Seismic risk evaluation of coupled systems of industrial plants often needs the implementation of complex finite element models to consider their multicomponent nature. These models typically rely on significant computational resources. Moreover, the relationships between seismic action, system response and relevant damage levels are often characterized by a high level of nonlinearity, thus requiring a solid background of experimental data. Furthermore, fragility analyses depend on the adoption of a significant number of seismic waveforms generally not available when the analysis is site-specific. To propose a methodology able to manage these issues, we present a possible approach for a seismic reliability analysis of a coupled tank-piping system. The novelty of this approach lies in the adoption of artificial accelerograms, FE models and experimental hybrid simulations to evaluate a surrogate meta-model of our system. First, to obtain the necessary input for a stochastic ground motion model able to generate synthetic ground motions, a disaggregation analysis of the seismic hazard is performed. Hereafter, we reduce the space of parameters of the stochastic ground motion model by means of a global sensitivity analysis upon the seismic response of our system. Hence, we generate a large set of synthetic ground motions and select, among them, a few signals for experimental hybrid simulations. In detail, the hybrid simulator is composed by a numerical substructure to predict the sliding response of a steel tank, and a physical substructure made of a realistic piping network. Furthermore, we use these experimental results to calibrate a refined ANSYS FEM. More precisely, we focus on tensile hoop strains in elbow pipes as a leading cause for leakage, monitoring them with strain gauges. Thus, we present the procedure to evaluate a numerical Kriging meta-model of the coupled system based on both experimental and finite element model results. This model will be adopted in a future development to carry out a seismic fragility analysis.