Computational fluid dynamics (CFD) is a powerful analytical tool for design and engineering teams. Many manufacturing and operations teams leverage the power of CFD to experiment with equipment and process modifications in a simulated environment to validate their capital investments prior to purchase.
CFD complements theoretical reasoning and physical experimentation; together, these synergistic approaches drive the philosophical study of fluid mechanics. Think of it like x-ray vision. When project teams face engineering problems involving fluid flows that they can’t solve through theory or experimentation alone, they can turn to CFD to identify, visualize, and interrogate a solution in three-dimensional space and time.
Like all fluid dynamics, CFD is governed by the fundamental conservation laws of physics—that is, the conservation of mass, momentum, and energy. With these three equations alone, it’s possible to describe all problems of fluid dynamics. That level of versatility is a beautiful thing for CFD teams, and it’s a pathway to faster, more reliable outcomes for project owners.
The basic CFD process involves three main steps:
Step 1. Pre-processing: The CFD team defines the engineering problem, builds a computer-aided design (CAD) model depicting the geometric properties of the physical domain or area of interest, then carefully discretizes the domain into a computational mesh. This step also involves selecting the appropriate physics to be modeled and applies the correct boundary conditions.
Step 2. Processing: Using the appropriate CFD code, the team solves the governing equations of mass, momentum, and energy for everywhere in the discretized CAD model’s domain. This is when the team chooses to use a steady state or transient solver (see the definition below) and establishes important convergence criteria.
Step 3. Post-processing: The team extracts and analyzes data from the simulation, using it to support detailed design assessments and conduct model verification and validation.
When executed well, this process—and the insights it yields—can significantly improve speed, quality, and cost control for manufacturers across different industries, from the life sciences to food and beverage manufacturers.
Whatever your own area of focus, this article will help you and your team unlock those benefits by exploring CFD from three angles:
Computational fluid dynamics (CFD) is a powerful analytical tool which enables project teams to identify, visualize, and interrogate solutions to engineering problems before putting real-world resources at stake.
Under the governance of the conservation laws of physics (that is, the conservation of mass, momentum, and energy), CFD can describe all problems of fluid dynamics in three-dimensional space and time. This makes it an especially effective approach in the philosophical study of fluid mechanics, alongside theoretical studies and physical experimentation.
As the name suggests, steady-state CFD analysis is suitable for scenarios that are not contingent on a specific interval of time, such as overall airflow dynamics in a room. Transient CFD analysis, on the other hand, generates solutions in scenarios that are time-dependent, such as the impact of incrementally escalating the agitation rate inside a bioreactor.
This is the digital environment in which team members will eventually carry out their CFD solution.
Understanding the density, pressure, and velocity of a fluid is key to running accurate “what if” simulations in the CFD model. The flow field is a multi-dimensional vector and scalar field within the computational domain that describes the physics of a fluid flow in space and time.
Physical boundaries like a wall or an outlet directly impact the behavior of fluids in space and time. To get an accurate picture of that impact, the CFD team needs to identify and define appropriate boundary conditions inside the computational model.
During the pre-processing phase of a CFD study, the team discretizes their computational model into cells, forming a mesh of many control volumes.
The density of this mesh impacts the granularity of detail it provides and, as a consequence, the processing time required to run a simulation. That’s where a meshing strategy (sometimes called a discretization strategy) comes into play. In order to generate useful data that accurately resolves important flow characteristics, the CFD team must determine where to discretize further (for example, in high-gradient regions around an inlet or near a wall) and where to allow for a coarser mesh so as to reduce computational expense while still preserving important flow features.
Before moving into the processing phase, the CFD team numerically verifies and validates their computational model.
Validation is about ensuring that the team is solving the correct problem; verification is about ensuring that they’re solving the problem correctly. The ultimate goal of these steps is to establish that the model’s behavior is consistent with real-world results.
When it comes to commercial manufacturing facilities and operations, CFD is less costly than real-world experimentation and it provides a detailed, three-dimensional view of complex industrial environments. These strengths alone make CFD a valuable engineering tool, but its greatest advantage may be its versatility.
CFD is applicable across a wide variety of specific scenarios, including:
In each of these scenarios, engineering teams can leverage CFD to uncover solutions to:
To find these solutions, engineering teams have two broad approaches to CFD at their disposal: steady-state analysis and transient analysis.
The key difference between a steady-state analysis and a transient analysis approach is time. Take a cup of coffee as a simple example. A steady-state simulation will tell you how much cream to add in order to have your “ideal” cup. A transient simulation will tell you how vigorously to stir that cream into the coffee, and for how long. The first simulation provides a solution invariant of time; the second provides a solution that is time-dependent. Both have different and important use cases for manufacturers across industries.
Steady-state CFD advantages: Requires less processing power and is therefore capable of converging to a solution much more quickly.
Transient CFD advantages: Can yield simulations with a higher resolution and interrogate dynamic scenarios as they change over time.
Steady-state CFD considerations: Not appropriate in scenarios that involve studying variance over time.
Transient CFD considerations: These simulations run several iterations per interval (seconds, minutes, etc.), which means they require significant time and computational resources.
