💻 computer science

Accurate simulation of delamination with a resin-rich layer-dependent penalty stiffness based on structural cohesive elements

This paper enhances a higher-order structural cohesive element by introducing a resin-rich layer-dependent penalty stiffness formulation that distinguishes between normal and shear components, thereby significantly improving mesh convergence and the accuracy of compression and delamination predictions in shell-based simulations compared to conventional methods.

Original authors: Xiaopeng Ai, Christos Kassapoglou, Boyang Chen

Published 2026-08-07

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Xiaopeng Ai, Christos Kassapoglou, Boyang Chen

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are building a house out of thousands of tiny, colorful LEGO bricks glued together to make a super-strong wall. In the real world, engineers build similar structures using layers of carbon fiber and plastic resin to create wings for airplanes or parts for rockets. These materials are incredibly strong, but they have a secret weakness: the layers can sometimes peel apart from each other, like pages in a book being ripped out. This peeling is called "delamination," and if it happens unexpectedly, the whole structure could fail. To stop this, scientists use computer programs to simulate how these layers might separate. However, these programs often struggle with a tricky problem: they need to be incredibly detailed (using millions of tiny digital bricks) to get the math right, which makes the computer take forever to run the simulation. If the digital bricks are too big, the computer gets confused and predicts that the layers will squish together in impossible ways, leading to wrong answers.

This paper tackles that confusion by inventing a smarter way to tell the computer how the layers should behave when they are peeling apart. The authors, Xiaopeng Ai, Christos Kassapoglou, and Boyang Chen, propose a new mathematical "rulebook" for the glue between the layers. Instead of guessing how stiff the glue should be based on the whole wall's thickness, they look at the tiny, messy layer of extra resin (the glue) that naturally sits between the fiber layers. By using a clever trick based on how beams bend, they figured out exactly how the stress spreads through that thin glue layer. Their new method allows the computer to use much larger, coarser digital bricks without getting the math wrong. This means engineers can simulate complex peeling problems much faster and with greater accuracy, ensuring that the airplanes and rockets built with these materials are safe and reliable.

The Problem: The "Squishy" Glue Mistake

Think of a composite material like a giant sandwich made of many layers of bread (the fiber) and a tiny bit of extra butter (the resin) between each slice. When the sandwich starts to tear, the butter layer is the first to stretch and break. In the past, computer simulations treated this butter layer as if it were a simple, uniform spring. The problem is that this "spring" was often too stiff or too weak depending on how the computer sliced up the sandwich into digital pieces.

When the computer used big digital slices, it made a specific mistake: it predicted that the layers right in front of the tear were being squished together with huge force, like a car crash, when in reality, they should just be gently peeling apart. This "squishy" error meant that to get a correct answer, engineers had to use tiny, tiny digital slices, which made the simulations take days to run. It was like trying to draw a smooth curve using only tiny, jagged steps; you needed millions of steps to make it look right.

The New Idea: Listening to the Stress

The authors realized that the old way of calculating the "stiffness" of the glue was too simple. They knew that the stress (the internal force) doesn't spread evenly through the thickness of the sandwich; it changes as you move from the top layer to the bottom. To fix this, they didn't just look at the glue; they looked at the physics of how a beam bends.

Imagine bending a thick book. The pages on the outside stretch, the pages on the inside squish, and the pages in the middle stay neutral. The authors used this same logic to map out exactly how the stress changes through the thickness of the resin layer. They derived a new formula that acts like a "smart translator." It takes the complex, changing stress pattern and converts it into a single, accurate stiffness number that the computer can use.

Crucially, this new formula works even when the computer groups many layers of the same type into one big "super-layer" (an equivalent single-layer model). Previous methods required the computer to model every single layer individually, which was slow and inefficient. The new method allows the computer to group layers together without losing accuracy, making the simulation run much faster.

