Process Safety Progress Journal, June 2002, issue (Vo l. 21, No.

2) article

An Example of a Test Method for Vent SizingOPPSD/SPI Methodology

Charles S. Poteet III a and Marvin L. Banks b a Peroxy Engineering Specialist, GEO Specialty Chemicals, Peroxy Chemicals Group, 50 N. Market Street, Gibbstown, NJ 08027 b Program Manager, Hazards Testing, Energetic Materials Research and Testing Center, New Mexico Institute of Mining and Technology, Campus Station, Socorro, NM 87801

An alternative methodology to that of United Nations, Appendix 5, Sample Method, has been developed by the Organic Peroxides Producers Safety Division of the Society of the Plastics Industry (OPPSD/SPI) to assure the safe emergency venting of organic peroxides in an approved container. The methodology is applicable for designing a new container for an existing or new organic peroxide. Bench-scale testing with a 10 liter vessel determined the maximum pressure pulse (spike) generated during a thermal decomposition/overpressurization vs. the ultimate, structural strength of the container measured and/or calculated. The methodology has been tested full scale with a United Nations 6(c) fuel fire. Container deformation, so long as there is no fragmentation, is permitted with the proviso that the container be retired from service following an incident. The methodology, as developed, reflects advances in technology while harmonizing to a maximum extent with the United Nations, Appendix 5, Sample Method.
INTRODUCTION

overdesigned. They were fitted with a 22.5 inch diameter frangible disk with an integral three inch diameter relief device designed to melt at 300 F, thus providing early pressure release before the primary relief device was engaged. This real-life incident, while successful, raised certain questions concerning the expected performance of other peroxides in approved containers. As a result, the O rganic Peroxides Producers Safety Division of the Society of the Plastics Industry (OPPSD/SPI) developed a methodology to evaluate the safety and suitability of a given container for transporting a specific organic peroxide, to preclude the possibility of container failure f rom overpressurization which could result in fragmentation. Bench-scale and full-scale tests sponsored by the OPPSD/SPI validated the method.
APPROACH

On December 11, 1990, heavy fog conditions along a Tennessee Interstate Highway caused a multiple car and truck accident. In all, 75 vehicles were involved. By chance, the resultant ruptured fuel tank fire s engulfed a tractor-trailer transporting ten (10) GEO Specialty Chemicals, (formerly Hercules, Inc.), D.O.T. Spec. 57 containers of dicumyl peroxide. In the ensuing fire, the 421 gallon containers, each containing 3,500 pounds of the organic peroxide, underwent thermal decomposition and overpressurization. Subsequent investigations, which included metallurgical examinations of the containers, determined that all ten had vented the organic peroxide decomposition products successfully without loss of structural integrity, breach, or fragmentation. It was also determined that the GEO container had been heavily
Process Safety Progress (Vol.21, No.2)

Design and fabricate a high-pressure, bench-scale, 10 liter, test vessel that can safely withstand multiple tests and overpressurizations. Conduct thermal decomposition/overpressurization tests to develop and verify reproducibility of a standard operating procedure. Select a standard organic peroxide for subsequent testing. Verify traditional literature relationship of decomposition pressure pulse (spike) vs. the ratio venting relief area to working-heated volume (A/V). Identify the maximum pressure pulse for the standard organic peroxide and extrapolate response to an approved commercial container. Determine, by testing and mathematical methods, the ultimate structural strength of the full-scale commercial container.
June 2002 85

 Conduct a full-scale United Nations 6(c) Fuel Fire and evaluate the applicability of the bench-scale method for predicting results. Present the methodology for the determination of the suitability of a specific organic peroxide for an approved container, and conversely, the design of a new container for an existing or new org a n i c peroxide.
BENCH-SCALE - 10 LITER SUPPORT STUDIES

