From Punch Cards to the AI Age: How Modern Tools Breathed New Life into a 1972 FORTRAN Code for Curved Pipe Flow

Decades-old scientific code rarely survives the march of technology — but what if it didn't have to stay buried? This paper documents a compelling act of numerical archaeology, using modern AI tools to resurrect a 1972 FORTRAN 66 program and unlock enduring insights into laminar fluid flow.

Laminar flow in curved pipes is a foundational topic in fluid dynamics, shaped by Dean's pioneering theoretical work in the late 1920s and extended by subsequent researchers through increasingly sophisticated numerical solutions. Despite this rich history, the legacy computational codes underpinning many of these studies have remained largely inaccessible — trapped in obsolete formats and deteriorating archives. A recent paper by Nils Tångefjord Basse, published in Discover Mechanical Engineering, directly addresses this problem by reviving a 1972 FORTRAN 66 code for simulating laminar flow in curved pipes and bringing it up to modern Fortran standards.

Basse employed a suite of artificial intelligence tools — including Amazon Textract for optical character recognition, ChatGPT-4 for code conversion, and GitHub Copilot for refactoring — to digitize, translate, and debug the original code. This methodological approach not only preserves historically significant scientific software but demonstrates a replicable workflow for the broader research community. The effort produced two distinct deliverables: a minimal modern Fortran version faithfully replicating the original output, and a fully modernized version incorporating modules, structured loops, double-precision arithmetic, and dynamic file handling.

The revived code successfully reproduced established results, including the characteristic Dean vortex structures and their influence on streamwise flow. Doubling the computational grid resolution further refined predictions for key flow parameters, bringing them into closer agreement with the more accurate finite difference scheme of Collins and Dennis (1975).

The work carries clear value for both education and research. By making the updated code openly available, Basse provides a fast, transparent tool for studying canonical laminar flows in curved pipes. The paper also stands as a compelling example of what the author calls "numerical archaeology" — recovering and modernizing classical solutions to ensure their continued relevance in contemporary fluid dynamics.