| The hysteretic energy dissipation performance is an important index for the evaluation of structural seismic performance and is the key to whether the building structure can maintain a safe state under the action of an earthquake.At present,cumulative energy dissipation and equivalent viscous damping coefficient are usually selected as the index to evaluate the energy dissipation performance of reinforced concrete columns(RC columns),and experimental methods and finite element simulation methods are used to predict the energy dissipation performance of the structure.However,experimental methods require a lot of time and cost,and the finite element simulation method has low efficiency for complex structure analysis.The linear stiffness structure is different from the structure applied in actual engineering,which will lead to a large deviation from the predicted results.Therefore,it is of great significance to propose a reasonable and effective method for the evaluation and prediction of the energy dissipation performance of RC columns to improve the seismic performance of RC columns and enhance the overall seismic and deformation capacity of buildings.Based on the mathematical method of measuring the degree of fullness of the measuring curve and the learning method of the artificial neural network,this paper makes a new attempt at this problem.First of all,212 sets of rectangular RC columns were selected from the PEER column performance database as the research samples of this paper.Based on the hysteretic loop area analysis,the existing energy dissipation index,the equivalent viscous damping coefficient,was proposed,and two new energy dissipation indices,perimeter coefficient,and inertia moment coefficient were proposed based on the mathematical problem of "how to measure the degree to which an irregular closed curve approaches a circle".MATLAB was used to implement the calculation and comparison analysis of the various energy dissipation indices,and finally,1444 sets of sample data were obtained.The results show that the perimeter coefficient,the inertia moment coefficient,and the equivalent viscous damping coefficient change with the maximum loading displacement,which are close to each other,and all of them can reflect the energy dissipation capacity of RC columns well,which proves the feasibility of using the perimeter coefficient and the inertia moment coefficient as the evaluation index of the energy dissipation performance of RC columns.Secondly,based on the 1444 sets of sample data,the influence of 14 RC column performance parameters on the three kinds of energy dissipation indices,the equivalent viscous damping coefficient,the perimeter coefficient,and the inertia moment coefficient,were explored by SPSS typical correlation analysis method.Taking the correlation coefficient greater than 0.2 and the significance less than 0.01 as the evaluation standard,the analysis shows that 12 column performance parameters,such as the volume of steel bar,axial pressure ratio,and angle longitudinal tension yield strength,have a higher correlation with the equivalent viscous damping coefficient;7column performance parameters,such as the span height ratio,the middle longitudinal tension yield strength,and the angle longitudinal tension yield strength,have a higher correlation with the perimeter coefficient;and 6 column performance parameters,such as the span height ratio,the middle longitudinal tension yield strength and the spacing of steel bars,have a higher correlation with the inertia moment coefficient.Thirdly,a method for predicting the energy dissipation performance of reinforced concrete columns based on neural networks is proposed.BP neural network,GRNN neural network,and INFO-GRNN neural network were used to construct the prediction model of equivalent viscous damping coefficient,and the average absolute error,the average relative error,and the coefficient of determination were selected as the evaluation indicators of the prediction accuracy of neural network.After comparing and analyzing the prediction results of the three neural networks,it was found that the average absolute error of the prediction model of the equivalent viscous damping coefficient was 0.027,the average relative error was 18.21%,and the coefficient of determination of the test set was 0.811.Compared with the prediction model of the BP neural network and GRNN neural network,the prediction effect was significantly improved and had better generalization ability and higher feasibility.Fourthly,based on the INFO-GRNN neural network,the prediction models of perimeter coefficient and inertia moment coefficient were established and combined with the prediction model of equivalent viscous damping coefficient,and the prediction results of the three energy dissipation indices were compared and analyzed.The analysis shows that the average absolute error of the prediction model of the perimeter coefficient is 0.030,the average relative error is 23.18%,and the coefficient of determination of the test set is 0.92;the average absolute error of the inertia moment coefficient prediction model is 0.028,the average relative error is 22.24%,and the coefficient of determination of the test set is 0.94.It shows that it is reasonable and feasible to predict the energy dissipation performance of RC columns by INFO-GRNN neural network,which has the advantages of fast training speed,high data processing ability,high computing efficiency,and good prediction effect,which makes up for the shortcomings of experimental methods and finite element analysis,such as time-consuming,laborious,high cost and low efficiency.Finally,the Garson sensitivity analysis method was used to analyze the contribution of the input vector to the prediction results of the prediction models of various energy dissipation indices,and the stability of the results under different conditions was verified,which verified the rationality of selecting the influencing factors of energy dissipation indices by typical correlation analysis. |
PDF Full Text Request