【摘要】 以5 a、14 a、29 a和40 a的杂种落叶松（Larix kaempferi 5×L. gmelinii 9）最外侧5个生长轮早材和晚材的组织切片为研究对象，分析气干密度与拉伸弹性模量的相关关系，并利用表格先验数据拟合网络（Tabular prior-data fitted network,TabPFN）模型对拉伸弹性模量进行建模与预测。结果表明：4个树龄早材和晚材的拉伸弹性模量与气干密度均呈现正相关，晚材的决定系数（R~2）高于早材，两者数据联合拟合的结果显著优于单独拟合，R~2介于0.94至0.97之间。基于TabPFN构建的模型可实现落叶松早材与晚材拉伸弹性模量的预测，预测值与实测值平均相对误差（mean absolute percentage error,MAPE）为16.03%。研究结果可为落叶松木材拉伸弹性模量预测提供参考，为木材材质的快速评估提供理论依据。更多还原

【Abstract】 In this research, earlywood(EW) and latewood(LW) tissue slices were taken from the outermost five growth rings of hybrid larch(Larix kaempferi 5×L. gmelinii 9) at 5, 14, 29, and 40 years with an aim at revealing the variation of air-dry density and tensile modulus of elasticity(MOE), as well as their correlation. Furthermore, a tabular prior-data fitted network(TabPFN) model was employed to simulate and predict tensile MOE. Results showed that, at any given tree age, a positive correlation was observed between the tensile MOE and air-dry density for both EW and LW, with the coefficients of determination(R~2) being higher for LW than for EW. The combined fitting of EW and LW data yielded significantly better results than individual fittings, with R~2 ranging from 0.94 to 0.97. A prediction model constructed based on TabPFN deep learning enabled the prediction of tensile MOE of EW and LW in hybrid larch, with a mean absolute percentage error(MAPE) of 16.03% between predicted and experimental values. These findings could support the prediction of tensile MOE of larch wood for rapidly evaluating wood quality.更多还原