The Three Waves of Commercial CFD
In recent years, numerous papers on the history of fluid simulation have been published. Many early CFD pioneers (such as Brian Spalding, David Tatchell, Ferit Boysan, and Michael Engelman) have recalled the memorable past through interviews or articles. Various historical materials, technical information, and personal recollections consistently describe the development of engineering simulation software—from initial academic research codes to the modern CFD products we know today, with participation and support from multinational software companies on an industrial scale. The development of CFD is closely linked to the continuous improvement of computing hardware performance. In the early stages, it was mainly used in aerospace and defense research and development, and later gradually expanded to civil industries. Looking back, the development of CFD software in industrial applications can be divided into three major stages:
Hanna and Parry (2011) provided an excellent overview of this topic with a detailed bibliography. Runchal (2008) and Tatchell (2009) also published vivid first-hand accounts.
Figure 1: Fluid Simulation in the 1980s (Hanna & Parry, 2011)
First Wave: The Emergence of Commercial CFD Software
In the first stage, the codes written by CFD software engineers actually originated from the work carried out by the T-3 Fluid Dynamics Research Group at the Los Alamos National Laboratory (USA) since 1958, and the research activities led by Professor D. B. Spalding at Imperial College London (UK) in the 1960s-1970s.
In the late 1960s, CHAM (Concentration, Heat and Momentum Ltd.), founded by Professor D. B. Spalding and originally based at Imperial College London, began offering consulting services. In 1974, CHAM moved to new offices in New Malden near London, marking the start of the era of commercial CFD software. Initially, CHAM's core business was developing custom CFD codes for clients, but this work was found to be time-consuming, labor-intensive, and inefficient. Therefore, the company decided to develop a general-purpose CFD software package for internal consulting work and launched it as a commercial product in 1981, named PHOENICS. This marked the official birth of the CFD software industry (CHAM Ltd, 2008). Other companies quickly followed suit. Fluid Dynamics International (USA) launched the finite element-based CFD package FIDAP in 1982, while Creare.Inc (USA) released the finite volume method-based CFD code Fluent in 1983. In 1980, Dr. C. W. Hirt founded Flow Science (USA) through the asset restructuring of Los Alamos National Laboratory and released Flow-3D in 1985. More CFD software packages emerged subsequently, including Flow3D launched by the UK's Harwell Atomic Energy Research Centre in 1987 and TASCflow released by Canada's Applied Scientific Computing in 1989 (these two are now integrated into ANSYS CFX). Another professor from Imperial College London, David Gosman, co-founded Computational Dynamics/ADAPCO (UK/USA) and released StarCD in 1989.
In the early 1990s, Silicon Graphics, a workstation manufacturer, listed as many as 18 commercial CFD software packages compatible with its hardware products in its software catalog, competing for the then $30 million CFD market (Boysan et al., 2009). Most of the basic technologies behind these CFD packages originated from the research results or scientific publications of former employees or visiting scientists from the two aforementioned research institutions (London and Los Alamos). However, CFD technology also developed elsewhere: in the 1980s, the former Soviet Union adopted another CFD simulation method in its military and civil aerospace projects. Due to the world political situation at that time, this method largely went unnoticed by Western scientific communities. The technical tasks of CFD simulation in the former Soviet Union were very similar to those in the West, but their solutions had more limited computing resources. Nevertheless, due to the high political priority of these research projects, the former Soviet Union conducted a large number of experiments on fluid flow and heat transfer (especially in near-wall regions) and recorded rich experimental data. This enabled the vigorous development of alternative CFD methods. These methods were based on the well-known Cartesian grid method published in Western scientific publications, combining numerical, analytical, and empirical data. This innovative approach produced high-quality simulation results for almost any complex computational domain while maintaining low resource requirements and the efficiency of the Cartesian grid method. In the late 1980s, as the former Soviet Union gradually liberalized its economy, several scientist teams commercialized this CFD technology and sold their products and services in Europe and Asia in the early 1990s. The most well-known of these products include Aeroshape-3D launched by Professor V. N. Gavriliouk and his team (Petrowa, 1998 & Alyamovskiy, 2008), and FlowVision by Dr. A. A. Aksenov and his team (Aksenov et al., 2003).
