研究了基于可见–近红外光谱技术的制动液品牌混掺比例快速无损检测方法。全波段建立的 偏最小二乘回归(PLSR)和最小二乘支持向量机(LS-SVM)模型均得到了较好的预测结果。这两个模型的建模集和预测集的确 定系数(r2c和r2p)均在0.98以上。采用连续投影算法(SPA)挖掘特征波长，最终选择了439 nm、 443 nm、459 nm、519 nm、570 nm、717 nm、896 nm和902 nm共8个波长作为最优变量。基于SPA选择的变量建立的PLSR 和LS–SVM模型的r2c和r2p均在0.97以上，能够满足实际应用的需要。研究结果表明，可见–近红外光谱 可以用于制动液品牌混掺比例快速无损检测。

A fast and non-invasive method for determining the mixture percentages of brake fluid based on visible and near infrared spectroscopy (Vis–NIRS) was proposed. Both a partial least square regression (PLSR) model and a least-square support vector machine (LS–SVM) model were established according to the spectra obtained in the whole wavelength range. With those two models, good prediction results were obtained. The determination coefficients of their calibration and prediction sets (r2c and r2p) were greater than 0.98. The successive projection algorithm (SPA) was used to select the effective variables. Finally, eight variables of 439 nm, 443 nm, 459 nm, 519 nm, 570 nm, 717 nm, 896 nm and 902 nm were selected as the optimal variables to be input into the PLSR and LS–SVM models. The r2c and r2p of both two models were greater than 0.97 which was adequate for practical application. It was concluded that Vis–NIRS could be used to fastly and non-invasively determine the mixture percentages of brake fluid.