| With the progress of society and the development of economy,the spatial structure has developed rapidly and has become a common structural form widely used in construction engineering.The connection area of the joint is an important area for force transmission between the bar and the ball joint of the spatial structure.Its stiffness changes greatly,local stress concentration and force transmission are complex.There is also a risk of fatigue failure under the repeated action of wind load and crane load.If the damage disease cannot be monitored in time,it will lead to connection failure and promote structural damage.The reasons for the failure of the joint connection area mainly include two aspects: on the one hand,the whole joint connection area bears the role of transmitting the internal force of the rod.Whether the rod is tensioned or compressed,the increasing internal force of the rod will cause the joint connection area to be overstressed,which will cause the stress changes of the connecting parts such as bolts,sleeves and sealing plates.In extreme cases,it will lead to the failure of the connecting parts,which will pose a potential threat to the safe operation of the joint connection system and even the whole spatial structure;on the other hand,in the daily service process of the bolt-ball joint connection system,sleeves,sealing plates,screws and other components may also cause damage such as spalling of the protective layer and local cracks due to environmental corrosion,wind load and crane fatigue load.The emergence and development of these diseases will also bring security risks to the normal force transmission and service of the joint area.In view of the above two different types of damage diseases in the connection area of the bolt-ball joint,namely the change of the internal force of the bar and the crack propagation,a health monitoring method for the connection area of the bolt-ball joint of the spatial structure based on the piezoelectric intelligent sensing technology is proposed.A full-scale model of the bar specimen in the connection area of the bolt-ball joint is designed and manufactured.Through theoretical analysis and model test,the effectiveness of the proposed method is verified,including:(1)A monitoring method of member internal force change based on time reversal technique is proposed.Using a universal testing machine,seven test conditions of different tensile forces were designed by applying different tensile forces to the specimens to simulate the internal force changes of the members.Two piezoelectric ceramic plates(PZT)are used,one of which is pasted on the end of the rod on the side of the ball node as the actuator,and the other is pasted on the bolt ball as the sensor.Based on the time inversion technology,the peak value of the stress wave is used as the monitoring index,and the change of the internal force of the rod is monitored according to the change of the peak value of the stress wave.The effectiveness of the proposed method is verified by full-scale model test.Further,the repeatability test and the anti-interference test with human disturbance are carried out.The results show that the method has certain effectiveness,repeatability and anti-interference.(2)An internal force monitoring method based on piezoelectric impedance technology and neural network is proposed.Because the impedance method has high sensitivity,especially for small damage,compared with the seven test conditions set in the time reversal test,the impedance method test sets thirteen test conditions with different tensile forces with small changes,in order to realize the monitoring of small changes in the internal force of the rod.The internal force change of the member is simulated by applying different tensile forces to the specimen.A PZT plate is pasted on the rod end on the side of the near-spherical node,and the admittance data of the PZT plate is extracted based on the impedance method as the original data for monitoring.Then,BP neural network method and convolutional neural network(CNN)are used to process the monitoring data respectively,in order to achieve accurate monitoring of small changes in the internal force of the bar,including:(1)The BP neural network method is used to identify the internal force of the bar.In order to construct the database of BP neural network model,the test is repeated 25 times for each working condition.Since the sampling frequency of the original admittance data obtained from each test is801 data points,considering the large amount of admittance data,the original data is reduced by manual feature extraction.For each test,the original admittance data is divided into 10 sub-bands,and the original data dimension is reduced to 10 by calculating the RMSD values of each sub-band.For each test,the 10 RMSD values are used as the input parameters of BP neural network.The corresponding working condition tension is used as the output parameter of the network.Through the training and testing of the network model,the results show that the test accuracy of the internal force change monitoring model based on BP neural network reaches 86%;(2)In order to more accurately and comprehensively reflect the true internal force state of the specimen under different working conditions,the convolutional neural network(CNN)with strong data processing ability is used to identify the internal force change of the member.The 801 original admittance data obtained from each test are directly used as the input parameters of the CNN network,and the corresponding working condition tension is used as the output parameters of the network.The test is repeated 25 times for each working condition to construct the database of the CNN network model.Through the training and testing of the network model,the results show that the accuracy of the test of the internal force change monitoring model based on convolutional neural network(CNN)reaches 100 %.Therefore,compared with BP neural network,CNN has higher recognition accuracy,stronger data processing ability,and monitoring results are closer to engineering practice.(3)A crack propagation monitoring method based on piezoelectric impedance technology and neural network is proposed.Five different depth crack conditions are set up on the sleeve of the connection zone by manual cutting.A PZT sheet is pasted at the end of the rod on the side of the near-sphere node.The admittance data of the PZT sheet extracted by the impedance meter is used to input the RMSD data after the artificial feature extraction and the admittance data of the original PZT sheet into the BP neural network and the convolutional neural network(CNN),respectively.The test was repeated 25 times for each working condition to construct the database of the neural network.Through the learning and testing of the network model,the test accuracy of the two monitoring models reached 100 %.(4)Based on piezoelectric impedance technology and neural network,a method for simultaneous monitoring of internal force change and crack propagation in the connection zone of bolt-sphere joints is proposed.At the same time,five kinds of crack depth changes and five kinds of internal force changes are considered,and a total of5×5=25 working conditions are set.BP neural network and convolutional neural network(CNN)are used to establish monitoring models respectively,and 25 damage conditions are identified.The results show that BP neural network has over-fitting phenomenon,which leads to the accuracy of model test is only 70 %,while the accuracy of model test based on convolutional neural network is 100 %,which realizes the simultaneous monitoring of internal force change and crack propagation double damage in the node connection area.The superiority of convolutional neural network in structural health monitoring is further illustrated,which provides a new idea for the health monitoring of spatial structure. |
PDF Full Text Request