Mathematical Rigor and Where Approximations Live

Rigor means knowing which steps are exact, which are approximate, and how large the approximation error can be.

Exact, approximate, and empirical

Most engineering answers are built from three kinds of step: exact algebra, controlled approximations (truncated series, linearizations, discretizations), and empirical correlations fit to data. A rigorous result labels every step by kind, because the failure modes differ. Algebra fails by mistake; approximations fail when a small parameter stops being small; correlations fail outside the data they were fit to.

Bounding the approximation, not hoping about it

When we linearize or truncate, we estimate the neglected term rather than assume it vanishes. When we discretize a field, we test that the answer stops moving as the grid refines. The goal is a stated bound: this approximation contributes at most X to the final number. Without that bound, a clean-looking result is just an assertion.

Notation and units discipline

Ambiguous notation hides errors. We fix symbol conventions, carry units through every expression, and reject any intermediate result whose units do not close. Dimensional consistency is cheap and catches a surprising fraction of mistakes before they reach a figure.

Mathematical rigor also means resisting false precision. Reporting a derived quantity to more digits than the inputs justify is itself an error; the significant figures we print are chosen to match the uncertainty, not to look authoritative. A number that reads 3.424 is stated to that precision because the analysis supports it, not because more digits look better. When the rounding, the notation, and the bounded approximations are all in order, the mathematics stops being a place errors can hide and becomes a place a reviewer can trust to be exactly as accurate as it claims.