Convolutional Neural Network-Based Laser Absorption Spectroscopy Combustion Field Tomography

Two-dimensional（2D）measurement of the temperature field for the combustion chamber of a marine engine is of great significance for improving the combustion efficiency of the engine,increasing the service life,and achieving energy saving and emission reduction.In response to measurement requirements such as non-contact and high temporal and spatial resolution,this paper developed a set of two-dimensional temperature field measurement system based on tunable semiconductor laser absorption spectroscopy（TDLAS）technology and combined with convolutional neural network（CNN）algorithms.Two-dimensional reconstruction of the combustion temperature field in the combustion chamber of the engine.Firstly,a two-dimensional reconstruction algorithm based on CNN was developed.Firstly,the absorption spectral lines of water vapor（H₂O）at 6807.83cm^-1and 7185.6cm^-1were determined as the thermometric line pair.Subsequently,the forward measurement model of the combustion field is studied,and the typical symmetrical unimodal temperature field and the non-uniform bimodal temperature field are simulated to obtain the training set and the test set respectively through numerical simulation,and the CNN model was established.Finally,the reconstruction effect of the CNN algorithm is evaluated through numerical simulation,and compared with the BP neural network algorithm and the Tikhonov regularization algorithm.The average reconstruction errors of the three are less 1.93%,6.71%,and5.07%,respectively.Secondly,Integrated two-dimensional temperature measurement system based on TDLAS combustion field.Firstly,the direct absorption spectroscopy scheme is used to integrate a set of 2D temperature field measurement,which consists of two parts:a host and a mechanical equipment.The host uses the data acquisition card to output two sawtooth waveforms so that the 1391nm and 1468nm distributed feedback lasers are scanned in a time-division multiplexed manner.The lasers of the two wavelengths are coupled in one fiber and then divided into 14 channels,which are arranged in 5×9 orthogonal light path to cover the entire combustion field by the designing and manufacturing mechanical equipment.The optical signal absorbed by the combustion product H₂O is received by 14 photodetectors and then sent to the data acquisition card for collection.Data collection and processing programs are designed to realize the 2D temperature distribution of the combustion field reconstruction.Finally,the performance of 2D temperature field reconstruction system is evaluated.In a laboratory environment,2D reconstruction of the temperature field of the candle flame is carried out.The results show that the reconstruction results are consistent with the measurement results of shearing interferometry and thermocouples,which proves that the method is suitable for the measurement of non-uniform temperature field.Then the system was applied to diesel engines for on-site testing.However,due to the harsh environment of high temperature,high pressure,and multiphase flow（mixing of gas,oil droplets and soot）,the measurement light path experienced large fluctuations.It is impossible to carry out the expected measurement.Therefore,the problems and shortcomings of the field experiment are summarized to provide help for the subsequent improvement of the experimental system.By improving the stability of the entire experimental system,the 2D reconstruction of the temperature field of the diesel engine combustion chamber exit can be realized.