Computational Fluid Dynamics Modeling in Water Infrastructure: Best Practices

Computational fluid dynamics (CFD) is a branch of fluid mechanics that uses numerical analysis and data structures to calculate, analyze, and visualize fluid (liquids, gases, and dissolved gases) flows. This chapter provides general introductions to best practices for CFD modeling in water infrastructure for practitioners, particularly those new to CFD modeling, which is becoming a widely used tool in the design and retrofitting of water, wastewater, and stormwater infrastructure. The method serves as an alternative or complement to physical modeling. The full CFD modeling process is divided into four major groups: problem formulation, preprocessing, model simulations, and postprocessing. The chapter also presents some of the key concepts discussed in this book. The book covers problem formulation and selection of multiphysics models. It describes the steps in choosing the most adequate numerical model and turbulence schemes. The book presents meshing techniques and types and discusses initial and boundary conditions.

Problem formulation is the first major step when building and executing a CFD model. This step consists of formulation of goals and expectations, definition of the geometry, and understanding of process models and temporal scales involved. A detailed, yet succinct description of the key elements associated with problem formulation, model domain, selection of process models and CFD software needed when building a CFD is provided. During the problem formulation phase the CFD modeler should seek to clearly state the objective, domain geometry, operating conditions, dimensionality of the mesh, appropriate temporal modeling technique, simulation time span required to properly model the flow phenomena, number of fluid phases and constituents, and viscous nature of the flow. Physical, chemical, and biological processes may be involved alone or in combination in water related problems. The main subprocesses associated with each of such processes typically available in CFD modeling today are listed.

Meshing is the process of spatial discretization of model geometry. A successful CFD model requires a high-quality mesh to ensure a stable and numerically accurate model. This chapter addresses mesh quality, focusing on the wide variety of scales and flow regimes that are seen in civil and environmental engineering. Meshing is a trial-and-error process but can be guided by some basic rules of thumb. In terms of cell connectivity, a mesh can be classified into structured, unstructured (also called flexible), or hybrid mesh. A high-quality mesh is a strong driver of overall model accuracy and repeatability. The size of the mesh depends on the specific geometry, type of flow, and phenomena being modeled. The goal of the model can serve as a guide to the meshing process. Visualizing the general flow field prior to modeling can be of significant help.

Definition of initial and boundary conditions is part of the preprocessing step. To resolve the spatiotemporal variations of hydrodynamic variables such as pressure, velocity, and turbulence quantities within a modeled area, a solution of the governing equations of mass, momentum, and energy transport are required at the boundaries of the computational domain. These are special and problem-specific conditions commonly known as boundary conditions. Different types of boundary conditions can be implemented in a CFD model. In addition to the boundary conditions related to velocity and pressure, CFD models require definition of the turbulent quantities for the turbulence closure model. For civil and environmental engineering applications, cell zones can be utilized to model flow through porous media by adding a momentum sink term that creates a pressure drop that is proportional to the fluid velocity in a computational cell by introducing viscous and inertial resistance coefficients.

The Navier-Stokes equations are nonlinear, continuous partial differential equations. To obtain a practical solution, continuous equations are discretized in space and time, resulting in algebraic equations that are then solved using iterative methods. Of the three common methods, finite difference or finite element are generally limited to groundwater modeling or for simple geometries that allow for structured meshing. The finite volume method is most used for practical CFD work. Discretization results in a set of algebraic equations, which are solved iteratively using segregated and coupled solution methods. Simply put, a numerical scheme refers to the method the equations are set up to solve for the primitive variables within RANS equations. There are many types of numerical schemes; however, these schemes are used to treat the temporal differential or the spatial differential terms. Several considerations should be given when choosing the different numerical schemes. These considerations are stability, accuracy, convergence, and computational cost.

