复混肥成分的快速、 原位检测对化肥的生产过程、 产品质量控制具有重要的意义。 在化肥企业生产中, 实验室进行分析, 检测时间长, 无法实现线上检测。 与复混肥成分现有检测方法相比, 激光诱导击穿光谱(LIBS)检测时间只需几分钟、 一次测量可完成复混肥成分检测、 几乎无需样品预处理, 将该技术用于复混肥成分快速、 现场检测非常合适。 搭建LIBS系统, 激光器(100 mJ, 1 064 nm, 1 Hz)输出的激光束经45°反射镜由水平转为垂直方向, 经焦距为40 mm的透镜聚焦至旋转台上的复混肥样品表面, 产生激光等离子体。 激光器的调Q信号控制光纤光谱仪(Avantes, 195～500 nm)采集信号, 设置光谱延迟时间为1.28 μs, 采集时间为1.05 ms, 最终获取复混肥样品LIBS光谱。 20个复混肥样品由安徽徽隆集团提供, 磷元素的参考值由企业采用国家标准方法测量。 将复混肥样品粉碎过筛取3 g, 采用压片机在8 MPa下压制成形。 实验中, 使用小型风扇吹扫复混肥样品表面, 形成稳定气流, 每个样品重复测量10次, 每次测量平均20个脉冲, 以减小样品不均匀性。 其中, 15个样品用于定标回归模型的建立, 五个样品用于检验定标模型的适用性。 复混肥是一种成分复杂的混合物, 其中氮、 磷、 钾均以化合物存在。 传统的LIBS定量方法是基于待测元素单条谱线强度, 未考虑其他元素影响, 降低了定量结果的准确性。 将LIBS技术和多元线性回归法结合用于分析复混肥中磷元素浓度。 选取磷元素的三条特征谱线即213.6, 214.9和215.4 nm。 磷矿中硅元素含量基本不变, 且硅元素在磷的谱线附近存在多条谱线, 如212.4, 220.8, 221.1和221.7 nm。 分别采用一元、 二元、 三元和四元线性回归法建立校准曲线。 结果表明, 采用P Ⅰ: 214.9 nm谱线强度作自变量建立一元线性回归, LIBS预测值与参考浓度的相关系数仅为0.083, 无法满足磷元素的定量分析要求。 当采用P Ⅰ: 214.9 nm谱线强度和三条特征谱线之和(P Ⅰ: 213.6, 214.9和215.4 nm)作自变量建立二元线性回归拟合时, 相关系数提高到0.856, 平均绝对误差由1.32%减小到0.16%。 在二元线性回归中引入Si Ⅰ: 212.4 nm谱线强度, 建立三元线性回归, 相关系数为0.869。 为进一步提高磷元素浓度测量的准确性, 建立四元线性回归方程, 将Si Ⅰ: 212.4, 220.8, 221.1和221.7 nm谱线强度之和作为自变量加入三元线性回归, 相关系数提高到0.980, 且15个定标样品的相对误差范围为0.06%～1.31%, 而验证样品为0.13%～1.26%, 这说明采用四元线性回归定标法可提高复混肥中磷元素浓度测量的准确性。

The rapid and in-situ detection of compound fertilizer components was of great significance to the production process control and product quality control of chemical fertilizer production enterprises. In the production of fertilizer companies, compound fertilizer samples were collected on the production line and sent to the laboratory for analysis. The time required for detecting was long, which can’t meet the compound fertilizer company’s production line for testing. Compared with the existing detection methods of compound fertilizer components, the detection time of laser-induced breakdown spectroscopy (LIBS) was several minutes. And the measurement of compound fertilizer components can be completed in one measurement, with almost no need for compound fertilizer samples. LIBS technique was very suitable for the rapid and on situ detection of compound fertilizer components. In the LIBS detection system, the laser beam output from the solid-state pulse laser (100 mJ, 1 064 nm, 1 Hz) was converted from horizontal to vertical by a 45°-mirror. The laser beam was focused onto the target using a lens (focal length=40 mm). The sample was placed at a rotating platform. The light emitted by the plasma was collected by an optical fiber spectrometer (Avantes, 195～500 nm) with a delay time of 1.28 μs and an integration time of 1.05 ms. The spectrometer was triggered by the signal of laser’s Q-switched. The LIBS spectrum of the compound fertilizer sample was finally obtained. In the experiment, 20 compound fertilizer samples were provided by Anhui Huilong Group. The reference concentration of phosphorus element was measured by the enterprise using the national standard method. The samples were ground into powder and sieved. Three grams of every sample was pressed pellets under a pressure of 8 MPa. During the experiment, a small fan was used to continuously purge the surface of sample to form a stable airflow environment. Each sample was repeatedly measured 10 times, and each spectrum was formed by 20 shot average to reduce the heterogeneity of the sample. 15 samples were selected as calibration set to train regression model, and 5 samples were chosen to test the model. The compound fertilizer was a complex mixture of components, in which the nitrogen, phosphorus, and potassium were all present as compounds. The traditional quantitative analysis method of LIBS was based on the intensity of a single characteristic line of the measured element. The influence of other elements was not considered, which greatly reduced the accuracy of the analysis results. In this paper, LIBS technology and multiple linear regression calibration were used in combination to determine the phosphorus concentration in compound fertilizer. Three feature spectral lines of P element detected by LIBS were 213.6, 214.9 and 215.4 nm. As the concentration of silicon in the phosphate rock was relatively stable. And near the spectral feature line of P element, there were many characteristic lines of Si element, such as 212.4, 220.8, 221.1 and 221.7 nm. The analysis was carried out based on unary, binary, ternary, and quaternary linear regression calibration curves, respectively. It turned out that the unary linear regression method hardly served the quantitative analysis for compound fertilizer sample only using the intensity of P Ⅰ 214.9 nm as variable, and the correlation coefficient between the LIBS prediction value and the reference concentration was only 0.083. When using the intensity of P Ⅰ: 214.9 nm and the sum of three characteristic lines (P Ⅰ: 213.6, 214.9 and 215.4 nm) as input variables to establish the binary linear regression, the correlation coefficient increased to 0.856 and the average absolute error decreased from 1.32% to 0.16%. When introducing the line intensity of Si Ⅰ: 212.4 nm into the binary linear regression equation, the ternary linear regression was established, and the correlation coefficient was only increased to 0.869. In order to further improve the accuracy of phosphorus concentration measurement in compound fertilizers, the quaternary linear regression equation was established. The sum of the Si Ⅰ: Si 2.4: 222.4, 220.8, 221.1 and 221.7 nm line intensities was used as an independent variable to add into ternary linear regression equation. The correlation coefficient was increased to 0.980. The relative error ranges were 0.06%～1.31% and 0.13%～1.26% for 15 calibration samples 5 validation samples, respectively. The results demonstrated that using the quaternary liner regression calibration method can improve the accuracy of phosphorus concentration measurement in compound fertilizers.