Steady-state CFD sample scenarios:
Transient CFD sample scenarios:
Sometimes, a steady-state scenario leads to a transient one.
For example, our CFD team recently ran a steady-state analysis for a biopharma client in order to understand the risk of contamination from their material airlock (MAL) room to their cleanroom. The analysis revealed a potential air-recirculation zone between the two spaces when the doors are open. This is bad for air quality since the vortex of air can transport particles from the dirtier space into the cleaner space. To understand the scope of that risk and test potential solutions in sufficient detail, our team suggested a transient analysis where the time evolution of the flow-field is resolved.
Steady state analysis: The simulation revealed a zone of recirculation near the door between the two spaces.
Transient analysis: By artificially introducing a pollutant in the simulation and watching it advect with the flow field over time, we were able to see the impact of that recirculation zone and show that relocating a diffuser in the MAL room solved the problem.
CFD teams frequently collaborate with architecture, engineering, and construction experts to determine which of these CFD scenarios is best suited to answer a broad range of mission-critical questions. For example:
Q: Are there any vulnerabilities in our existing process that might impact product quality?
Recently, a pet food client worked with CRB’s CFD team to understand and address a lack of uniformity among products emerging from enclosed steam cooker ovens.
We built a three-dimensional model of those ovens and ran a series of steady state CFD simulations, giving the project team a view inside the ovens without the expense or time required to physically take them apart.
Our study resulted in a few simple modifications that had the effect of inducing greater turbulent kinetic energy, thereby convecting more heat towards the conveyor belt and the food it transported. These adjustments cost relatively little but helped restore aging equipment to a high level of performance.
Q: What are the safety implications of our proposed building design?
While preparing to upgrade outdoor diesel generators at an existing facility, a confidential pharma manufacturer asked our CFD team to assess two risk factors:
We used a steady-state plume dispersion CFD model coupled with an atmospheric boundary layer model to simulate the behavior of generator exhaust for various different exhaust stack heights. This gave us the data necessary to visualize the generator’s predicted gas and temperature dispersion under prevailing wind conditions, which led us to the precise height and equipment combination that would protect the health and safety of workers inside the facility.
Q: It’s time to upgrade our HVAC. How can we ensure that our investment will pay off?
The global pharma company engaged our design and engineering team to help protect the temperature-sensitive raw materials and pharmaceutical products inside a 55,000-square-foot warehouse.
Our team designed a new HVAC system to replace the aging rooftop units. The client needed assurance that this new design was capable of maintaining warehouse temperatures within a narrow three-degree range. When their HVAC vendor suggested a simpler, less expensive design, they needed to know if the alternative design would perform reliably.
Our CFD team answered these questions. We built a high-fidelity model of the warehouse, including its racks, trucks, operators, and other boundaries, and we simulated both our proposed design and the vendor’s alternative design at peak summer and peak winter temperatures. We used these steady-state simulations to compare the temperature and airflow dynamics of each design. Using experimentally determined diffuser throw distances, we undertook a minor validation of the model to ensure that the inlet supply air evolved realistically as it spread into the space.
Our simulations qualitatively and quantitatively confirmed that the vendor’s alternative approach introduced a significant risk of temperature excursions while our design would maintain the target range. As a result, the pharma client was able to invest in its new system with confidence in its value and reliability.
Q: How can we improve design outcomes as we scale from the lab to commercial-volume production?
When project owners at a lab-scale biopharma client recently approached CRB with plans to expand, their most pressing challenge had to do with scaling their bioreactor equipment. They had been operating 2-liter benchtop reactors; how would a transition to 3,000-liter bioreactors impact factors like their process design and their material inputs?
To help them find answers while optimizing their scale-up strategy, we modeled the geometry of their current and proposed bioreactors and ran transient simulations of each one’s hydrodynamics. This allowed us to collect data and proactively develop an appropriate plan for scaling their operation in response to variables such as:
What you get out of a CFD study will only be as good as what you put into it. Poor or incomplete data will generate a poor or incomplete model, which will impact the quality of your results and limit your ability to apply those results to the real world.
This second “must have” element of a strong CFD project piggybacks on the first one. Your CFD team needs to understand the meaning and significance of the physical laws and equations that govern the conservation of mass, momentum, and energy, and they need experience in deploying that understanding to create an appropriately detailed computational domain—one that’s supported by the correct numerical models with consistent boundary conditions.
Teams that are new to CFD analysis may assume that sophisticated software will compensate for their lack of understanding or experience, but software alone isn’t capable of making the decisions that are required as part of a CFD study. The engineer must know how to select the correct boundary conditions and physical models, make the right limiting assumptions, generate a high-quality mesh that accurately resolves important flow characteristics, choose appropriate numerical algorithms, and on and on.
To get each of these decisions right, teams must rely on a deep well of knowledge and experience. This is the only way to understand what is behind the “black box” of the CFD solver.
CFD is not a tool to use in isolation. Its success depends on its synergistic relationship with the other two philosophical approaches of fluid mechanics, namely, theory and experimentation. A skillful CFD engineer looks for opportunities to use these three tools in combination, creating the best possible conditions for finding, analyzing, and applying accurate solutions to problems.
