As a non-intrusive, fast-response and highly sensitive and diagnostic tool, Wavelength Modulation Spectroscopy (WMS) has been extensively applied in accurate retrieval of gas properties, e.g. species concentration and temperature. Using the calibration-free WMS (CF-WMS) strategy, the first harmonic normalised second harmonic signal, e.g. 2f/1f, of the modulated laser transmission is extracted, and then fitted to calculate the path-integrated absorbance. However, the fitting process mainly suffers from (a) noise in the fitting results introduced by the shift of the centre wavelength of the laser, and (b) a relatively high computational cost due to the least square optimisation. To improve the measurement precision and efficiency, this paper proposes a machine learning regression algorithm to calculate the gas properties. The proposed method employs artificial neural networks (ANN) to compute the path-integrated absorbance rapidly with a high signal-to-noise ratio, which was experimentally validated by calculating the absorption of water vapour at the wavelength of 1391.2 nm. In comparison with the traditional fitting method, the proposed machine learning based WMS is two times more noise-resistant with high capability to compute 100 sets of 2f/1f signals in approximately 0.4 s, denoting its potential applicability in real-time and rapid trace gas sensing.

Chemical Species Tomography (CST) using Tunable Diode Laser Absorption Spectroscopy (TDLAS) is an in-situ technique to reconstruct the two-dimensional temperature distributions in combustion diagnosis. However, limited by the lack of projection data, traditionally computational tomographic algorithms are inherently rankdeficient, causing artefacts and severe uncertainty in the retrieved images. Recently, data-driven approaches, such as deep learning algorithms, have been validated to be more accurate and stable for CST. However, most attempts modelled the phantoms using two-dimensional Gaussian profiles to construct the training set, enabling reconstruction of only simple and static temperature fields and can seldom retrieve the dynamic and instantaneous temperature imaging. To address this problem, we use Fire Dynamics Simulator (FDS) to simulate the dynamic and fire-driven reacting flows for training set construction. Based on this training set, a Convolutional Neural Network (CNN) is designed. This newly introduced method is validated by numerical simulation, indicating good accuracy and sensitivity in monitoring dynamic flames.

Tunable diode laser absorption spectroscopy tomography (TDLAST) has been widely applied for imaging two-dimensional distributions of industrial flow-field parameters, e.g., temperature and species concentration. Two main interested imaging objectives in TDLAST are the local combustion and its radiation in the entire sensing region. State-of-the-art algorithms were developed to retrieve either of the two objectives. In this paper, we address the both by developing a novel multi-output imaging neural network, named as Spatially Progressive Neural Network (SpaProNet). This network consists of locally and globally prioritized reconstruction stages. The former enables hierarchical imaging of the finely resolved and highly accurate local combustion, but coarsely resolved background. The later retrieves a fine-resolved image for the entire sensing region, at the cost of slightly trading off the reconstruction accuracy in the combustion zone. Furthermore, the proposed network is driven by the hydrodynamics of the real reactive flows, in which the training dataset is obtained from large eddy simulation. The proposed SpaProNet is validated by both simulation and lab-scale experiment. In all test cases, the visual and quantitative metric comparisons show that the proposed SpaProNet outperforms the existing methods from the following two perspectives: a) the locally prioritized stage provides ever-better accuracy in the combustion zone; b) the globally prioritized stage shows turbulence-indicative accuracy in the entire sensing region for diagnosis of heat radiation from the flame and flame-air interactions.

No abstract

Fast and continuous data acquisition (DAQ)with well resolved spectral information is essential for highspeed and high-fidelity measurement of thermophysical parameters of industrial processes using laser absorption spectroscopy tomography (LAST). However, the state-ofthe-art DAQ systems suffer a) inability to collect raw spectral data in real time due to the very high data throughput; b) degradation of spectral integrity when excessive on-chip down-sampling is implemented to reduce data throughput. In this work, we designed a star-networked and reconfigurable DAQ system for real-time LAST imaging at kilo-Hz frame rate. The DAQ system is embedded with a new field programmable gate array (FPGA)-accelerated digital lock-in (DLI) technique, whereby a cascaded integrator-comb (CIC) filter is implemented for down-sampling of the raw signal with well-maintained spectral information. Furthermore, a customized data-encapsulation protocol is developed to enable continuity of real-time data communication between the front-end DAQ hubs and back-end processor. Performance of the developed DAQ system is experimentally validated by flame temperature imaging at 1 kHz, providing the necessary temporal resolution to penetrate turbulent flow and related industrial processes such as reaction propagation.