wind turbines and collected one of the largest turbine datasets available.
Measure was incredibly excited to work with AES and Google to employ Google's AutoML Vision technology to automatically identify defects and improve the speed of data analysis.
How it Works
Measure collects about 300 images per turbine, and with wind farms typically having dozens, or even hundreds, of turbines, we’ll usually end up with thousands of images that need to be processed and analyzed after each inspection. Our data engineering team had already been working to transform tedious processes from manual to automated, using both in-house and third-party solutions. To start, Measure developed a patent-pending method of automatically sorting and organizing large sets of turbine inspection images to make data processing more efficient.
Working with AES and Google, Measure provided raw data sets and annotated inspection imagery from hundreds of past AES turbine inspections, which were used to train Google’s Machine Learning/Computer Vision models. Measure’s expert data analysts worked with Google to validate and refine the model through each stage of development. Measure is providing on-going expertise regarding turbine defect identification and classification to continuously advance the model’s results.
The Benefits of Machine Learning
Combining Measure’s data set and wind turbine expertise with Google’s Auto ML has sped up the machine learning development process significantly. With machine learning, images with and without defects are identified and sorted automatically, vastly reducing the number of images requiring manual review and cutting the time between data collection and inspection results delivery. As we further refine the ML algorithm, it will also automatically classify the severity of defects, which means Measure’s blade experts will only need to examine a few images with specified defects and recommend the best course of action instead of the entire set of thousands of images.
For AES, robotics and AI are a key part of their mission to accelerate a safer and greener energy future. Drones coupled with advancements in machine learning enable safer, faster turbine inspections and open new opportunities for employee advancement. Learn how AES’s partnerships with Measure and Google are helping to advance the energy future in this video:
Working with AES and Google’s AutoML has been a great opportunity for Measure. The ability to process and analyze data more quickly will both bring down costs and speed up data delivery time. And we have just scratched the surface. Measure’s deep experience in drone operations has resulted in quality datasets not just in wind, but also in solar, electric distribution, and other energy generation infrastructure. With continued work, we will be able to expand machine learning to improve data products across the energy industry.