Gas Explosions in Vessels, Pipes, Channels and Tunnels
When we analyse an internal explosion, we will find that the gas cloud size is the main parameter determining pressure build-up. The geometrical conditions nearly always support flame acceleration and pressure build-up. So, if a large cloud is formed within equipment it is likely that there will be a severe explosion if it ignites.
An internal explosion may result in loss of containment. The subsequent event can then be strong blast waves from high pressure reservoirs, fires or toxic releases.
In the chemical and hydrocarbon process industries, we will find a large variety of cases where internal gas explosions may occur. Such explosions can be caused by uncontrolled leaks, or simply by accidental purging with air (and thereby formation of fuel-air mixtures). There is limited information available in the open literature about these aspects of gas explosions. It is beyond the scope of this chapter to present detailed methods for analysing gas explosions in such systems.
The objective of this chapter is:
- To present the basic physics of internal gas explosions.
- To point out which phenomena can occur during an internal gas explosion.
- To indicate worst case scenarios.
9.1 Closed Vessels
A closed vessel often has very small openings, such as connected pipes, rupture disks or relief valves through which pressure can be relieved during a gas explosion. In this case, the relief process is often too slow to relieve the pressure fast enough, and the vessel may behave like a fully closed vessel with regard to pressure build-up. The pressure build-up will mainly depend on type and concentration of fuel, the initial pressure, the filling ratio in the vessel, the burning rate, the venting and the oxidiser.
In the first part of this discussion we will assume that the flame is a slow deflagration with a velocity of less than 20% of the initial speed of sound (i.e. in fuel-air at 1 atm. and 25°C less than 70 m/s), hence local high pressures due to high burning rate is neglected. For a slow deflagration in a homogeneous gas mixture, the pressure in the vessel will gradually increase as the flame consumes the gas mixture. As shown in Figure 5.6 the maximum pressure will be reached when the combustion has been completed. For most hydrocarbon fuels, a Stoichiometric fuel-air cloud with initial pressure of 1 atm. will give 8-10 bar pressure (See Table 4.5), when burning under constant volume conditions. Figure 9.1 shows the pressure for a constant volume combustion as function of percentage fuel in air for homogeneous methane- and ethylene-air mixtures. The highest pressure is found for slightly rich mixtures, i.e. slightly higher concentration than the Stoichiometric mixture which is 9.5% for methane and 6.54% for ethylene. When the fuel concentration approaches the flammability limits, the explosion pressure will be reduced, but even close to the flammability limits, the theoretical values for constant volume combustion will be in the 4-5 bar range. Even a cloud near the flammability limit can, in an explosion, cause significant pressure build-up in a closed vessel.
||Explosion pressure predicted by STANJAN for constant volume combustion of ethylene- and methane-air at 1 bar and 25°C (Salater 1991).|
The initial pressure is a parameter which influences the explosion pressure at constant volume conditions. By increasing the initial pressure, the energy content, i.e. heat of combustion, per unit volume will increase. Bartknecht (1981) has given some measurements of explosion pressure for slow deflagration for propane in a 7 litre spherical vessel. These results are shown in Figure 9.2. There is a nearly linear relation between initial pressure and explosion pressure. For Stoichiometric fuel-air the pressure increase at constant volume will be approximately 8 times the initial pressure. For other oxidisers than air, such as pure oxygen, oxygen enriched air or chlorine, higher constant volume explosion pressures can be expected. To estimate constant volume explosion pressure for specific fuel-air or fuel/oxidiser mixtures at given initial conditions, programs like STANJAN can again be used.
||Explosion pressure versus initial pressure for Stoichiometric propane-air in a 7 litre vessel (Bartknecht 1981)|
In many situations only a portion of the vessel will be filled with combustible gas. The explosion pressure for a partly filled vessel is shown in Figure 9.3. Note that in this figure, it is assumed that the cloud has a Stoichiometric composition even though it occupies only part of the vessel. If about 15% of the closed vessel is filled with a Stoichiometric cloud and this cloud burns, the pressure in the vessel will be doubled. Even 1-2% filling ratio may cause problems for large vessels or tanks operating at atmospheric conditions. They are often very weak and may rupture at a few hundred mbar overpressure. This shows that even low filling ratio can cause significant increase of pressure in closed vessels. To avoid this problem, controlled venting is recommended as a mitigation device.
||Approximate values for pressure increase versus filling ratio in a closed vessel.|
Lees (1980) states that detonations occur in pipelines, but are very improbable in vessels. In an empty vessel there are no obstructions causing turbulence and flame acceleration. Transition to detonation is therefore not likely in vessels, unless the gas is very detonable (small cell size), the gas cloud is large, the cloud is jet ignited or the vessel contains obstacles.
