ACCIDENT INVESTIGATION BY USING CALORIMETRIC TECHNIQUES

L. Gigante, A. Lunghi, C. Pasturenzi, B. Rubiliani, M. Dellavedova, P. Cardillo

(Stazione Sperimentale per i Combustibili, San Donato Milanese)

ABSTRACT
The chemical reactors should be equipped with emergency relief systems (generally bursting disk
or relief valves) to minimize the risks of explosion in the case of a remarkable incident as for
example a runaway reaction or a fire. These so-called protective measures start to operate only
when pressure becomes higher than a fixed value and are usually well defined only for gas
generating reaction systems. Reactions carried out at atmospheric pressure under reflux
conditions are normally considered not critical for overpressure risks.
Recently SSC studied a process that lead an explosion due to the overpressure generated for the
decomposition of a final reaction mixture containing a residual of hydrogen peroxide, in spite the
reactor was connected, via distillation column, to an atmospheric vent.
The desired process was therefore studied by using reaction calorimetry to understand the reasons
of the accident and adiabatic tests were run to calculate the vent sizing of the production reactor.
Other Thermoanalytical techniques were used to characterize the thermal behaviour of reactants,
intermediates and products in order to obtain a complete risk assessment: DSC and adiabatic
Calorimetry were extensively used and the results of these test were analysed.

INTRODUCTION
Recently Stazione Sperimentale per i Combustibili was charged to investigate about the accident
occurred in a chemical industry in the North of Italy, during the first batch production of a new
product. As the investigation is going on, all details cannot be provided. The reaction involved the
oxidation of an intermediate (named, in the present paper, as Intermediate A) with hydrogen
peroxide 130 V. The original recipe used an excess of 4% of H2O2 which should be destroyed with
potassium hydroxide. Reaction should be performed at 65C, in isothermal conditions. After the
reaction there was to perform a distillation step, so the reactor was connected, via distillation
column, to an atmospheric vent. It was not equipped with a relief system. The accident occurred
during the destruction of hydrogen peroxide at 80C, by adding the potassium hydroxide to the
reaction mass. The anomalous formation of foam was noticed by operators. The temperature of
reactor started to increase: the last value registered was 120C. Before the explosion the reactor
trembled and a hissing sound was heard. After few seconds the reactor exploded. Fortunately in
the accident only two workers were slightly injured.
The aim of the present work is to find out the causes of the accident and to define a new recipe to
optimize the safety of the reaction.
The study involved:
the study of the process reaction by the reaction calorimetry technique in the conditions of
the final recipe;
the study of the thermal stability of intermediates and products;
adiabatic tests performed with a low phi calorimeter in order to measure the pressure
evolution and calculate the right venting sizes of the reactor.

METHODOLOGY
The study of the desired reaction
As the investigation is going on, all details cannot be provided, particularly on the chemical
structure of intermediate A e final product P.
An RC1 test was performed following the original recipe, in isothermal conditions, at 65C, in two
steps. During the first step hydrogen peroxide was added to a mixture of water, potassium
hydroxide and intermediate A. As indicated by the recipe, the reaction mass remained at 65 C for
3 hours after the end of the addition for the completion of the reaction. An analysis on a sample of
the reaction mixture showed the presence of 4% of H2O2. During the second step KOH was added
in order to destroy the excess of hydrogen peroxide. Fig. 1-2 show the RC1 plots of both steps.
st

300

0,7

250

0,6

200

0,5

150

0,4

100

0,3

50

0,2

0,1

-50

calibration

calibration

Addition, kg

Heat Flux, W

1 step: H2O2 addition

-100

-0,1
2

2,5

3,5

4,5

5,5

6,5

time, h

Figure 1: RC1 plot of the addition of hydrogen peroxide (first step).

nd

step: KOH addition

300

0,4

250

0,35
0,3
0,25

150

0,2
100
0,15
50

Addition, kg

Heat Flux, W

200

0,1

0,05

-50

0
calibration

calibration
-100

-0,05
6

6,5

7,5

8,5

9,5

10

10,5

11

time, h

Figure 2: RC1 plot of the addition of potassium hydroxide (second step).

