Periodic Reporting for period 2 - Windrone Zenith (Autonomous & Intelligent UAV-based Wind Turbine Inspection System for Cost-effective, Reliable, Safe and Actionable Blade Fault Detection and Prediction)
Période du rapport: 2020-09-01 au 2021-08-31
Wind energy is at a critical point as it strives to be economically competitive against traditional energy generation sources without the support of subsidies. While the value chain that has been established over the past 15 years has delivered significant CAPEX reductions, these have plateaued, and the key now is to optimize OPEX costs which have a huge cumulative impact over the lifespan of a wind turbine. A key part in these OPEX costs is the inspection of the blades, which is currently carried out using unscalable, costly manual labour, or limited resolution ground imaging.
WindDrone is our autonomous drone-based single-blade inspection solution. WindDrone has successfully completed over 3,900 commercial inspections to date, proving value proposition, strong willingness to pay, ability to execute, and core technology, and providing us with critical and direct market information leading to the development of our next-gen product, the WindDrone Zenith.
The objective of Phase 2 is to bring to market our WindDrone Zenith, a breakthrough autonomous inspection solution enabling for the first time complete 3-blade inspection in a single flight, and incorporating machine learning and large data analytics to move beyond inspection and onto predictive maintenance and complete asset management. This solution reduces inspection downtime by a factor of 6, direct blade inspection costs by over 50%, and average annual OPEX by €2,800/on-shore and €8,850/off-shore turbine.
WindDrone is our autonomous drone-based single-blade inspection solution. WindDrone has successfully completed over 3,900 commercial inspections to date, proving value proposition, strong willingness to pay, ability to execute, and core technology, and providing us with critical and direct market information leading to the development of our next-gen product, the WindDrone Zenith.
The objective of Phase 2 is to bring to market our WindDrone Zenith, a breakthrough autonomous inspection solution enabling for the first time complete 3-blade inspection in a single flight, and incorporating machine learning and large data analytics to move beyond inspection and onto predictive maintenance and complete asset management. This solution reduces inspection downtime by a factor of 6, direct blade inspection costs by over 50%, and average annual OPEX by €2,800/on-shore and €8,850/off-shore turbine.
WORK PACKAGE 1
Milestone 1 is on route to be achieved, as a first production-ready automated inspection tool was deployed in our platform BladeInsight in September 2021. This tool has the ability to identify damages of types: Leading Edge Erosion and Cracks. Although the results are already over 50%, and commercially acceptable, the tool is still being improved by our internal blade experts using pilot data. We expect to release this tool commercially in late 2021.
WORK PACKAGE 2
During the first half of the project, the team working on WP2 has been mainly focused on 3 areas: Perception, Navigation, and Infrastructure – tasks 2.2 and 2.3.
On Perception, the focus was on the ability to interpret the data collected by sensor chosen to help with navigation, a 3D LiDAR. This sensor uses laser beams to sweep the environment and collect data points, that together create a point cloud that maps the obstacles around the system. This feature is particularly important to understand where the turbine’s tower and blades are located and how they are shaped, and to gather additional parameters that will be essential for path planning and image localisation.
On Navigation, having understood the environment around the system, the team then focused on designing algorithms to plan the path of the drone around the turbines. This work is mainly focused on finding the perfect balance between following the fastest flight course and ensuring obstacle avoidance, whilst ensuring 100% blade coverage. As the system is designed to be agnostic to the turbine/blade models, this work entails a strong effort on the mathematical models as they are put against the oddest real-life scenarios.
On Infrastructure, the focus was on creating a simulation framework that the team could use to rapidly test each iteration of the software developed. This foundational work was key to optimise resources and be for almost a year without needing any field test, while still making tremendous progress towards the desired solution. Moreover, the team implemented a set of online continuous integration methodologies that proved to be essential when everyone started working from home in early March. Without it, the development pace would have been significantly slower until finishing building such framework.
Lastly, this effort culminated in a field test done on June 17th, 2020 at Parque Eólico do Vergão 2, where our partners Generg made 1 turbine available for our team for a whole day. The scope of the test was to do a Proof of Concept on the navigation strategy chosen, as well as to test various photographic sensors.
Since this day, on top of leveraging the lessons learned on the field test, the team has also been focused on exploring the intricacies of the drone model – DJI M300 RTK – chosen to carry the new payload. This recently released model is expected to be capable to cope with demands that the new payload has, as far as flight time, carry capacity and connectivity are concerned.
