Image-based State Recognition And Dynamic Analysis Of Lifting Equipment Braking System

In recent years,the construction of transportation infrastructure has been developing rapidly.As an important infrastructure in traffic engineering,reinforced concrete beam components need to be prefabricated on site during the construction of bridges.Lifting and transportation equipment for bridge construction such as girder lifting machine,girder transporting vehicle and bridge erecting machine is indispensable.Therefore,the construction safety of such lifting and transportation equipment has received more and more attention.The condition monitoring and fault diagnosis of the braking system of the lifting mechanism of the lifting equipment for bridge construction play an important role in construction safety.Therefore,it is necessary to study the safety monitoring and early warning theory and engineering application of the braking system of the lifting mechanism of such equipment.Computer vision technology combined with traditional machine vision methods can effectively identify the state of key parts of the brake system,measure the position information of parts with high precision,and obtain relevant mechanical parameters,which provides a reliable basis for the safety state identification and early warning of the brake.At the same time,the use of visual measurement can improve the intelligent level of safety monitoring of lifting equipment,and reduce the professional requirements of equipment operators and the attention of operating equipment.The main contents of this thesis are as follows :(1)The kinematics,statics and nonlinear dynamics models are established by using the operating characteristics and force characteristics of the brake system mechanism.The key parameters of the brake state-brake shoe thickness,brake oil pressure,braking force,disc spring stiffness and brake disc spring compression are related.The viscoelastic damping and Stribeck friction damping are introduced into the dynamic model of the braking system to study the dynamic behavior of the braking system,and the frequency response characteristics and stable state of the system are obtained.The correctness of the calculation results is verified by comparing the analytical solution with the numerical solution.The influence of braking force on braking system is analyzed,and the theoretical safety range of braking force and the relationship between braking force and system dynamic response are solved.(2)A dual task of image classification and semantic segmentation is established to separate explicit faults and implicit faults.Based on the U-Net + + semantic segmentation framework,the residual structure with group convolution is used to eliminate the gradient explosion problem under high learning rate.Through the multi-scale feature aggregation structure,each scale feature is reconstructed for the classification task,and the compression of the network model is realized through network structure multiplexing.The multi-task loss function is constructed to realize the parallel training of two tasks,so that the final effect of classification and segmentation is more balanced.Combined with the actual situation of the brake,a multi-task fault diagnosis data set is constructed.The explicit fault images are separated after labeling,and the non-explicit fault images are semantically segmented and labeled for semantic segmentation tasks.The multi-task fault diagnosis data set is constructed for fault diagnosis network training and model verification.(3)Establish a visual measurement model.Based on the results of multi-task neural network and visual prior knowledge,a visual measurement model is established.The noise in image segmentation is filtered out by morphological filtering method of machine vision.The proportion of window area is transformed into displacement by pixel distribution law,which improves the measurement accuracy.At the same time,by comparing with the traditional sensor,the correctness of parameter identification such as brake disc spring deformation,brake shoe clearance,brake oil pressure and braking force is verified.