Abstract:The calibration of multi-parameter constitutive models is both the key and the challenge for accurately characterizing the complex stress-strain relationships of geotechnical materials. To address the issues of numerous parameters and difficult calibration in complex constitutive models, this study improved the dung beetle optimization algorithm by integrating dynamic range adjustment mechanisms, T-distribution variation strategies, Levy flight global search, and multi-strategy population update mechanisms. By optimizing typical benchmark functions and comparing the optimization results with those of the dung beetle optimization (DBO), gray wolf optimization (GWO), sparrow search algorithm (SSA), whale optimization algorithm (WOA), and northern goshawk optimization (NGO), the improved dung beetle optimization algorithm was verified to achieve significant improvements in both global optimization capability and parameter calibration efficiency. Based on a three-dimensional multi-mechanism bounding surface model, a bidirectional data interaction mechanism program between the optimization algorithm and the constitutive model was established. Parameter calibration was performed using triaxial test data of sandy soil. The comparison of the parameter calibration results between the improved dung beetle algorithm and the traditional dung beetle algorithm demonstrated that the optimal fitness value of the improved algorithm was significantly reduced, and the average invalid operation rate decreased from 8.75% to 4.17%, effectively overcoming premature convergence. The stress-strain curve obtained after parameter calibration showed good agreement with the experimental data in key aspects such as peak deviatoric stress, residual stress, and overall trends, providing methodological support for the multi-parameter calibration of complex constitutive models.
