Recovery processes and seismic risk assessment represent a critical and challenging frontier in engineering risk analysis under uncertainty. Despite growing attention, the problem remains inherently complex, shaped by nonlinear system behaviours and high-dimensional stochastic spaces. These difficulties are compounded by the limited availability and often confidential nature of recovery data, highlighting the urgent need for modelling approaches that are not only efficient, but also flexible enough to adapt to real-world constraints.

In this work, we introduce a novel framework that explicitly integrates recovery into state-dependent seismic risk assessment. The approach combines fragility modelling, recovery processes, and hazard evaluation into a cohesive structure, enabling holistic and reliable risk analysis. Designed for flexibility, the framework draws from the state-of-the-art in different disciplines, such as structural engineering, recovery modelling and probabilistic seismic modelling, and focuses on balancing adaptability and computational efficiency.

At the core of the methodology lies its rigorous treatment of uncertainty propagation. By employing polynomial chaos expansion (PCE) and bootstrap techniques, the framework delivers robust probabilistic estimates even in the presence of sparse or noisy data. This capability is demonstrated through an application to a full-scale industrial steel moment-resisting frame structure from the European SPIF project. The structure was tested under realistic seismic conditions by means of the shake tables at EUCENTRE in Pavia. This case study demonstrates the framework's ability to enhance seismic risk assessments by seamlessly integrating recovery processes, while preserving both accuracy and computational efficiency.