Most flows of interest in the water and wastewater industry are turbulent. The unsteady and irregular behavior of turbulent flows are their defining characteristic. Thus, the complete solution of the Navier-Stokes equations is inherently transient. However, most of the time the CFD engineer can capture sufficient information from a CFD analysis using mean quantities of the flow field. This chapter describes three turbulent resolving strategies for the N-S equations. Each strategy is introduced by presenting the assumptions and changes employed in the solution of the Navier-Stokes equations. The chapter briefly describes applications of RANS closure models. Recommendation for a generic turbulent closure model is difficult because it requires an understanding of the relevant flow features of the problem of interest. In the activated sludge model, RANS simulation is used to predict the distribution of water flow velocities, pressure, and volume fractions in the activated sludge tank.

Many factors contribute to the results of CFD. The simulation should be designed in a way such that the targeted observation is free of or has the minimum impact of numerical configuration, such as mesh and boundary condition. One of the most important factors in this consideration is the computational mesh. It will be demanded to eliminate the impact of grid resolution and configuration on the targeted physics. A common practice to detect the presence of such impact is called grid independence. In general, different resolutions and/or configurations of grids are used to repeat the same simulation. The key to determining if a CFD solution is grid independent is to use significantly different grid resolutions to evaluate the targeted observation. Wang et al. performed a series of CFD studies to investigate the penetration of starting buoyant jets. The computational domain and the boundary conditions are shown.

As the use of CFD models continues to raise to support the analysis of complex water infrastructure projects, they need to be adequately verified, calibrated, and validated; processes typically termed for simplicity as VV. Verification and validation are the primary means to assess accuracy and reliability in computational simulations. Such processes are key components of the best CFD modeling practices involving, among others, convergence and mesh independency tests, whereas they are antecedent to postprocessing visualization, documentation, and reporting steps. In CFD, the amount of uncertainty involved depends directly on the solution of the Navier-Stokes equations, which are technically parabolic nonlinear, nonhomogeneous, partial differential equations. Once model verification is carried out, the model needs to be calibrated. Calibration ensures that the matching between analytical or experimental values and model results is accurate within established standards via the iterative adjusting of model parameters.

The careful and deliberate documentation of modeling methods used, as well as the reporting of a holistic set of modeling results, is critical for the implementation of CFD in water-related projects. Practitioners run the risk of attempting to impress a crowd or reviewer with beautiful graphics while major underlying model errors go unnoticed, unreported, or yield flawed final designs. Technical reports are required to document all the information and data that were used to develop the model and identify findings of the CFD analysis. Practitioners need to highlight validation results to establish confidence in the model and provide explanation when validation is missing for key variables. It is imperative that practitioners do not oversell the accuracy of a CFD model. Although CFD is an incredible tool, there are limitless possibilities for unrepresentative results to be obtained.

Quality control is defined as a set of measures and procedures to follow to ensure the quality of a product. Quality control for CFD modeling projects provides procedures to minimize errors and consistently meet or exceed the requirements of a project. Reviewers need to have at least practice level experience performing CFD analyses and should be involved from beginning to end of the CFD project. Quality control procedures should be followed at each of the steps, from conceptualization and throughout implementation and execution, to avoid carrying errors going forward. A key to CFD modeling best practices is that quality control overarches the entire framework and is necessary to eliminate or at least reduce errors at each step of the modeling process. Utilizing quality control with CFD is essential in informing decisions and avoiding a potential impact of misguided design based on flawed results.

The use of computational fluid dynamics to assist in design and retrofitting of water infrastructure projects is continuously expanding. Nevertheless, it is recognized that no rigorous standards and guidelines have been developed for CFD applications in water projects. The proposed guidelines in this publication promote practices such as clear documentation of all efforts corresponding to problem formulation, as well as model development, preprocessing, model simulations, verification and validation, postprocessing, and presentation of results. Experienced and new practitioners are encouraged to apply these guidelines to best utilize CFD tools without overpromising capability or accuracy. Comprehensive quality control needs to accompany the effort to ensure that proper methods have been applied and results are representative of reality.