Research On Damage Detection Method For Span Spatial Lattice Structure

Large span space structure with reasonable force, high stiffness, light weight, lowcost, rich and novel structure type, lively and vivid, which can highlight the beauty ofstructure and rich artistic performance, is widely used in the stadium, exhibition hall,restaurant, garage, waiting hall and other types of building. These large and complexstructures are built with a large number of new materials, new technology, newtechnology. In the process of using, with the passage of time, the aging of materials,environmental erosion, uneven settlement of foundation and complex factors such asload and the coupling effect between the structure system will inevitably lead to theaccumulation of damage and the decay of resistance, which will lead to catastrophicemergency in extreme cases and caused huge losses of people’s lives and property.Therefore, there is important significance for timely and accurately distinguishing,accessing and repairing the damage of structure. This paper uses the time seriesanalysis and neural networks to distinguish the damage of the large span spacestructure, which has accuracy and robust. The paper mainly includes the followingcontents:(1).According to the present situation about the research on damage identificationof structures at home and abroad, summarize the damage identification approach basedon structural vibration response, and point out their advantages, disadvantages andapplication scope(2).Using optimal sensor placement method based on particle swarm optimizationalgorithm to determine the location of the acceleration sensor. Calculate the structuralvibration mode, select the appropriate fitness function, and according to the fitnessvalue to determine the location of the acceleration sensor. The method has theadvantages of good operability, simple and easy to realize and no needing to adjust theparameters. Numerical simulation results show that this method is stable and accurate.(3).Establish the time series model based on the acceleration response information.Make the difference between the coefficients of time series model as the damageidentification parameters. Overcome the disadvantages of low recognition accuracy,the incomplete test information by using modal parameters for damage identification.To further improve the identification accuracy.(4).Using the span space grid structure damage identification method based on time series analysis and neural network to detect the damage and the location ofdamage. Firstly, based on the MATLAB system identification toolbox, use the timeseries analysis method to identify the damage; and then, based on MATLAB neuralnetwork toolbox, the neural network method is used to identify the structure damagelocation. The method has the following advantages:①.Don’t need to identify the modal parameters, use the acceleration responsesignal dircetly, and without excitation;②. Avoid the dependence on structure model in the process of structure damageidentification;③.Time series analysis model is established without considering how muchsystem input, system with external contact, system complexity and other factors. Thereis the advantage of high recognition accuracy, sensitive to small damage, lowerenvironmental requirements, strong operability and so on.④.Neural network has the ability to filter out the noise and draw the rightconclusions in the noisy case;⑤.Neural network has the feature of parallel, adaptive, associative memoryfunction, robustness and fault tolerance.(5).The damage identification of single damaged component and two damagedcomponents are simulated with a single Kiewitt latticed shell structure，the white noiseand ambient excitation are used to motivate the structure respectively, and theinfluence of3%measurement noise is considered. The numerical simulation resultsshow that,①.The span space grid structure damage identification method based ontime series analysis and neural networks can identify the damage location and hascertain ability to resist noise;②.The number and location of the acceleration sensorand the number of neurons in hidden layer of BP network have great influence on therecognition accuracy. Use the intelliencent particle swarm Optimization algorithm tofix up the acceleration sensor in symmetrical areas of the structure. According to thetest algorithm, determine the number of neurons in hidden layer of BP network.Damage identification results show that: the method is feasible and effective.