Research On Damage Identification Of Steel Structure Communication Tower Based On Neural Network

With the continuous updating of China's mobile communication technology,network devices are constantly updated.In the current situation which multiple standard networks are operated together,equipment installed on steel towers is gradually increasing.Therefore,it is necessary to detection the structural health of the steel structure communication tower.At present,the detection method of steel structure communication tower mainly relies on manual inspection.Manual detection has shortcomings such as long detection period,late detection of damage and unstable detection quality.To solve the above problems,It is necessary to study the automated steel structure communication tower detection method.The core of automated detection technology is the automatic identification of damage.Therefore,this paper mainly studies the damage identification of steel structure communication tower based on neural network.This paper proposes a multi-step damage identification method based on BP-RBF neural network.Using ANSYS software to extract the modal data which can reflect the damage location and damage degree of the steel structure communication tower.The damage identification sample library is constructed by using the damage marker quantity based natural frequency,for neural network damage identification.Using BP neural network for damage location.Using RBF neural network to identification of damage degree.For the specific damage identification task,select the modal data that can reflect the damage information to form the damage sample library,and designed neural networks.By decompose the problem,Using multiple neural networks to identifying damage.The complexity of the damage identification problem is reduced,and the feasibility and reliability of the method are improved.After completing the preliminary design of the damage identification neural network.This paper selects a triangular steel tube tower in Shaanxi Province as an example to carry out the damage identification study.The BP neural network based on LenenbergMarquardt algorithm is used to identify the tower segment where the damage is located.The average recognition accuracy rate is 96.66%.The BP neural network based on elastic gradient descent algorithm is used to judge the type of damage bar.The average recognition accuracy rate is 96.66% The damage degree of the bar was identified by RBF neural network.90% of the sample identification errors are within 5%.By combining the identification information of the above three networks,the identification of the damage of the steel structure communication tower bar can be better accomplished.The feasibility of the neural network-based steel structure communication tower damage identification method is verified.Considering the influence of modeling accuracy and measurement factors on damage identification.In this paper,the damage tower segment identification is taken as an example,and the performance of the neural network under error is studied by using the sample with error.It has been verified that the average recognition success rate of damaged tower sections is over 80% under the error level of 5%.The neural networkbased steel structure communication tower damage identification method has good robustness.Finally,based on the above research results,this paper proposes a framework for automated non-destructive deceting of steel structure communication towers based on neural network damage identification,and further studies on automated detection systems.