Test Site The 10 liter test apparatus was set up at the Little Eagle explosion-proof bunker of the Energetic Materials Research and Testing Center at New Mexico Institute of Mining & Technology (EMRTC/NMT) in Socorro, New Mexico. Tests were witnessed by OPPSD/SPI personnel. A video system permitted real-time monitoring of the tests progress during overpressurization and subsequent venting. Pre s s u re and temperature telemetry was sent to a remote location for acquisition and analysis. Results, along with descriptions of test apparatus and pro c e d u res are contained in EMRTC/NMT Report FR-98-07. Test Apparatus Test Ve s s e l . (See Figure 1b) A 10 liter cylindrical vessel was constructed with an 8 inch internal diameter stainless steel pipe having a wall thickness of 1 inch. The base plate and top plate were cut from a 1 inch thick stainless steel plate. An 0.5 inch stainless steel plate was welded into the bottom of the pipe, and an O-ring was installed at the top of the pipe to complete the pre s s u re vessel. The vessel is bolted together with six 0.625 inch bolts of number 8 hardness. The pressure rating for the vessel was calculated at 2,500 psig. H e a t e r s . Sixteen 0.5 inch diameter heater wells w e re drilled into the walls to a depth of 11 inches f rom the bottom. Eight 2,000 watt cartridge heaters (Watlow J10A64-D12H) and eight 1,500 watt cartridge heaters (Watlow J10A33-D12H) were installed using A/B/A/B pattern. An 0.125 inch diameter, 3,000 watt cable heater was spiral wound in the 0.25 inch space between the welded bottom and the 1 inch thick base plate. Because the 240 volt heaters were run using 208 volts, the total heater power was nominally 23,250 watts (21,000 watts in the walls and 2,250 watts on the bottom). Two layers of ceramic fiber insulation were wrapped around the side and bottom of the vessel. Thermocouples. Three 0.063 inch diameter K-Type, stainless steel sheathed, ungrounded thermocouples (Omega Engineering, Inc., CASS-116-U-12) were used to measure the temperatures inside the vessel. One thermocouple was located inside the vapor phase, the second was 0.5 inches below the surface of the liquid, and the third was located 1.5 inches from the bottom. The thermocouples were installed as shown in Figure 1 utilizing Swagelok fittings. Omega SA1-K stick-on K-Type thermocouples were used on the outside and bottom walls of the vessel to control the heaters.
86 June 2002

Rupture Disks & Orifice Plates. The Type P rupture disks had a burst pre s s u re of 20 psig at 90 C and were purchased from the Fike Corporation. Elevated temperatures have very little effect on the burst press u re. Burst pre s s u res from 21 to 23 psig were observed. The holder assembly was Type G, bolted style, 15-# RF, CS/CS with an open throat. The top ring of the assembly was 0.5 inches larger than the rupture disk diameter. Orifice plates were machined from 0.25 inch thick stainless steel plate and machined to varying orifice diameters. Diameters range from 0.75 to 4.0 inches in 0.25 inch increments. Load Cell. An Omega Engineering, Inc., LCDA-150 Load Cell was used to record the rate of material loss f rom the apparatus. Although the load cell had a capacity of 150 pounds, the fixture was mounted on a fulcrum arrangement, which allowed for a total load capacity of 300 pounds. The load cell and fulcrum were installed as depicted in Figure 1b. P re s s u re Measure m e n t s . Two Setra Model 280E variable capacitance transducers were used to obtain p re s s u re data. They were connected to the vessel using 14 inches of 0.25 inch outer diameter stainless steel tubing, as depicted in Figure 1b. The manufacturers listed response time for the Model 280E is one millisecond. Each transducer was ported to a separate computer system, reducing the possibility of data loss. Test Procedures Vessel Temperature Control. The heaters were cont rolled using a Pentium desktop computer with a Strawberry Tree, Inc., Model ACPC-12-8 Analog/Digital I/O board. Thermocouple signals from the vessel wall were recorded and used to control a 12 volt DC control signal from the computer to switch the heaters on and off. This signal was connected to six Omega Engineering, Inc., 240v 45a solid-state relays to control the 208 volt AC directed to the heaters. The system sample rate of 12 measurements per second was more than adequate. Pressure Data Collection. The signal from one of the pressure transducers was fed to a Strawberry Tree, Inc., Dynares 8-channel Analog/Digital I/O board installed in a Pentium Computer. This system was programmed to collect data at 1 KHz, which matched the response time of the transducer. Because of the high sample rate, the system was run at 12 bit resolution. In order to limit the size of the data file, this system was activated as soon as the rupture disk had burst and was shut down a few minutes after the reaction. This method of operation limited the amount of data collected, but retained a profile of the high-pressure pulse. Pressure, Temperature & Load Cell Data Collection. The liquid and vapor temperatures inside the vessel, the second pressure transducer, and the load cell data were recorded using a third Pentium computer and a D y n a res 8-channel Analog/Digital I/O board fro m Strawberry Tree, Inc. The sampling rate of this system was set at 100 Hz. In order to maintain this sample rate, the pressure and load cell signals were recorded as voltages. The temperatures were converted by the software to the actual temperatures.
Process Safety Progress (Vol.21, No.2)

