DEFLAGRATION TO DETONATION TRANSITION (DDT) IN JET IGNITED HYDROGEN-AIR MIXTURES: LARGE SCALE EXPERIMENTS AND FLACS CFD PREDICTIONS Prankul Middha 1, Olav R. Hansen 1 and Helmut Schneider 2 1 GexCon AS, P.O. Box 6015, Postterminalen, NO-5892 Bergen, Norway; e-mail: prankul@gexcon.com 2 Fraunhofer Institut Chemische Technology, Joseph von Fraunhoferstr. 7, 76327 Pfinztal, Germany As a result of a long history of model development and experimental validation, FLACS is established as a CFD-tool for simulating hydrocarbon gas deflagrations with reasonable precision. FLACS is widely used in petrochemical industry and elsewhere for explosion predictions for input to risk assessments and design load specifications. In recent years the focus on predicting hydrogen explosions has increased. A dedicated project was carried out between 2001 and 2004 to improve the modelling and validation of hydrogen explosions wherein many small and largescale experiments were simulated [1]. With the latest release of FLACS, the validation status for hydrogen explosions is therefore considered good. For hydrogen explosions, deflagration to detonation transition (DDT) can be a significant threat. Recently, FLACS has been extended to indicate the possibility of DDT in realistic situations. As a part of the study, four practical scenarios were simulated and the simulation results were found to compare well with experimental data [2]. The model has now been developed further and used to simulate the experimental investigations performed by Fraunhofer Institute of Chemical Technology. These concerned the transition of a deflagration into a detonation in jet ignited hydrogen air mixtures within a partial confinement [3]. The background for this project was the investigation of the potential hazards for a nuclear power plant, whose process heat is used for the operation of an adjacent chemical plant (e.g. for the gasification of coal), which should be located close to the nuclear plant to minimize heat losses. The test set up consisted of a rectangular container (3 m 1.5 m 1.5 m) with an opening on its front side. The container was followed by 2 parallel walls at a distance of 3 m with a length of 12 m and a height of 3 m. The whole volume was filled with a hydrogen air mixture, enclosed within a very thin PE-foil. The mixture was ignited at the rear side of the container. The experiments observed very high pressures and transition to detonation due to the high turbulence generated by a jet flame shooting into a large, reactive gas cloud followed by reflections of the high speed combustion front from the ground and the walls. The experiments observed DDT for 21% hydrogen concentration, but not for mixtures less sensitive than that [4]. The modeling results are able to capture the experimental observations, including pressure traces and locations of DDT, reasonably well. The possibility of DDT is indicated in terms of a spatial pressure gradient across the flame front. The effect of geometrical dimensions on the observation of DDT is also discussed by comparison with the detonation cell size. The flame speeds of the detonation front are somewhat lower than those observed in the experiments but the development of a shock ignition model is ongoing which is expected to resolve this difference. KEYWORDS: DDT, hydrogen, CFD, FLACS, large scale experiments INTRODUCTION As a result of a more than 25-year history of model development and validation on the basis of experimental results at GexCon, FLACS is established as a CFD-tool for simulating hydrocarbon gas deflagrations with reasonable precision. An extensive knowledge database has been compiled using both experimental and theoretical studies under the aegis of a series of Gas Safety programs (GSPs) that started in 1980. This information has been implemented in the CFD tool FLACS, which was first released in 1986. Today, FLACS is used widely in petrochemical industry and elsewhere for explosion predictions for input to risk assessments and design load specifications. In recent years, there has been a lot of focus on predicting gas explosions involving hydrogen. This is driven by an increasing interest from the nuclear industry, and the ongoing development of hydrogen-fuelled vehicles, which are deemed to be the symbol of a future hydrogen economy. A dedicated project was carried out between 2001 and 2004 to improve the modelling and validation of hydrogen explosions in FLACS wherein many small and large-scale experiments were carried out, combined with simulations and model improvements (Hansen, et al., 2005). Therefore, the validation status of FLACS for hydrogen explosions is considered good. For hydrogen explosions, deflagration to detonation transition (DDT) can be a significant threat. Transition to detonation can occur in a variety of situations, many of which are commonly employed in industrial settings. These include flame acceleration as a result of repeated obstacles (e.g. Peraldi, et al., 1986) and jet ignition (e.g. 1
