_version_ 1866901698328395776
author Srebranig, Steven
author_facet Srebranig, Steven
contents <p><strong>Nonparametric Simulation Input Modeling, Drift Stability, and Low-Sample Optimization: A Unified Framework</strong><br><strong>Version 1.1 (2025)</strong></p> <p>This whitepaper presents a unified, end-to-end framework for simulation under uncertainty, designed for real industrial environments where data are incomplete, drifting, or historically constrained. The workflow integrates empirical distribution reconstruction, autocorrelation-based drift diagnostics, efficient experimental design (DOE), surrogate modeling, and constrained optimization into a coherent methodology suitable for both legacy systems and modern digital-twin toolchains.</p> <p>The framework emphasizes <strong>lightweight, interpretable, and reproducible methods</strong>—including histogram-only CDF reconstruction, small-sample drift detection, 2×2 factorial DOE, quadratic surrogate modeling, and linear programming with simulation-based validation. These techniques require no complex statistical assumptions and are easily implemented in Python (pandas, statsmodels, Pyomo), Excel, or legacy simulation environments.</p> <p>The appendices provide full numerical worked examples:</p> <ul> <li> <p><strong>Appendix A:</strong> Empirical CDF reconstruction from raw samples and histogram bins, including exponential tail extension.</p> </li> <li> <p><strong>Appendix B:</strong> Autocorrelation analysis, correlogram interpretation, and independence restoration through subsampling.</p> </li> <li> <p><strong>Appendix C:</strong> 2×2 factorial DOE with hand-computed effects and ANOVA decomposition.</p> </li> <li> <p><strong>Appendix D:</strong> Surrogate-model-based optimization for a five-product manufacturing system, including simulation validation.</p> </li> <li> <p><strong>Appendix E:</strong> Archival SIMAN and MathCAD code from the original 1994–95 study.</p> </li> </ul> <p>Although the earliest version of this methodology was developed in 1994 under tight data and compute constraints, the structure aligns naturally with 2025 practices in <strong>drift-aware ML simulation, digital-twin modeling, predictive maintenance, and robust industrial scheduling</strong>. The result is a practical and durable framework that remains relevant across decades, toolchains, and application domains.</p>
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_17731303
institution Zenodo
language
publishDate 2025
publisher Zenodo
record_format zenodo
spellingShingle Nonparametric Simulation Input Modeling, Drift Stability, and Low-Sample Optimization: A Unified Framework
Srebranig, Steven
Simulation, Input Modeling, Empirical Distribution Reconstruction, Histogram CDF, Tail Modeling, Autocorrelation, Drift Detection, Independence Testing, Correlogram, Subsampling, Design of Experiments (DOE), Factorial Design, ANOVA, Surrogate Modeling, Quadratic Regression, Response Surface Modeling, Optimization, Linear Programming, Monte Carlo Validation, Digital Twin, Predictive Maintenance, Nonparametric Methods, Simulation Optimization, Industrial Engineering, Operations Research, Stochastic Systems, Low-Sample Modeling, Manufacturing Systems, Simulation Drift, Data Scarcity, Python Implementation, pandas, statsmodels, Pyomo
<p><strong>Nonparametric Simulation Input Modeling, Drift Stability, and Low-Sample Optimization: A Unified Framework</strong><br><strong>Version 1.1 (2025)</strong></p> <p>This whitepaper presents a unified, end-to-end framework for simulation under uncertainty, designed for real industrial environments where data are incomplete, drifting, or historically constrained. The workflow integrates empirical distribution reconstruction, autocorrelation-based drift diagnostics, efficient experimental design (DOE), surrogate modeling, and constrained optimization into a coherent methodology suitable for both legacy systems and modern digital-twin toolchains.</p> <p>The framework emphasizes <strong>lightweight, interpretable, and reproducible methods</strong>—including histogram-only CDF reconstruction, small-sample drift detection, 2×2 factorial DOE, quadratic surrogate modeling, and linear programming with simulation-based validation. These techniques require no complex statistical assumptions and are easily implemented in Python (pandas, statsmodels, Pyomo), Excel, or legacy simulation environments.</p> <p>The appendices provide full numerical worked examples:</p> <ul> <li> <p><strong>Appendix A:</strong> Empirical CDF reconstruction from raw samples and histogram bins, including exponential tail extension.</p> </li> <li> <p><strong>Appendix B:</strong> Autocorrelation analysis, correlogram interpretation, and independence restoration through subsampling.</p> </li> <li> <p><strong>Appendix C:</strong> 2×2 factorial DOE with hand-computed effects and ANOVA decomposition.</p> </li> <li> <p><strong>Appendix D:</strong> Surrogate-model-based optimization for a five-product manufacturing system, including simulation validation.</p> </li> <li> <p><strong>Appendix E:</strong> Archival SIMAN and MathCAD code from the original 1994–95 study.</p> </li> </ul> <p>Although the earliest version of this methodology was developed in 1994 under tight data and compute constraints, the structure aligns naturally with 2025 practices in <strong>drift-aware ML simulation, digital-twin modeling, predictive maintenance, and robust industrial scheduling</strong>. The result is a practical and durable framework that remains relevant across decades, toolchains, and application domains.</p>
title Nonparametric Simulation Input Modeling, Drift Stability, and Low-Sample Optimization: A Unified Framework
topic Simulation, Input Modeling, Empirical Distribution Reconstruction, Histogram CDF, Tail Modeling, Autocorrelation, Drift Detection, Independence Testing, Correlogram, Subsampling, Design of Experiments (DOE), Factorial Design, ANOVA, Surrogate Modeling, Quadratic Regression, Response Surface Modeling, Optimization, Linear Programming, Monte Carlo Validation, Digital Twin, Predictive Maintenance, Nonparametric Methods, Simulation Optimization, Industrial Engineering, Operations Research, Stochastic Systems, Low-Sample Modeling, Manufacturing Systems, Simulation Drift, Data Scarcity, Python Implementation, pandas, statsmodels, Pyomo
url https://doi.org/10.5281/zenodo.17731303