Damage Identification And Test For Detection Of Space Frame Structures Based On Neural Network

Space frame structure as one of the basic form of steel structures is widely used inmodern engineering. In the use process, the structure will inevitably produce a minor injury.With the gradually developed of damage, the use secure of structure will be threatened.Therefore, the detection methods of structural damage need to be established in order toquickly and accurately identify injury. At present, studies of using vibration method fordamage identification and location of steel structures mainly concentrated in the single-spanbeam and truss structure. Study on space the frame is less, and more reasonable the labeledamount of damage on identifying space damage is not systematically proposed. Therefore,the space frame damage detection needs further study.The subject manufactures a small experimental frame model, adopts DASP2006software to conduct experimental modal analysis and gets the modal parameters of naturalfrequencies, mode shapes, etc. This topic creates a finite element model of the experimentalframe using ANSYS software, and gets the simulation modal parameters of naturalfrequencies, mode shapes, etc by calculating. The correctness of the finite element model isverified by comparing.In this paper, vibration change rate, curvature mode and unit modal strain energychange ratio are used as the damage labeled amount to compare and study the damagelocation and identification for the damage degree of finite element model. This article drawsthat the axial vibration information of component is more suitable for the identification ofstructural damage than the lateral vibration of the component by comparing the change rateof vibration and curvature mode itself, and gets that the sensitivity of curvature mode and theunit modal strain energy change ratio for structural damage are superior to the vibrationchange rate and be able to effectively identify multi-location injury. In the actual project,Vibration-type measurement is relatively simple, easy to construct the parameters of modalchange rate and curvature mode, etc, and easy to achieve in actual damage detection. Butthere are shortcomings of larger data volume on space problem. Strain energy belongs to thelabeled amount of energy areas, does not directional and is highly sensitive, and is suitable asthe labeled amount of the space frame damage identification. However, the requirements ofmeasuring instruments is high and measurement cost is high, which make strain energy isnot suitable to promote in the actual project. According to comparative analysis, this textproposes a new damage labeled amount which is axial vibration mode differential changerate. Axial vibration mode differential change rate has advantages of avoiding space problems, small amount of work in processing data, high sensitivity, lower equipmentrequirements and cost, and is suitable for promotion in the actual project.The damage identification of experimental frame uses curvature modal and axialvibration mode differential change rate as the damage labeled amount. Experimental resultsshow that the damage labeled amount which adopt curvature mode and axial vibrationdifferential change rate can identify a single damage and multiple damage workingconditions of the experimental frame, and can promote in field measurement. And curvaturemode and axial vibration differential change rate is two more ideal damage labeled amountof space rigid frame.Finally, this article establishes neural network model of the space frame damageidentification, uses axial vibration mode differential change rate as the training sampleparameters to train the network, and achieve the effect of identifying damage location anddamage degree of space rigid frame, which provide a viable program for Structural damageidentification of Space rigid frame.