Knystaus, et al., 1979). There has been a strong debate on the mechanisms underlying the transition to detonation and it is still an active research area. Detailed description of all processes following ignition that may lead to DDT is extremely challenging. This is due to a complicated interaction of compressible flow, chemical reaction, and turbulence that needs to be described at very high spatial and temporal resolution. Much theoretical effort has been focused on development of criteria for DDT (Breitung, et al., 2000) but these criteria and scaling arguments are difficult to apply in a process facility. Although the validation of the current version of FLACS in simulating flame acceleration and high-speed deflagrations is good, a detonation model is lacking. We are currently involved in an activity sponsored by Norwegian Research Council that aims at predicting the extent to which DDT may be expected using FLACS. As a part of this work, FLACS has been extended to indicate the possibility of DDT in realistic situations. Previously, many practical scenarios have been simulated and the simulation results have been found to compare well with experimental data (Middha, et al., 2006). The model has now been developed further and used to simulate the large scale experiments carried out by Fraunhofer Institut Chemische Technology (Fh-ICT) which concerned the transition of a deflagration into a detonation in jet ignited hydrogen air mixtures within a partial confinement (Pförtner and Schneider, 1984; Schneider, 2005). Comparisons have been performed in terms of pressure traces, location and time of DDT, and the possibility of propagation of a detonation front. BRIEF DESCRIPTION OF FLACS FLACS is a computational fluid dynamics (CFD) code that solves the compressible Navier-Stokes equations on a 3-D Cartesian grid. The basic equations used in the FLACS model as well as the explosion experiments to develop and validate FLACS initially have been published (Hjertager, 1985; Hjertager, et al., 1988). A model for development of the flame that describes how the local reactivity changes with parameters like concentration, temperature, pressure, turbulence, etc. is implemented. A good description of geometry and the coupling of geometry to the flow, turbulence, and flame is one of the key elements in the modelling. The real flame area is described properly and corrected for curvature at scales equal to and smaller than the reaction zone. All flame wrinkling at scales less than the grid size is represented by sub-grid models, which is important for flame interaction with objects smaller than the grid size. FLACS uses a standard k-1 model for turbulence. However, some modifications are implemented, the most important being a model for generation of turbulence behind sub-grid objects and a model for flame folding around them (Arntzen, 1998). With the close coupling between sub-grid modelling and turbulence model, it is not believed that using a more advanced turbulence model with more equations and constants will give much added value for the typical simulations carried out with FLACS. The representation of geometry using a distributed porosity concept is one of the key advantages of FLACS compared to several other CFD tools. The geometry is represented with area and volume porosities, as well as wake generating sub-grid object areas in all flow directions. FLACS can therefore be used to simulate all kinds of complicated geometries using a Cartesian grid. Large objects and walls will be represented on-grid; smaller objects will be represented sub-grid. Sub-grid objects will contribute to flow resistance, turbulence generation, and flame folding (for explosions). More details on FLACS can be found elsewhere (Hansen, et al., 2005). DESCRIPTION OF EXPERIMENTS This section provides a short description of the experiments that were conducted at Fh-ICT in 1984 (Pförtner and Schneider, 1984). The background for this project was the investigation of the potential hazards for a nuclear power plant, whose process heat is used for the operation of an adjacent chemical plant (e.g. for the gasification of coal), which should be located close to the nuclear plant to minimize heat losses. The test set up consisted of a driver section that was a rectangular container (3 m 1.5 m 1.5 m). In the front side of the driver section there is a square spaced opening with blocking ratio 0.1 (tests IA1, IA2, IA3) and 0.3 (tests IA4 and IA5). The container was followed by a lane which consisted of 2 parallel walls at a distance of 3 m with a length of 12 m and a height of 3 m (see Figure 2 for details). The whole volume was filled with H 2 -air mixture, enclosed within a very thin PE-foil. The mixture was ignited at the Figure 1. For explosion and dispersion studies representation of the detailed geometry is important for the quality of the predictions. In FLACS this is handled with a porosity concept 2
