A group of heat-treatment-free Al-7Si-0.3Mg-xFe (x = 0.15, 0.3, 0.45, 0.6 wt%, wt% is mass fraction) die-cast alloys were prepared, and the effects of the Fe content and natural aging (NA) on the microstructures and tensile properties were investigated using machine learning (ML). The intermetallic phases of blocky Mg2Si and chunky α-Al15(Fe, Mn)3Si2 (~0.5–1 μm) were found in the as-cast microstructure, while striped β-Al13(Fe, Mn)4Si0.25 with a width of ~2–3 μm appeared until the Fe content exceeded 0.45 wt%. As the Fe content increased from 0.15 wt% to 0.6 wt%, the yield strength (YS) increased from (128.6±2.7) MPa to (132.1±2.5) MPa and the elongation (EL) decreased from (14.35±1.60)% to (10.56±1.32)%. The underlying mechanism was attributed to the increase in the number density of α-Fe from (1.03±0.09)×1010/m2 to (1.11±0.08)×1011/m2, along with the increase in that of β-Fe from zero to (1.73±0.16)×109/m2. The relationships among the Fe content and the as-cast YS (MPa), ultimate tensile strength (UTS) (MPa), and EL (%) were quantitatively determined (YS = 129.03−4.54x+16.53x2, UTS= −22.76x2+21.64x+262.94, and EL=−8.2x+15.59). The investigated alloys showed a YS enhancement of 5.8–9.8 MPa after NA for 30 d, which was attributed to the appearance of β″ nanoscale precipitates. The adaptive boosting (AdaBoost) ML model performed the best among the four investigated ML models, with a high R2 of 0.92 and low mean absolute error of 0.53 when predicting the YS of the Al-7Si-0.3Mg-xFe die-cast alloys. The quantitative relationships among the Fe content, NA time, and YS/UTS/EL values were determined using the AdaBoost model (YS = 137.74−29.07exp(−x/4.35t(d)/3.91), UTS=265.1+7.73x+0.22t−5.95x2, and EL=15.52−4.89/((1+exp((0.16−x)/0.16))(1+exp((46.55−t)/(−6.34))))). A high-performance and recyclable Al-7Si-0.3Mg-0.5Fe alloy was predicted using AdaBoost, with YS, UTS, and EL values of (137.8±1.8) MPa, (272.6±1.8) MPa, and (12.03±1.18)% after NA for 30 d. The results of this study provide an effective method to intelligently develop high-performance and recyclable heat-treatment-free die-cast alloys.

Effects of Fe content and natural aging on the microstructures and mechanical properties of recyclable heat treatment-free Al-7Si-0.3Mg die-cast alloys determined using machine learning

期刊论文
发表日期：

2026年5月19日

查看全文

Qian Liu,
Xi-Xi Dong,
Wen-Jing Han,
Qi-Xiu Han,
Hai-Lin Yang,
Shou-Xun Ji

摘要

A group of heat-treatment-free Al-7Si-0.3Mg-xFe (x = 0.15, 0.3, 0.45, 0.6 wt%, wt% is mass fraction) die-cast alloys were prepared, and the effects of the Fe content and natural aging (NA) on the microstructures and tensile properties were investigated using machine learning (ML). The intermetallic phases of blocky Mg₂Si and chunky α-Al₁₅(Fe, Mn)₃Si₂ (~0.5–1 μm) were found in the as-cast microstructure, while striped β-Al₁₃(Fe, Mn)₄Si_0.25 with a width of ~2–3 μm appeared until the Fe content exceeded 0.45 wt%. As the Fe content increased from 0.15 wt% to 0.6 wt%, the yield strength (YS) increased from (128.6±2.7) MPa to (132.1±2.5) MPa and the elongation (EL) decreased from (14.35±1.60)% to (10.56±1.32)%. The underlying mechanism was attributed to the increase in the number density of α-Fe from (1.03±0.09)×10¹⁰/m² to (1.11±0.08)×10¹¹/m², along with the increase in that of β-Fe from zero to (1.73±0.16)×10⁹/m². The relationships among the Fe content and the as-cast YS (MPa), ultimate tensile strength (UTS) (MPa), and EL (%) were quantitatively determined (YS = 129.03−4.54x+16.53x², UTS= −22.76x²+21.64x+262.94, and EL=−8.2x+15.59). The investigated alloys showed a YS enhancement of 5.8–9.8 MPa after NA for 30 d, which was attributed to the appearance of β″ nanoscale precipitates. The adaptive boosting (AdaBoost) ML model performed the best among the four investigated ML models, with a high R² of 0.92 and low mean absolute error of 0.53 when predicting the YS of the Al-7Si-0.3Mg-xFe die-cast alloys. The quantitative relationships among the Fe content, NA time, and YS/UTS/EL values were determined using the AdaBoost model (YS = 137.74−29.07exp(−x/4.35t(d)/3.91), UTS=265.1+7.73x+0.22t−5.95x², and EL=15.52−4.89/((1+exp((0.16−x)/0.16))(1+exp((46.55−t)/(−6.34))))). A high-performance and recyclable Al-7Si-0.3Mg-0.5Fe alloy was predicted using AdaBoost, with YS, UTS, and EL values of (137.8±1.8) MPa, (272.6±1.8) MPa, and (12.03±1.18)% after NA for 30 d. The results of this study provide an effective method to intelligently develop high-performance and recyclable heat-treatment-free die-cast alloys.