Research On Tower Crane Structure Damage Diagnosis Method Based On Deep Learnin

Tower crane(hereinafter referred to as tower crane)is an indispensable construction machinery in the construction industry,mainly composed of on-site assembly of steel structures.Once its structure is damaged,it is highly likely to cause a major accident of tower crane collapse.Therefore,it is very important to identify whether there is damage to the structure of the tower crane.Traditional diagnosis methods mainly rely on the strong professional knowledge of experts to determine.In the case of increasingly complex tower crane structural damage and a large amount of operational data,there are problems of complex and inefficient manual diagnosis processes.Based on this,this thesis proposes a tower crane structural damage diagnosis method combining deep learning and data signal processing,which can achieve rapid and effective intelligent diagnosis of tower crane structural damage.Due to the complex operating conditions of tower cranes,the structural data used in this thesis have irresistible noise,making it difficult to extract damage features.In order to effectively extract the characteristic information of tower crane structural damage,empirical mode decomposition(EMD)is introduced to process the original data,and experimental research is conducted based on the cumulative contribution rate,correlation coefficient,spectrum change,etc.to determine the number of IMF components to be selected.The results show that the proposed method can achieve the effect of feature information extraction for structural damage.An intelligent damage diagnosis method based on One-Dimensional Convolutional Neural Network(1DCNN)is proposed.First,the decomposed EMD data was labeled,and the training and testing data sets were divided.Secondly,the damage diagnosis model of tower crane structure is established,and the measured data of tower crane structure is introduced into the model to get the accuracy of training and prediction.Finally,the influence of the selection of super parameters such as optimization algorithm,iteration number,convolution kernel and learning rate of 1DCNN diagnostic model on the accuracy was studied,and the experimental results of different network parameters were analyzed.The effectiveness of the proposed method is verified by comparing with the main methods at present.The influence of 1DCNN model generalization method on diagnosis results was studied.Dropout technology,Batch Normalization(BN)and L2 regularization are introduced into the model.Based on the experimental results,the accuracy and training time changes of 1DCNN structure before and after optimization are studied and analyzed.The experiment verifies that the introduced method can improve the accuracy and adaptability of 1DCNN model,that is,improve the generalization ability of the model.By combining EMD and 1DCNN,a Multi-Channel One-Dimensional Convolutional Neural Network(MC-1DCNN)model is constructed.The actual structural damage data of tower crane is decomposed by EMD method,and the proposed method is verified by the decomposed data.The influence of network parameters and network structure optimization on the diagnosis results was analyzed.The results show that the proposed method can accurately diagnose the structural damage of tower cranes,and the diagnostic accuracy is higher than that of the single-channel1 DCNN model and other deep learning methods.