During the second period of the project the team's main achievement were:
• Improved in the software-in-the-loop framework allowing the test of complex perception and navigation algorithms, within minutes after changing the codebase;
• Developed an automatic testing framework to create randomized test scenarios for every pull request made to the code. This system significantly improved the overall reliability of the algorithms and helped to identify logic, performance, and compatibility issues in the very early stage;
• Reached 95% success rate on surveying wind turbines in the simulation environment using SLAM (simultaneous localization and mapping) algorithms;
• Defined and implemented over 80% of the interactions and user journeys between the operator and the drone;
• On the infrastructure and system robustness sides, the team developed key supporting features such as data packing and processing, system monitoring and logging, or low-level system failures checks;
• Accomplished numerous smaller improvements to miniaturize the payload’s components, PCBs and electronics enclosures, in an effort to reduce the overall weight of the payload and increase the maximum flight of the drone.
As of September 2021 the team is looking forward to closing the final details of the first prototype so that it can be taken to the field and tested on a real turbine during the second half of October 2021.
Milestone 1 is on route to be achieved, as a first production-ready automated inspection tool was deployed in our platform BladeInsight in September 2021. This tool has the ability to identify damages of types: Leading Edge Erosion and Cracks. Although the results are already over 50%, and commercially acceptable, the tool is still being improved by our internal blade experts using pilot data. We expect to release this tool commercially in late 2021.
WORK PACKAGE 2
During the first half of the project, the team working on WP2 has been mainly focused on 3 areas: Perception, Navigation, and Infrastructure – tasks 2.2 and 2.3.
On Perception, the focus was on the ability to interpret the data collected by sensor chosen to help with navigation, a 3D LiDAR. This sensor uses laser beams to sweep the environment and collect data points, that together create a point cloud that maps the obstacles around the system. This feature is particularly important to understand where the turbine’s tower and blades are located and how they are shaped, and to gather additional parameters that will be essential for path planning and image localisation.
On Navigation, having understood the environment around the system, the team then focused on designing algorithms to plan the path of the drone around the turbines. This work is mainly focused on finding the perfect balance between following the fastest flight course and ensuring obstacle avoidance, whilst ensuring 100% blade coverage. As the system is designed to be agnostic to the turbine/blade models, this work entails a strong effort on the mathematical models as they are put against the oddest real-life scenarios.
On Infrastructure, the focus was on creating a simulation framework that the team could use to rapidly test each iteration of the software developed. This foundational work was key to optimise resources and be for almost a year without needing any field test, while still making tremendous progress towards the desired solution. Moreover, the team implemented a set of online continuous integration methodologies that proved to be essential when everyone started working from home in early March. Without it, the development pace would have been significantly slower until finishing building such framework.
Lastly, this effort culminated in a field test done on June 17th, 2020 at Parque Eólico do Vergão 2, where our partners Generg made 1 turbine available for our team for a whole day. The scope of the test was to do a Proof of Concept on the navigation strategy chosen, as well as to test various photographic sensors.
Since this day, on top of leveraging the lessons learned on the field test, the team has also been focused on exploring the intricacies of the drone model – DJI M300 RTK – chosen to carry the new payload. This recently released model is expected to be capable to cope with demands that the new payload has, as far as flight time, carry capacity and connectivity are concerned.
During the second period of the project the team's main achievement were:
• Improved in the software-in-the-loop framework allowing the test of complex perception and navigation algorithms, within minutes after changing the codebase;
• Developed an automatic testing framework to create randomized test scenarios for every pull request made to the code. This system significantly improved the overall reliability of the algorithms and helped to identify logic, performance, and compatibility issues in the very early stage;
• Reached 95% success rate on surveying wind turbines in the simulation environment using SLAM (simultaneous localization and mapping) algorithms;
• Defined and implemented over 80% of the interactions and user journeys between the operator and the drone;
• On the infrastructure and system robustness sides, the team developed key supporting features such as data packing and processing, system monitoring and logging, or low-level system failures checks;
• Accomplished numerous smaller improvements to miniaturize the payload’s components, PCBs and electronics enclosures, in an effort to reduce the overall weight of the payload and increase the maximum flight of the drone.
As of September 2021 the team is looking forward to closing the final details of the first prototype so that it can be taken to the field and tested on a real turbine during the second half of October 2021.
The project is still on route to contribute to the expected impacts.
Given the global pandemic, we would risk saying that the pressure to decrease the cost of wind power production is even higher now, increasing the market need for a solution like the one this project is delivering.
From an employment perspective, the project is and will continue to be a huge success, as the team is close to triple the number of employees since the project start. Also, like many tech start-ups with an initial male-only team, the company is investing heavily in increasing its gender diversity, both in tech and non-tech roles.
Given the global pandemic, we would risk saying that the pressure to decrease the cost of wind power production is even higher now, increasing the market need for a solution like the one this project is delivering.
From an employment perspective, the project is and will continue to be a huge success, as the team is close to triple the number of employees since the project start. Also, like many tech start-ups with an initial male-only team, the company is investing heavily in increasing its gender diversity, both in tech and non-tech roles.