Research on the TDLAS detection of oxygen concentration in an airborne dynamic environment based on FPSOGSA-EM-BP neural network

Objective Aircraft safety issues have always been a major concern, and the combustion and explosion of aircraft fuel tanks are one of the important causes of catastrophic aircraft safety accidents. The detection of oxygen concentration in airborne inerting systems is of vital importance. With the advancement of laser manufacturing and signal processing technologies, Tunable Diode Laser Absorption Spectroscopy (TDLAS) technology has been widely applied in environmental gas monitoring. Due to the irregular changes in temperature and pressure in the airborne environment, these changes will indirectly affect the accuracy of gas detection by influencing factors such as the absorption spectral lines of gases. The existing compensation measures for a single temperature or pressure change do not apply to complex detection scenarios where both temperature and pressure change simultaneously. Considering the limitations of previous studies, this paper proposes a TDLAS oxygen concentration detection scenario in an airborne dynamic environment and explores solutions to the influence of the environment on the concentration detection accuracy.Methods This paper chooses to build a simulation system for the preliminary study. Firstly, the influence of the TDLAS gas detection principle, as well as temperature and pressure changes on gas detection, was analyzed theoretically. Meanwhile, A ground experimental platform was built for the preliminary detection of oxygen concentration. Meanwhile, referring to the parameters of the experimental equipment, the TDLAS oxygen concentration simulation detection system was constructed by using Simulink. The detection results of the experimental platform further verified the reliability of the simulation, and then the influence of temperature and pressure changes on oxygen concentration detection was compared. Finally, aiming at the compensation problem of TDLAS in an airborne dynamic environment, a Fractional Particle Swarm Optimization Gravitational Search Algorithm-Entropy Metric-Back Propagation (FPSOGSA-EM-BP) is proposed. This algorithm introduces fractional calculus in the particle velocity update, utilizes historical velocity information to construct dynamic memory characteristics, and enhances the global search ability. Meanwhile, diversity is monitored by real-time calculation of the population entropy value. FPSOGSA-EM-BP and PSO-BP neural networks were utilized for compensation and intercomparison.Results and Discussions Concentration inversion was carried out using the TDLAS gas detection system. The Voigt line types in the airborne dynamic environment have maximum errors of 82.2% respectively, and average errors of 23.7% (Fig.8). The FPSOGSA-EM-BP neural network based on the hybrid optimization algorithm and the other neural networks all have certain correction effects. However, the particle swarm optimization of the PSO-BP algorithm is prone to fall into local optimum and has insufficient convergence accuracy; FPSO-BP lacks a dynamic balance mechanism from exploration to development, and its convergence speed is slow. The search trajectory of GSA-BP lacks continuity and may oscillate near the optimal solution. It is precisely because of these defects that the fitting accuracy of these methods has always lagged behind that of the FPSOGSA-EM-BP method (Fig.10). The maximum error of FPSOGSA-EM-BP is reduced to 0.98% compared with 2.45%, 1.52% and 1.31% of GSA-BP, PSO-BP and FPSO-BP respectively, meeting the requirement that the oxygen detection error of the airborne fuel tank is less than 1%. The average error decreased from 0.54%, 0.34% and 0.30% of GSA-BP, PSO-BP and FPSO-BP to 0.27% respectively (Tab.3). At the same time, the compensation effects were compared with those of other relevant literatures. The temperature-pressure compensation model proposed in this paper has a wider temperature and pressure test range in gas concentration detection, and a better compensation effect (Tab.4). This means that this model not only maintains the robustness of PSO-BP in complex environments but also improves the compensation accuracy to a new level.Conclusions This paper explores the O2 gas detection and temperature and pressure compensation methods based on TDLAS detection technology, and analyzes the influence of the airborne dynamic environment on the O2 concentration correspondence. Considering the relatively large error brought by the airborne dynamic environment to the detection of oxygen concentration, the temperature and pressure compensation model of FPSOGSA-EM-BP was proposed. Compared with other models, the FPSOGSA-EM-BP model improves the measurement accuracy after compensation. The maximum error and the average error are 0.98% and 0.27% respectively. Furthermore, compared with the existing temperature and pressure compensation methods, when using FPSOGSA-EM-BP, the temperature and pressure environment in this paper not only has a wider range but also a better compensation effect. The accuracy and reliability of TDLAS technology in detecting O2 gas in the fuel tank under airborne dynamic environments have been further enhanced.