Large Eddy Simulation - Based Mixing Tee Analysis and Comparison with in-Service Failure Data

The mixing process of hot and cold fluids in a tee junction is chaotic (turbulent) in nature and can result in high cycle thermal fatigue of the junction.  This random quasi steady state phenomenon of hot and cold shocks can lead to fatigue cracks and possible leakage, causing a structural integrity and process safety concern.  Standard Computational Fluid Dynamics (CFD) two parameter turbulence models (Reynolds Averaged Navier Stokes Models) cannot predict this phenomenon, and advanced analysis, specifically Large Eddy Simulation (LES) is required to obtain meaningful results.  This paper presents a comprehensive analysis of a petrochemical industry mixing point, including LES, a standard two parameter CFD turbulence model, finite element stress analysis, fatigue and fatigue crack growth approaches.  This investigation is driven by extensive analysis of operational data, and is rigorously benchmarked against the observed in-service failure of the mixing tee.

The tee junction under consideration introduces cold hydrogen at variable flow rates into the primary hot hydrocarbon stream.  After approximately 15 years of service, multiple cracks were identified downstream of the mixing point on the inside surface of the run pipe.  In order to predict the observed cracking, a 3D model of the tee junction is created and discretized using a commercial CFD software, STAR-CCM+. The model contains nearly half a million polyhedral cells.  Temperature fluctuations at the pipe wall due to turbulent thermal mixing at steady average flow conditions are assessed by resolving large eddies using Large Eddy Simulations.   Temperature fluctuations caused by variations in the operating conditions (global variations) are assessed using a realizable k-ε turbulence model, a conventional two equation Reynolds Averaged Navier Stokes (RANS) turbulence model in which eddies are in general not resolved.  The temperature histories from the LES are post processed to determine the severity of temperature fluctuations at the recorded locations. Post processing of the temperature histories reveals that the temperature fluctuations of 100^oF at the frequency of 5 - 10Hz are typical at steady average flow conditions.

Stresses due to these thermal fluctuations are estimated using the commercial Finite Element Software, ABAQUS.  The number of cycles to initiate a crack is determined using ASME design fatigue curves for stainless steel. The propagation of the initiated crack is modeled using Paris law as per ASME FFS-1/API 579.  For crack propagation, through wall stresses due to turbulent thermal mixing at constant average flow conditions and global variations are considered to predict typical crack sizes observed in the field. The methodology and results of this investigation can be used to assess the design and safety of both new and current mixing tee systems.

Keywords: T-junction, turbulent thermal mixing, Large Eddy Simulations (LES), Fatigue Crack Growth, Computational Fluid Dynamics (CFD)