以某直列六缸柴油机为研究对象，设计台架测试DOE获取试验数据，基于试验结果分别构建了原机NO_x和Soot排放的二阶和三阶多项式回归模型，并进行对比分析。结果显示：二阶多项式原机NO_x排放模型在训练集上的决定系数R²为0.999 3，均方根误差（RMSE）为13.27×10^－6，Soot模型在训练集上的R²为0.949 1，RMSE为0.343 2 mg/m³，模型拟合效果较好；三阶多项式模型在训练集上的R²比二阶模型的更高，但对训练集外工况点的预测精度较差。基于此，应用二阶多项式模型对全球统一瞬态试验循环（WHTC）和高原环境下的排放进行了预测，结果显示，热态WHTC的原机NO_x和Soot比排放预测值的相对误差分别为－0.51%和－1.13%；在2 200 m高原环境下， NO_x排放万有相对误差在±10%以内的工况点占比为86.3%。该二阶多项式回归模型精度满足工程预测需求，且具备高原排放预测泛化能力。

Taking a certain in-line six-cylinder diesel engine as the research object, the design of experiments (DOE) for bench testing was conducted. Based on the experimental data, second-order and third-order polynomial regression emission models for NO_x and soot from the original engine were constructed and compared. The results indicated that the second-order polynomial NO_x emission model achieved a coefficient of determination (R²) of 0.999 3 and a root mean square error (RMSE) of 13.27×10^－6 on the training set, while the soot model achieved R² of 0.949 1 and RMSE of 0.343 2  mg/m³ on the training set, which demonstrated a good model fitting. The third-order polynomial model exhibited a higher R² value on the training set, but it had poorer prediction accuracy for operating points outside the training set. Based on the results, the second-order polynomial models were applied to predict the emissions of world harmonized transient cycle (WHTC) and plateau conditions. The results showed that the prediction errors for NO_x and soot specific emissions of hot WHTC were －0.51% and －1.13% respectively. At an altitude of 2 200 m, the proportion of operating points for the relative error of NO_x specific emissions within ±10% is 86.3%. The accuracy of second-order polynomial regression model met the needs of engineering predictions and had generalization capability for emission predictions at high-altitude conditions.