Search

In order to comprehensively evaluate the working state and safety performance of the bridge plate rubber bearing (PRB), the basic principle and method of SVM in machine learning theory (Support Vector Machine) was introduced in this paper. Firstly, the performance indices (such as the average compression stress, tangent value of the shear angle and the rotation angle of the PRB) were analyzed and calculated through its finite element model considering the different working conditions (axial compression, shear deformation and the rotation deformation of the PRB). Subsequently, considering the relevant indices affecting the safety performance of the PRB, the Pearson Correlation Coefficients (PCC) were calculated by SVM method for assessing the internal correlation of the performance indices of the PRB for different cases, respectively. The threshold values of the influential factors for different cases considering the PCC and the weight coefficient were obtained for grading the working state of the PRB as five levels. Moreover, in order to verify the validity and suitability of the evaluation of the working state of the plate rubber bearing by the SVM method, three apparent indices for grading the performance level of the PRB such as the crack, the undistributed spalling deformation and the dislocation of the PRB were analyzed and compared with the qualitative descriptions and the quantitative values. Finally, the corresponding threshold values for grading the performance levels of the PRB working states were proposed. It was shown that the SVM based evaluation method for the PRB is feasible and convenient, especially for the bridges with real-time monitoring devices to get the response data of the PRB. [ABSTRACT FROM AUTHOR]