Electromechanical impedance(EMI)damage identification technology is widely used in structural health monitoring because of its high sensitivity to local damage.However,the change of ambient temperature will shift and change the amplitude of the impedance spectrum,and even cover up the damage information of the structure,resulting in misjudgment of damage identification.To eliminate the influence of temperature variation on damage identification,singular spectrum analysis(SSA)method is used to process the impedance signal to separate the signal components which are not affected by the temperature variation.An unsupervised machine learning method combining t-distribution stochastic neighbor embedding(t-SNE)and k-means clustering algo-rithm is proposed to further process the signal components to realize damage identification.In order to verify the feasibility of this method,an aluminum plate connected with the bolt group is taken as the experimental object to carry out damage identification experiment of the bolt loosening using EMI under changing temperature condi-tions.The experimental results show that the signal component processed by SSA method can effectively iden-tify bolt loosening under the influence of changing temperature,and the recognition accuracy of each working condition is more than 98％,which proves the effectiveness of the method in eliminating the influence of tem-perature variation.
维普
万方数据