The Proof: Testing the New Rulebook

To see if their new idea actually worked, the team ran a series of tests using standard "benchmark" problems that engineers use to check their math. These included:

The Double Cantilever Beam (DCB): A test where you pull two layers apart like opening a book.
The End-Notched Flexure (ENF): A test where you slide layers past each other.
The Mixed-Mode Bending (MMB): A test that combines pulling and sliding.
The Reinforced DCB (R-DCB): A more complex test with extra layers added to the outside, simulating a real-world reinforced structure.

In every test, they compared their new "smart" stiffness against the old "dumb" stiffness. The results were clear:

Mesh Convergence: When they used coarse (big) digital meshes, the old method failed. It gave different answers depending on how big the pieces were, and it predicted that the layers were squishing together way too hard. The new method, however, gave the same accurate answer whether the digital pieces were big or small. It was like having a map that looked just as good whether you zoomed in or out.
Accuracy: In the DCB test, the new method reduced the error in the load-displacement curve (how much force is needed to peel the layers) significantly. For a 7.5 mm mesh, the old method had an error of over 50%, while the new method kept it very low.
Real-World Comparison: In the complex R-DCB test, which was compared to real physical experiments, the new method predicted the crack's path and the force needed to break it much better than the old method. Specifically, it predicted the second peak of the breaking force to be 14% lower than the old method, which matched the real-world data much more closely.

What This Means

The paper doesn't claim to have solved every problem in the universe of composite materials. It specifically focuses on improving the "penalty stiffness" for structural cohesive elements, which are the digital tools used to simulate peeling. The authors show through simulations and comparisons with experiments that their new formula is superior to the conventional one.

By using this new method, engineers can simulate delamination in complex 3D structures with coarser meshes. This means they can get accurate results in a fraction of the time it used to take. It's like upgrading from a slow, pixelated video game to a smooth, high-definition one without needing a supercomputer. The authors suggest that this approach could be extended in the future to predict other types of damage, like cracks inside the layers themselves, but for now, they have successfully fixed the "squishy glue" problem, making the digital testing of composite materials faster and more reliable.

Technical Summary: Accurate Simulation of Delamination with a Resin-Rich Layer-Dependent Penalty Stiffness Based on Structural Cohesive Elements

Problem Statement
Fiber-reinforced polymer (FRP) composites are critical in aerospace structures, yet predicting delamination remains a challenge. While cohesive zone models (CZM) using cohesive elements (CE) are standard for simulating delamination, they suffer from a well-known limitation: the requirement for element sizes smaller than the cohesive zone length (CZL) to capture high-stress gradients, which leads to high computational costs.

To address this, higher-order structural cohesive elements (based on Kirchhoff-Love shell theory) were previously developed to replace solid elements, significantly improving efficiency. However, these elements rely on a penalty stiffness formulation to define the constitutive behavior. Conventional formulations (e.g., Turon et al.) define penalty stiffness based on the total laminate thickness and out-of-plane modulus. This approach fails to accurately predict through-thickness stress distributions, particularly overestimating compression ahead of the crack tip and failing to achieve mesh convergence.

Previous attempts to improve accuracy involved resin-rich layer-based formulations. While accurate, these required a layer-wise modeling approach where every ply interface was explicitly modeled. This negated the computational efficiency gains of equivalent single-layer (ESL) models, where multiple plies with the same orientation are homogenized into a single shell element. There was a need for a penalty stiffness formulation that retains the physical accuracy of the resin-rich layer approach but is applicable to ESL models.

Methodology
The authors propose a new penalty stiffness formulation for structural cohesive elements that bridges the gap between resin-rich layer accuracy and ESL efficiency. The core of the method involves deriving analytical expressions for the through-thickness distributions of out-of-plane normal stress ( $\sigma_z$ ) and transverse shear stress ( $\tau_{xz}$ ) based on beam theory and the Kirchhoff-Love hypothesis.

