Key Takeaways & Executive Findings
- •• • Increasing cooling rate from 1 to 20 °C/s at 1.5×10−3 mL/g initial hydrogen reduces average equivalent porosity diameter from 51.40 to 19.16 μm, directly enhancing fatigue life in automotive structural components by minimizing stress concentrators. • • The KNN model predicts porosity percentage with R2 = 0.94, RMSE = 0.035, and MAE = 0.022 on the test set, outperforming SVM, RF, and GBM, enabling reliable virtual quality assurance and reducing costly physical trial casts. • • Experimental validation via optical microscopy shows prediction errors within 20% for average equivalent porosity diameter and porosity percentage, meeting industrial tolerance thresholds for safety-critical castings. • • The coupled CA-ML framework generalizes across cooling rates from 0.25 to 50 °C/s, providing a scalable database that can be integrated into casting simulation workflows to cut computational time by orders of magnitude compared to pure physics-based models.
China Advanced Materials & Deep-Tech Radar
Get verified English translations, SEM micrographs & open-access PDF alerts from China's leading state key laboratories delivered to your inbox every Monday at 08:00 EST.
Abstract
A coupled three-dimensional cellular automata (CA) model was employed to predict hydrogen porosity in Al−Si alloy castings as a function of thermal boundary conditions. Simulations quantified porosity distribution across cooling rates from 0.25 to 50 °C/s at an initial hydrogen content of 3.0×10−3 mL/g, generating a comprehensive porosity defect database. Four machine learning algorithms—support vector machine (SVM), random forest (RF), K-nearest neighbors (KNN), and gradient boosting machine (GBM)—were trained and compared for each porosity characteristic to identify the optimal model. For porosity percentage prediction, the KNN model achieved a determination coefficient (R2) of 0.94, root mean square error (RMSE) of 0.035, and mean absolute error (MAE) of 0.022 on the test set. Experimental validation via optical microscopy confirmed that average equivalent porosity diameter and porosity percentage predictions fell within 20% error. The model demonstrates superior performance compared to single nonlinear function fits and other simulation approaches, offering a pathway to reduce simulation time while enhancing prediction accuracy for porosity size distribution in large casting components. The database and coupled CA-ML framework provide a robust tool for mapping porosity defects in industrial Al−Si castings, addressing a critical need for reliable quality control in automotive and aerospace applications.
1. Introduction
Al−Si alloys dominate automotive and aerospace castings due to their castability and mechanical performance, yet hydrogen porosity remains a persistent defect that degrades fatigue life and structural integrity. Conventional trial-and-error process optimization and physics-based simulations are either too slow or lack predictive accuracy for industrial-scale components, where cooling rate gradients and initial hydrogen content vary unpredictably. Existing two-dimensional cellular automata models fail to capture the three-dimensional nature of porosity growth, and single nonlinear regression fits cannot handle the multivariate complexity of real casting conditions.
This study addresses the bottleneck by coupling a three-dimensional cellular automata model with machine learning to construct a porosity defect database for Al−Si alloys. By simulating cooling rates from 0.25 to 50 °C/s at 3.0×10−3 mL/g initial hydrogen, the authors generate high-fidelity data that trains four ML algorithms. The KNN model achieves R2 = 0.94 for porosity percentage, with experimental validation confirming errors within 20%. This hybrid approach enables rapid, accurate porosity mapping in large castings, offering a practical tool for reducing simulation time and improving quality control in industrial foundries.
Loading authentic research manuscript (Pages 1–5)...
Qing-huai HOU, Xue-long WU, De-cai KONG, Hai-bo QIAO, Xiao-ying MA, Xiang CI, Wen-bo WANG, Yu-ling LANG, Shi-wen XU, Zhong-yao LI, Yi-sheng MIAO, Xing-xing LI, Jun-sheng WANG (2026). Constructing porosity database for Al−Si alloy castings through 3D cellular automata model and machine learning. Transactions of Nonferrous Metals Society of China (中国有色金属学报). https://doi.org/10.1016/S1003-6326(26)67056-2
Research & Educational Purpose Only: The translations, structured abstracts, analytical annotations, and data reports provided by SinoTechIntelare intended exclusively for academic research, internal corporate R&D, and educational benchmarking. They do not constitute formal engineering, chemical safety, legal, or professional advice.
Copyright & Intellectual Property Notice: Original copyright of the underlying source articles and experimental data remains with the respective authors, institutions, and original publishing journals. SinoTechIntel claims intellectual property only over its proprietary translations, analytical syntheses, and AEO structured enhancements in accordance with international fair use and academic citation principles.
Frequently Asked Questions
What is the quantitative impact of cooling rate on porosity size, and how does it affect industrial casting design?
At an initial hydrogen concentration of 1.5×10−3 mL/g, increasing the cooling rate from 1 to 20 °C/s reduces the average equivalent porosity diameter from 51.40 to 19.16 μm. This 63% reduction directly improves fatigue resistance, allowing thinner wall sections and weight savings in automotive components without compromising durability.
