Power line transmission inspection using deep learning and a convertible drone

Objective:

This project aims to develop and implement a deep learning-based algorithm for inspection of electric transmission lines with cameras embedded in a convertible UAV.

A poor condition of power lines carries consequences of accidents involving total energy transmission loss and destruction of electric towers. For these reasons, the inspection of such electric lines encompasses factors such as: detection of a poor conductor connection, physical condition of conductors, structural state of the towers, and detection of the corona effect.

Figure 1. Convolutional neural network model.

Currently, the inspection is carried out by specialized technicians, representing a risk for these personnel. For solving that, several research efforts have been taken looking for ways to improve and optimize the power line inspection. For instance, some actual works [1], [2] use artificial intelligence (AI) algorithms to detect automatically common faults presented in electrical transmission lines, as well as track them to analyze hard-to-reach locations; in Fig. 1 it is presented the neural network model for this purpose.

On the other hand, the use of semantic segmentation with AI [3] is used to detect specific objects to characterize the potential problems in the electrical transmission lines. In [4], [5], and [6] authors use deep learning algorithms trained with thousands of images related with power lines issues. In [7], the authors use thermal images to detect poor connections; Fig. 2a depicts an example of a thermographic inspection image.

Fig. 2a. A picture from a thermal camera of a transmission power line.

Fig.2b. A fixed-wing aircraft inspecting a transmission line.

Adapting AI to drones endowed with cameras, allows to develop a variety of applications concerning object detection and inspection. The use of a convertible drone relies on the fact that it can perform both flight modes: hover and cruise, depicted in Fig. 3 [8]. Hence, the application of inspection can be expanding no only to analyze a particular area with hover flight mode, but also to cover a big area of interest due to the convertible drone capability of move at high velocity during cruise flight mode.

In our lab, two different convertible drone platforms are used for research. In Fig. 4 it is shown a tail-sitter UAV used currently in the laboratory.

Fig. 3. The three flight modes of the tail-sitter drone: hover, cruise and transition.

Fig. 4. Tail-sitter convertible UAV developed at the LAPyR-CIO Lab.

References

[1] C. Sampedro, J. Rodriguez-Vazquez, A. Rodriguez-Ramos, A. Carrio and P. Campoy, "Deep Learning-Based System for Automatic Recognition and Diagnosis of Electrical Insulator Strings," in IEEE Access, vol. 7, pp. 101283-101308, 2019. doi: 10.1109/ACCESS.2019.2931144

[2] P. Pienroj, S. Schönborn and R. Birke, "Exploring Deep Reinforcement Learning for Autonomous Powerline Tracking," IEEE INFOCOM 2019 - IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS), Paris, France, 2019, pp. 496-501. doi: 10.1109/INFCOMW.2019.8845212

[3] L. Wang, Z. Chen, D. Hua and Z. Zheng, "Semantic Segmentation of Transmission Lines and Their Accessories Based on UAV-Taken Images," in IEEE Access, vol. 7, pp. 80829-80839, 2019. doi: 10.1109/ACCESS.2019.2923024

[4] X. Miao, X. Liu, J. Chen, S. Zhuang, J. Fan and H. Jiang, "Insulator Detection in Aerial Images for Transmission Line Inspection Using Single Shot Multibox Detector," in IEEE Access, vol. 7, pp. 9945-9956, 2019. doi: 10.1109/ACCESS.2019.2891123

[5] V. N. Nguyen, R. Jenssen and D. Roverso, "Intelligent Monitoring and Inspection of Power Line Components Powered by UAVs and Deep Learning," in IEEE Power and Energy Technology Systems Journal, vol. 6, no. 1, pp. 11-21, March 2019. doi: 10.1109/JPETS.2018.2881429

[6] B. Tian et al., "Transmission Line Image Defect Diagnosis Preprocessed Parallel Method Based on Deep Learning," 2018 3rd International Conference on Mechanical, Control and Computer Engineering (ICMCCE), Huhhot, 2018, pp. 299-303. doi: 10.1109/ICMCCE.2018.00068

[7] D. G. Caetano et al., "Design and Implementation of an Automatic Vehicle for Thermographic Inspections in Electric Distribution Network Using Deep Learning Based Software," 2018 3rd International Conference on Computational Intelligence and Applications (ICCIA), Hong Kong, Hong Kong, 2018, pp. 145-150. doi: 10.1109/ICCIA.2018.00034

[8] A. Flores, A. M. de Oca and G. Flores, "A Simple Controller for the Transition Maneuver of a Tail-Sitter Drone," 2018 IEEE Conference on Decision and Control (CDC), Miami Beach, FL, 2018, pp. 4277-4281. doi: 10.1109/CDC.2018.8619303

State of the project

Project started in 2017; it is currently being developed.

Students involved in the project:

At the present time (October 2019) one PhD student is involved in this project.

Publications from the team:

  1. Quad-tilting rotor convertible mav: Modeling and real-time hover flight control. G Flores, J Escareño, R Lozano, S Salazar. Journal of Intelligent & Robotic Systems 65 (1-4), 457-471, 2012.
  2. PID Switching control for a highway estimation and tracking applied on a Convertible mini-UAV. G Flores, L.R. Garcia, G. Sanahuja and R. Lozano. IEEE Conference on Decision and Control (CDC 2012). Maui, Hawaii, USA, 3110-3115, 2012.
  3. Lyapunov-Based Switching Control for a Road Estimation and Tracking Applied on a Convertible MAV. G Flores, R Lozano, G Sanahuja. AIAA Guidance, Navigation, and Control (GNC) Conference, 2013.
  4. Transition flight control of the quad-tilting rotor convertible MAV. G Flores, R Lozano. 2013 International Conference on Unmanned Aircraft Systems (ICUAS), 789-794, 2013.
  5. Lyapunov-based controller using singular perturbation theory: An application on a mini-UAV. G. Flores and R. Lozano. 2013 American Control Conference, 1596-1601, 2013.
  6. A nonlinear control law for hover to level flight for the quad tilt-rotor UAV. G. Florer and R Lozano. IFAC Proceedings Volumes 47 (3), 11055-11059, 2014.
  7. 6-DOF hovering controller design of the quad tiltrotor aircraft: Simulations and experiments. G Flores, I Lugo, R Lozano. 53rd IEEE Conference on Decision and Control, 6123-6128, 2014.
  8. A simple controller for the transition maneuver of a tail-sitter drone. A Flores, AM de Oca, G Flores. 2018 57th IEEE Conference on Decision and Control (CDC), 4277-4281, 2018.
  9. Transition maneuver longitudinal control for the convertible unmanned aerial vehicle. G. Flores, Submitted. 2019.
  10. Tail sitter UAV full control. A. Flores and G. Flores. In preparation. 2020.

Financial support:

CIO

This thesis is developed in collaborations with PhD, master and undergraduate students who are LAB members.