| Steel structures are widely used in the field of construction.Pipe trusses are important components of steel structures,and most of them need to be welded and assembled on site.The welding quality inspection is inconvenient,time-consuming and laborious,and the operation is dangerous.Therefore,this research designs and develops a welding seam quality inspection robot for steel structure pipe truss.The main research contents are as follows:Robot structure and control system design.According to the spatial structure characteristics of pipe truss and the requirements of robot walking and obstacle crossing,the main structure of the robot with wheeled walking and clamping attachment is selected,the clamping,steering and turning mechanisms are designed,and the robot motion control and image acquisition system is developed.Structural optimization and transient characteristic analysis of clamping mechanism.The structure of the clamping mechanism is optimized.Taking the motion constraint and strength constraint of the clamping mechanism as the main constraints,the mechanical and kinematic models of the main parts of the clamping mechanism are established,and the size of the clamping mechanism is optimized by using the penalty function method.Static analysis was carried out on the outrigger,and topology optimization was adopted to further reduce the weight of the outrigger.The weight of the robot after topology optimization and reconstruction was reduced by 16.16%.The transient characteristics of the clamping mechanism are analyzed,and the sensitivity of the clamping mechanism is simulated and optimized in MATLAB.Appropriate parameters are selected and dynamic absorbers are added to improve the transient performance of the clamping mechanism,and dynamic tests are carried out.The errors of transient performance indexes between the test results and the simulation results are less than 5%,and the test results are consistent with the simulation results.Kinematics and stability analysis of robot obstacle crossing.A mathematical model with job loss efficiency as the objective function is proposed,which optimizes the main parameters of the flip joint,and obtains the optimal angular velocity of the flip joint through data fitting and multiple iterations.The stability test of the prototype is carried out to test the slippage of the robot under different turning angular velocities.The test results show that the stability of the robot in horizontal state is better than that in vertical state,and the slippage in horizontal state is much smaller than the structural size of the whole machine,and the overall stability is good;In the vertical state,with the increase of angular velocity,the stability of the robot becomes worse.When the turning angular velocity reaches 0.785 rad/s,the robot slips.The experimental results are consistent with the kinematic simulation results.Research on inspection system of weld surface quality.The types and characteristics of weld surface defects are analyzed,the weld image is denoised by adaptive median filter,and the weld area is segmented by a region segmentation algorithm combining anisotropic filter and watershed algorithm.The Defective-Resnet model is designed to classify the image by combining residual learning with VGGNet.The classification accuracy of Defective-Resnet algorithm on the test set is 93.68%,and the classification effect is better than the traditional algorithm. |