RUNAWAY STUDIES OF COMPLEX REACTION SYSTEMS
M. Papadaki1*, E.Marqués-Domingo[2], Jun Gao1 and T. Mahmud1
1. Chemical Engineering, The University of Leeds, Leeds LS2 9JT, UK
* Author to whom correspondence should be addressed
Tel:+44 (0)113 343 2420, Fax+44 (0)113 343 2405 e.mail:
ABSTRACT
A number of runaway scenarios of the excess of hydrogen peroxide used during the N-oxidation of alkylpyridines, under closed and open conditions, were examined. It was found that, in most cases, if the volume of the liquid hydrogen peroxide solution occupies more than 10% of the total volume of a closed system (e.g. reactor and vent line between reactor and blockage), the production of gases raises the pressure so quickly that evaporation is completely suppressed. Higher that 70% filling levels result in complete expansion of the liquid. The MTSR(t) of the system falls rapidly if the normal process temperature is high, but if a runaway occurs exactly at the end of dosing, MTST will be very high and secondary decompositions will rapidly develop. The results of this study are currently being used to critically assess the current approaches and to further the study of inherently safer designs.
Keywords: reaction runaway, hydrogen peroxide, alkylpyridines, N-oxidation
INTRODUCTION
A reaction thermal runaway begins when the heat produced by the reaction exceeds the heat removed. The temperature rises, the reaction rate increases causing a further increase in the rate of the heat generation, until limiting reactant is consumed. The temperature can be very rapid leaving a limited time for correction. More violent secondary decomposition reactions may then initiate resulting in higher temperature and pressure and a possible explosion, which can also be followed by release of toxic or flammable gases. In order to deal with thermal chemical-reaction hazards it is necessary to identify them, to assess the likeliness of their occurrence and the seriousness of their consequences (Barton& Nolan 1989)
According to Marco, Peña, & Santamaría (1997) if emergency actions are necessary the timing of them is critical. If they are taken too early the result may be unjustified reactor shutdown, which is always expensive and dangerous, and a delayed response could lead to an accident of extremely severe consequences. It is therefore necessary to develop methods capable of predicting the onset of the runaway reaction, as it can be a computer-based modelling of reactions based on well-known kinetics.
In this work, we present the predictions of selected runaway scenarios of the decomposition of the excess of hydrogen peroxide used in the N-oxidation of alkylpyridines. These predictions have been produced by a model that utilises results of a thorough calorimetric kinetic study of the N-oxidation of 2-methylpyridine (Sempere, Nomen, Rodriguez & Papadaki, 1998).
REACTION AND KINETICS EMPLOYED
Hydrogen peroxide is one of the most powerful oxidizers known. Through catalysis, H2O2 can be converted into hydroxyl radicals with reactivity second only to fluorine. Hydrogen peroxide is generally supplied in aqueous solutions. This study focuses on industrial solutions, which range in strength from 27.5% to 70%.
Hydrogen peroxide oxidative power, and the gas and heat generated by its decomposition, are responsible for the large majority of hazardous situations that arise during its storage, handling and use. Fires and runaway decompositions are the most common incidents, many of the latter causing pressure-bursts of equipment.
As reported by Mackenzie (1990), the normal rate of decomposition for commercially supplied hydrogen peroxide is extremely low, well under 1% loss per year. However, significantly faster breakdown can be triggered in the user's plant by even minute amounts (less than 1 ppm) of soluble impurities, or by contact with incompatible (e.g., mild steel, copper, brass) rough surfaces.
Hydrogen peroxide plays an important role in the industry and further research onto its runaway behaviour is imperative.
This work is part of an on-going research project involving the semi-batch N-oxidation of alkylpyridnes. The N-oxidation is performed in the liquid phase using an excess of hydrogen peroxide, which is added linearly in the reactor. The reaction is homogeneously catalysed by phosphotungstic acid and it is performed isothermally at a temperature near its boiling point. An overhead condenser is used to condense the produced vapours during the reaction course. The macroscopically observed reactions are the N-oxidation synthesis and the unwanted hydrogen peroxide decomposition. (Sempere et al., 1998 & Papadaki, Stoikou, Mantzavinos, & Rodriguez-Miranda, 2002)A global kinetic model for this reaction system has been previously developed for 2-methylpyridine (Sempere etal., 1998). As shown in our previous research (Sempere et al., 1998 & Papadaki et al., 2002), the N-oxidation reaction essentially finishes soon after the end of dosing while the decomposition of hydrogen peroxide takes place for a few more hours. Typical concentration profiles of the reactants are shown in Figure 1.
