Research On Damage Identification Of Lattice Structure Based On Support Vector Machine And Moving Window Least Square Method

The spatial lattice structure has been widely used in large-span structures.However,due to the less resistance redundancy of the structure,it is easy to be partially unstable,resulting in frequent accidents.Effective monitoring of it in time to identify its damage has become an urgent engineering problem.Therefore,this paper firstly designs and made a full-scale 2×20 square pyramid lattice structure and thirteen kinds of damage units as the research object.The dynamic test and data acquisition of the lattice structure are carried out by impact load,and repeat sampling for each condition,to create a library of data samples suitable for machine learning methods.Then,based on the moving window least squares fitting damage identification method(MWLSQ),the related theory of it about damage identification in spatial lattice structure is deduced and improved.The experimental data of truss structure and full-scale large-span spatial grid structure are used to study the applicability and effectiveness of this method in large-span spatial grid structures.Then,a damage identification method which is characterized by autoregressive model parameters(AR parameters),and classified by support vector machine learning(SVM)is established.The data sample library composed of acceleration signals is used to train and test support vector machine.The training and testing are carried out to verify the correctness and superiority of the method.At the same time,the parameters optimization of SVM and its identification of different damage locations,damage levels,excitation sources and noise conditions,and AR parameter selection are studied and analyzed.The analysis results show that,on the one hand,the MWLSQ method has better effect on damage identification with larger damage and near mid-span position,and the recognition effect on the side span and small damage is general.On the other hand,the learning accuracy of the support vector machine in damage identification is extremely high(more than 98%),the recognition accuracy obtained by the corresponding parameter research is also extremely high.,and the optimization algorithm can continue to improve the recognition accuracy.In short,through the more complete research on the damage identification of spatial lattice structure,the damage identification method based on support vector machine learning which is using AR algorithm as the feature can effectively make the identify of spatial lattice structure's damage very accurate.