Figure 2: Results from Aeroshape-3D (Parry et al., 2012)
By today's standards, the development of CFD software in the first stage was characterized by primitive user interfaces for data input, simple graphics, and very limited computing resources—especially insufficient memory, which severely restricted the expansion of model sizes. These limitations imposed high requirements on users in terms of geometric and physical modeling: any real-world scenario had to be analyzed, simplified, and then input into the software, a time-consuming and laborious process. Due to significant uncertainties in the selection and configuration of physical models and high risks of input errors, comprehensive evaluation and testing of simulation results became routine steps in the workflow. These operations required not only extensive numerical skills but also in-depth understanding of the basic principles and limitations of physical models, as well as their potential impact on CFD models. Therefore, users of CFD technology at that time were almost exclusively scientists or scientifically trained engineers, who had to partially or fully validate almost every simulation result through experimental research.
Another feature of this period was that due to limited experience in applying CFD simulation technology to industrial projects and increasingly fierce competition in the CFD software market, suppliers often overstated their products' capabilities in solving industrial problems. Such exaggerations, combined with early industrial users' mixed evaluations of CFD costs and result quality, earned CFD simulation a reputation of being "too slow, too expensive, too vague, and impractical." This negative reputation persisted in the general engineering community for over two decades but improved as CFD software development made significant progress in the current third stage, making CFD simulation a daily task for a new generation of users.
From the early 1990s onwards, CFD software and simulation conditions underwent earth-shaking changes. Significant progress was made in the performance of computer hardware, mathematical methods, and physical models. CPU speed and memory capacity grew rapidly while prices dropped, bringing industrial users newly accessible hardware resources such as UNIX workstations and workstation PCs. Later, with the emergence of affordable workstation clusters, high-performance computing (HPC) became more accessible. These new hardware capabilities naturally promoted software prosperity. Various numerical methods suitable for complex geometric calculations (such as unstructured finite volume methods, multigrid methods, and sliding grids) were optimized for HPC, and more flexible, reliable, and widely applicable physical models became available. Since then, CFD has entered new application fields, become more practical, and for the first time provided truly usable model sizes for actual industrial application scales. Hanna and Parry (2011) analyzed the development of this stage and found a direct correlation between Moore's Law on computing power and the size of CFD simulation models (e.g., models used in racing CFD simulations). These new capabilities heralded a new stage in the application of commercial CFD software—its full entry into the product R&D departments of major enterprises.
Figure 3: CFD Development Relies on Hardware Progress – CFD Trends from 1990 to 2010 Illustrated by Formula 1 Racing (Hanna & Parry, 2011)</div>
Second Wave: CFD Software Enters R&D Departments of Major Enterprises
A pioneer in selling CFD software dedicated to industrial applications was Flomerics, founded by David Tatchell and Harvey Rosten in Kingston upon Thames, UK, in 1988. Using typical first-stage technologies, it released its first software package, FloTHERM, in 1989. The two founders, who previously held senior positions at CHAM, left to establish Flomerics with the vision of "contributing to the technological advancement of industries" (Tatchell, 2009). FloTHERM represented the first paradigm shift in the CFD industry: instead of focusing on complex CFD technologies, it made solving industrial engineering tasks its core goal. This shift meant that product development engineers, in addition to scientists, would become the main target user group for such CFD software. However, the then CFD technology level, computer hardware, and operating systems to some extent limited this innovative approach. Therefore, Flomerics initially focused on two application areas: electronic thermal management (product: FloTHERM) and building HVAC systems (product: FloVENT). These fields had relatively clear and, more importantly, feasible requirements for engineering-oriented CFD software.
Figure 4: Early Version of FloTHERM (Hanna & Parry, 2011)
This concept opened up new market opportunities by catering to a larger user group with shared industry backgrounds and application needs. It was during this period that many product development engineers, lacking expertise in numerical methods and extensive CFD experience, first began to use CFD simulation confidently as a development tool. The focus shifted to solving technical engineering tasks, with CFD technology more or less serving as a means to an end.
Other CFD providers obviously recognized this paradigm shift, especially the new business opportunities it brought, and launched their own products. For example, MixSim, a dedicated interface for the Fluent solver released in 1996, was used to model industrial mixing processes. Fluid Dynamics International entered the electronic thermal management CFD market with Icepak (based on the FIDAP solver), while CD Adapco launched various dedicated engineering tools for the automotive industry. Emerging companies such as Exa Corporation (product: PowerFlow) and Blue Ridge Numerics (product: CFdesign) also targeted new market opportunities and entered the market with new CFD products designed specifically for industrial applications. Overall, all CFD software suppliers invested heavily in developing better user interfaces, powerful solvers, and reliable physical models, aiming to ensure CFD's entry into the R&D departments of large industrial enterprises and attract a new generation of CFD users.
This second wave of CFD software development, targeting industrial applications, lasted from the early 1990s to the early 2000s. It was characterized by lower-cost, more powerful computing hardware, which spawned a variety of practical CFD simulation tools. This in turn stimulated rapid growth in market demand for CFD simulation (especially among large enterprises), further accelerating the popularization of CFD technology by software suppliers. However, at the same time, many users (understandably) noticed another trend with mixed feelings: the CFD software industry began market consolidation through mergers, acquisitions, and withdrawals. Many established CFD systems became outdated and required significant R&D investment. For major CFD software suppliers, the previous high growth rates faded. Rising development costs and increasing market competition prompted companies to integrate resources, maintain competitive advantages, and address future challenges. During this period, the foundation for corporate mergers was laid, resulting in many CFD companies being acquired by large software companies with thousands of employees, which came to dominate the CFD software market.
Figure 5: FIDAP Software User Interface in the Late 20th Century (University of Delaware, 2007)
Since the early 2000s, CFD software has successfully entered large industrial companies and been widely used as a regular tool for verifying and optimizing product design, functionality, processes, and physical effects. Its poor reputation among engineers has significantly improved. Hundreds of case studies have shown that with careful modeling, commercial CFD software paired with powerful hardware can greatly reduce R&D time and costs. As a result, market demand for CFD simulation has surged, especially among small and medium-sized companies urgently needing to reduce the cost of physical prototypes often had to outsource their previous physical prototypes due to prohibitively high expenses. Nevertheless, for the CFD simulation industry at the beginning of this century, cost remained a very serious constraint, as the expenses associated with relevant physical experiments were persistently high. The primary determining factor causing this situation was staff costs, particularly those related to hiring and training qualified users; other factors included a relatively long learning curve required after job training before becoming an expert, a highly labor-intensive modeling process (especially when involving complex geometries), and relatively high software licensing costs. Another crucial aspect was the need for user companies to integrate CFD simulation into their routine product development processes, while these companies often lacked dedicated simulation departments. That is to say, qualified engineers from product development or design teams needed to conduct simulations themselves; simultaneously, for simulation results to genuinely guide the improvement of design solutions, it was essential to enhance the efficiency of simulation projects, ensuring that CFD simulation results could keep pace with the product design cycle.
Figure 6: CFD User Pyramid (Hanna & Parry, 2011)
Industrial-grade geometry processing also played a key role. By then, these geometry processing results were available as 3D CAD data, and ideally, minimal simplification and modification of this data should be required for subsequent (preferably fully automated) mesh generation processes. The CFD software market responded positively to these needs by launching many new and improved products, marking the start of the third wave of CFD software development targeting industrial product design, which continues to this day.
Third Wave: CFD Becomes a Fundamental Element of Product Design Processes
Major suppliers of CAD and PLM systems played a key role in this third stage. Since the 1990s, they had successfully introduced product lifecycle management (PLM) concepts, including CAE. As a result, customers increasingly pressured commercial CFD software suppliers to comply with this concept and take steps to integrate their products into major PLM systems. Therefore, in the early 2000s, almost all CFD software providers upgraded their systems to at least offer CAD import interfaces. Many developed bidirectional links with major CAD/PLM systems, and a few even embedded their CFD technology directly into 3D CAD systems.
