Tomographic measurements of temperature and species concentration via deep learning and wavelength modulated absorption spectroscopy

To achieve the 2-D nonintrusive measurements of temperature and species concentration in combustion field, a new framework, combining calibration-free wavelength modulation absorption spectroscopy (CF-WMS) with designed convolutional neural network (CNN), was developed. The principle CF-WMS, along architecture CNN net, training, performance network, has been investigated. region interest discretized into 24 × pixels2, 48 probing beams six targeted frequencies were used to verify feasibility WMS 2f/1f signal for reconstruction. 20 000 samples water vapor distributions are randomly fabricated, featuring three positioned Gaussian distributions. Reconstructed images phantoms agreed well original relative error about 5.0%–9.2% 8.0%–12.4% using 17 training datasets different beam arrangements H2O species, respectively. Several representative limited number have examined compared. arrangement BA2 BA3 exhibited better than BA1 average errors 5% 8% Such method can provide an effective way spatially temporally resolved, real-time, situ monitoring practical environments.