Physical Intuition

To master the Finite Element Method (FEM) and Finite Element Analysis (FEA), we must establish a solid foundation using the simplest yet most physically revealing elements: the one-dimensional (1D) spring and axial bar element. Real-world structures—whether they are bridge trusses, aircraft wings, or engine crankshafts—deform in three dimensions. However, their complex behaviors are governed by the same mathematical and physical principles that we can observe in a simple 1D axial member. By stripping away geometric complexities, we focus purely on the core mechanics: discretization, linear interpolation, local vs. global coordinate systems, energy minimization, and global system assembly.

In this chapter, we transition from continuous systems—where physical fields (like displacements, strains, and stresses) are described by differential equations at every single point—to discrete systems, where the behavior is approximated at a finite number of points called nodes. We will explore how shape functions act as a mathematical bridge between discrete nodal values and the continuous space between them. We will then derive the stiffness matrix using both energy principles and weighted residual methods, and finally, we will solve a multi-step axial assembly under complex loading conditions.

Core Concepts

1. Continuous vs. Discrete Systems: The Discretization Paradigm

In classical solid mechanics, a bar is treated as a continuous elastic body. If you apply a force at the end of a bar, every infinitesimal slice of the bar deforms. The displacement field $u(x)$ is continuous, meaning it has an infinite number of degrees of freedom (DOFs) because we need an infinite number of values to describe the displacement of every single point along the length. The exact behavior is governed by the strong-form differential equation:

where $A$ is the cross-sectional area, $E$ is Young's Modulus, and $q(x)$ is the distributed axial load.

For simple geometries and uniform loads, finding an analytical solution to this differential equation is straightforward. However, when the cross-sectional area $A$ varies arbitrarily, the material properties $E$ change along the length (e.g., in functionally graded materials), or multiple point loads are applied, finding a closed-form analytical solution becomes mathematically intractable.

The Finite Element Method resolves this by discretizing the continuous bar into a finite number of sub-domains called elements. Instead of solving for the displacement at an infinite number of points, we solve for the displacements at a finite number of discrete points called nodes. The displacement within each element is then approximated by interpolating the values at its boundary nodes. Through this discretization, we convert a differential equation (infinitely many variables) into a system of algebraic equations (finite number of variables):

where $[K]$ is the global stiffness matrix, $\{U\}$ is the vector of nodal displacements, and $\{F\}$ is the vector of nodal forces.

2. Linear Interpolation and Shape Functions

Since we only calculate displacements at the nodes, we need a way to estimate the displacement at an arbitrary point $x$ inside an element. We use interpolation functions, commonly referred to as shape functions.

For a 1D bar element defined by two nodes (local Node 1 at $x=0$ and local Node 2 at $x=L$), the simplest approximation is a linear variation of displacement. The displacement field $u(x)$ within the element is represented as:

where $u_1$ and $u_2$ are the axial displacements at Node 1 and Node 2, respectively, and $N_1(x)$ and $N_2(x)$ are the shape functions.

Shape functions are not arbitrary; they must satisfy strict mathematical and physical criteria:

For our 2-node element:

This property ensures that when we evaluate the displacement field $u(x)$ exactly at Node 1 ($x=0$), we get $u(0) = N_1(0)u_1 + N_2(0)u_2 = 1 \cdot u_1 + 0 \cdot u_2 = u_1$.

If this property is violated, the element cannot represent rigid body motion. If we translate the entire element by a constant displacement $u_0$ (i.e., $u_1 = u_2 = u_0$), the displacement at any point inside must be $u(x) = u_0$. Since $u(x) = (N_1(x) + N_2(x)) u_0$, we must have $N_1(x) + N_2(x) = 1$. If $N_1(x) + N_2(x) \neq 1$, the interpolation would falsely predict that a rigid translation induces strains and stresses in the material.

3. Local vs. Global Coordinate Systems and Node Mapping

In structural analysis, elements can be oriented in various directions. To make the calculations manageable, we define two coordinate systems:

For a 1D assembly where all bars lie along the same straight line, the local and global coordinates are aligned. However, we still need a mapping mechanism to assemble local element matrices into the global system. This is done using an element connectivity table (or mapping matrix). For example, Element 3 might connect global Node 3 to global Node 4. Its local Node 1 maps to global Node 3, and its local Node 2 maps to global Node 4. This mapping ensures that internal forces and displacements are balanced at the shared joints.

4. The Axial Bar Stiffness Formulation: Analogy to a Spring

Let us relate the physical behavior of a solid bar to a simple mechanical spring. A spring is characterized by its stiffness constant $k$, which relates force $F$ to elongation $\delta$:

Now consider an elastic bar of length $L$, cross-sectional area $A$, and Young's Modulus $E$, fixed at one end and pulled with an axial force $F$ at the other. From classical strength of materials:

Equating the stress expressions:

Comparing this to the spring equation $F = k \delta$, we see that a bar behaves exactly like a spring with an equivalent axial stiffness:

The stiffness depends on:

5. Assemblies of Springs: Series vs. Parallel Networks

When modeling complex structures, individual elements are connected in networks. We can evaluate their combined behavior using series and parallel spring equivalents:

Springs in Series (End-to-End Connection)

Consider two springs with stiffnesses $k_1$ and $k_2$ connected in series under an axial load $F$.

Using Hooke's Law:

Since $F = k_{eq} \delta_{total}$, we have:

In series systems, the equivalent stiffness is always *smaller* than the smallest individual stiffness.

Springs in Parallel (Side-by-Side Connection)

Consider two springs with stiffnesses $k_1$ and $k_2$ connected in parallel, sharing the same load $F$ and undergoing the same displacement $\delta$.

Using Hooke's Law:

Comparing to $F = k_{eq} \delta$, we get:

In parallel systems, stiffnesses add up directly.

6. Physical Interpretation of Stiffness Coefficients

A key concept in FEA is the physical meaning of the elements within a stiffness matrix. For a system of equations $[K]\{U\} = \{F\}$, the entry $K_{ij}$ in the stiffness matrix represents the reaction force at Node $i$ due to a unit displacement applied at Node $j$, while all other nodes are held fixed (zero displacement).

For a single bar element:

This physical interpretation is universal across all structural finite elements, including 2D plates, shells, and 3D solids.

7. Essential vs. Natural Boundary Conditions

A stiffness matrix $[K]$ assembled for a structure without any supports is singular (its determinant is $|K| = 0$). Mathematically, this means the matrix cannot be inverted. Physically, it indicates that the structure is free to move as a rigid body without any resistance. To solve for displacements, we must apply constraints to eliminate rigid body modes.

Boundary conditions in FEM are classified into two types: