Deposit-resolved Taguchi design, kinetics, and spectroscopy datasets for fulvic-acid recovery from coal gangue via H₂O₂ oxidation

Description

This dataset contains the experimental design matrix, replicate fulvic-acid (FA) yields, kinetic time-series tables, Arrhenius-derived temperature dependence, and supporting characterization summaries for deposit-resolved FA recovery from four coal-gangue sources (GLC, NEC, SGC, SMC) using single-oxidant H₂O₂ oxidation. The study compares deposit-specific oxidation windows using a Taguchi L9(3⁴) orthogonal design with confirmation runs, links yield to acidity and surface functionalization, and reports pseudo-first-order kinetic parameters and apparent activation energies across 40–80 °C. The files enable independent verification of yield optima (including the order-of-magnitude yield contrast across deposits), reproduction of kinetic/Arrhenius calculations, and reuse of derived structural indices (fluorescence, UV–vis, XPS C1s fractions) for meta-analysis and process–structure modeling. This dataset contains design matrices/yield tables, kinetics/Arrhenius tables (where provided), response-surface modeling outputs (coefficients, diagnostics, OriginPro NLFit reports/templates), and supporting documentation used in a coal-gangue → fulvic-acid (FA) recovery study based on single-oxidant H₂O₂ oxidation. Files are organized into data and report folders for direct reuse and verification of modeling/optimization steps.

Steps to reproduce

Load the Taguchi L9 replicate dataset (screening stage). Open data/01_design_yields/CGFA_L9_design_replicates_long.csv. This table contains the Taguchi L9(3⁴) factor settings and replicate FA yields for each gangue (n = 27 per gangue = 9 runs × 3 replicates). In this deposit, Run = 1–9 indexes the L9 array (corresponding to the L9 run block reported in the manuscript). Compute run means (and confirm consistency with exported means). Group by Deposit and Run and compute the mean and SD of Yield_pct. Compare to the provided summaries: data/01_design_yields/CGFA_L9_design_replicates_summary_by_run.csv (mean/SD/min/max per run) data/01_design_yields/CGFA_L9_design_replicates_mean_sd_wide.csv (wide mean ± SD) A QC comparison against the workbook mean table is provided in docs/QC_L9_means_compare.csv (differences are rounding-level). Fit the reduced quadratic regression model (Eq. 5; no interaction terms). For each gangue, fit the reduced quadratic model used in the manuscript: Y = β0 + βT·T + βR·R + βC·C + βD·D + βTT·T² + βRR·R² + βCC·C² + βDD·D² (Eq. 5) where Y is FA yield (%), T is temperature (°C), R is H₂O₂/CG mass ratio, C is H₂O₂ concentration (wt%), and D is time (min). (Templates and the OriginPro NLFit setup are in data/04_models/OriginPro_NLFit_Template_RSM.xlsx; the compiled OriginPro reports are in reports/RSM_NLFit_quadratic_reports_GLC_NEC_SGC_SMC.docx.) Reproduce model adequacy metrics exactly as defined. Compute the residual sum of squares RSS = Σ(yi − ŷi)², residual variance s² = RSS/(n − p), and RMSE = √(RSS/(n − p)), with p = 5 for the linear model and p = 9 for the reduced quadratic model (Eq. 5). Compare your outputs to data/04_models/RSM_fit_diagnostics_and_Ftests.csv. Reproduce the curvature test (linear vs reduced quadratic). Compute the residual-variance F statistic exactly as in the manuscript: Fcal = s²linear / s²quadratic, with df1 = n − 5 = 22 and df2 = n − 9 = 18. Confirm significance using the reported p-value / Fcrit values in data/04_models/RSM_fit_diagnostics_and_Ftests.csv. Coefficient estimates and term p-values are in data/04_models/RSM_quadratic_coefficients_by_gangue.csv. Kinetic/Arrhenius reproduction (requires the kinetic tables). The manuscript’s pseudo-first-order kinetic fitting uses a fixed sampling grid ti = 0, 5, 10, 20, 30, 45, 60, 90, 120, 150, 180, 210 min (see docs/Temperature_time_sampling_grid_for_kinetic_runs.docx). To fully reproduce the kinetic results and Fig. 6, you must also use the kinetic tables (Run, T, t, Y, X, k) referenced in the manuscript supplement; then: compute conversion X from yield (as defined in your workflow), estimate k by nonlinear least-squares fitting of the integrated PFO form (equivalently X = 1 − exp(−k t) or ln[1/(1−X)] = k t), compile kapp(T) across temperature levels and fit Arrhenius using ln(k) vs 1/T to obtain Ea and reproduce Fig. 6.

Institutions

• Shanxi University

Categories

Funders

• Professor Gao Lizhen （Shanxi Provincial Department of Education）

  Grant ID: YDZJSX20231B002

Licence