Tomographic absorption spectroscopy for the study of gas dynamics and reactive flows

The computed tomography of chemiluminescence (CTC) technique was applied for the first time to a real highly turbulent swirl flame setup, using a large number of CCD cameras (Nq=24 views), to directly reconstruct the three-dimensional instantaneous and time-averaged chemiluminescence fields. The views were obtained from a 172.5° region (in one plane) around the flame, and the CTC algorithm [Floyd et al., Combust. Flame158, 376 (2011)CBFMAO0010-2180] was used to reconstruct the flame by discretizing the domain into voxels. We investigated how the reconstructions are affected by the views’ arrangement and the settings of the algorithm, and considered how the quality of reconstructions should be assessed to ensure a realistic description of the capabilities of the technique. Reconstructions using Nq≤12 were generally better when the cameras were distributed more equiangularly. When Nq was severely low (e.g., 3), the reconstruction could be improved by using fewer voxels. The paper concludes with a summary of the strengths and weaknesses of the CTC technique for examining a real turbulent flame geometry and provides guidance on best practice.

Instantaneous 3D imaging of highly turbulent flames using computed tomography of chemiluminescence.

Measurements from flame chemiluminescence tomography of forced laminar premixed propane flames

Flame chemiluminescence tomography (FCT) plays an important role in combustion monitoring and diagnostics due to the easy implementation and non-intrusion. However, on account of the high data throughput and the inefficiency of the conventional iteration methods, the 3D reconstructions in FCT are typically conducted off-line and time-consuming. In this work, we present a 3D rapid FCT reconstruction system based on convolutional neural networks (CNN) model for practical combustion measurement, which has the ability to reconstruct 3D flame distribution rapidly after training process. First, the numerical simulation has been performed by creating three cases of phantoms which are designed to mimic the 3D conical flame. Next, after the evaluation of loss function and training time, the optimal CNN architecture has been determined and certificated quantitatively. Finally, a real time FCT system consisting of 12 color CCD cameras is realized and multispectral separation algorithm is adopted to extract CH* and C2* components. Certificated by practical measurements testing, the proposed CNN model is able to reconstruct 3D flame structure from real time captured projections with credible accuracy and structure similarity. Furthermore, compared with conventional iteration reconstruction method, the proposed CNN model shows better performance on obviously improving reconstruction speed and it is expected to achieve 3D rapid monitoring of flames.

Three-dimensional rapid flame chemiluminescence tomography via deep learning

Combustion science is an interdisciplinary study that involves nonlinear physical and chemical phenomena in time and length scales, including complex chemical reactions and fluid flows. Combustion widely supplies energy for powering vehicles, heating houses, generating electricity, cooking food, etc. The key to studying combustion is to improve the combustion efficiency with minimum emission of pollutants. Machine learning facilitates data-driven techniques for handling large amounts of combustion data, either through experiments or simulations under multiple spatiotemporal scales, thereby finding the hidden patterns underlying these data and promoting combustion research. This work presents an overview of studies on the applications of machine learning in combustion science fields over the past several decades. We introduce the fundamentals of machine learning and its usage in aiding chemical reactions, combustion modeling, combustion measurement, engine performance prediction and optimization, and fuel design. The opportunities and limitations of using machine learning in combustion studies are also discussed. This paper aims to provide readers with a portrait of what and how machine learning can be used in combustion research and to inspire researchers in their ongoing studies. Machine learning techniques are rapidly advancing in this era of big data, and there is high potential for exploring the combination between machine learning and combustion research and achieving remarkable results.

Machine learning for combustion

Flame chemiluminescence tomography (FCT) has been widely used in flame diagnostics for three-dimensional (3D), spatially resolved measurements of instantaneous flame geometry and, to some extent, of species concentrations. However, in most studies, tomographic reconstructions are based on a traditional parallel projection model. Due to the light collection characteristics of a lens, a parallel projection model is not appropriate for the practical optical setups that are used for emission imaging, particularly at small F-numbers. Taking the light collection effect of the lens into account, this Letter establishes a complete and novel tomographic theory for a multi-directional tomography system consisting of a lens and CCD cameras. A modified camera calibration method is presented first. It determines the exact spatial locations and intrinsic parameters of the cameras. A 3D projection model based on the lens imaging theory is then proposed and integrated into the multiplicative algebraic reconstruction technique (MART). The new approach is demonstrated with a 12-camera system that is used to reconstruct the emission field of a propane flame, thereby resolving space and time.

Multi-directional 3D flame chemiluminescence tomography based on lens imaging.

A model for arc plasma's optical diagnosis by the measurement of the refractive index

This research proposes an issue that has previously been omitted in flow field temperature reconstruction by optical computerized tomography (OCT). To prove that it is not reasonable to always assume an isobaric process occurs when OCT is adopted to obtain the temperature distributions of flow fields, a propane–air flame and an argon arc plasma are chosen as two practical examples for experiment. In addition, the measurement of the refractive index is achieved by moire deflection tomography. The results indicate that the influence of pressure distribution on temperature reconstruction is a universal phenomenon for various flow fields. Hence, the condition that can be introduced to estimate when an isobaric process can no longer be assumed is presented. In addition, an equation is offered to describe the temperature reconstruction imprecision that is caused by using the supposed pressure instead of the practical pressure.

Zhen-hua Li

Papers

Multi-directional 3D flame chemiluminescence tomography based on lens imaging.

A model for arc plasma's optical diagnosis by the measurement of the refractive index

Influence of pressure distribution on flow field temperature reconstruction.

A uniform description of the gas and plasma flow fields' refractive index

Flow field's temperature partition reconstruction based on its phase distribution