Dataset and MATLAB analysis scripts for surface roughness and tool wear study in C45 steel turning

Description

Experimental dataset and MATLAB analysis scripts supporting the study: Surface Roughness Prediction and Tool Wear Analysis in Steel Turning Using Machine Learning and Bayesian Optimisation. Includes: (1) DOE measurements (N=40, Rz and Ra), (2) time-series Rz data (424 measurements at 6 cutting speeds), (3) flank wear VB data (42 measurements), and (4) complete MATLAB pipeline (S0-S7) for DOE analysis, ML modelling, Bayesian optimisation, and tool wear analysis.

Steps to reproduce

EXPERIMENTAL DATA COLLECTION Surface roughness (Rz, Ra) was measured using a Mitutoyo Surftest SJ-301 contact profilometer (ISO 4287, evaluation length lm = 12.5 mm, cut-off lambda_c = 2.5 mm, Gaussian filter). Workpiece material: C45 steel (EN 10083-2, AISI 1045), cylinder diameter 60 mm. Cutting tool: KNUX 190408-EL P20 insert (ISO 1832) in MWLNR holder, nose radius r_eps = 0.80 mm. Machine: centre lathe MUS 25 (SU 50). All experiments dry (no cutting fluid). DOE experiment: 2^3 full factorial design, factors: cutting speed vc (8.792 / 351.680 m/min), feed f (0.10 / 0.50 mm), depth of cut ap (0.10 / 3.00 mm), 5 replicates per run, N = 40 total measurements (sheet: 5_DOE_RzRa). Time-series experiment: fixed f = 0.10 mm, ap = 0.20 mm, 6 cutting speeds vc = {6, 25, 70, 110, 140, 240} m/min. Rz measured after each pass (424 records, sheet: 2_Rz_TimeSeries). Flank wear VB measured optically at 20x magnification after each pass, ISO 3685, VBcrit = 0.30 mm (42 records, sheet: 3_VB_TimeSeries). DATA ANALYSIS — MATLAB SCRIPTS Requirements: MATLAB R2021b or later, Statistics and Machine Learning Toolbox, Global Optimization Toolbox. Run scripts in this order: 1. S0_LoadData.m — loads DATA_Final.xlsx, saves DATA.mat 2. S1_DOE_Analysis.m — ANOVA, effects, regression models 3. S2_DOE_ML.m — GPR, ANN, SVR with LOO-CV (~5 min) 4. S3_DOE_Optimization.m — Bayesian opt., GA, Pareto (~5 min) 5. S4_Rz_TimeSeries.m — exponential model Rz = A*exp(b*tau) 6. S5_VB_Analysis.m — power-law VB, Taylor equation 7. S6_Rz_VB_Correlation.m — VB-Rz correlation analysis 8. S7_Figures.m — composite publication figures All figures saved as PDF (vector) and PNG (300 DPI) in subfolder 'figures/'. Random seed fixed (rng(42)) for reproducibility. Total runtime ~15-20 minutes. NOTE: vc = 25 m/min excluded from Taylor analysis (R2 = -0.34, built-up edge regime). vc = 140 m/min excluded (experiment ended at VB = 0.15 mm < VBcrit).

Institutions

• University of West Bohemia in Pilsen

  Plzeň Region, Pilsen

Categories

Licence