CHAPTER 1: INTRODUCTION TO COMPUTATIONAL TECHNIQUES

1.1 INTRODUCTION

Physical problems in engineering are inherently represented via complex mathematical equations. For instance, the general differential equation governing the axial deformation \( u(x) \) of a bar under distributed load \( q(x) \) is given by:

The accurate solution of these complex engineering problems has always been the primary aim of engineers and scientists. Computational Techniques are methods used to find approximate mathematical solutions to complex physical problems, typically governed by such higher-order partial differential equations (PDEs), using the processing power of computers.

Necessity in Civil Engineering

In Civil Engineering, structures and systems are rarely simple. The necessity of computational techniques arises from the following limitations of classical analytical methods:

• Complex Geometry: Real-world structures (like arch dams, irregular building footprints, and bridge nodes) have complex shapes that cannot be represented by simple mathematical domains.
• Complex Boundary Conditions: Supports are rarely purely "fixed" or "pinned." Soil-structure interaction requires modeling semi-infinite domains and complex, yielding support conditions.
• Non-linear Material Properties: Materials like concrete, soil, and yielding steel do not obey linear Hooke's law under all stress levels. Analytical methods fail easily when material non-linearity is introduced.
• Dynamic & Arbitrary Loadings: Earthquakes, wind gusts, and moving vehicle loads are dynamic and arbitrary, making closed-form mathematical integration nearly impossible.

1.4 VARIOUS SOLUTION TECHNIQUES

The formulation for structural and civil engineering analysis is based on three fundamental mechanics relations: equilibrium, constitutive (material behavior), and compatibility (continuity). Solution approaches are broadly divided into two categories:

1. Analytical (Exact) Approach: Leads to closed-form equations. Effective only for simple geometries and linear elastic materials.
2. Numerical (Approximate) Approach: Used when geometries and conditions are complex. It discretizes the problem to find approximate values at distinct points.

Types of Numerical Methods

1. Finite Element Method (FEM)

Concept & Working: FEM divides a complex continuous domain into finite, smaller, and simpler parts called "elements". The governing differential equations are approximated over each element to formulate an element stiffness matrix \([k]\), relating nodal forces \(\{f\}\) to nodal displacements \(\{d\}\):

Suitability: Highly suited for structural mechanics, solid mechanics, heat transfer, and problems involving highly complex geometries and mixed boundary conditions.

Merits: Handles complex geometries easily; easily accommodates varied boundary conditions and material non-linearities.

Demerits: Requires substantial computational memory. It struggles with extreme deformations where the mesh becomes excessively distorted.

2. Finite Difference Method (FDM)

Concept & Working: FDM replaces the continuous partial derivatives in the governing differential equations directly with algebraic difference quotients evaluated at discrete grid points. Based on Taylor Series expansion, the first derivative is approximated as:

Suitability: Fluid dynamics, groundwater seepage analysis, and heat conduction.

Merits: Very simple to mathematically formulate and code. Highly efficient for regular, structured grids.

Demerits: Struggles heavily to accurately model complex, curved, or irregular boundaries.

3. Boundary Element Method (BEM)

Concept & Working: BEM mathematically transforms the governing PDEs defined over a volume domain \(\Omega\) into integral equations defined only on the boundary \(\Gamma\). For example, a potential problem is transformed to:

Suitability: Infinite or semi-infinite domains (e.g., soil masses around a deep foundation), acoustics, and fracture mechanics.

Merits: Reduces problem dimensionality by one (a 3D volume problem becomes a 2D surface problem).

Demerits: The underlying mathematics are very dense. Resulting system matrices are fully populated and non-symmetric.

4. Discrete Element Method (DEM)

Concept & Working: A mesh-free method computing the motion and effect of discrete interacting particles. It solves Newton's equations of motion for every single particle independently by considering contact forces and collisions.

Suitability: Granular materials, rock mechanics (highly jointed rocks), powder dynamics, and landslides where continuum mechanics fail.

5. Smoothed Particle Hydrodynamics (SPH)

Concept & Working: A mesh-free Lagrangian computational method where a continuous fluid or solid domain is represented by a set of discrete particles. These particles possess material properties and interact within a defined "smoothing length" (\( h \)) using a kernel function.

Suitability: Fluid dynamics with highly deforming boundaries (e.g., dam break, wave impact), and hyper-velocity impacts.

DEEP DIVE: FINITE ELEMENT METHOD (FEM)

Basic Steps / Algorithm in Finite Element Analysis

1. Discretization (Meshing): The continuous physical domain is divided into elements connected at nodes.
2. Selection of Displacement/Shape Functions: A simple mathematical approximation is assumed to represent the variation of the displacement field \( u(x) \) within each element using shape functions \( N_i \):
   $$ u(x) \approx \sum_{i=1}^{n} N_i(x) d_i $$
3. Derivation of Element Stiffness Matrix & Equations: The relationship between local nodal forces and displacements is established: \([k^e]\{d^e\} = \{f^e\}\).
4. Assembly of Global Equations: The individual matrices are topologically assembled to form the global stiffness matrix and global force vector for the entire structure:
   $$ [K] = \sum_{e=1}^{E} [k^e] \quad \Rightarrow \quad [K]\{D\} = \{F\} $$
5. Application of Boundary Conditions: Supports are applied to the global system to prevent rigid body motion and make the global matrix \([K]\) non-singular (invertible).
6. Solution for Primary Unknowns: The resulting system of simultaneous linear algebraic equations is solved (using techniques like Gaussian elimination) to find the unknown nodal displacements \(\{D\}\).
7. Post-processing: Secondary quantities like element strains \( \varepsilon \) and stresses \( \sigma \) are computed:
   $$ \sigma = E \cdot [B] \{d^e\} $$

DISCRETIZATION AND TYPES OF ELEMENTS

What is Discretization?

Discretization is the foundational first step of FEM. It is the process of dividing a complex, continuous physical domain into an equivalent mathematical system consisting of smaller, discrete, and finite parts called elements. These elements interconnect only at specified points called nodes.

Example Problem: Discretizing a Plate with a Circular Hole

(Addressing the specific exam question 2074 Magh Q4b regarding the square plate with a central hole under horizontal tension 'P').

Step-by-step strategy for discretizing the plate with a hole:

1. Exploiting Symmetry: Because the geometry, the circular hole, and the tension load (P) are perfectly symmetrical about both the X and Y axes, we only need to model and analyze one quarter of the plate.
2. Applying Boundary Conditions: On the cut symmetry lines, we apply roller supports.
3. Variable Mesh Density (Finer near the hole): To capture rapid stress gradients (stress concentration) accurately, we use a finer mesh (smaller elements) immediately around the hole.
4. Coarse Mesh away from hole: As we move towards the outer boundary where stress distribution is uniform, we smoothly transition to a coarser mesh to save computational memory.