Finite Element Analysis in the AI era: Insights from scientific publishing trends

Introduction 

Will AI make traditional engineering tools like finite element analysis (FEA) obsolete? With the rise of machine learning and data-driven methods, it's a fair question. But if we look at scientific publishing trends over the past 20 years, the answer is clear: FEA is not going away — in fact, it's growing. In this article, we explore how FEA is evolving in the age of AI and why physics-based simulations remain a key part of the engineer’s toolbox.

A brief history of Finite Element Analysis 

Finite Element Analysis has been a cornerstone of engineering simulation for decades. While AI is now making waves in many disciplines, FEA has a long tradition of solving complex, real-world problems based on physics and mathematics. Here are some key moments that shaped its development: 

• 1956 – The foundation is laid: A scientific paper is published that many consider the first paper on the finite element method. 

• 1960s–70s – Research labs lead the way: FEA becomes essential in several disciplines like aerospace and civil engineering, with early software developed in research labs and universities. 

• 1980s – Commercial tools emerge: The first versions of well-known packages like ANSYS, Abaqus, and NASTRAN gain traction, enabling widespread industrial use. 

• 1990s–2000s – Mainstream adoption: Improved user interfaces, faster computers, and integration with CAD make FEA accessible to engineers across disciplines. 

• 2010s–2020 – Performance and automation: Advancements in solver technology as well as the availability of high-performance and cloud computing help scale FEA to increasingly complex problems. 

This long and steady evolution highlights how FEA has evolved and adapted over time — a trend that continues in the age of AI. 

Why FEA remains essential in the age of AI 

Finite Element Analysis remains a core part of the engineering toolbox. It provides accurate, physics-based insights that are difficult to replicate using data alone. Moreover, artificial intelligence depends on large volumes of data, which are often unavailable or too costly to obtain. FEA fills that gap by enabling engineers to simulate and explore systems based on physical laws, even in the absence of extensive datasets. 

From stress analysis in aerospace components to design of biomedical devices, FEA is trusted across industries for its ability to simulate complex physical systems. It plays a crucial role in safety validation, design optimization, and regulatory compliance — areas where precision and reliability are non-negotiable. 

Despite the growing interest in AI, scientific publications suggest that FEA is not being replaced. In fact, its usage in research continues to rise.

A look at the data 

To understand how interest in FEA has evolved, we analyzed scientific publication trends using data from ScienceDirect, one of the largest scientific databases with mainly publications from Elsevier. The graph below shows the number of research articles published each year that explicitly mention the term “finite element analysis.” This data spans from 2005 to 2024 and reflects articles across various disciplines. 

As the chart shows, the number of FEA-related papers has steadily increased over the past two decades. This trend reinforces the idea that FEA remains central to research and innovation — even in the face of rapidly emerging technologies like AI.

While the increase in FEA-related publications is clear, one might wonder whether this simply reflects a broader rise in scientific publishing overall. According to a study analyzing the overall increase in scientific publications, the total number of Elsevier publications grew by approximately 41% between 2016 and 2022. However, our data shows that publications mentioning “finite element analysis” increased by 73% over the same period. This suggests that the growth of FEA in academic research is outpacing the general rise in publishing volume. Of course, this is a rough comparison, and care should be taken when interpreting such trends, but it does provide useful context for understanding FEA’s continued relevance.

In addition to overall FEA trends, we also analyzed publication counts mentioning specific FEA software packages: Abaqus, ANSYS, COMSOL, and LS-DYNA. All four have seen growth in scientific mentions over the past two decades, but some interesting differences emerge. Between 2005 and 2024, Abaqus experienced a rapid increase in academic publications. While ANSYS and Abaqus were referenced in a similar number of papers in 2005, there is now a wide gap between them with Abaqus appearing in significantly more publications in recent years. COMSOL has also shown strong growth, especially in the last few years, signaling rising adoption in certain academic fields.

It’s important to note that these trends reflect academic publishing patterns, not overall market share. This may differ due to several factors like commercial pricing. Nevertheless, the data provides an interesting window into how different FEA tools are used and cited in scientific research.

FEA meets AI 

There are many possible synergies between FEA and AI. One of the most promising ways AI and FEA come together is through surrogate modeling. Running full FEA simulations can be time-consuming and computationally expensive, especially for complex geometries. By using a large dataset of FEA results, machine learning models can be trained to act as fast approximations, or “surrogates”, of the full simulation. These surrogate models can produce results in real time, making them ideal for design optimization and interactive applications. While they don’t fully replace traditional FEA, they can significantly accelerate workflows by providing quick, informed estimates grounded in physics-based data.

The growing interest in combining FEA with machine learning is clearly reflected in recent publication trends. In 2015, only about 1% of scientific articles mentioning “finite element analysis” also mentioned “machine learning.” By 2024, that number has risen to 17%, showing a sharp increase in research focused on integrating physics-based and data-driven modeling. This upward trend suggests that more researchers are actively exploring the synergies between these two approaches — using machine learning to enhance, accelerate, or extend the capabilities of traditional FEA methods.

Meshing insights 

An essential part of any finite element analysis is the mesh — and the choice between hexahedral and tetrahedral elements remains a key consideration. Let’s use a similar approach to gain some insights based on scientific publications. An analysis of publication data from 2005 to 2024 shows a steady increase in scientific articles referencing both "finite element analysis" alongside either "hexahedral" or "tetrahedral." Interestingly, the two curves are nearly identical, suggesting that both element types are used with similar frequency and that this balance has remained stable over time.

This mirrors what many engineers and researchers already know: each element type has its strengths, and the best choice often depends on the specific application. For those interested in a deeper dive, we wrote an article on the trade-offs of both element types and how high-quality hexahedral meshes can be generated using Python scripting. As meshing tools continue to improve, access to automated workflows for generating structured meshes will likely make hexahedral elements more practical for a wider range of problems. 

Limitations 

While the publication trends presented in this article offer useful insights into how finite element analysis is evolving in the AI era, there are several limitations to keep in mind: 

• The data was sourced exclusively from ScienceDirect, which primarily features journals published by Elsevier. As a result, it does not capture the full spectrum of scientific literature across other publishers and databases. 

• The analysis of FEA software focused on a select group of tools — Abaqus, ANSYS, COMSOL, and LS-DYNA. Other used packages, both commercial and open-source, were not included. 

• Articles that mention both “finite element analysis” and “machine learning” may not necessarily integrate the two technologies. In some cases, the terms may appear in separate contexts within the same paper, such as in literature reviews or future work sections. 

• Publication counts reflect interest and usage in academic research but do not necessarily correspond to industry trends. 

  
  

Despite these limitations, the trends provide a valuable window into the ongoing role of FEA in research and how it continues to evolve alongside emerging technologies like AI. 

Conclusion 

Despite the rapid rise of AI and machine learning, our data shows that finite element analysis remains a crucial tool in engineering and scientific research. Its steady growth in academic publishing underscores its continued importance — not just as a legacy method, but as an evolving discipline that’s adapting to new technologies. Rather than being replaced, FEA is increasingly being enhanced by AI, with exciting developments like surrogate modeling.

As engineering challenges become more complex, combining physics-based and data-driven methods will be key to innovation. Tools like FEA will continue to play a vital role in that future.

If you're interested in staying up to date with the latest insights in 3D modeling and simulation, we invite you to subscribe to HelloTriangle.