MATLAB Code for Accelerated Lifetime Prediction of Oil-Aged 3D-Printed Dual-Curable Silica/XSBR Rubber Composites

Description

This code-only record provides the MATLAB implementation used for accelerated lifetime prediction of oil-aged, 3D-printed dual-curable silica/XSBR rubber composites. The workflow fits reference-temperature response data with a power-law model, estimates a temperature acceleration factor by horizontal time shifting and pooled profile fitting, and predicts times to specified response thresholds. Both a deterministic bounded optimizer and a fixed-seed Artificial Bee Colony implementation are provided, together with input validation, diagnostic plotting, run-manifest generation, and synthetic numerical tests. Experimental raw data, spreadsheets, precomputed results, and manuscript figures are intentionally excluded from this repository.

Steps to reproduce

Method and reproducibility notes Deterministic workflow The primary workflow minimizes a one-dimensional profile objective using `fminbnd` on the logarithm of the acceleration factor. Optimization in log space enforces positive factors and gives balanced numerical resolution across the allowed interval. Artificial Bee Colony option `abc_optimize_1d.m` preserves the Artificial Bee Colony approach used during method development. It is supplied as an optional sensitivity/provenance implementation. A fixed seed is used by default, but stochastic estimates should still be checked against the deterministic bounded optimum. Boundary diagnostics The default acceleration-factor interval is 1–50. Results close to either endpoint are labelled lower-bound or upper-bound limited. Such values indicate that the available data do not resolve an interior optimum within the imposed interval. Extrapolation Threshold lifetimes are analytical extrapolations from the fitted reference-temperature power law. Predictions far beyond the observation period should be reported with suitable uncertainty and sensitivity analysis rather than as directly observed service lives. Software validation The included tests use generated synthetic power-law observations with a known acceleration factor. They verify parameter recovery, threshold ordering, output creation, and manifest generation without using article data.

Institutions

• Fujian Institute of Research on the Structure of Matter

  Fujian, Fuzhou
• University of Chinese Academy of Sciences

  Beijing, Beijing

Categories

Licence