Tab. 1 summarizes the main results of the experiment.
Table 1: RC1 results on the desired reaction
1st step

2nd step

Maximum heat flow, W

250

200

H, kJ

-258

-525

H norm. to the mass at the end of

addition, kJ/kg

-134

-237

cp at the end of addition, J/kg K

2957

2741

T ad, C

45

87

MTSR, C

65+45=110

65+87=152

During the first step it was measured a thermal accumulation of 27% and the reaction seems to be
completed (from a calorimetric point of view) 2 hours after the end of the dosage.

The study of the decomposition reaction

An ARC test was performed on a sample of final reaction mass at the end of RC1 experiment,
using HWS (Heat-Wait-Search) mode and Hastelloy bomb. Decomposition of the mixture starts at
85C. Tab. 2 summarizes the main results and Fig. 3 illustrates the temperature and pressure of
the exotherm data as a function of time.
Table 2: ARC results on the final reaction mass
Initial temperature of reaction, To (C)

85

Initial self-heating rate (C/min.)

0,03

Thermal inertia factor,

2,40

Maximum self heating rate (C/min.) corrected by

0,9

Temperature at the maximum self-heating rate (C)

104

Final Temperature, Tf (C)

110

Observed temperature rise (C)

25

corrected temperature rise (C)

62

corrected final Temperature (C)

85+62 = 147

Reaction enthalpy (J/g)

-283,47

Initial Pressure at To (bar)

1,5

Pressure at Tf (bar)

2,1

Basic ARC Data: Temperature and Pressure vs. Time Plot

Final reaction mass

48

Temperature

115

46
44

105

42

95

40
38

85

36
34

75

32

65

Pressure

30

55

28

45

26
24

35
25
550

Pressure (psia)

Temperature (C)

125

22
600

650

700

750

800

850

20
900

Time (Minutes)

Figure 3: Temperature and Pressure vs time exotherm data of the final reaction mass
DSC tests both in air and nitrogen atmosphere, in stainless steal sealed medium pressure
crucibles, were performed on different samples:
sample A: H2O2 130 V;
sample B: reaction mixture at the end of the addition of H2O2;
sample C: reaction mixture 3 hours after the end of the addition of H2O2;
sample D: reaction mixture at the end of the reaction, after the addition of KOH;
sample E: reaction mixture at the end of the reaction, after the addition of KOH and after
overheating at 80-85C for 2 hours (to totally destroy the excess of H2O2).
Fig. 4-8 show the DSC plots on different samples.

^exo

^exo
Sample: H2O2 (azoto), 2.6900 mg
Method: DSC/azoto/30-300/2/acciaio/4sec
30.0-300.0C 2.00C/min

Method: Scre enDSC/N 2/30-28 0/5/ac ciaio

30.0- 280.0C 5.00C /min

Pa n: Stee l/Vito n Mediu m Press ure 120 ul

Pa n: Stee l/Vito n Mediu m Press ure 120 ul

10
Wg^ -1

10
Wg^ -1

Integral
3813.65 mJ
normalized 921.17 Jg^-1
Onset
141.86 C
Peak
144.13 C
Endset
145.51 C
Left Limit
90.54 C
Right Limit
166.97 C

I ntegra l
2 067.68 mJ
norma lized 7 68.65 J g^-1
O nset
72.37 C
P eak
82.41 C
E ndset
90.05 C
L eft Li mit
59.41 C
R ight L imit
98.60 C

40

60

80

100

120

140

160

180

200

220

240

260

METTLER TOLEDO STA Re System

Stazione Sperimentale per i Combustibili: SSC

40

60

80

100

120

140

160

180

200

220

240

260

METTLER TOLEDO STA Re System

Stazione Sperimentale per i Combustibili: SSC

Figure 4: sample A - H2O2 130 V

Figure 5: sample B - reaction mixture at the

end of the addition of H2O2

^exo

^exo
Method: ScreenDSC/N2/30-280/5/acciaio
30.0-280.0C 5.00C/min

Method: ScreenDSC/N2/30-280/5/acciaio
30.0-280.0C 5.00C/min

Pa n: Stee l/Vito n Mediu m Press ure 120 ul

Pa n: Stee l/Vito n Mediu m Press ure 120 ul

10
Wg^ -1

10
Wg^ -1

Integral
4654.51 mJ
normalized 850.92 Jg^-1
Onset
138.02 C
Peak
143.27 C
Endset
146.73 C
Left Limit
86.50 C
Right Limit
164.70 C