Stress Distribution Derivation:
- The 3D shell problem is reduced to a 2D beam problem.
- Under Kirchhoff-Love assumptions, axial stress varies linearly, while transverse shear stress follows a quadratic distribution.
- Equilibrium equations are used to derive a cubic distribution for the out-of-plane normal stress ( $\sigma_z$ ).
- Boundary conditions for specific loading modes (Mode I/DCB and Mode II/ENF) are applied to determine the coefficients of these stress distributions.
Penalty Stiffness Formulation:
- The total opening (Mode I) or sliding (Mode II) displacement of the cohesive element is calculated by integrating the strains across the thickness of the homogenized laminate.
- Strains are derived from the stress distributions and the properties of the resin-rich layer (modulus $E_{rr}$ or $G_{rr}$ and thickness $h_{rr}$ ).
- The penalty stiffness ( $K_n$ for normal, $K_s$ for shear) is formulated as a function of the resin-rich layer properties and the sum of the stress ratios at each interface layer relative to the maximum stress.
- This allows the formulation to account for the cumulative compliance of all resin-rich layers within a homogenized ply block, rather than treating the block as a single homogeneous layer.

Key Contributions

New Formulation: A resin-rich layer-dependent penalty stiffness formulation that distinguishes between normal and shear stiffnesses.
ESL Compatibility: The method extends the resin-rich layer approach from layer-wise modeling to the equivalent single-layer framework, enabling the use of coarser meshes without sacrificing physical accuracy.
Analytical Derivation: The derivation of through-thickness stress distributions using beam theory to determine the specific stress ratios required for the stiffness calculation.
Mode-Dependent Application: The formulation is applied to Mode I, Mode II, and mixed-mode scenarios using appropriate boundary conditions for DCB, ENF, and Mixed-Mode Bending (MMB) configurations.

Results and Validation
The proposed method was validated against four benchmark problems: Double Cantilever Beam (DCB), End-Notched Flexure (ENF), Mixed-Mode Bending (MMB), and Reinforced DCB (R-DCB).

Mesh Convergence: In DCB and MMB benchmarks, the proposed formulation demonstrated significantly improved mesh convergence compared to the conventional formulation. Accurate load-displacement curves and traction distributions were achieved even with coarse meshes (e.g., 7.5 mm), whereas the conventional method failed to converge or required much finer meshes (2 mm) for similar accuracy.
Stress Prediction: The proposed method significantly reduced the overestimation of compressive stress ahead of the crack tip observed in conventional shell-based cohesive elements. The predicted traction profiles showed better agreement with solid element models and analytical solutions.
Mode II Performance: In ENF benchmarks, the proposed shear stiffness outperformed other shell-based formulations (e.g., Bazilevs et al.), which significantly underestimated initial stiffness.
3D Complexity (R-DCB): In the Reinforced DCB benchmark, which involves complex 3D delamination and reinforcement, the proposed method provided better agreement with experimental data regarding the second peak load and crack front evolution. It correctly captured the softening behavior in reinforced regions, which the conventional method predicted as excessively stiff.
Computational Efficiency: By enabling the use of coarser meshes while maintaining accuracy, the method reduces the computational time required for delamination simulations.

Significance
The paper claims that this work resolves a critical limitation in the application of higher-order structural cohesive elements: the inability to use equivalent single-layer models without losing accuracy in through-thickness stress prediction. By incorporating the physical properties of the resin-rich layer into the penalty stiffness derivation via analytical stress distributions, the method achieves a balance between computational efficiency and predictive accuracy.

The authors state that this approach allows for accurate prediction of out-of-plane delamination and stress distributions in 3D composite laminate models based on the ESL method. The improved mesh convergence and reduced discrepancy with solid element models make the method particularly valuable for simulating complex structures with reinforcements. The paper concludes by noting that future work will focus on extending this formulation to predict in-plane damage mechanisms (matrix cracking and fiber failure) by combining in-plane shell stresses with the out-of-plane stresses provided by the cohesive elements.

The Problem: The "Squishy" Glue Mistake

The New Idea: Listening to the Stress

The Proof: Testing the New Rulebook

What This Means

More like this