Figure 2. Test facility, with details of sensors and cameras. The tubes were not installed in test IA1 rear side of the container with 5 pyrotechnic igniters, distributed over the area. In tests IA2 IA5, 2 vertical tubes (diameter 14 cm) were installed, respectively, with a distance of 5 cm from the wall each in the middle and at the end of the lane. It was ensured that the mixture is homogeneous by mixing and sampling. We simulated all experiments, except IA3, and the relevant scenario parameters are presented in Table 1. More information, including all results is presented in Pförtner and Schneider (1984). The experiments observed very high pressures and transition to detonation due to the high turbulence generated by a jet flame shooting into a large, reactive gas cloud followed by reflections of the high speed combustion front from the ground and the walls. The experiments observed DDT for 21% hydrogen concentration, but not for less sensitive mixtures, with the exception of test IA1 where no tubes were installed. Also, DDT was observed to occur near the tube for test IA2 but near the ground for test IA4. No detonation was seen for test IA5. RESULTS This section presents some of the key results of the simulations, and comparisons with experimental data. The simulation domain was resolved by around 3.25 million grid cells with a grid resolution of 5 cm. The grid was according to the FLACS guidelines for explosion Table 1. Relevant scenario parameters for the four tests considered Parameters IA1 IA2 IA4 IA5 Ambient Temp. (K) 279.2 281.5 293 293.4 Ambient pressure (bar) 0.988 0.991 0.993 0.996 H 2 concentration in 21.9 20.8 22.3 19.9 driver unit (%) H 2 concentration in lane (%) 21.0 21.1 22.5 20.0 simulations. The simulations were carried out on a LINUX PC with 1 2 processors and 3 4 GB RAM, and took 2 days to complete. The possibility of DDT is indicated in terms of a spatial pressure gradient across the flame front (DPDX) as it is hypothesized that this parameter is able to visualize when the flame front captures the pressure front, which is the case in situations when fast deflagrations transition to detonation (Middha, et al., 2006). The effect of geometrical dimensions on the observation of DDT is also discussed by comparison with the detonation cell size. Figure 3 presents the comparison of simulated pressure traces for selected sensors at different distances inside the geometry with experimental observations for test IA2. It is seen that the simulations agree reasonably well with measurements, and similar agreement was observed for other sensors. The arrival times of the pressure peaks were consistent with those seen in the experiments, while some discrepancies were seen in peak pressures. Similar comparisons were seen for other tests. Detailed results were not presented for all tests due to lack of space. The maximum simulated pressure in test IA2 was 10.2 barg at sensor 12, compared to 9.2 barg in the experiments also at sensor 12. The flame arrival times were calculated to be 50 and 90 ms at photo transistors F2 and F3, compared to the observed values of 50 and 80 ms. For test IA4, the maximum simulated was 9 barg at sensor 8, compared to around 12 barg in the experiments also at sensor 8. In this case, the flame arrival times at F2 and F3 were calculated to be 35 and 52 ms, compared to observed values of 38 and 49 ms. The maximum pressure in the driver section was also found to compare very well with experimental results for all tests. 2D snapshots of the pressure field at the ground, along with the flame and DPDX for test IA2 at different times with a detailed description of various different stages during the simulation are shown in Fig. 4. The pressure is seen to rise very quickly as the hot flame jet shoots out into the lane, creating a lot of turbulence and mixing of hot products with the unburnt mixture. The hot products act as ignition sources, and a 3
Figure 3. Comparison of simulated pressure traces (left) for selected sensors with experimental data (right) for test IA2. Similar agreement was seen for other monitor points large amount of the unburnt mixture is simultaneously ignited. In the shear layer near the wall, still higher turbulence levels and reflections are seen, with very high pressures and maximum likelihood of DDT. The shock wave is seen to sustain a pressure of 20 barg before it decays as seen in the last picture in Fig. 4. A more pertinent parameter for this work is the parameter DPDX, shown in the bottom part of each snapshot (a value 10 indicates a strong likelihood of DDT if dimension of high DPDX region is significant compared to detonation cell size). For test IA2, the maximum value of DPDX (maximum likelihood of