PSI - Issue 44

Gianluca Quinci et al. / Procedia Structural Integrity 44 (2023) 251–258 Gianluca Quinci et al. / Structural Integrity Procedia 00 (2022) 000–000

255

5

Fig. 2. Mock-up photographs at the Eucentre laboratory.

In total four tanks were installed, two vertical tanks on the first level and two horizontal tanks on the second level. Furthermore, one electrical cabinet was placed on the first level. The nominal pipe diameters for the pipes connecting pipes of the tanks are DN 100. The suspended pipes on the third story will be installed with DN 80. Moreover, bolted flange joints BFJs were necessary to connect several parts of the pipe. More precisely, the loose flanges adjacent to the tanks were conceived as moving flanges, while the others were simple flanges welded to pipes. A Low-Fidelity (LF) Finite Element Model (FEM) was developed in Midas Gen software in order to implement the numerical analysis with a reasonable computational effort and predict the seismic response of both primary structure and secondary element. In particular, the max acceleration value recorded on the vertical tank at the first floor is carried out and used as input for the ANN-based surrogate model training. Both modal and non-linear time history analyses were run, using the accelerograms selected in the previous section for the latter. The FEM described below, was validated by comparing the numerical results with experimental results in terms of modal behavior and seismic response. The primary structure is composed of steel columns and beams: these elements have been modeled as beam elements, as shown in Figure 3. Columns are hinged at the base and the connections with the primary beams allow all the stresses to be transferred. The simply supported secondary beams sustain tanks as well as the other live loads. No floor diaphragm constraints are inserted. Tanks are modeled with mass-concentrate nodes and are fixed to the primary structure by rigid link. The piping system is mainly made of straight segments, elbows, and tee joints. The straight segments of pipes are modeled by linear beam elements, while elbows are modeled using equivalent diagonal beam elements, as suggested in Bursi et al. (2015). Pipes have a hollow section with a diameter of 114.3 mm and a thickness of 3.6 mm, as well as elbows and tees joints. The pipes boundary conditions were chosen considering the most common industrial pipes connection used in the industrial plant. At the ground level, the pipes are all fixed to the base just as the rest of the structure. Meanwhile, the connection between the pipes and the supporting structure is achieved with hinges that allow all the relative rotations and one relative movement in the longitudinal direction of the pipe, to catch the real behavior of the boundaries. The FEM described above was used to carry out the seismic response of the vertical tank located at the first floor. To be consistent with the limit states thresholds defined in Section 2, a correlation between the strain values in the vertical tanks and the peak acceleration on the wall of the tanks is carried out from the experimental result of the test campaign, due to the impossibility to carry out the tank strain value from the LF model. The acceleration threshold values correlated to the limit states considered are summarized in Table 2.