CAD system manufacturers strongly supported these developments, aiming to provide customers with complete solutions within their PLM system frameworks by supporting external professional module developers. During this period, products such as Fluent for CATIA (Fluent Inc), CFdesign (Blue Ridge Numerics), and FloWorks (NIKA GmbH) emerged. At the same time, new CFD technologies were developed to support these new requirements—some from scratch, others by improving existing technologies. For example, since 1999, CD-Adapco has successfully used an innovative object-oriented approach to develop STAR-CCM+. The German-Russian joint venture NIKA GmbH (founded in 1999) is a typical example of a new commercial CFD software supplier in the early third wave. Based on the aforementioned Aeroshape-3D technology, NIKA exclusively developed CAD-embedded CFD software, which is now available as dedicated versions for several mainstream 3D CAD systems (Figure 7).
Figure 7: FloEFD for Creo Software – Provided by Mentor Graphics
To adapt to changing market conditions, Blue Ridge Numerics adjusted its CFdesign software package to function as a "front-end CFD" system. Several PLM suppliers also actively participated in CAD-integrated CFD software development through acquisitions to better support customers' product development processes. Representative companies in this niche market include Dassault Systèmes (product: SIMULIA Abaqus/CFD) and Siemens PLM Software (products: NX Advanced Flow and Femap Flow Solver). Autodesk also enriched its CFD software portfolio by acquiring Blue Ridge Numerics' CFdesign software to complement its Algor suite.
The current third wave has also created opportunities for new entrants in other fields to enter the CFD market, bringing breakthrough technologies that have refreshed the market. One example is XFlow from Spain's Next Limit Technologies, which not only introduced another CFD technology from the film industry but also brought a user interface closer to animation software to the engineering field. Autodesk's Project Falcon, on the other hand, introduced game elements into the CFD world.
Figure 8: XFlow Software User Interface – Provided by Next Limit Technologies (MSC Software, 2011)
These examples illustrate a potential future trend: the CFD software market will become increasingly diverse, adopting more innovative and unconventional methods, especially in improving user experience and product usability. Of course, all these share a common focus: industrial users remain the priority, and CFD software has become an essential tool for digital prototyping. Their needs for CFD software include ease of use, task orientation, automation, efficiency, reliability, and accessibility. The underlying reasons lie in the continuous changes in product development processes and the evolving role of simulation engineers. Process integration, reliability, modeling security, and repeatability are becoming key focuses and influencing users' purchasing decisions for CFD software. Future CFD software development centered on these needs will bring exciting new technologies and products to the market. Thus, a new fourth wave is just around the corner...
Outlook – Vision
Hanna and Parry (2011) described their outlook for the future as follows: "In the authors' view, the ultimate goals for CFD—real-time, push-button, automated, easy-to-use, CAD-embedded, bidirectional, and multiphysics—remain to be achieved. Currently, some CFD codes are approaching these ideals, and over the next 20 years, various factors including advances in computing hardware, algorithms, physical modeling, and coupling will help us reach this ideal state."
However, such a long-term goal can only be achieved gradually, and there are still many challenges along the way, as observed by the authors. Of course, this ultimate goal may also need to be adjusted over time, as the design environment evolves—after all, CFD is iterative! Below, from today's perspective, we discuss several possible milestones in achieving this ultimate goal.
An important aspect of achieving CFD's ultimate goal is to more realistically represent complex physical realities, removing the artificial "boundaries" that exist today, formed by the independent development of CFD, computational structural mechanics, multibody dynamics, kinematics, and other disciplines using different numerical technologies. A sign of progress in this area is what is commonly referred to as "multiphysics" simulation. However, this often means applying the results of one simulation (e.g., thermal analysis) as initial or boundary conditions for another (e.g., thermal stress analysis).
