EvoProto: Python Software Companion for Evidence-Gated, Evolution-Informed Engineering Design

Abstract

This software deposit provides the English-language computational companion to EvoProto, a Python reference implementation of an evidence-gated framework for evolution-informed engineering design. It accompanies the manuscript “From Evolutionary Biology Data to Technological Prototypes: An AI-Driven Engineering Design Framework with an Evidence-Gated Validation Protocol.” The project investigates how evolutionary biological evidence can inform the generation and refinement of technological prototypes through traceable design constraints, explicit uncertainty assessment, and engineer-supervised decision-making. Its purpose is to support methodological inspection and computational experimentation rather than to claim experimentally demonstrated improvements in engineering performance. Art-2.docx The computational walkthrough covers provenance-aware data handling, an evolutionary–engineering knowledge graph, phylogenetic signal and convergence analysis, biological analog ranking, parametric design exploration, and constrained multi-objective optimization. The knowledge graph links organisms, environments, traits, functions, physical mechanisms, and engineering parameters, while representing homology, convergence, trade-offs, and ancestral-state changes as distinct relations. Computational components include Blomberg’s K, Pagel’s λ, Stayton’s C1, an NSGA-II-style search with constraint domination, and metrics for Pareto-front quality and design diversity. Together, these components expose the methodological path from biological evidence to candidate engineering solutions. A central feature is an evidence-gated decision mechanism that combines the support available for a biological-to-engineering traceability chain with the uncertainty of transferring a mechanism across scales, materials, and loading conditions. The decision logic distinguishes RECOMMEND, ABSTAIN, and REJECT outcomes. Insufficient evidence or excessive transfer uncertainty produces an abstention report identifying the relevant evidence gap or uncertainty component; failed numerical verification produces rejection. Recommendations remain subject to final engineer approval. Art-2.docx The principal worked example is an analytically modeled end-of-arm-tool bracket intended for additive manufacturing. It supports exploration of lightweight structural designs within a constrained parametric space. A synthetic three-arm demonstration compares a template-based stand-in for expert biomimetic design, a uniform-prior generative baseline, and a search variant using a manually specified analog-shaped prior. The associated protocol supports feasibility assessment, hypervolume and diversity analysis, iterations-to-specification reporting, and statistical comparisons. These demonstrations exercise the computational workflow; they do not constitute evidence that evolutionary context improves real-world design outcomes. The archive contains an English-language Jupyter/Google Colab notebook, an HTML reading copy, a linear Python export, setup and reproducibility documentation, citation metadata, support-function tests, validation reports, and SHA-256 checksums. The core evoproto package is an external dependency and is not bundled. No biological corpus, trained AI models, physical-test measurements, or freshly generated scientific results are included. Notebook outputs are cleared, and complete scientific execution in hosted Google Colab was not verified during archive preparation. The release is intended for research, peer review, teaching, and further development of transparent, evidence-aware engineering design workflows. Keywords: research software; Python; biologically inspired design; evolutionary data; knowledge graphs; phylogenetic analysis; generative design; multi-objective optimization; evidence gating; transfer uncertainty; additive manufacturing; computational reproducibility. Resource type: SoftwareVersion: 0.1.0-en.1Language: English