基于TDLAS检测西林瓶内氧气浓度的多光束干涉抑制方法

应用可调谐半导体激光吸收光谱(TDLAS)技术开放单光路短光程检测西林瓶内氧气浓度，因玻璃瓶壁造成入射光多次反射和透射，形成多光束干涉，严重影响信号波形和检测精度。本文提出了一种改变激光入射角度来抑制瓶壁光学干扰的方法，理论分析了入射角度对透射光强分布的影响，详细推导了使两相干光束叠加部分在接收端探测范围之外的入射角度计算公式，并根据现场参数得到理论最佳入射角度。对氧气浓度1%的样瓶进行多次测量，将二次谐波信号峰值的平均值作为信号，峰值的标准差作为噪声，以信噪比(signal to noise radio, SNR)最大作为系统入射角角度的优化指标，实验获得系统的实际最佳入射角度。与决定系数较高的入射角度进行浓度预测对比，交互验证后的最小二乘拟合结果显示：相关系数分别为0.995 9和0.988 9，前者相比后者提高了0.7%，预测的均方根误差(root mean square errors of prediction, RMSEP)分别是0.003 1和0.005 3，前者相比后者降低了41.5%，说明本文方法所确定的最佳入射角，能有效抑制玻璃瓶壁引起的多光束干涉影响，改善系统检测精度。

The Multi-Beam Interference Suppression for Measuring Penicillin Vial’s Oxygen Concentration Based on Tunable Diode Laser Absorption Spectroscopy

Penicillin vial’s oxygen concentration detection based on tunable diode laser absorption spectroscopy (TDLAS) was conducted in the open, on a single and short optical path. In this system, the multiple reflections and transmissions were caused by parallel glass walls, and the multi-beam interference impacted the signal wave form and the accuracy of the detection. A method was created tochange the incident angle of laser, and the influence of incident angle was the oretically analyzed to the distribution of transmission intensity. The calculation formula of incident angle was deductedin order that the superposition part of two coherent beamslocated in the outside of the detection range, and then the theoretical optimal incident angle was determined according to the field parameters. The 2f signal of sample(concentration 1%) was severally measured for many times in different time, and the signal to noise ratio (SNR) could be use as the optimization index of optimal angle. Finally, the actual optimal incident angle acquired in this system was determined. To compare with the optimal incident angle and the incident angle of high coefficient of determination, the concentration prediction result of cross validation showed that the correlation coefficients were 0.995 9 and 0.988 9 respectively. The former was 0.7% larger than the latter, and the root mean square errors of prediction (RMSEP) were 0.003 1 and 0.005 3 respectively. In addition, the former was reduced by 41.5% than the latter. This validates the effectiveness of the proposed method which can suppress the Multi-beam Interference and further improve the accuracy.