40

40

60

80

100

120

140

160

Integra l
1 291.60 mJ
norma lized 2 56.27 Jg^-1
Onset
117.3 0 C
Peak
125.5 3 C
Endset
131.9 0 C
Left Li mit
97.85 C
Right L imit
164.4 4 C

Integral
172.72 mJ
normalized 34.27 Jg^-1
Onset
64.73 C
Peak
83.03 C
Endset
92.50 C
Left Limit
60.51 C
Right Limit
96.34 C

180

200

220

240

260

METTLER TOLEDO STA Re System

Stazione Sperimentale per i Combustibili: SSC

Figure 6: sample C - reaction mixture 3

hours after the end of the addition of H2O2

60

80

100

120

140

Stazione Sperimentale per i Combustibili: SSC

160

180

200

220

240

260

METTLER TOLEDO STA Re System

Figure 7: sample D - reaction mixture at the

end of the reaction, after the addition of KOH

^exo
Method: ScreenDSC/N2/30-280/5/acciaio
30.0-280.0C 5.00C/min
Pa n: Stee l/Vito n Mediu m Press ure 120 ul

10
Wg^ -1

Int egral
921. 21 mJ
n ormaliz ed 193. 13 Jg^- 1
Ons et
12 0.63 C
Pea k
13 0.25 C
End set
13 6.23 C
Lef t Limit
96 .83 C
Rig ht Limi t
15 3.04 C

40

60

80

100

120

140

Stazione Sperimentale per i Combustibili: SSC

160

180

200

220

240

260

METTLER TOLEDO STA Re System

Figure 8: sample E - reaction mixture at the

end of the reaction, after the addition of KOH
and after overheating at 80-85C for 2 hours
It is possible to underline some evidences. The final product, in the reaction system, decomposes
starting from 95 C, with a peak at 130C and with an energy release of -200 J/g (sample E). In the
presence of a high amount of H2O2 (sample B and C) the decomposition of the final product and
the decomposition of hydrogen peroxide are superimposed: there is a double exothermic effect
starting from 90 C with a higher energy release (-900J/g). After the addition of KOH (sample D)
there is a smaller amount of hydrogen peroxide and in the new conditions of the reaction system,
H2O2 decomposes before the final product and two different peaks can be seen. During

overheating at 80-85 C hydrogen peroxide decomposes and in the reaction mixture remains only
the final product (sample E).
Some residual products picked up from the reactor and the area around the explosion were
analysed by DSC test. All of them presented no exothermic picks as they totally decomposed
during the accident and no further information can be achieved by this analysis.
Analyzing RC1 and ARC/DSC test the reaction performed in the conditions of the recipe, seems to
be out of the conditions of intrinsic safety, it is the process is a potential runaway reaction. In this
case, if it is not possible to increase the safety of the process adopting preventive measures, it is
necessary to adopt protective ones.
In this process, the main cause of the explosion of the reactor was the increase of pressure due to
the hydrogen peroxide decomposition, so the possibility of a vent system was studied.

PhiTec tests for the vent sizing

After the reaction step there was to perform a distillation step, so the reactor was atmospheric
through the distillation column. It was not equipped with protective apparatus as a bursting disk.
Facts demonstrated that protective measures must be taken also for atmospheric reactors.
In the present work for the calculation of the area of the bursting disk, the following methodology,
normally adopted by SSC, was used. The calculation is based on the DIERS [1] methodology,
takes the system in the accidental worst case scenario and uses experimental data obtained with
low Phi adiabatic calorimeters.
The first step consists in the definition of the system. It should be vapour, gassy or hybrid. The
present system seems to be an hybrid system using both a bibliographic approach [2-3] and the
experimental approach, plotting lnP vs 1/T.
The second step consists in the definition of the flow regime, that should be laminar or turbulent.
Using a simplificative approach, as the system has a viscosity lower than 1 cP, the flow can be
defined in a turbulent regime.
The third step consists in the definition of the type of the flow, that should be one phase liquid, one
phase gassy or two phases. In the accident description the formation of foam was underlined. So,
applying a conservative approach, the system seems to be a two phases flow.
Finally the system is an aqueous one and can be defined as a temperate system.
For the calculation of the of the area of the bursting disk, the ratio between the mass flow W (kg/s)
and the flow per unit area G (kg/s m2) must be defined. Pset is 1,5 bar and Pmax is 1,65 bar.
For the calculation of W, the appropriate Leung-Fauske equation [1] was used; for the calculation
of G, the appropriate simplified Leung equation [4-6] was used. This equation calculates the
average G between the gassy (Gg) and the vapour (Gv) system. Gg was calculated by the Omega
method, Gv was calculated by the Fauske - ERM method.
Two different PhiTec II tests were performed in different conditions to measure the critical
parameters (dP/dT and SHR al Pset and Pmax) in the worst case scenario to insert in the previous
equations.
The first test was performed using the reaction mass with an amount of 4% of H2O2. These were
the conditions of the reactor just before the accident. Fig. 9 shows Temperature and Pressure plot
vs time of the exothermic effect of decomposition. In this case, calculations lead to a diameter of

