Study On Optimal Sensor Placement Methods For Damage Detection Of Reticulated Shell

Large span space structure is widely used in recent decades with many advantages,such as rich and beautiful shape,big span and light weight.However,the structural formation and force conditions of large span space structure is complicated,the material performance degradation will happen during the period of service coupled with the effect of the wind, snow, rain, thunder and lightning and so on.Structure will inevitably produce effects of fatigue,material aging,which will create a threat to the safety of the structure.So it is necessary to take measures for the health monitoring of the important buildings and real-time monitor the health status of the buildings.The primary problems of the real-time monitor the health status of the buildings is that we should calculate the optimal placement of sensor.On the one hand,considering the economic reasons,the acceleration sensor unit is expensive for large-span spatial structure,on the other hand,it has the characteristics of a great amount of bar and node number, degree of freedom and the low accuracy of the finite element model which has serious modal localization and transition phenomenon.So how to make a limited number of sensors in reasonable positions should be related to the area of research in order to improve the structure of the signal collection efficiency and the damage identification accuracy.The impact loading is studied based on the damaged element of a K6 single-layer spherical reticulated shell by particle swarm optimization algorithm to optimize the acceleration sensor location and damage identification of bar.We can use the BP neural network by the L-M algorithm to acquire the acceleration response information based on particle swarm algorithm,so we can get the test condition for identification of damage.At the same time, the optimal placement of the sensor method based on particle swarm optimization algorithm will be studied.The main research work is as follows:( 1) The commonly used sensors technology at home and abroad were summarized as well as the damage identification methods, and the paper introduces their respective advantages and disadvantages.(2)The paper in detail introduces the basic principle and implementation process respectively based on the particle swarm optimization algorithm and LevenbergMarquardt algorithm.We can know that the particle swarm algorithm has the characteristics of simple operation,fast convergence speed,the needlessness to adjust the number of parameters and the high operation efficiency;The L-M algorithm has the characteristics of the high training speed and the hardness to fall into local minimum.On the other hand,the implement process of the net shell structure based on the particle swarm optimization algorithm to ensure the optimal placement of the sensor was introduced,and the realization method of damage identification methods based on L-M algorithm was introduced.(3)Using a scale ratio of 1:10 the damaged element of a K6 single-layer spherical reticulated shell as the experimental model,and we can build the damage of 9 kinds of experiment conditions to achieve the scheme of the optimal placement of the acceleration sensor based on particle swarm optimization algorithm,then we can get the optimal placement of the acceleration sensor based on particle swarm algorithm using the single-point multi-point response sampling method by the impact loading test,using the results of the analysis,then we can get the acceleration response value of each measuring point in different damage conditions,and we can test damage identification by L-M algorithm and research the optimal placement of the sensor method based on particle swarm optimization algorithm.The specific work includes:① Making a comparison between the test model and the finite element model of the top 20 order frequency,we can find that the error is within 4%,and the modal analysis model of the first 4 order formation in line confirms the net shell structure vibration characteristics which indacates that the use of net shell test scale model has certain reliability.②According to the modal vibration of different order and the different sensors number,we can achieve different optimal placement of the sensor position based on particle swarm algorithm,but the result has the symmetry and the inheritance,and this is consistent with symmetry of net shell structure itself and the theoretical logic of optimal layout of measuring points;When the formation order in calculation are different and the number of sensors are the same,the fitness function achieves the early convergence and stay until the maximum number of iterations.When the formation order in calculation are the same and the number of sensors is not the same,the fitness value will decrease with the increase of the number of sensors,and this is in accordance with the general rule when a sensor is arranged,namely,the more the number of sensors,the more comprehensive structural response information will be got.Based on the above analysis results,the method of optimal placement of the sensor based on particle swarm algorithm is verified the rationality,convergence and efficiency.③Considering the 6 acceleration sensor of the first 16 order formation,we can get the optimal placement of the sensor location based on particle swarm optimization algorithm,the L-M algorithm is used to identify the damage of the neural network,and the working condition of the damage can be successfully identified;Consider three different kinds of sensor placement solutions,including: different sensor placement position,different sensor number and maximum stress position,damage identify of the neural network based on L-M algorithm,but the test results are not completely identified successfully,which can further verify the optimal placement of the sensor method based on particle swarm algorithm’s feasibility and validity.