DNS dataset for modelling homogeneous ignition processes of clustering solid particle clouds in isotropic turbulence

Abstract

This dataset is being published to enable the development of models for igniting and combusting solid particles in isotropic turbulence using flamelet tabulated chemistry. This dataset is generated using the forced homogeneous isotropic turbulence in order to investigate the effect of active turbulent forces on the ignition of particles, and it is used as the supplementary material for the manuscript "Modeling homogeneous ignition processes of clustering solid particle clouds in isotropic turbulence", which was accepted for publication in the special issue of Fuel Journal for the proceeding of the 4th international Oxyflame workshop. The particles are chosen to have near unity Stokes numbers, which promote particle clustering to study the ignition phenomenon for particle clusters, which is common during ignition and combustion of particle clouds in large industrial solid fuel-powered burners. Since the ignition process of clustering solid particle clouds is transient, different time instances during the ignition process are presented to facilitate the modelling effort for the transient ignition behaviour of the particles. The dataset consists of gas-phase data, particle data and the most important routines required to process the dataset.   

This database is a valuable resource for users developing solid fuel ignition and combustion in turbulent conditions. 

It should be noted that this dataset is a reduced version of the full dataset in order to size limitations in the sharing platforms. More information and full dataset can be provided upon request. For more information, contact: [e-mail]

Technical details

The data provided in this dataset contains gas phase data, particle data, and some post-processing scripts for visualization of the data. The data is generated in forced homogeneous isotropic turbulence (HIT) with an initial preferential concentration of the particles in a hot atmosphere to study the impact of particle clustering on ignition. Simulations were performed within a region with the physical size of 12.8mm * 12.8mm * 12.8mm with periodic boundary conditions in all directions. The domain size is discretized with a three-dimensional cartesian mesh with a resolution of Δx = 50 μm. A forced isotropic turbulent field with Re_λ=30 and the Kolmogorov length scale η =100 microns has been chosen. The dispersed phase consists of 10,000 particles of Colombian coal with D_p= 20 microns and T0=300K and with an apparent density of 700kg/m3. Non-reactive particles are first randomly distributed in the box filled with air with 20% oxygen and an initial gas temperature of T = 1500K, which is relevant to practical PCC applications. These conditions lead to an initial Stokes number of around 5, for which a clustering behaviour in particle cloud motion is expected. The employed forced isotropic turbulence ensures maintaining the same turbulence statistics during non-reactive and reactive simulations, as summarised in the following table:

time [ms]

Re_λ

Re_Turb

η[m]

l_t[m]

t_η[ms]

t_l[ms]

St

0
30.7
141.8
1.04e-4
4.27e-3
4.47e-2
5.33e-1
6.199

0.46
30.9
143.9
9.94e-5
4.13e-3
4.16e-2
4.99e-1
6.207

0.5
30.5
140.1
9.96e-5
4.06e-3
4.27e-2
5.05e-1
6.333

0.55
29.4
129.8
1.01e-4
3.88e-3
4.34e-2
4.94e-1
6.032

0.6
28.5
122.1
1.02e-4
3.76e-3
4.46e-2
4.92e-1
5.775

0.65
27.7
115.2
1.04e-4
3.66e-3
4.48e-2
4.81e-1
5.365

η and t_η are the respective Kolmogorov length and time scales, and l_t and t_l correspond to the integral length and time scales.

Gas phase and particle data:

The gas phase data has HDF5 formation, which contains selected scalars relevant to model developments. Since the ignition process is a transient process, two different time instances, t=0.5ms with the maximum number of ignited regions during the ignition process and t=0.65ms at the end of the ignition process, are chosen. Before starting the reactive simulations, a non-reactive simulation for 20.5ms was performed to form the particle clusters, and then the reactive simulation was started. Therefore, the data.out_2.100E-02.h5 corresponds to t = 0.5ms and data.out_2.115E-02.h5 corresponds to t = 0.65ms. Each HDF5 dataset has the following structure:

Group Flow:

U, V, W

Group scalar: 

CO, CO2, H2O, C2H2, O2, N2, OH, progress variable (PV), temperature(T), enthalpy(h), heat capacity (Cp), density(RHO), pressure(P), mixture fraction(Z), dissipation rate, heat conductivity

Group C_dot:

Molar production rate of CO, CO2, O2, OH

Group ST:

Mass fraction rate of OH

Group var (data required for subfilter analysis and PDF modelling):

RHO * h

RHO * h * h

RHO * PV

RHO * PV * PV

RHO * RHO

RHO * RHO * RHO

RHO * T

RHO * T * T

RHO * Z

RHO * Z * Z

These gas phase data can be used to study the Eulerian field, which is impacted by the particles through the source terms, which are obtained from the Lagrangian framework. For the particle data, two CSV files containing information about each particle's position and temperature are provided. 

In each CSV file, which corresponds to the same time instance as the gas phase data, this structure can be found:

Points_0, Points_1, Points_2: particle position x,y,z

coal_defaultT: particle temperature

Scripts:

Different scripts are provided for a reader to enable the first-time use of the data, such as visualising the gas phase quantities, calculating the conditional mean and statistical analysis, and clustering analysis for the particles. Here is a brief information regarding the scripts:

hdf5_2D_Con_Mean_Plots.m: This script can be used for loading the HDF5 file and visualising the field, joint PDF, and joint correlation of different quantities with respect to different parameters. This script requires another module to calculate the conditional mean of the data.

Compute_ConditionalMean_histograms_2D.m: This module calculates the conditional average of a 2D array.

clustering_voronoi_3D.m: This script can analyse the particle data and calculate the clustering limit based on the Voronoi algorithm. It can also filter the clustered particles and separate different clusters using the nearest neighbour and DB-SCAN methods.