the bursting disk of 60 cm, absolutely unacceptable. We can say that in the conditions defined by
the recipe there are no protective measures that can be adopted.
It is evident that to perform the reaction in safer conditions it is necessary to reduce the residual of
H2O2 after the first step of the reaction, it is to decrease the excess of hydrogen peroxide to dose.
So the second test was performed using the reaction mass with an amount of 1% of H2O2. Fig. 10
shows Temperature and Pressure plot vs time of the exothermic effect of decomposition. In this
case, calculations lead to a diameter of the bursting disk of 9,5 cm, an acceptable value and a
value compatible with the diameter of the pipe connected to the distillation system.

Test n. 76: H2O2 4%

Test n. 75: H2O2 1%

100

18

Pressure
95

10

78

20

77

Temperature

16

10

Temperature

8
80

75

6
5

74

Pressure

73

6
4

75

70
30

35

40

3
72
2

71

70

45

Time, min

Figure 9: Phi Tec II test on the reaction

mixture at 4% H2O2

Pressure, bar

85

7
Temperature, C

12

Pressure, bar

Temperature, C

76
14

90

1
0
30

35

40

45

50

55

Time, min

Figure 10: Phi Tec II test on the reaction

mixture at 1% H2O2

CONCLUSION
Using the report of the accident made by workers and data on reaction acquired by calorimetric
and thermoanalytical technique we can say that the main cause of the accident was the lack of
knowledge on the thermochemistry of the process. It is that this new process was scaled up by
similitude with an old process, without studying safety implications and preventive or protective
measures based on real data.
Analyzing RC1 and ARC/DSC test the reaction, performed in the conditions of the recipe, seems to
be out of the conditions of intrinsic safety: the process is a potential runaway reaction. In this
process the loss of control of the reaction leaded to the decomposition of hydrogen peroxide, the
formation of foam and the explosion of the reactor. In this case, if it is not possible to enhance the
safety of the process adopting preventive measures it is necessary to adopt protective measures.
But as the reactor was an atmospheric one, no relief systems were adopted.
So in this work was studied one of the chance to increase the safety of the process, introducing a
little change in the original recipe. The original recipe lead, at the end of the first step, to an excess
of 4% of H2O2 in the reaction mixture. In this case, for a decomposition of the mass, the calculated
diameter of the bursting disk was unacceptable. Just a little modification in the recipe so that the
hydrogen peroxide residual is reduced to 1%, leads to a drastic reduction of the diameter of the
bursting disk, that should have avoided the explosion of the reactor.

BIBLIOGRAPHY
[1] J. Etchells, & J. Wilday, Workbook for chemical reactor relief system sizing, 1998, Sheffield, UK:
HSE BOOKS
[2] H.K. Fauske, Properly size vents for nonreactive and reactive chemicals, Ch. Eng. Prog., 2000
[3] J.K. Wilberforce, Emergency venting of hydrogen peroxide tanks, CEFIC Hydrogen peroxide
safety conference, Gotherburg, Sweden (Sep. 22, 1988)
[4] J.C. Leung, Two-Phase Flow Discharge in Nozzles and Pipes A Unified Approach, J. Loss
Prev. Process Ind., 3 (1), 27-32, January 1990.
[5] J.C. Leung, The Omega method for discharge rate evaluation, International Symposium on
runaway reactions and pressure relief design, 367-393, AIChE, 1995, ISBN 0-8169-0676-9.
[6] L. Gigante, A. Lunghi, P. Cardillo, R. Rota, Metodi semplificati per il dimensionamento di
dispositivi di scarico di emergenza con efflusso bifase, Riv. Comb, 58 (2), 46-58, 2004.