How does the KNN model's predictive accuracy compare to other machine learning algorithms, and what are the implications for production?
The KNN model achieves R2 = 0.94, RMSE = 0.035, and MAE = 0.022 for porosity percentage on the test set, outperforming SVM, RF, and GBM. This accuracy enables reliable virtual screening of casting parameters, reducing the need for costly physical trials and accelerating time-to-market for new alloy components.
What is the experimental validation error, and does it meet industrial acceptance criteria?
Optical microscopy validation shows prediction errors within 20% for both average equivalent porosity diameter and porosity percentage. This falls within typical industrial tolerances for safety-critical castings, where a 20% margin is often acceptable for initial process qualification, though tighter control may be required for aerospace applications.
How does the coupled CA-ML framework address the limitations of traditional simulation approaches?
Traditional 2D CA models and single nonlinear fits cannot capture 3D porosity growth or multivariate interactions. The 3D CA-ML framework generalizes across cooling rates from 0.25 to 50 °C/s, reducing simulation time by orders of magnitude while maintaining accuracy, as evidenced by the KNN model's performance and experimental validation.
What are the scalability bottlenecks for deploying this model in large industrial castings?
The primary bottleneck is the computational cost of generating the 3D CA training database for new alloy compositions or boundary conditions. However, once trained, the ML models predict porosity in seconds, enabling real-time process control. The current database covers Al−Si alloys at 3.0×10−3 mL/g hydrogen; extending to other alloys requires additional CA simulations, but the framework is transferable.
Related Chinese Research & Cross-Citations
Assessment of zinc migration behavior and toxicity characteristics in redox smelting of zinc leaching residue
The redox smelting of zinc leaching residue (ZLR) was investigated to determine the migration behavior and toxicity characteristics of zinc under varying anthracite addition, temperature, and holding time. The ZLR, containing 10–20 wt.% Zn, 0.5–5 wt.% Pb, and 0.1–0.5 wt.% Cd, generates TCLP leachate concentrations of Zn up to 4589.0 mg/L, far exceeding regulatory limits. Experimental results reveal that CaSO4 in the residue promotes the transformation of ZnFe2O4 into a ZnS–FeS eutectic, which hinders zinc recovery and elevates environmental risk due to its lower thermodynamic stability relative to (Fe,Zn)2SiO4, ZnFe2O4, and (ZnO)slag. At temperatures above 1573 K, the ZnS–FeS eutectic is oxidized by O2/(O)slag to ZnO(s), subsequently dissolved into the slag as chemically dissolved Zn, and finally reduced to Zn(g) by CO. Pre-desulfurization or increased oxygen potential enhances zinc volatilization. Under optimized conditions, the zinc recovery ratio reached 99.13%, and the residual zinc content in the slag decreased to 0.22 wt.%, substantially below the industrial range of 1.0–3.0 wt.%. A novel strategy integrating desulfurization pretreatment with redox smelting is proposed, which lowers the required smelting temperature and improves zinc recovery efficiency, offering a more economical and environmentally sustainable solution for ZLR treatment.
Effects of oxidation roasting on surface characteristics and flotation behavior of bastnaesite
Oxidation roasting of bastnaesite was conducted to evaluate its impact on surface characteristics and flotation behavior. Experiments varied temperature, time, and O2 concentration. Increasing temperature promoted thermal decomposition, yielding Ce7O12, RE2O3, and REF3 as main phases. The Ce oxidation degree and REO grade of roasted products exceeded 85.00%. Roasting induced long, narrow, nearly parallel cracks within particles, increasing porosity and causing partial fragmentation. During flotation, dissolved rare earth ion concentration increased significantly, and surface hydrolysis formed rare earth hydroxyl compounds. Complete decomposition raised the required collector dosage to achieve recovery above 85.00%. This increase is attributed to enhanced particle wettability, altered collector adsorption, and deeper penetration into the porous structure. The findings provide a basis for optimizing flotation circuits treating roasted bastnaesite, particularly in iron-bearing rare earth deposits where pyrometallurgical pretreatment is employed.
Single Crystal NCM811 Cathode Material Prepared by Rapid Solvothermal Method
Polycrystalline LiNi0.8Co0.1Mn0.1O2 (NCM811) cathodes undergo intergranular cracking and structural collapse during extended cycling, limiting their commercial viability. This study reports single-crystalline NCM811 synthesized via a rapid ethanol–water solvothermal method. The solvothermal duration was varied, and the 60 min sample (NCM-60) exhibited optimal electrochemical performance. X-ray diffractometry confirmed an α-NaFeO2 structure with R-3m space group and high crystallinity. NCM-60 delivered a reversible capacity of 157.28 mA·h/g at 1C and a capacity retention of 55.06% after 200 cycles, significantly outperforming polycrystalline NCM (PC-NCM). Cross-sectional scanning electron microscopy revealed no apparent cracks in NCM-60 after 200 cycles, whereas PC-NCM exhibited severe intergranular fracture. The results demonstrate that shortening solvothermal time reduces precursor particle size and crystallinity, but 60 min yields the best balance. Pre-oxidation of the carbonate precursor before lithiation is recommended to mitigate CO2 evolution and lithium–nickel disorder during high-temperature sintering. This rapid solvothermal route offers a scalable pathway to single-crystal NCM811 with enhanced cycling stability and mechanical integrity.
