Multi-objective inverse identification of meso-scale parameters of bonded particle model of frozen particle systems

Calibrating bond parameters in Bonded-Particle Model (BPM) simulations remains a fundamental challenge due to the mesoscale nature of these parameters. Neither macroscopic mechanical tests nor microscopic characterization techniques can directly measure their properties. Furthermore, the macroscopic response is governed by strongly coupled parameters, leading to non-unique inverse solutions. This study introduces a hybrid inverse calibration framework with Tikhonov regularization based on a scalarized multi-objective formulation that integrates global parameter-space exploration to determine bond parameters. The methodology simultaneously considers uniaxial compression, three-point bending, and shear tests, ensuring consistency across loading modes. The framework is demonstrated on saturated spherical glass-bead specimens as a controlled model frozen particle system, and is applied across multiple temperatures and two representative loading rates, identifying condition-specific bond-parameter sets whose trends are consistent with the physical behaviour of ice bonds.