可调谐二极管激光吸收光谱(TDLAS)技术用于气体浓度检测时, 会受到谐波检测中基线漂移及噪声的影响, 因此如何去除系统噪声一直是研究的热点。 分析了连续截断信号和构造hankel矩阵两种不同方法下, 奇异值分解(SVD)对TDLAS系统检测的理论意义。 将二次谐波信号分别用该方法进行矩阵化排列和奇异值分解, 选取适当阈值将部分奇异值置零并重构矩阵, 得到了这两种方法对基线纠漂和去噪的不同效果。 实验证明, 奇异值分解方法不需加入额外系统部件、 不需通零气扣除背景, 就能够快速有效地去除TDLAS系统噪声, 而构造hankel矩阵的方法适用于去除高频噪声, 连续截断信号的方法适用于进行基线纠漂。 将该方法应用于实际TDLAS系统氨气检测时的二次谐波, 系统噪声去除率达80%。

Detection of gas concentration with tunable diode laser absorption spectroscopy (TDLAS) techniques is affected by baseline drift and high-frequency noise. Therefore, how to remove the systematic noises has been a hot spot. This paper analyzes the significance of singular value decomposition (SVD) in TDLAS detection system with two different methods of constructing a matrix, and it discusses the differences of processing results for different noises. The second harmonic signal is arranged in a matrix and decomposed. We select the appropriate threshold and putthose singular values smaller than the threshold into zero, then reconstruct the matrix. Experiments show that SVD method does not require additional system components or pass into the zero gas to subtract background. This method is able to remove noises of TDLAS system quickly and effectively. We found that the method of constructing a hankel matrix is suitable for removing high-frequency noise. However, the method of constructing a continuous-cutoff-signal matrix is suitable for removing baseline drift. For example, we set up a TDLAS system to measure the concentration of NH3 while the noise removal rate of the second harmonic curve is up to 80% with this method.