Efficient separation of heavy metals from gypsum residue and secondary zinc oxide fume based on synergistic sulfidation
Synergistic sulfidation roasting of heavy metal gypsum residue and secondary zinc oxide fume was proposed by using the research idea of 'waste to treat waste'. Thermodynamic studies indicated that the sulfidation of zinc oxide could be effectively enhanced by increasing the dosage of calcium sulfate and carbon powder in the range of 500−800 °C. The synergistic sulfidation experiments of heavy metal gypsum residue with secondary zinc oxide showed that the sulfidation rate of zinc reached 90.39% and the grain size of ZnS increased from 5 to 10 μm under the conditions of temperature 700 °C, carbon powder 30%, Na2CO3 10%, mass ratio of gypsum residue to secondary zinc oxide 1.4:1, roasting time 2 h and cooling rate 1 °C/min. Meanwhile, 76.32% F, 72.11% Cl and 93.41% As were removed. TG/DTG−DSC, 3D FTIR spectra and SEM analysis showed that the conversion of CaSO4 to CaCO3 and the avoidance of CO2 and SO2 production were achieved under optimized conditions. This study achieves efficient sulfidation of zinc as well as growth of ZnS grains, laying the theoretical and technological foundation for subsequent recovery of ZnS by flotation.
Low-Ammonium Synergistic Leaching of Ionic Rare Earth Ore with Acetic Acid–Ammonium Sulfate System
Conventional ammonium sulfate leaching of ionic rare earth ores generates 4–6 t of ammonia-nitrogen wastewater per ton of rare earth and drives mining-area soil pH to 3.5–4.0, creating an acute environmental compliance risk. This study evaluates a low-ammonium synergistic lixiviant comprising 0.020 mol/L (NH4)2SO4 and 0.010 mol/L acetic acid (HAc) at pH 4–5, 30 °C, and 1 h contact time. Comparative leaching experiments establish a rare earth element (REE) leaching efficiency of 88.92%, a 13.36% absolute increase over single 0.020 mol/L (NH4)2SO4 leaching. To achieve the same ~90% efficiency benchmark, the conventional single-salt system requires 0.030 mol/L (NH4)2SO4; the synergistic system therefore reduces ammonium consumption by 33.3%. Surface characterization indicates a dual mechanism: H+ attenuates electrostatic interactions between RE3+ and silicate surfaces, enhancing NH4+–RE3+ exchange, while CH3COO− forms soluble RE3+/Al3+ complexes that prevent Al(OH)3 passivation and sustain surface reactivity. The protocol offers a directly deployable route to cut reagent cost and ammonia-nitrogen load without sacrificing recovery, addressing the principal bottleneck restraining sustainable ionic rare earth ore exploitation under China's dual-carbon and rare earth total-amount control policies.
Achieving strength-ductility tradeoff in near alpha titanium alloy via multi-stage heat treatment-induced nano-martensite phase transformation
A multi-stage heat treatment (MSHT) strategy, comprising a high-temperature short-duration water quench (WQ) followed by low-temperature long-duration furnace cooling (FC), was applied to a near-alpha Ti-0.3Mo-0.8Ni-2Al-1.5Zr alloy to overcome the strength-ductility tradeoff. The WQ state produced lath nano-martensite alpha-prime, residual beta-prime, and equiaxed recrystallized alpha. Subsequent FC decomposition transformed alpha-prime/beta-prime into homogeneously dispersed nano-scale alpha+beta precipitates, while equiaxed alpha coarsened via grain boundary migration. The WQ condition exhibited an ultimate tensile strength (sigma_UTS) of 610 MPa and elongation to failure (epsilon_f) of 18.2%. The WQ+400FC condition achieved a peak sigma_UTS of 791.5 MPa with epsilon_f = 16.7%, yielding a strength-ductility product (sigma_UTS * epsilon_f) of 13.2 GPa*%, a 19% improvement over the WQ state. Texture analysis revealed a duplex texture in WQ: weak {0001}//Z0 and strong {0110}//Y0, inherited after FC. The 400FC sample showed the highest lattice strain inhomogeneity, with peak kernel average misorientation (KAM) of 1.5 degrees and grain orientation spread (GOS) of 0.96 degrees, correlating with the excellent sigma_UTS. Non-basal slip systems exhibited higher Schmid factor (SF) values after heat treatment, contributing to ductility. Burgers orientation relationship (BOR) reconstruction confirmed variant selection during beta to alpha-prime transformation, with only four predominant alpha-prime variants instead of the twelve theoretically possible.