High-performance Cu–Cr–X copper alloys face an inherent trade-off between mechanical strength and electrical conductivity. This study develops a physics-informed, interpretable machine-learning framework for their optimization and inverse design. A literature-derived dataset of 3017 experimental records was compiled, with 2302 records retained after preprocessing and reliability filtering. Metallurgically motivated descriptors, including Aging Severity, Cold Work Index, Cr–Aging interaction, Zener-pinning and Hall–Petch proxies, and source reliability, were incorporated. Random Forest, Extra Trees, XGBoost, and CatBoost were evaluated using a leakage-free 10-fold out-of-fold stacking framework. The stacking model achieved [Formula: see text], MAE [Formula: see text], and RMSE [Formula: see text], compared with [Formula: see text] values of 0.781, 0.802, 0.837, and 0.857, respectively, for the four base learners. SHAP analysis highlighted Cu, Cr, source reliability, aging-related variables, and precipitation-related descriptors as important predictors. Pareto-based inverse screening identified candidates with predicted Strength Index values exceeding 520 and approaching 580, with conductivity reaching approximately 95%IACS. The framework provides a data-driven route for narrowing the Cu–Cr–X design space, while the predicted candidates require experimental validation.
This study investigates the wear performance of carburized EN39B alloy steel, a very rarely explored aerospace gear material, using a hybrid Machine Learning-Particle Swarm Optimization (PSO) approach for predictive modelling and process optimization. An L16 orthogonal array-based experimental design was adopted, and wear data were obtained using a Pin-on-Disc tribometer. The experimental results revealed minimum and maximum wear values of 93 and 418 microns respectively. The results of ANOVA revealed that speed is a significant influencing input parameter with a contribution of 77.67%, followed by load with a contribution of 16.26% on the variation of output characteristics. The Gradient Boosting (GB) model achieved an R 2 regression coefficient of 0.9977, indicating the model's adequacy and efficiency in the present study. The PSO algorithm was used to optimize wear parameters, resulting in an optimal configuration (Speed = 556.47 RPM, Load = 29.92 N, Time = 85.59 min, Track Diameter = 25.57 mm), which resulted in a predicted wear value of 92.4 microns. 3D surface plots and heatmaps revealed nonlinear interactions between the parameters, with speed and load being the major influencing parameters. Microstructural characterization of the selected samples and optimized run revealed transitions in wear mechanisms. An integrated analytical framework was developed where multi-objective optimization is combined with multidimensional parameter sensitivity and thermodynamic energy-flow modelling, which mapped complex input-output relations resulting in a unified P.V (Contact pressure vs Sliding Velocity) phase map that clearly delineates distinct wear regimes using a non-dimensional Severity Index. These findings provide a comprehensive framework for predictive wear analysis that can be employed on conventional materials used for aerospace applications.
Cancer biomarkers, glucose molecules, and low-refractive-index analytes require sensitive and reliable optical sensing platforms. This study proposes a structurally optimized surface plasmon resonance (SPR) biosensor developed using COMSOL Multiphysics, where multilayer structural parameters are systematically varied to enhance electromagnetic-field confinement and resonance tuning. A onedimensional convolutional neural network (1D-CNN) is integrated to predict sensor performance under varying structural conditions. The optimized sensor demonstrates stable resonance behavior across a refractive-index range of 1.36–1.401 RIU, achieving a maximum angular sensitivity of 480°/RIU, a peak figure of merit (FOM) of 93 RIU -1 , and quality factors ranging from 10.301 to 11.370. Strong plasmonic coupling occurs at 79°, confirming enhanced light–matter interaction at the sensor–analyte interface. A highly linear resonance-angle response is obtained with R 2 > 0.98 and minimal prediction error. Overall, the proposed platform offers high sensitivity, stable performance, and machine-learningassisted optimization for diverse analyte detection.
The uneven temperature distribution and deformation in the thickness direction during the hot rolling process of heavy plates lead to uneven microstructures across different thicknesses, seriously affecting product quality. Clarifying the microstructural evolution behavior in the thickness direction of heavy plates is particularly important for product quality control. In the presented work, the temperature field and strain field models under different deformation processes were established using the finite element model (FEM), and the distribution patterns of Plastic Equivalent Strain (PEEQ) at different thicknesses were analyzed. The Plastic Equivalent Strain-Efficient Calculation Model (PEEQ-ECM) is established through a symbolic regression algorithm to achieve rapid calculation of PEEQ during the rolling process. The square residual coefficient ([Formula: see text]) of the PEEQ-ECM calculation results reaches 0.9356, and the root mean square error (RMSE) is 1.55. The calculation time is only 8.3% of the time required for FEM calculations. On this basis, the microstructural evolution behavior during the rolling process of heavy plates is calculated by combining a microstructural evolution and mechanical property calculation model of the hot rolling process. The calculated result shows an absolute error of ±2% for phase fraction and ±2[Formula: see text][Formula: see text]m for ferrite grain size.
Electric Discharge Machining (EDM) is widely used for machining hard materials but faces challenges with inefficient debris removal and unstable discharge behavior. Vibration-assisted EDM improves flushing and spark stability, but a unified process vibration metric and accurate prediction framework are lacking. This study presents an interpretable, Response Surface Methodology (RSM)-driven optimization framework combined with Deep Learning (DL) for vibration-assisted EDM of AA7050–SiC/Graphite hybrid composites. Leveraging internship experience at Titan Engineering and Automation Ltd, the research integrates RSM for statistical modeling and a hybrid system combining Kolmogorov–Arnold-Informed Neural Network (KAINN) to capture complex discharge-vibration interactions. Partial Dependence Plots (PDPs) offer interpretable feature effects, while Nutcracker Optimization Algorithm (NOA) enables multi-response optimization. Factors such as peak current, pulse on time, vibration amplitude, and frequency are optimized to predict Vibration Energy Density (VED), Modulated Frequency Ratio (MFR), Material Removal Rate (MRR), and Tool Wear Rate (TWR). Experimental results show response variations with VED ranging from 0.15[Formula: see text]J/mm 3 to 4.69[Formula: see text]J/mm 3 and MRR up to 4.6[Formula: see text]mm 3 /min. The hybrid model achieved high prediction accuracy ([Formula: see text]) and minimal error (0.05[Formula: see text]MSE). NOA optimization identified settings yielding high VED (3.68[Formula: see text]J/mm 3 ), MRR (3.27[Formula: see text]mm 3 /min), and low TWR (0.22[Formula: see text]mm 3 /min), enhancing machining efficiency.