Some software suppliers (such as ANSYS and COMSOL) have made multiphysics a core concept of their products and offer a wide range of simulation capabilities. However, the current focus of multiphysics applications remains on mastering the functionality of various components and addressing the technical challenges of their collaboration, as each component may have a unique history and technical background, leading to incompatibility. Solving this problem may require software architectures to provide the infrastructure for component collaboration. These architectures can be developed internally by multiphysics software suppliers or provided as middleware by independent third-party developers. An example of this is the Fraunhofer MpCCI architecture.
Figure 9: MpCCI Visualizer Software – Provided by Fraunhofer SCAI (Fraunhofer SCAI, 2012)
Another constraint of current multiphysics methods is accurately representing the complex actual physical scenarios of various individual solver modules used in a given simulation project. To ensure that the results of one simulation can be used as input for another, a "white-box" model is often needed, capable of capturing various geometries without simplification and simulating all relevant physical effects in full detail, including associated simulation overhead. "Black-box" models can offer high simulation efficiency but are limited to specific aspects of a problem (e.g., thermal models of electronic components) and thus are not suitable for this mode.
Currently, the selection of appropriate simulation modules, configuration, and workflow arrangements are entirely the user's responsibility, and the actual workflow is determined by the different needs of various solver modules rather than the physical fields of the actual engineering task. Therefore, "multinumerics" may be just a descriptive term.
For this approach to succeed in the future, a prerequisite is not just linking independent solvers but merging them into a single unified solution methodology. This would allow users to focus on the product's physical fields (complex but unified) while the simulation environment accommodates any required numerical technologies with consistency. This approach must be complemented by user experience-oriented design methods, shifting focus from simple product feasibility to efficiently solving engineering tasks as the most important criterion.
If one insists on seeking a universal physical solver just because it is possible, an inevitable question arises: how to unify multiple distinct and incompatible numerical methods. Diverse methods are certainly useful because product behavior involves multiple physical properties, each inherently different, and each has one or more corresponding numerical methods that offer the desired combination of result accuracy, computational resource requirements, and solution efficiency.
Abandoning this significant advantage to develop a single process for all possible physical applications is not desirable, as these applications may span many different fields. A single process would be far less efficient than using the best method for each application. Instead, the ideal goal should be to develop a solver infrastructure that can automatically apply the best method for each scenario, integrate them into a single simulation model, and achieve bidirectional cross-boundary coupling. This means integrating highly diverse methods: in fluid dynamics alone, these include discretization methods (e.g., finite volume methods for internal fluid simulation), particle methods (e.g., smoothed particle hydrodynamics (SPH) for multiphase flow and phase change regions), and one-dimensional methods for large flow systems. For many elements of this approach, mature and reliable components are already available on the market. The current task is to end the historical isolation of solver modules and develop a single simulation engine that can integrate the best methods for any simulation task. The greatest advantage of this approach is providing workflows focused on engineering tasks and their solutions, allowing engineers to be completely relieved of the burden of defining the entire numerical workflow. From this perspective, we believe this is a viable path to achieving CFD's ultimate goal.
There is no doubt that the needs of engineer users will drive the further development of future simulation software. Software must adapt to users' work environments, needs, and knowledge levels, not the other way around. This affects the overall software concept, every detail, and the product specifications and code implementation processes used by software suppliers. Currently, many software companies have adopted modern product development processes such as agile development. This naturally supports user-centered design processes and is a prerequisite for effectively implementing product usability requirements, with the sole goal of providing and maintaining excellent user experience. Wise investment in this area will undoubtedly become a highly attractive unique selling point in the CFD software market.
The work environment of development engineers and designers is constantly changing. New input technologies that better reflect natural human movements are being developed, while others are gradually being applied in the workplace. For example, augmented reality or touchscreen operations are particularly noteworthy. Similarly, new visualization technologies will be used for ergonomically accurate demonstrations of physical simulation scenarios. For instance, while 2D sketches and printed materials have long been the medium of communication between engineers, technicians, and workers, 3D printing has now been added. Such technological advancements will continue, and engineers will remain the core decision-makers in product development processes for the foreseeable future. This trend has undoubtedly been keenly noticed and actively responded to by the simulation software industry, increasing its importance. In an environment increasingly reliant on virtual prototypes (for cost-effective product development), the visualization and communication of simulation results are becoming more important, closely linked to the growing responsibility of simulation engineers for their conclusions.