Theory
Experimentation and validation
Modeling and simulation
In this step, the CFD team defines the engineering problem and creates the computational model in which it will interrogate and solve that problem.
Define the problem, establish the modeling objectives, and gather data.
In this initial phase, CFD team members work in collaboration with the project owner to understand the problem or the analytical opportunity that will be the focus of their study. At this early stage, the team confirms that CFD is the appropriate technique, and they determine whether a steady-state or transient analysis is most suitable (see explanation above).
Team members then establish their modeling approach and its objectives. This is a period of important due diligence; the team conducts research, gathers necessary files, and collects the data they’ll need to build and refine their model.
In some cases, such as a renovation or a process optimization exercise, the project owner may have that data available. Occasionally, the CFD team conducts a site visit to gather physical measurements in the absence of a CAD model.
In the case of a greenfield project, the CFD team works closely with the project architects and engineers to develop assumptions and generate the necessary data based on available drawings and designs.
Build the model.
The CFD team translates all of this project data into a computational domain, called a CAD (computer-aided design) model. This is the digital environment in which team members will eventually carry out their solution. Sometimes this entails cleaning and refining an existing CAD model; in more complex or custom scenarios, the team may have to build a model from the ground up.
Defining the model’s boundary conditions is an important component of this work. The CFD team must identify any physical boundaries in the domain, such as a wall, a roof, a person, or an inlet or outlet, and then formulate those boundaries into data that’s propagated throughout the model’s geometry. This helps the team build a picture of velocity, pressure, and/or temperature gradients across the computational domain.
An example of a computational domain with boundary conditions, followed by a visualization of the temperature gradients within those conditions.
Discretize the model.
This is one of the most important steps in the CFD process. To get correct results from their simulations, CFD team members must start by establishing the correct level of discretization.
Discretizing (also called meshing) is the process of dividing the model into discrete cells or control volumes, which transform the computational domain into a mesh. Like the pixels in a TV screen, the number of cells in a mesh determines the level of detail or resolution available to the observer.
The mesh does not have to be uniformly discretized, though; areas that require more detail because of their flow features, such as high-gradient regions around an inlet or near a wall, are typically locally refined into a finer mesh. If done correctly, the mesh will provide the CFD team with enough resolution to run accurate and useful simulations, without adding more complexity than required and thereby extending the processing time unnecessarily.
In this close-up taken from a CFD model we created for a client in the pet food manufacturing industry, you can see the impeller equipment before and after discretization. Smaller cells yield greater detail but require more processing time.
If team members discretize the computational domain beyond what’s required to solve their problem, they’ll lose valuable time for a diminishing return. On the other hand, not discretizing the domain finely enough will impair their ability to validate their model against reality, which leaves them vulnerable to errors.
To strike a balance between accuracy, integrity, and computation time, CFD teams need a carefully formulated meshing strategy or adaptive mesh refinement:
In this step, the team runs the simulation, refines the model, and repeats the process.
Team members have defined the problem they’re going to solve. They’ve gathered all the necessary data. They’ve built and discretized their model. Now comes the time to run the CFD solver.
Success at this stage depends on team members’ experience and due diligence during the pre-processing stage. They need to have selected the correct CFD solver and solution algorithms based on the particular problem they’re studying. Choosing wrong could lead to errors that invalidate the results of a simulation; at best, it could mean spending more time processing the solution than necessary. Knowing how to navigate these options to achieve the best, most accurate results possible in the shortest amount of time is a function of experience, ability, and a deep understanding of CFD.
In this step, the team extracts data from the simulation.
The CFD team will repeat the pre-processing and processing cycle multiple times as they monitor and refine the quality of their model and move iteratively towards a solution. Numerical validation and/or verification of the model is an important component of this phase.
The goal of model validation is to ensure that the team is solving the correct problem. Experimental data and hand calculations provide a basis for assessing the model’s “truth,” and for comparing its predicted results with the behavior of flow variables in a real-world space.
Model verification has a slightly different goal: to ensure the team is solving the problem correctly. In this step, the team quantifies the degree of uncertainty in the simulation’s predicted results. This step is always crucial, but it becomes especially important in cases where validation isn’t possible because of a lack of experimental data.
Experienced CFD team members know how to enlist their modeling software as a tool in the validation and verification process. The top multiphysics simulation software brands offer pre-validated code for many common industrial problems, with the associated margin of error already quantified. For example, the software may offer a CFD model for solving a common HVAC problem with a 5% error rate; CFD team members using that model can reasonably expect a similar error rate when they run that solver on a different problem, as long as they do it correctly.
As their final step, the team extracts the appropriate results (temperature, velocity, thrust, etc.) from their model and uses them to create detailed visualizations of the flow field at any point in space or time (depending on whether they used a steady-state or transient analysis). These results support a thorough evaluation of the design problem at hand.
Ready to optimize your next capital project with a computational fluid dynamics study?
Reach out to our team of CFD experts and find answers to your most complex engineering questions.
A comprehensive overview on how modeling and simulation can be leveraged to improve operations and meet business targets
Read More
