Working Through Budnick's Numerical Methods
I ran into trouble trying to solve a boundary value problem for a heat equation on an irregular domain. Budnick's Applied Mathematics By Frank S Budnick walks through finite difference discretization, but the treatment assumes uniform grids. My domain was a tapered beam, and the standard approach blew up near the edges. I ended up switching to a transformed coordinate system that mapped the irregular geometry onto a rectangular grid, then reapplied the difference scheme from chapter 8. It took about three extra hours of setup, but it converged in under a minute instead of oscillating. The book is structured around numerical algorithms with engineering examples, not abstract proofs. You'll find derivations of error bounds, but they're usually tucked into appendices. The main text focuses on implementation. For instance, when covering Newton's method for nonlinear systems, it shows a spreadsheet-friendly iteration table rather than a rigorous convergence theorem. That makes it practical for quick engineering calculations, but it skips the edge cases where the method stalls. One thing beginners miss is that Budnick often presents algorithms as ready-to-code recipes, but many require scaling or preconditioning for real-world inputs. The section on matrix inversion, for example, works fine for 5x5 systems with moderate condition numbers. Try it on a 50x50 stiffness matrix from a structural model, and round-off errors dominate unless you normalize the rows first. I learned that the hard way when my eigenvalue routine returned garbage for a simply supported plate under distributed load. The fix was to subtract the mean from each row before inversion, which I found in a footnote on page 214 that most people overlook.
Another counter-intuitive point is that the book's chapters on numerical integration don't stress adaptive quadrature enough. Budnick explains Simpson's rule well, but for integrals with singularities at endpoints, like those in fracture mechanics stress intensity factors, fixed-step methods waste steps or diverge. I use Gauss-Jacobi quadrature instead, which weights the endpoints appropriately. The book mentions it in problem 12-7, but doesn't expand it. That gap cost me two days of debugging a fatigue-life calculation once. Here's a concrete workflow if you're tackling applied problems with this text: start by identifying whether your problem is stiff, ill-conditioned, or has sharp gradients. Then pick the chapter that matches the numerical technique, but don't copy the code verbatim. Test it on a known solution first—like a simple harmonic oscillator for ODEs—and compare the error against the analytical result. If the error grows with time step, you likely need an implicit method or smaller step size. Budnick's examples are concise, so you'll spend more time filling in the gaps than copying them. Common pitfalls include ignoring stability criteria for explicit methods. The heat equation example on page 302 uses a forward-time central-space scheme, which is fine for demonstration but unstable for large Fourier numbers. In practice, I switch to the Crank-Nicolson method from page 318, even though it's a bit more code. It cuts computational time from hours to minutes for transient simulations because it allows larger time steps without sacrificing accuracy. Also, the book's treatment of iterative solvers like Jacobi and Gauss-Seidel is overly optimistic about convergence rates. For larger systems, I use successive over-relaxation with an optimal relaxation factor estimated from the spectral radius, which Budnick doesn't derive.
When Budnick's approach falls short, consider supplementing it with more modern references. If you're doing finite element analysis, the book's coverage is superficial; go to Zienkiewicz instead. For spectral methods, Trefethen's work is clearer. The strength of Applied Mathematics By Frank S Budnick is its breadth across classical numerical techniques, but it doesn't dive deep into any single area. That's fine for a reference, but inadequate for research-level applications. I keep a copy on hand for quick look-ups on basis functions or interpolation formulas, but for actual problem-solving, I write custom scripts that adapt the algorithms to my specific constraints. The book gives you the foundation, not the finished tool. Use it to understand the why, then build the how around it.
Get the Full Details
