New Publication · Computers & Mathematics with Applications
I’m proud to share our paper introducing the NURBS-Divergence-Meshless (NDM) method — a meshless formulation that applies the divergence theorem in NURBS parameter space to discretise governing equations.
Originally posted 28 July 2025 · Updated October 2026
NURBS-Divergence-Meshless (NDM) method
N. Rasin, H. Hirol, A. R. Zainal Abidin, M. H. Mokhtaram, M. A. Mohd Noor, A. Y. Mohd Yassin
Computers and Mathematics with Applications, Vol. 196, pp. 151–171, 2025 · Online 22 July 2025
The idea in a nutshell
Meshless methods free us from building and maintaining a finite element mesh, but they leave two practical headaches behind: describing curved geometry accurately, and turning the governing equation into exactly the right number of algebraic equations without heavy domain integration.
NDM tackles both with one move. The geometry is described by NURBS, so curved boundaries are represented exactly, as in isogeometric analysis. The governing equation is then integrated over small cells, and the divergence theorem turns each area integral into a line integral around the cell boundary — the same trick that underpins the finite volume method:
∫Ω ∇·q dΩ = ∮Γ q·n dΓ
The key detail is where this happens. NDM performs the integration in the NURBS parameter space, which is rectangular. Every cell there is a simple rectangle with four straight edges, so the line integrals become very easy to write down and evaluate — even though the corresponding cells in the physical domain may be curved.
The field variable itself is approximated with standard meshless interpolation: either moving least squares (MLS) or radial basis functions (RBF).
Physical domain (curved) → NURBS mapping → parameter space (rectangular cells) → divergence theorem per cell → line integrals on straight edges → algebraic equations
What NDM borrows, and from where
One way I like to explain NDM is that it combines three ideas many of us already know, each doing the job it does best:
| Ingredient | Borrowed from | Role in NDM |
|---|---|---|
| NURBS geometry | Isogeometric analysis | Exact domain representation and a rectangular parameter space |
| Divergence theorem over cells | Finite volume method | Turns the governing equation into one equation per cell via line integrals |
| MLS / RBF approximation | Meshless methods | Approximates the field variable from scattered nodes |
Why I think it matters
- The right number of equations, naturally. Because the line integration is built around cells, it delivers exactly the number of equations the system needs, which simplifies the discretisation considerably.
- Boundary conditions with MLS, imposed directly. MLS shape functions lack the Kronecker delta property, which is why Galerkin meshless methods usually need Lagrange multipliers, penalty terms or special corrections to enforce essential BCs. NDM allows these BCs to be imposed directly. For anyone who has fought with BC enforcement in EFG, this is the part to look at first.
- Curved geometry without the usual pain. Exact NURBS geometry combined with straight-edged integration in parameter space means accuracy in the geometry does not come at the cost of complicated boundary integration.
What the paper shows
This paper is deliberately foundational. To establish the concept clearly, we discretised and solved a simple scalar problem using both MLS and RBF interpolation, then carried out convergence studies on the variants.
The clear winner was the non-overlapping NDM-MLS variant, which showed all the favourable characteristics we were hoping for.
Being equally clear about scope: these are results for a scalar problem. Extending NDM to elasticity and other vector problems, multi-patch NURBS geometries, three dimensions, and comparisons against FEM, IGA and established meshless methods on engineering benchmarks are the natural next steps — and exactly the kind of work I would love to do with motivated students.
Where this fits in our work
NDM continues a line of work on bringing exact NURBS geometry into meshless methods. Some earlier posts on this blog give useful background:
- N-RPIM for structural analysis — our earlier coupling of NURBS geometry with RPIM interpolation
- Radial Basis Function (RBF) — the basics of RBF interpolation, with MATLAB
- Point Interpolation Method (PIM) — polynomial point interpolation, with MATLAB
- State-of-the-art review on meshless methods for crack problems — where meshless methods are heading in fracture mechanics
Interested in meshless methods or NURBS-based analysis?
I welcome students (final-year project, MSc and PhD) and collaborators who would like to extend NDM to new problems, or work more broadly on meshless methods, isogeometric analysis and computational mechanics. Drop me an email with a short note about yourself and your interests.
Cite this paper
Rasin, N., Hirol, H., Zainal Abidin, A. R., Mokhtaram, M. H., Mohd Noor, M. A., & Mohd Yassin, A. Y. (2025). NURBS-Divergence-Meshless (NDM) method. Computers and Mathematics with Applications, 196, 151–171. https://doi.org/10.1016/j.camwa.2025.07.011
@article{Rasin2025NDM,
title = {NURBS-Divergence-Meshless (NDM) method},
author = {Rasin, N. and Hirol, H. and Zainal Abidin, A. R. and
Mokhtaram, M. H. and Mohd Noor, M. A. and Mohd Yassin, A. Y.},
journal = {Computers and Mathematics with Applications},
volume = {196},
pages = {151--171},
year = {2025},
doi = {10.1016/j.camwa.2025.07.011}
}
Keywords: divergence theorem · finite volume · isogeometric analysis · meshfree formulation · meshless methods · NDM method · NURBS
Recent Comments