Figure 1.
In this work we are considering the consequences of a potential runaway of the hydrogen peroxide decomposition reaction owing to a cooling failure just after the end of dosing (when the synthesis reaction has practically finished but concentration of hydrogen peroxide is maximum).
The kinetics of the decomposition, taken from our previous work (Sempere et al. 1998) are cited here for the readers' convenience and they consist of the following set of equations:
H2O2 + Z H2O2·Z [H2O2 Z]=K [H2O2][Z] (a)
[ Z ] = (1)
[ H2O2·Z ] = K [ H2O2 ] [ Z ] (2)
2 H2O2·Z 2 H2O + O2 + 2Z (b)
(3)
H2O2 + H2O2·Z 2H2O + O2 + Z (c)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
where Zo are the initial moles of catalyst, Z are the free moles of catalyst, V is the volume of the mixture, k1 and k2 are kinetic constants, K is the equilibrium constant, r1 and r2 are the rates of elementary reactions (b) and (c), respectively, T is the absolute temperature of the reaction mass, [H2O2] and [H2O2 Z] are the concentration of hydrogen peroxide and the concentration of the intermediate between hydrogen peroxide and catalyst, HR is the heat of decomposition reaction (dq1/dt) and (dq2/dt) is the power generated by reactions (b) and (c) respectively.
The values of the constants reported in our previous work (Sempere et al 1998) have been employed. It has been verified that the same set of equations and similar constant values dictate the N-oxidation of -picoline. Similar power generation profiles are obtained for the N-oxidation of lutidines indicating that a similar set of kinetic equations can be employed (Jun Jao & Papadaki 2004). Moreover, the catalytic decomposition of hydrogen peroxide follows these kinetics but its decomposition in the absence of alkylpyridines is much more vigorous (Papadaki 2004).
MODELLING OF RUNAWAY SCENARIOS: ASSUMPTIONS AND PROBLEM DEFINITION
Work has been based on a 50% filled 10L batch reactor containing a 35% aqueous solution of hydrogen peroxide and 0.0025 moles of phosphotungstic acid (catalyst). Unless otherwise stated, the process temperature is assumed to be 373K before reactor looses thermal control. In an open system the reaction mixture will either continue reacting and then boil with the overhead condenser condensing all the produced vapours, or react and evaporate if the condenser also suffers a cooling failure. In a closed system the reactor temperature and pressure will rise almost adiabatically. The temperature and pressure history of the latter scenario have been examined for different reactor filling levels, hydrogen peroxide strength, catalyst concentrations and rate constant values.
The results of two modelled scenarios are shown here:
a)open vessel-evaporation scenario with condenser cooling system failure.
b)c) closed vessel scenarios.
These models were based on the solution of a system of differential equations that were derived from the mass and energy balances. In all cases heat losses were ignored. Moreover, only the thermal mass of the reaction mass was taken into consideration. Boiling of reactor contents was considered. In order to simulate this possible evaporation, thermodynamic properties of the reactant are required. PVT data of water were used for this purpose. Water approximates well the system under consideration and has two main advantages. On one hand, whereas there is a lack of PVT data for the majority of chemical substances, the properties of water are easily available. On the other hand, the assumption of the adequacy of PVT data of water can be further tested as it enables verifications, such as adiabatic calorimetry measurements. Those can be also used to validate the kinetic model and the simulation developed in this work.
