Four-Dimensional Measurement of Thermoacoustic Oscillations

自阿波罗计划初期以来，热声振荡就一直困扰着燃烧工程师。今天，人们越来越认识到大幅减少污染物排放的必要性，这对用于地面燃气轮机和航空发动机的热力涡轮机械变得重要，旨在实现气候中性航空和可持续燃料。最先进的技术提高了这些低排放机器对热声振荡的敏感度。在FWF和DFG之前资助的一个项目中，我们在格拉茨和德累斯顿的联合小组证明了所谓的“基于相机的激光干涉测振仪”(CLIV)可以定量记录这些密度和热释放振荡。但是，非对称火焰需要多方向的观测，以便能够三维重建局部密度结构、声音产生和对流速度。这种需要与CLIV的有限视场(FOV)相冲突。多向背景纹影法(3D-BOS)可以解决这一问题，但需要用激光干涉测振法进行标定。此外，采用多次曝光技术的密度标记测速仪(DTV)可以增加火焰动力学的信息，从而增加第四维的信息。利用特殊训练的深度神经网络(DNN)来解决这种具有有限角度投影和缺失数据的四维不适定问题，最有希望的数值方法是应用特殊训练的深度神经网络(DNN)。该项目的基本假设是，基于CLIV与3D-BOS、DTV和DNN的组合的局部热声振荡的四维检测将揭示燃烧、声学和流体动力学的局部和耦合信息。方法：从时间进度的角度--设计或重新设计，施工和采购，测试和测量，以及最终的实验资格--以传统的方式组织项目。组织水平：这种新方法得到了最近数码相机技术范式转变和人工智能(深度学习)快速发展的支持，这两个方面都是由于可用计算能力的增加。DNN方法还为其他更成熟的技术的组合增加了一个创新元素，并将促进实验燃烧研究，以及空气声学和流体动力学。

Thermoacoustic oscillation troubles combustion engineers since the early days of the Apollo program. Today there is an increasing awareness regarding the necessity of a dramatic reduction of pollutant emissions, becoming important to thermal turbomachinery used in ground-based gas turbines and aero-engines aiming towards climate-neutral aviation and sustainable fuels. State-of-the art technology increases the susceptibility to thermoacoustic oscillations in these low-emission machines. In a previous project funded by FWF and DFG, our joint groups in Graz and Dresden proved that a so-called “camera-based laser interferometric vibrometer” (CLIV) can quantitatively record these density and heat release oscillations. But, non-symmetric flames require multidirectional observation in order to enable three-dimensional reconstructions of local density structures, sound production and convection velocities. This need is in conflict with the limited field of view (FOV) of CLIV. The use of multidirectional Background-Oriented-Schlieren method (3D-BOS) could solve this problem, but needs calibration by laser interferometric vibrometry. Additionally, a density tagging velocimetry (DTV) by multiple exposure technique could add information on the flame dynamics and thus the fourth dimension. A most promising numerical approach to solve this four-dimensional ill-posed problem with limited-angle projections and missing data is the application of specially trained, deep neural networks (DNN).HYPOTHESIS: The underlying hypothesis of this project claims that four-dimensional detection of local thermoacoustic oscillations, based on the combination of CLIV with 3D-BOS, DTV and DNN as binding element, will reveal local and coupled information on combustion, acoustics and fluid dynamics. METHODS: The project is organized in a conventional way from the point of view of time schedule - design or redesign, construction and procurement, testing and measurement, and final experimental qualification.LEVEL OF ORGINALITY: This new approach is supported by the recent paradigm shift in digital camera technology and the rapid development of artificial intelligence (deep learning), both due to the increased computational power available. The DNN approach also adds an innovative element to the combination of the other, more established techniques, and will booster experimental combustion research, as well as, aeroacoustics and fluid dynamics.