Abstract
To reduce the influence of interference fringes in gas detection based on tunable diode laser absorption spectroscopy (TDLAS) and improve measurement accuracy, an interference fringe filtering method based on a back propagation neural network (BPNN) was developed in this study. First, a set of numerical simulation data of interference fringes was generated by periodic sinusoidal functions with different frequencies and phases. The topology, input, and output parameters in the BPNN were determined and optimized by the simulation data. A varying temperature WMS experimental system with a dual optical path was built, by which signals and reference concentrations of CH4 in different concentration ranges were measured. The experimental results verified the excellent performance of the BPNN in suppressing interference fringes with different frequencies and phases. When the ratio of the fringe spectral range to the absorption line width was 1.37, the standard deviation of BPNN was reduced by 4.9 times, and the relative error was reduced by 4.5 times compared with the traditional least squares (LS) method. The obtained results sufficiently demonstrated that the BPNN can effectively reduce interference fringe noise in TDLAS.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have