Complete loss of thermal control: Open vessel and evaporation (condenser cooling failure)
This model considers the runaway which takes place in an open system. The reaction is performed at Tprocess and following the cooling failure the heat of reaction initially produces an adiabatic rise of the temperature in the reactor. When the boiling point (100ºC) is reached all the heat generated by the reaction is consumed for the evaporation of the reaction mixture. The heat of reaction results in evaporation of both hydrogen peroxide and water. The condenser cooling system is not functioning. The system will thus behave as gassy with vapours being produced simultaneously with the production of oxygen by the reaction. Although under those conditions neither the pressure nor the temperature of the system change, the evaporation of the liquid results in an increase of the catalyst concentration thus in reaction acceleration. The reaction continues at constant temperature and pressure until all the reactant has been consumed and/or has evaporated. No evaporation was considered at temperatures lower than the boiling point of the system. For this scenario the model calculates the gas and vapour production as a function of time. The programming algorithm for this model is shown in Figure 2:
Closed vessel scenarios
Here the reactor is a closed vessel. The reaction produces oxygen, which results in a substantial increase in pressure depending on the filling level of the reactor. Evaporation of the mixture will take place if pressure is less than its vapour pressure. So, unlike the first model where the pressure in the reactor is not relevant, here it plays an important role. By comparing the pressure at time t with the vapour pressure of water at the current temperature (Tt) is possible to find if evaporation can take place. Then, using the enthalpy of evaporation and the specific volume of the liquid at the respective conditions the mass evaporated and the new reaction mixture volume are determined.
Expansion of the liquid caused by heating has been taken into account. The influence of the filling level has also been examined.
It has to be pointed out that the most likely runaway scenario when working with hydrogen peroxide in chemical industry would lie between those two models. Reactors employed in this type of reactions are always equipped with relief systems that maintain the pressure inside the vessel to an acceptable level. Considering this, the latter model would represent the worst-case scenario of a blockage in the vent or by a runaway occurring in a sealed storage container.
The programming algorithm for this model is shown in Figure 3.
Figure 2 Figure 3
RESULTS AND DISCUSSION
Figure 4 shows the cumulative oxygen production and mass evaporation for an open reactor with no condenser system. Initial Tprocess is 95oC. Other conditions are: Reactor filling level 50%, amount of catalyst is 0.0025 mol. As can be seen, it takes approximately 80 h for all hydrogen peroxide to react. By that time, almost all the reactor contents have evaporated.
Figure 4.Figure 5
Figure 5 shows the pressure and temperature history of the reactor for different hydrogen peroxide strengths. Other conditions are: Reactor filling level 50%, Tprocess 373K, amount of catalyst 0.0025 mol. As can be seen for 70% and 50% solutions total liquid expansion occurs thus rising the pressure to very high values before decomposition is complete. The maximum pressures shown are the ones reached when the reactor was filled up with liquid. No further calculations were performed.
Figure 6
Figure 7
Figure 6 shows the pressure and temperature history of the reactor for different filling levels. Other conditions are: hydrogen peroxide strength 35%, Tprocess = 373K, amount of catalyst is 0.0025 mol. As can be seen for a filling level of 70% total liquid expansion raises the pressure to a few thousand atmospheres. A 5% filling level reacts in a short time. It does not result in an alarming temperature rise. However, the pressure rise still is significant. This scenario can be envisaged as one where the reactor vent system has a blockage far down the vent line, so that the void volume of the system is 18 times larger than the volume of the liquid solution. In all other cases MTSR is greater than 540K, and the resulting pressures are much higher too.
Figure 7 shows the pressure and temperature history of the reactor for different catalyst concentrations. Other conditions are: Reactor filling level 50%, Tprocess = 373K, hydrogen peroxide strength is 35%. As can be seen, if 0.005 mole of catalyst are employed the onset of the runaway occurs approximately 10 hr after the cooling failure (embedded graph) while for higher concentrations it develops within minutes.
Figure 8
Figure 8 shows the influence of an error in the evaluation of constants k1, k2, K and their combination (20% change of all constants) on the pressure and temperature history of the reactor Other conditions are: Reactor filling level 50%, Tprocess =373K, amount of catalyst 0.0025 mol, hydrogen peroxide strength 35%. A 20% error has been considered. As can be seen, although 20% error on k1 or K has an insignificant effect on the assessed temperature and pressure profiles, a 20% error on k2 has a significant effect. If all constants are changed by 20% simultaneously, the combined effect is the same as the one induced by a 20% error in k2. A larger than 20% error is possible during the process of the evaluation of constants (Papadaki, Emery, Serra, Nomen, & Sempere 2002a).
