Enhanced TDLAS ammonia concentration detection with an OSP-SVR optimization algorithm for water vapor interference suppression

In complex industrial environments, the tunable diode laser absorption spectroscopy (TDLAS) technique for quantitative detection of ammonia (<i>N</i><i>H</i>₃) is frequently constrained by water vapor interference. This constraint leads to misinterpretation of second harmonic peak values and inaccuracies in concentration inversion. To address this challenge, this study proposes a hybrid algorithm that integrates orthogonal subspace projection (OSP) and support vector regression (SVR), aiming to eliminate the spectral effects of background gases and achieve high-precision <i>N</i><i>H</i>₃ detection under multicomponent interference conditions. First, the second harmonic signal extracted from the <i>N</i><i>H</i>₃-<i>H</i>₂<i>O</i> mixed spectrum is projected onto the complementary orthogonal subspace of the water vapor component to eliminate its spectral contributions; then, based on the purified <i>N</i><i>H</i>₃ second harmonic feature points, an SVR concentration inversion model is constructed to suppress residual nonlinear noise and baseline drift issues. Experimental results show that the enhanced TDLAS ammonia detection method based on the OSP-SVR algorithm can effectively eliminate water vapor interference. Under the water vapor-interfering experimental conditions, the OSP-SVR algorithm reduces the water vapor-induced secondary harmonic interference peak by up to 98.54% compared with the conventional single-peak detection method. Meanwhile, for the residual standard error, it achieves reductions of 62.65% and 61.44% in the 0.2-0.9 and 1-9 ppm concentration groups, respectively. These results demonstrate that this method is more suitable for measuring <i>N</i><i>H</i>₃ concentrations in dynamic and complex environments.