How we validate against LPILE and RSPile (and why you should demand it)

A pile-analysis tool is only as good as your trust in its numbers. New software has to earn that trust, and the way you earn it in geotechnical engineering is well established: cite your method provenance, and benchmark against the codes the profession already relies on. Here is exactly how PileCalc is validated — and how you can check it yourself.

1. Method provenance

Every model in PileCalc traces to a named, published source — the same sources LPILE and RSPile cite:

• Governing equation: the beam-column on a nonlinear (Winkler) foundation, from COM624P (Wang & Reese, 1993, FHWA-SA-91-048).
• p-y curves: Matlock (1970); Reese, Cox & Koop (1974, 1975); Welch & Reese (1972); API / O'Neill & Murchison (1983); Reese (1997) weak rock.
• Axial & shafts: t-z / q-w load transfer; NAVFAC DM-7.02; FHWA-IF-99-025 (O'Neill & Reese, 1999); Vesić (1977) settlement.

2. Equation-level agreement

The RSPile Laterally Loaded Piles Theory Manual documents the governing equation and every p-y model. We checked the PileCalc formulation against it term by term — the Matlock pu expression, the y₅₀ = 2.5·ε₅₀·b relationship, the Reese sand wedge with its tan⁸ term, the API tanh form, the Reese-Nyman weak-rock reduction. They are identical. RSPile uses finite elements and PileCalc uses finite differences, but that's a discretization choice; the physics is the same.

3. Numerical benchmarks

Where a reference reports numbers, we run PileCalc on the same problem and compare:

A note on honesty: most published verification manuals present results graphically — the second curve is plotted on top of the first to show agreement, with no numbers table. The reference values above for those cases are read off the charts and carry a few-percent reading tolerance. Where an exact answer exists — a closed-form solution or a printed output table — the engine matches it to a fraction of a percent; against chart-read references the criterion is “reproduces the reference curve,” and it does. The slope-stabilization row is the one genuine method-level difference in the published suite, at ~4%.

4. Two independent codes, not one

We deliberately benchmark against both LPILE and RSPile. They're independent implementations — different teams, different numerical methods, even opposite sign conventions for moment and soil reaction. If PileCalc matched only one, you couldn't rule out that it had inherited that program's idiosyncrasies. Matching both, plus closed-form solutions, is much stronger evidence that the solver is right.

What this does and doesn't mean

Validation means the math is faithful to the published methods and reproduces accepted benchmarks. It does not remove engineering judgment. The p-y method itself has limits, soil parameters carry uncertainty, and you remain responsible for reviewing results for design. PileCalc's job is to make every assumption and intermediate value visible so that review is actually possible — and then to get out of your way.