The influence of the initial process temperature has also been assessed. It was found that any temperature higher than 50oC results in temperatures near the "onset" of the N-oxide decomposition. However, N-oxidation does not practically occur at temperatures lower than 80oC, so 50oC cannot be considered as a process temperature. At lower temperatures, runaway develops so slowly that may pass untraced until it is too late.
Table 1 summarizes the different values for MTSR, Pmax and t1 obtained in all aforementioned cases.
CONCLUSIONS
It can be seen that the parameter that has a wider influence on all the final values is the filling level. It can be seen in table 1 that the "onset" of the N-oxide decomposition is exceeded in all cases but the one that corresponds to a filling level of 5%. In other words, if any of those scenarios occurs during the N-oxidation reaction, it will provoke the subsequent N-oxide decomposition. In industry a reactor that uses only 5% of its maximum capacity is not economically viable, but if this scenario is seen as one caused by a blockage in the reactor vent system, the calculated MTSR will be valid if the reactant mass fills a 5% only of the available volume.
Examining the values of MTSR and Pmax, we conclude that, from a safety point of view, working with concentrations of hydrogen peroxide greater than 70% should be avoided.
Table 1 shows that the amount of catalyst employed has a marked influence on the rate of decomposition of hydrogen peroxide. To minimize these effects, stabilizers are normally added to hydrogen peroxide. They protect it from chance contamination with small amounts of catalysts during handling and use, and inhibit the activity of minute amounts of catalytic impurities present in the material when it is produced. But, if high concentrations of catalyst are needed to power the N-oxidation reaction, we should know that stabilizers are not going to be effective as expected (Mackenzie 1990).
The last parameter to consider is the initial temperature. If the reactor operates at higher than atmoshperic pressure, temperatures over 120ºC can cause runaways that develop in every short time, such that it is impossible to take any kind of emergency measure.
The results obtained for the different cases studied show that for any of these reactions, venting facilities should be implemented.
Acknowledgements:
The financial support of the Engineering and Physical Sciences Research Council, UK (research grant GR/R14095/01) is greatly acknowledged.
Nomenclature
K / equilibrium constant of reversible reaction (b)k1 / Rate constant of elementary reaction (c)
k2 / Rate constant of elementary reaction (d)
ko / Preexponential factor for Arrhenius equation
nH202reacted / Number of moles reacted (function of time)
nZo / Initial moles of catalyst
/ Thermal power generated by reaction [W]
q1 / Heat produced by elementary reaction (c)
q2 / Heat produced by elementary reaction (d)
ri / rate of reaction i
t / Time
TR,process / Reaction temperature under normal conditions
T / Absolute temperature of the reaction mass
V / Volume of the reaction mixture(function of time)
Z / Free catalyst sites [moles]
Zo / Total catalyst sites [moles]
dq1/dt / thermal power produced by reaction (c)
dq2/dt / thermal power produced by reaction (d)
dq/dt / Total thermal power produced
Tadiab / Adiabatic temperature rise
REFERENCES
Barton, J.A., & Nolan,P,F(1989). Incidents in the chemical industry due to thermal-runaway chemical reactions. IchemE Symposium Series No.115, 3-13.
Mackenzie.J, (1990). Hydrogen Peroxide without Accidents. Chemical Engineering, June, 84-90.
Marco.E, Peña,J.A., & Santamaría,J (1997). Early detection of runaway reactions in systems with gas evolution using on-line mass spectrometry. Chemical Engineering Science, 52 (18), 3107-3115.
Papadaki, M., Stoikou, V., Mantzavinos, D. & Rodriguez-Miranda, J.L.(2002), Towards improved reaction runaway studies: kinetics of the N-oxidation of 2-methylpyridine using heat-flow calorimetry. Process Safety and Environmental Protection, 80(July), 186-196 (2002).