PSI - Issue 60

Balaji Srinivasan et al. / Procedia Structural Integrity 60 (2024) 418–432 Balaji Srinivasan et al./ StructuralIntegrity Procedia 00 (2019) 000 – 000

419

2

1. Introduction In the realm of pressure vessel design, engineers frequently confront the difficult task of assessing the interface loads at the intersection of the pressure vessel equipment and the linked piping system. It is essential to accurately measure these interface stresses since they significantly influence the pressure vessel's overall structural integrity and performance. When the pressure vessel is in operation, various mechanical and thermal forces come into play, transferring loads and stresses to different parts of the vessel. These loads can potentially cause significant strains in the pressure vessel's shell and the nozzle's neck, which links to the piping system. Insufficient nozzle-shell design or failure of nozzles can result in hazardous situations, environmental problems, and financial losses due to unexpected downtime and maintenance costs. Engineers either take external nozzle loads from allowable standard nozzle loads [Koves et al. 2011] or extract them from the reaction summary obtained during the detailed piping analysis for designing nozzles. ASME Section VIII (2021) and various other publications describe the procedure to design and evaluate nozzle-shell junctions for external loads. The ASME STP-PT-074 (2015) document "Local Stress in Nozzles, Shells and Formed Heads from External Loads", provides the guidelines to determine the local stresses in nozzles and shells. In this paper, a comparative study is undertaken to study the Machine Learning (ML) models’ (Zahi M. et al. 2022 and NozzlePro 2018) applicability for the detailed dimensions and Finite Element Analysis (FEA) result data of wide range of nozzles (branch) – shell (header) configurations of STP-PT-074. The ML predictions are presented and contrasted with the FEA results. The paper focuses on exploring dimensional limitations in the Dm/T ratios (7 2500), dm/Dm ratios (0-0.7), and t/T ratios (0.1-10). Interestingly, this study demonstrates that it is possible and has been explored in literature to use machine learning models to predict local stresses in nozzles, shells, and formed heads based on external loads.

Nomenclature D m

mean diameter of header at a point of nozzle attachment

d m

mean diameter of nozzle header/vessel/shell

H

STP

ASME Standards Technology Project (STP)-PT-074

T

shell thickness

t

nozzle wall thickness

B

branch/nozzle

WRC

welding research council

2. Existing Local Stress Evaluation Methods 2.1. Manual Analytical methods and limitations

Evaluation of localized external stresses on pressure vessel nozzles may be done using a variety of techniques. The first approach uses manual analytical techniques including PD 5500 Appendix G (2021), EN 13445-3 clauses 16.4 and 16.5 (2021), WRC 107 (1965), WRC 297 (1984), and WRC 497 (2004) as well as FEA. However, manual analytical methods are known to be excessively conservative. WRC-368, entitled "Stresses in Intersecting Cylinders Subjected to Pressure" was released in 1991. WRC 368 provides an approximate method of calculating the maximum stress intensities due to internal pressure at cylinder nozzle intersections. However, the WRC techniques (WRC 107, WRC 297, and WRC 497) have limitations, such as diameter-to-thickness ratios. WRC 107 only calculates stresses in the shell, not the nozzle neck, which may be crucial for relatively thin neck thicknesses. WRC 297 is suitable for pressurized nozzles in shells and helps calculate stresses due to external loads. WRC 497 has its own set of limitations and should not be used beyond its applicable range.