Solving PDEs with the Finite Element Method and Meshing
This article explores the Finite Element Method (FEM), a numerical technique used to predict how complex structures react to physical forces. It covers why traditional partial differential equations (PDEs) cannot be solved directly on irregular geometries, how the meshing process partitions these structures into manageable elements, and how localized equations are reassembled to calculate an accurate global solution.
The Challenge of Partial Differential Equations
In physics and engineering, continuous phenomena like heat transfer, fluid flow, and structural stress are governed by partial differential equations (PDEs). These equations describe how physical quantities change across space and time.
While it is possible to solve PDEs analytically for simple shapes like a basic sphere or a flat plate, real-world engineering structures—such as an airplane wing, a car chassis, or a suspension bridge—have highly irregular geometries. Because a continuous mathematical formula cannot easily account for every curve, hole, and material change in a complex object, finding an exact analytical solution is impossible. Instead, engineers rely on the Finite Element Method to find a highly accurate approximation.
Discretization: Building the Mesh
The core philosophy of the Finite Element Method is "divide and conquer," known mathematically as discretization. To bypass the impossibility of solving a PDE across a complex structure all at once, FEM partitions the continuous object into thousands or millions of smaller, simpler geometric shapes.
- Finite Elements: These are the individual building blocks of the partitioned structure. In two-dimensional models, elements are typically triangles or quadrilaterals. In three-dimensional models, they are usually tetrahedrons (four-sided pyramids) or hexahedrons ( six-sided bricks).
- Nodes: Elements are connected to one another at specific points called nodes, usually located at the corners or midpoints of the shapes.
- The Mesh: The complete network of connected elements and nodes is called a mesh. It acts as a simplified geometric approximation of the original engineering structure.
By increasing the density of the mesh—using smaller and more numerous elements—the approximation becomes closer to the true geometry of the structure, especially around complex curves or high-stress areas.
From Calculus to Algebra: Shape Functions
Once the mesh is generated, the impossible task of solving a complex PDE over a weird shape is replaced by the manageable task of solving simplified equations over basic triangles or bricks.
Within each individual finite element, the complex physics are approximated using simple polynomial equations called shape functions (or basis functions). These functions interpolate how a physical quantity, such as temperature or displacement, varies between the nodes of that specific element. Because the geometry of a single element is uniform and simple, the calculus of the PDE can be mathematically transformed into a set of basic algebraic equations for that single piece.
Global Assembly and Solving the Matrix
Solving the equations for a single isolated element does not tell the engineer how the whole structure behaves. Because the elements share nodes, a force applied to one element directly affects its neighbors.
To find the solution for the entire structure, the FEM software performs a process called global assembly. It combines the local algebraic equations from every individual element into one massive system of simultaneous equations. This system is represented as a large matrix, often called the global stiffness matrix.
The equation generally takes the form of \([K]\{u\} = \{F\}\), where:
- \([K]\) is the global matrix representing the material properties and geometry of the connected mesh.
- \(\{u\}\) is the unknown vector of values (such as displacement or temperature) at every node.
- \(\{F\}\) is the vector of known external forces and boundary conditions applied to the structure (like gravity, wind load, or fixed supports).
Using advanced linear algebra solvers, a computer processes this matrix to calculate the exact values at every single node. Once the nodal values are known, the shape functions are used one last time to map the data across the elements, providing a complete, continuous visualization of how the entire complex structure responds to the applied forces.