Experimental and Numerical Characterization of the Self-Excited Dynamics Behavior of a Technically Premixed Burner

Abstract

In this paper, the numerical findings of a high fidelity CFD model will be compared with the experimental data of a test campaign devoted at characterizing the performance of a technically premixed industrial burner regarding the thermoacoustic instabilities. The data are retrieved at relevant gas turbine conditions in a test bench where the flame tube can change its length during the test execution allowing its fundamental acoustic frequencies to be modified and, in case, triggered. Mimicking the test configuration, several Large-Eddy Simulations are performed with different lengths of the flame tube in order to verify the ability of the numerical model to reproduce the excited dominant frequency and the corresponding limit cycle amplitude measurements. The numerical model demonstrates the ability to correctly reproduce the frequency triggered during the test and to reach different limit cycle amplitudes along different flame tube lengths in agreement with the tests, as well. However, it is found that the amplitude of the acoustic pressure fluctuation during the limit cycle is generally under-predicted. Despite this, the proposed approach demonstrates to be a robust tool for the characterization of a given design, allowing to dramatically reduce the computational cost of the analysis, at least in the early design phase. Since the numerical model can correctly reproduce the behavior of the investigated design, a deep post-processing of the solutions is performed to shed light on the physical mechanisms sustaining the thermo-acoustic instability. Among the numerical techniques employed at this purpose, the Phase-Locked Average and the Extended-POD are applied trying to correlate the fluctuations of the different quantities inside the premixed channel of the burner and the primary zone as well.

E-POD investigations of turbulent premixed flame dynamics approaching lean blow-out conditions

“…In any case, this analysis suggests that, despite the flame profile is wrinkled by the identified vortices, there is still a strong anchoring region represented by the high temperature core sustaining the heat release inside the IRZ. As clearly visible from Figure 8 and differently from the operating conditions far from the LBO limits characterized by the presence of hot products also in the ORZ, 3,13,[21][22][23][24][25][26][27][28][29][30] in this case the operability of the flame is ensured exclusively by the IRZ. So, in order to retrieve some clues of the destabilizing mechanism leading to the loss of flame, the discussion will be focused on the analysis of the IRZ and its interaction with the velocity field.…”

“…The Proper Orthogonal Decomposition represents one of the most used algorithm for the detection of coherent structures in turbulent flows, 3,13,25 allowing a given data set of acquisitions D(x, t), describing only the fluctuating part of a phenomenon, to be decomposed into a combination of orthogonal basis, or modes, function exclusively of space ϕ k (x) and time ψ T (t), each one with a different energy content σ k :…”

“…The Proper Orthogonal Decomposition represents one of the most used algorithm for the detection of coherent structures in turbulent flows, 3,13,25 allowing a given data set of acquisitions D false( x , t false), describing only the fluctuating part of a phenomenon, to be decomposed into a combination of orthogonal basis, or modes, function exclusively of space ϕ k false( x false) and time ψ T false( t false), each one with a different energy content σ k: These quantities are here calculated according to the so-called snapshot-method introduced by Sirovich et al 26 In this approach, once N different numerical or experimental samplings have been collected, the problem is reduced to find the eigenvalues σ k and the eigenvectors ψ T false( t false) of the cross-correlation matrix K , 0.25em built according to Eq. 2. Especially when the dimension of the data base is extremely large, it can be decided to compute only the first N false0 false2 ⌣ eigenvalues of the cross-correlation matrix: this truncation leads to a percentage error in the reconstruction of the fluctuating part of the solution that can be limited when N false0 false2 ⌣ false→ N...…”

Hydrogen Combustion for an Industrial Gas Turbine Burner: OH* Chemiluminescence Preliminary Measurements

“…Fuel system is organized in such a way that each line to the fuel burner is equipped with a mixing pipe where different species can be fed; hydrogen and ammonia have their own skid for staged expansion, heating, flow controlling and safety. The tested fuel burner is a modification of the standard component of NovaLT ™ 16 and NovaLT ™ 12 and it is already described in previous papers [5,6]; it is a fully premixed type made by four channels integrally printed in Additive Manufacturing with the main fuel injection in the premixer channel (Line 1) and a secondary injection downstream close to the burner tip (Line 2).…”