Figures 1a and b. 10 liter vessels for venting tests.

P ro c e d u r a l . The test vessel was preheated to just above the storage temperature listed in the MSDS for the material before loading. The 10 liter volume vessel was filled with 9.0 liters of the sample material, 99.7% dicumyl peroxide. The internal temperature ramp rate was set at 6 C/min/min, the calculated American Petroleum Institutes API 520/521 heat-up rate for a United Nations 6(c) fuel fire regarded as the worst case scenario for an organic peroxide shipping container. The rupture disk was installed to seal the vessel, and heating was begun within 10 minutes. Both the heater control and the data collection computers were started simultaneously, and the fast data collection computer was started when the rupture disk failed. The heaters were run at the designated ramp rate until several minutes after the decomposition reaction had occurred.
Process Safety Progress (Vol.21, No.2)

A video re c o rding system allowed monitoring of the tests progress. An explosion-proof fan was used to remove residual vapors prior to personnel re-entry.
10 LITER TEST RESULTS

Temperature Ramp Rate vs. Peak Pressure Pulse It is commonly accepted in the literature that the Peak Pressure Pulse from an organic peroxide decomposition and vessel overpressurization is dependent on, and can be correlated with, the heating rate or temperature ramp rate of the material being tested for a constant venting area to working-heated volume ratio. There f o re, this relationship was investigated first. A 2.25 inch diameter disk was set as a constant. This yields a 1.67 inch2/gallon ratio of venting area to
June 2002 87

Table 1. Results of four temperature ramp tests.

Table 2. Results for peak pressure pulse vs. venting area with constant ramp rate.

working-heated volume, or 0.60 gallons/inch 2 , depending on preferred method of stating the ratio, A/V vs. V/A. Dicumyl peroxide was the test material. Table 1 presents the results for four ramp rates; 2, 6, 9, and 23 C per minute per minute with replicates. When graphed (Figure 2), the relationship is seen as near-linear, confirming the assumption and demonstrating reproducibility. Peak Pressure Pulse vs. Venting Area to Working Volume Ratio Next, the investigation turned to determining the relationship of peak pressure pulse (spike) vs. venting area to working-heated volume ratio (A/V) with the temperature ramp rate set constant at 6 C/min/min. This is the calculated API 520/521 heat-up or ramp rate for a fuel fire and is currently recognized by the United Nations in a U.N. 6 (c) full-scale burning test. Five disk or orifice size diameters were selected; 3.00, 2.25, 2.00, 1.75, and 1.5 inches, to bracket the equiva88 June 2002

lent frangible disk size on a D.O.T. Spec. 57 GEO Specialty Chemicals container. Results are presented in Table 2. P revious work on temperature ramp rate would predict approximately the 68 psig peak pressure pulse recorded for a 2.25 inch diameter disk. The relationship of peak pressure pulse (spike) vs. vent area to working-heated volume ratio is presented in Figure 3. For the disk size of interest which relates to the Spec. 57 container, 1.75" diameter, the peak pressure pulse (spike) was 232.5 psig. It happens that the vent area to working volume, as well as the working volume to vent area, for this size disk is 1.00. From the time the disk relieved to the time of peak pre s s u re was approximately half a minute. In all cases the peak temperature was approximately 350 C. Therefore, for scale-up to the D.O.T. Spec. 57 container used by GEO Specialty Chemicals, which has 421 working gallons of volume and contains 3,500
Process Safety Progress (Vol.21, No.2)