DDT) was seen at the first set of pipes. DDT could also occur before, but with lesser probability. This agrees with the experimental observation of DDT at the left pipe. For test IA4, DDT was predicted to occur at or near the ground next to the sidewalls, and no special likelihood was seen at the pipes. This was also consistent with the experimental result. The simulations also indicated a similar, but somewhat lower possibility of DDT for test IA1 compared to test IA4, as the H 2 concentrations in this test were actually higher than those in test IA2. The values and dimensions of the simulated highly tumultuous region for test IA5 were much smaller, and thus indicated a small chance of DDT. A comparison of the geometrical dimensions with the detonation cell size was also carried out, but that was found 4
IChemE SYMPOSIUM SERIES NO. 153 Figure 4. 2D snapshots of simulation results (test IA2): P (top), flame (middle), DDT indication parameter DPDX (bottom) to be much larger than the minimum required for possible propagation of detonation waves for all tests. The maximum flame speed was seen to be 1208 m/s in test IA2 and 1374 m/s in test IA 4 (compared to 1651 m/s and 1740 m/s in the experiments). These are somewhat lower than those observed in the experiments but the development of a shock ignition model is ongoing which is expected to resolve this difference. However, our calculations also indicated the possibility of DDT in test IA1, which was not seen in the experiments. But as mentioned above, DDT is a very complex phenomenon, and is extremely difficult to predict accurately. Also, the absence of a detonation model leads to the decay of the shock front, even after DDT is expected to happen. In general, the modeling results are able to capture the experimental observations, including pressure traces and locations of DDT, reasonably well. We hope that the current model, when coupled with the additional features currently under development, can be used by the process industry to get a fair idea of the danger of DDT. CONCLUSIONS Large-scale experiments carried out at Fh-ICT have been simulated using the CFD tool FLACS. In general, the modeling results are able to capture the experimental observations, including maximum pressures, arrival times, and locations of DDT, reasonably well. However, some discrepancies are seen, which may be attributed to experimental uncertainties and the very difficult nature of the simulations. The flame speeds of the detonation front are somewhat lower than those observed in the experiments 5
but the development of a shock ignition model is ongoing which is expected to resolve this difference. Also, the absence of a detonation model leads to the decay of the shock front, even after DDT is expected to happen. The support of Norwegian Research Council for this work is acknowledged. REFERENCES 1. Arntzen, B.A., 1998. Modeling of turbulence and combustion for simulation of gas explosions in complex geometries, PhD Thesis, NTNU, Trondheim, Norway. 2. Breitung, W., et al., 2000. Flame Acceleration and Deflagration to Detonation Transition in Nuclear Safety. State-of-the-Art Report, OECD Nuclear Energy Agency, Ref. NEA/CSNI/R/2000/7. 3. Hansen, O.R., Renoult, J., Sherman, M.P., and Tieszen, S.R., 2005, Validation of FLACS-Hydrogen CFD Consequence Prediction Model Against Large Scale H 2 Explosion Experiments in the FLAME Facility, Proceedings of International Conference on Hydrogen Safety, Pisa, Italy, September 2005. 4. Hjertager, B.H., 1985, Computer simulation of turbulent reactive gas dynamics. J. Model. Identification Control, 5: 211 236. 5. Hjertager, B., Fuhre, K., Bjorkhaug, M., 1988, Gas explosion experiments in 1:33 and 1:5 scale offshore separator and compressor modules using stoichiometric homogeneous fuel/air clouds, J. Loss Prevention Proc. Ind., 1: 197 205. 6. Knystautas, R., Lee, J. H., and Wagner, H. G., 1979, Direct initiation of spherical detonation by a hot turbulent gas jet. Proc. Comb. Inst., 17: 1235 1245. 7. Middha, P, Hansen, O. R., and Storvik, I. E., 2006. Prediction of deflagration to detonation transition in hydrogen explosions. Proceedings of the AIChE Spring National Meeting and 40 th Annual Loss Prevention Symposium, Orlando, FL, April 23 27, 2006. 8. Peraldi, O., Knystautas, R., and Lee, J. H., 1986, Criteria for transition to detonation in tubes. Proc. Comb. Inst., 21: 1629 1637. 9. Pförtner, H., Schneider, H., Tests with Jet Ignition of Partially Confined Hydrogen Air Mixtures in View of the Scaling of the Transition from Deflagration to Detonation, Final Report for Interatom GmbH, Bergisch Gladbach, Germany, Oct. 1984, Fraunhofer ICT Internal Report. 10. Schneider, H., 2005, Deflagration and deflagration to detonation transition within a partial confinement similar to a lane Proceedings of International Conference on Hydrogen Safety, Pisa, Italy, September 2005. 6