基于改进形状上下文算法的滞回曲线相似性评价

Aiming at the imperfection of indicators for evaluating the fitting degree between the hysteresis curve from experiments and the curve from numerical simulation, a hysteresis curve shape similarity indicator is proposed upon one shape context algorithm improved. The information representation of the shape context was simplified and a similarity algorithm for hysteresis loops is established. On this basis, the shape similarity algorithm for complete hysteresis curves is established by a projection-length-ratio-based vector similarity algorithm. The Bouc-Wen model, solved by fourth order Runge-Kutta method, was established for numerical simulation. First, the similarities of hysteresis loops, and the parameter errors of minimum circumscribed parallelograms were examined under a set of parameters. Then, the similarities of hysteresis curves, and the errors of peak points, energy dissipation and stiffness were analyzed under the variation of six parameters. The analysis results show that the shape similarities of hysteresis loops reflect the dissimilarity in shape, angle, and area. The shape similarities of hysteresis curves reflect their variation. Furthermore, they are negatively correlated with average relative errors of the force acting at the peak point, of the displacement at the peak point, of the energy dissipation area, and of the secant stiffness. Therefore, the rationality of the method was verified. Key words.