A significant challenge in health monitoring and prediction is the accurate assessment of the bearing capacity of existing engineering structures. To address the challenge of accurately assessing the stress state of engineering structures, a mechanical twin inversion method was developed based on the digital twin model. This approach employs the high-fidelity geometric twin model and the apparent damage twin model as inputs for force analysis, utilizing the precise spatial continuous geometric form and spatial continuous deformation mode as constraints. Following experimental research, the proposed framework has been demonstrated to have certain universality, thereby providing a reliable basis for the monitoring and prediction of the health of engineering structures.