Anisotropic yield functions for sheet forming are commonly selected from their agreement with directional yield stresses and Lankford coefficients. This study examines whether close calibration agreement also indicates stable coefficient identification, convexity, and accurate predictions outside the fitted loading conditions. A literature inventory contains 62 sheet-material records; eight provide the common seven-direction and balanced-biaxial measurements used for quantitative comparison. Seven yield-function families are represented by eight formulations, including Hill48, Yld89, Yld2000-2d, reduced Yld2004-18p, two BBC2008 variants, Cazacu (2018), and HomPol6. Coefficients are identified separately using transformed and conventional residuals with paired initial values. Greater flexibility reduces calibration errors but does not ensure identifiable coefficients or convex surfaces. Two exact polynomial perturbations explain why the selected observations cannot identify all fifteen normalized HomPol6 coefficients. Seven of eight BBC2008-p16 fits are numerically rank deficient under each objective; the remaining conventional-objective case is sensitive to the rank threshold. Prediction errors in loading conditions not used for calibration depend on the fitting measure: the Hill48 and Yld2004-18p ordering changes, while their exploratory material-level intervals include zero. A separate compression, plane-strain, shear, and flow assessment uses DP980, QP980, and AA5754-O. Hill48 gives the smallest mean radial stress error under both objectives (4.28–4.47
This study investigates the peeling behaviour of 3D-printed polylactic acid (PLA) re-entrant core sandwich structures (SWSs). The study aimed to characterize interfacial peel failure in 3D-printed PLA re-entrant core SWSs using a bilinear cohesive zone modelling (CZM) based finite element (FE) framework. Two process configurations, differing in facesheet thickness and resin-to-hardener ratio, were compared as a comparative case study. Cohesive parameters were identified using a conventional trial-and-error approach and a Proper Orthogonal Decomposition–Gaussian Process Regression (POD-GPR) assisted inverse framework, which combined FE simulations, dimensionality reduction, surrogate modelling, and optimization. Peel experiments were conducted on two case study samples fabricated using fused filament fabrication (FFF). FE model successfully reproduced the elastic response, peak load, and progressive debonding observed in experiments. For case 1, using the POD-GPR framework reduced the peak force prediction error from 0.2
Motivated by the automotive industry’s demands for energy efficiency and emission reduction, warm stamping has emerged as a promising solution to overcome the limitations of both cold forming and hot stamping of high-strength steels. This study proposes an innovative warm stamping process specifically designed for 1500 MPa-grade cold-rolled martensitic steel. The process is compatible with existing industrial hot stamping lines, enabling seamless integration with existing infrastructure, while achieving significant springback reduction without substantial strength loss. A systematic investigation was conducted to evaluate the effects of holding temperature and time on microstructural evolution, mechanical properties, and springback behavior. The optimal process parameters were determined to be a holding temperature of 500 °C for 3 min. Under these conditions, the sidewall and flange springback angles decreased by 71.2
Heavy forgings for power and petrochemicals, with extremely large dimensions, are made from hundreds-tonne steel ingots via multi-pass hot forging, where repeated thermo-mechanical loading promotes damage and raises cracking risk. Accurate cracking prediction for heavy forgings using finite element simulation is essential for forging safety and economics. Focusing on 2.25Cr1Mo0.25 V steel, this study systematically investigated the critical damage value (CDV) for the normalized Cockcroft and Latham model and its applicability, with emphasis on cracking prediction in industrial-scale simulations through experimental and theoretical approaches. A conventional hybrid experimental–numerical approach (HENA) was first used to calibrate the CDV, revealing a nonlinear decrease with increasing element size, an effect attributable to numerical errors in FE analysis. The HENA-based CDV is only suitable for scaled-down simulations of heavy forgings, as it inherently couples with small elements (0.2–1.6 mm). To overcome this limitation, a macroscopic equivalent damage method (MEDM) was proposed to determine the CDV for full-scale simulations, and its analytical expression was established. A characteristic element size was derived from the generalized Hillerborg fracture energy equation, which theoretically pairs the MEDM-based CDV with this size, yielding a value of 19.73 mm upon regression of experimental data. The applicability of the four CDVs for different-sized elements was preliminarily supported by three laboratory upsetting experiments and one heavy forging case, demonstrating their potential for crack prediction. A dual-scale strategy was thus established to determine the CDV according to simulation scale, thereby providing a pathway to extend cracking prediction of heavy forgings from laboratory research to industrial practice.
Chloride ion corrosion is considered one of the primary causes of durability deterioration in reinforced concrete structures exposed to marine environments or chlorine-containing deicing salts. This research introduces a significant soft computing model to assess the chloride penetration of metakaolin concrete by comparing the Artificial Bee Colony (ABC), Genetic Algorithm (GA), Grey Wolf Optimizer (GWO), Harris Hawks Optimization (HHO), Particle Swarm Optimization (PSO), and Salp Swarm Optimization (SSO) algorithm-based least square-support vector machine (LSSVM) models. For this purpose, two hundred one datasets were used to develop, train (161), and test (40) the models. The performance comparison showed that the SSO_LSSVM model assessed chloride penetration with a performance index (PI) of 1.8702, variance accounted for (VAF) of 95.9727, performance (R) of 0.9798, and root mean square error (RMSE) of 336.1474 C. In addition, several analyses, such as score, regression error characteristics curve, generalizability, uncertainty, computational cost, reliability, objective function criterion, and validation dataset, confirmed the reliability and capabilities of the SSO_LSSVM model. Conversely, the multicollinearity analysis revealed that the binder and fine aggregate significantly affected the performance of hybrid LSSVM models. The analysis of variance (ANOVA) supported the research hypothesis for the binder, aggregates, metakaolin content, and compressive strength features. The Anderson–Darling test indicated statistically significant departures from normality (p < 0.005) for the actual and model-predicted chloride penetration data; however, the SSO_LSSVM predictions exhibited an AD statistic closest to that of the actual observations, indicating comparatively similar distributional characteristics. The fundamental study of algorithms reveals that SSO exhibited superior optimization performance for the present problem, as evidenced by its favorable fitness-evolution trajectory and the lowest final convergence MSE. Additionally, it requires fewer parameters to tune than GA and PSO. The SHAP and LIME analyses revealed that the compressive strength (fc) is the most significant feature in assessing the chloride ingress of metakaolin concrete.