In addition to closed vessels, pipes (including channels and tunnels) are also typical simple geometries where internal explosions can occur. In pipes, the pressure generated by the flame has the possibility to propagate away from the combustion front. For long pipes or open ended pipes, a high flame speed is required to generate high explosion pressure. Figure 5.5 shows the relation between flame speed and explosion pressure. The planar case is applicable for pipes. The main mechanism causing the flame to accelerate in pipes, is turbulence. When the gas burns, it expands and pushes unburnt gas ahead of the flame front. The flow ahead of the flame will cause a turbulent boundary-layer to grow and the turbulence will enhance the burning rate. This is illustrated in Figure 9.4.
||Flame acceleration in a pipe, channel or tunnel.|
Bartknecht (1971) has measured flame velocities in a 1.4 m diameter pipe with methane-air at 1 atm. The pipe was 40 m long and the end was either closed or open. The results are shown in Figure 9.5. The highest flame speed was observed when the gas was ignited in the closed end and the other end was open. In that case the gas ahead of the flame was pushed through the pipe and a lot of turbulence was generated. When the pipe was closed in both ends, the flame accelerated fast in the beginning, but after 15-20 m the flame started to decelerate, because the flow ahead of the flame is obstructed by the closed end. Since the pipe is closed in both ends, the pressure will increase like in a closed vessel. In the third case the ignition is at the open end and the other end is closed. Here, the flow velocity and the turbulence level ahead of the flame are very low and the flame propagates at low velocities through the pipe. These experiments show the importance of boundary conditions for the flame acceleration in a pipe. The boundary conditions in a pipe will be similar to ignition at the closed end of a pipe if the gas cloud is ignited in the centre of the cloud. In that case the flame will propagate in both directions and there will be zero flow velocity where the ignition took place (i.e. symmetry plane).
||Flame speed in a 1.4 m diameter pipe with methane-air. (Bartknecht 1971)|
In a pipe the flame can continue to accelerate until it becomes a detonation (a supersonic combustion wave propagating at 1500-2000 m/s in fuel-air). As discussed in Chapter 6 we have only a qualitative understanding of the mechanism of transition from deflagration to detonation. We are therefore not capable of predicting this phenomenon. Experimental data is all that is available. The transition phenomenon is characterised by very high local pressures, pressures of 50 times the initial pressure have been measured when transition to detonation has occurred. In accidental situations, very strong damage can be observed at the location of transition to detonation . A case history from a gas explosion in a pipe is illustrated in Figure 9.6. At one particular location the pipe was expanded radially. That was the place where the transition to detonation took place. When the detonation propagated further down it stabilised at a so-called CJ-condition, which gives lower pressure. In the case history, the pipe did not rupture. If the pipe had ruptured, a high pressure reservoir would have been released. This shows that transition to detonation in pipes, channels and tunnels is a hazardous phenomenon which should be recognised as being possible.
||A case history. Transition to detonation deformed the pipe.|
The run-up distance, i.e. the distance from ignition to transition to detonation in pipes is an experimental value giving some indication of the likelihood of transition to detonation. Steen and Schampel (1983) have reviewed experimental investigations of the run-up distance of gaseous detonations in large pipes. The experimental conditions, i.e. pressure, temperature and gas mixture, are limited compared with the actual conditions in the industry. The data presented by Steen and Schampel are mainly for 1 atm. and fuel-air mixtures. Figure 9.7 shows the run-up distance for Stoichiometric ethylene and propane/air versus pipe diameter. The run-up distance increases with increasing pipe diameter. The turbulent boundary-layer ahead of the flame is filling a relatively larger portion of the tube in small pipes than in large pipes. The fuel concentration is also an important factor for the run-up distance. This can be seen in Figure 9.8.
Several other factors also influence the run-up distance. Experiments show that it decreases with
i) increasing initial pressure
ii) decreasing initial temperature, and
iii) increasing turbulence in the pipe (i.e. obstructions in the pipe).
In general we can say that the run-up distance depends on the reactivity and cell size. The smaller the cell size, and the more reactive the mixture (burning velocity), the shorter is the run-up distance.
||Run-up distance to detonation versus pipe diameter. (Steen and Schampel, 1983).|
||Run-up distance for ethylene- and propane-air. Pipe diameter 50 mm. (Steen and Schampel, 1983).|
9.3 Pressure Piling
In a process unit or an underground system, we will find that large volumes (i.e. tanks, unit operations and rooms) are interconnected by pipes and channels. In case of an internal explosion, these inter-connections can cause very strong pressure build-up. This phenomenon is often referred to as precompression or pressure piling. Pressure piling is a local dynamic effect which can cause high local explosion pressures.
||Experiments with pressure piling. (Heinrich, 1988).|
Figure 9.9 illustrates such a situation. Volumes 1 and 2 are interconnected by a pipe. Heinrich (1988) performed some laboratory tests with the geometry shown in Figure 9.9. The volume of the tanks was 12 l and the pipe was 50 cm long. When the cloud is ignited in Volume 1, the pressure will gradually increase and some unburnt gas from Volume 1 will flow into Volume 2. When the flame enters Volume 2, the gas is pre-compressed. The flame will now be a jet flame shooting into Volume 2. As discussed in section 5.7 such jet flames will cause fast pressure build-up. If the pressure relief back into Volume 1 through the pipeline is not sufficient, the pressure in Volume 2 will become much higher than in Volume 1, due to the pressure piling and the jet flame. In Heinrich's experiment the maximum pressure was 30% higher than that predicted from the constant volume conditions.
- Avoid combustible mixtures that can cause internal explosions. It is bad practice to rely solely on elimination of sources of ignition.
- To calculate constant volume conditions and detonation pressure tools like STANJAN can be used.
- For reactive mixtures take the possibility of transition to detonation into account.
- Do not design atmospheric vessels too strong. If they rupture, they should rupture at low pressure, not at several bars.
- Design vessels with relief valves and/or rupture disks.