Figure 10: User-Centered Design Activities - Seamlessly Applicable to Any Level of Agile Development Methodology (Limina Application Office, 2012)
However, user experience and product usability will play an increasingly important role as decision criteria for tool selection in the future—not just at the abstract and conceptual level, but every detail of the user interface requires attention. Many elements of current CFD software user interfaces (though many have been replaced with new and aesthetically pleasing ones) still bear traces of the early stages of software development. The problem is not just in interface details but often in deeper aspects of the software and its behavior. Since 1990, Jakob Nielsen has proposed the now well-known general principles of user interface design, namely the "Ten Heuristics for Usability Evaluation" (Nielsen et al., 1993). Below, we briefly comment on the application of these rules in the context of future simulation software needs, from both conceptual and practical detail perspectives:
➤ Visibility of system status
The system should always keep users informed about what it is doing, providing appropriate feedback within a reasonable time.
➤ Match between system and the real world
The system should use the user's language—familiar words, phrases, and concepts—rather than system-oriented terminology. Follow real-world conventions, presenting information in a natural and logical order.
➤ User control and freedom
Users often mistakenly select system functions, so there should be a clearly marked "emergency exit" to leave unintended states without unnecessary dialog boxes. Additionally, the system should support "undo" and "redo" functions.
➤ Consistency and standards
Users should not have to wonder whether different words, situations, or actions mean the same thing. Follow platform conventions.
➤ Error prevention
A careful design that prevents problems in advance is far better than timely error messages. Either eliminate error-prone conditions or check them and offer confirmation options.
➤ Recognition rather than recall
Make every object, operation, and option visible at all times, minimizing the user's memory burden. Users should not have to remember information from different parts of a dialog. System instructions should be either visible or easily retrievable.
➤ Flexibility and efficiency of use
Accelerators—unseen by novice users—can speed up interactions for expert users, allowing the system to meet the needs of both inexperienced beginners and experienced advanced users. Allow users to perform frequent operations.
➤ Aesthetic and minimalist design
Dialogs should not contain irrelevant or rarely needed information. Any extra information in a dialog severely distracts from relevant information, reducing its visibility.
➤ Help users recognize, diagnose, and recover from errors
Error messages should use plain language (not codes), accurately indicate the problem, and offer constructive solutions.
➤ Help and documentation
Although some systems may be usable without documentation, it is usually necessary to provide help and documentation. Such information should be easy to search, task-oriented (focused on users' tasks), list specific steps required, and not be overly lengthy.
Commercial CFD software dedicated to industrial applications has a 30-year history. Over these three decades, thousands of scientists, engineers, and university students have successfully made this technology an indispensable tool through various CFD simulation practices, gradually embedding it into product design processes across industries. While traditional CFD technology is quite mature, more exciting new concepts and technologies will continue to emerge to meet future CFD application challenges. Each of the first two waves in commercial CFD brought a paradigm shift; now, we are experiencing the third wave, another paradigm shift—integrating CFD software into design processes. It is believed that in the near future, CFD simulation software will usher in a fourth wave. The author expects the next wave to move closer to CFD's ultimate goals: real-time, push-button, automated, easy-to-use, CAD-embedded, bidirectional, and multiphysics. By then, traditional CFD software from the second wave will be far left behind, eventually becoming obsolete.
Contact:
Prof. Tian：WhatsApp：+86 15029941570 | Mailbox：540673737@qq.com
we specialize in CAE finite element calculation and simulation, virtual prototyping and simulation testing, and product design optimization for enterprises and institutions.
Copyright © 2025.Boye Engineering Technology All rights reserved. Yue ICP17017756Num-1