Indirect dynamic load identification provides a practical means of estimating unknown external loads from measured structural responses when direct load measurement is difficult or infeasible. However, the inverse nature of the problem makes load reconstruction highly sensitive to measurement noise, modeling errors, and sensor configuration. Conventional regularization methods typically rely on predefined analytical priors, which may provide limited capability in representing the complex characteristics of dynamic loads. This study presents a plug-and-play (PnP) regularization framework for indirect dynamic load identification by incorporating data-driven priors into the inverse solution process. Instead of explicitly prescribing a regularization functional, the proposed framework employs a learned denoising operator as an implicit prior, enabling prior knowledge extracted from representative load data to be incorporated into load reconstruction while retaining the physics-based structural model. In addition, the influence of sensor configuration on identification performance is systematically investigated. A Fisher-information-based sensor placement strategy is developed to optimize the measurement locations by minimizing the uncertainty associated with the reconstructed loads, thereby improving the informativeness of structural response measurements.  The proposed methodology is evaluated through a series of benchmark studies, representative engineering scenarios, and a real-world structural application. The results demonstrate that the data-driven prior improves the robustness and accuracy of dynamic load reconstruction, particularly under noisy and ill-conditioned conditions, while optimized sensor placement further reduces estimation uncertainty and enhances identification reliability. These findings highlight the potential of combining physics-based inverse modeling, data-driven priors, and information-based sensor design for reliable indirect load measurement of engineering structures.

Indirect Load Identification;,Data-Driven Priors,Optimal Sensor Placement,Regularization