Research On The Auxiliary Patrol System Of The Four-rotor UAV Slurry Pipeline

Technology of long distance slurry pipeline delivery has successfully solved the problem of mineral resource transportation in remote mountain areas.Since most slurry pipelines are laid in mountains and hills,significant economic loss can be caused in case of extreme weather.Making the rounds of the pipelines along the mountain roads,trackwalkers are unable to find out and confirm the accurate location of the fault pipeline in time,and therefore the fault cannot be recovered.As a result,it is significant to figure out a new method of fixing the fault position fast and meanwhile saving manpower and material resources.Aiming at solving the above problems,this paper designs an auxiliary line-tracking system which is based on small quad-rotor UAV to monitor slurny pipeline operation.Trackwalkers can track and inspect the pipelines automatically through the image processing and transferring devices of the UAV under the assist of line-tracking robot when it's unable to reach the pipelines.The main work and findings of the paper are as followings:(1)Build up a pipeline classification model based on Convolutional Neural Networks(CNN).This paper collects the simulated pipeline data under laboratory conditions for training the pipeline classification model of neural networks.This model can identify the pipelines in complex conditions by verifying on the test set,and the result shows that the classification accuracy basically meets the requirement.(2)Design a flight control system based on STM32F405 SCM.First,analyze features of the quadrotor by mathematical modeling to determine control strategy,and then obtain accurate attitude information by means of fitting correction of each sensor,including accelerometer,gyroscope and magnetometer.And finally design a flight controller based on cascade control algorithm.(3)Design the pipeline prediction software,flight control system software and hardware.The software of the model is designed by means of Python scripting language and TensorFlow deep learning framework,which is deployed on Jetson TK1 on-board computer.The software of flight control system,processing and calibration procedures of sensor data and control programs of attitude and position are all realized by C language and operate on STM32 SCM.(4)Install on-board computer and flight control system on the quadrotor and test it by tracking flight under the laboratory simulation pipeline.The quadrotor can fly autonomously by the simulation route which meets the requirements of line-tracking,successfully monitoring the operation condition of the slurry pipelines.