Figure 2. 10 liter results: Peak pressure pulse vs.

temperature ramp rate. pounds of dicumyl peroxide, with a 22.5 inch diameter frangible disk (1.00 A/V ratio), a peak pressure of 232.5 psig was predicted from the 10 liter studies, if the calculated heating rate assumption is accurate and verified in full-scale testing and is 6 C/min/min.
CONTAINER STRUCTURAL EVALUATION

Figure 3. 10 liter results: Peak pressure pulse vs. vent area to working-heated volume ratio.

Hydrostatic Testing The container was hydrostatically tested. While there was deflection of the flat heads, top and bottom, with increasing applied pressure, there was no structural failure, as defined by breach or fragmentation at 288 psig (19.6 barr), where the test was discontinued. Fragmentation, not deflection, is the true point of concern in any incident. Curvilinear Finite Element Testing The container was further evaluated from a structural-mathematical approach. Curvilinear finite element analysis (FEA), which is frequently used to define ASME pre s s u re vessels, was employed. The method, by which finite elements of the container are evaluated separately, confirmed the hydrostatic test results by predicting deflection at the heads, but no structural failure as defined by breach or fragmentation to pressures greater than 20 barg. Therefore, with a safety factor of permitting only 85% of the demonstrated hydrostatic pressure limit, the Spec. 57 could reasonably be expected to contain the predicted ~ 230 psig maximum pressure pulse from the 10 liter prediction for a full-scale burn with the standard peroxide; dicumyl peroxide and an A/V = 1.00.
FULL-SCALE UNITED NATIONS 6(C) BURN

Design Criteria Process vessels, for example, pressurized work tanks, reactors, etc., are designed based on material of construction yield stresses, or the point at which first elongation or deflection occurs. Then, to avoid approaching yield or elongation, which could lead to fatigue, a conservative working pressure of 25% of that pressure is chosen as an upper limit. For a vessel that routinely cycles in normal operation over several decades of expected life, this conservative approach is understandable. However, when applied to a shipping container, which may experience deflection or elongation once in its working lifetime due to an internal decomposition/overpressurization as the result of an emergency situation, for example, an external fuel fire, this approach is viewed as excessively conservative and restrictive. A more realistic approach for a shipping container such as the D.O.T. Spec. 57 portable tank and United Nations 31-AY Intermediate Bulk Container (IBC) is to permit deflection on a one-time basis during an emergency, with the proviso that the container be retired from service. Then the design bases can work off of material of construction true strength, the ultimate stress, or the point at which failure would occur. This approach was tested and proven to be valid. A D.O.T. Spec. 57 portable tank from the GEO Specialty Chemicals fleet, which had seen 20 years of active service and was similar in all design respects to the container planned for use in an EMRTC/NMT full scale U.N. 6(c) burn, was selected. The design basis for this atmospheric vessel based on yield stress was 4 barg, which, with a 25% factor, permitted it to be used at atmospheric pressure (1 barg).

Test Site Projections from the 10 liter studies were extended to a full-scale United Nations 6(c): Fuel Fire Test. The Energetic Materials Research and Testing Center at the New Mexico Institute of Mining & Te c h n o l o g y ( E M RTC/NMT) in Socorro, New Mexico, test range facilities were once again utilized. Tests were witnessed by D.O.T. officials, as well as OPPSD/SPI members, while a video system permitted it to be re c o rded. Pre s s u re and temperature telemetry was sent to a remote location for acquisition and analysis. Test results, along with experimental set-up and procedures, are presented in EMRTC/NMT Report FR-99-03.

Process Safety Progress (Vol.21, No.2)

June 2002 89

Figure 4. Experimental tests of full-scale United Nations 6(c) burn.

Experimental Test Material. GEO Specialty Chemicals Di-Cup R, dicumyl peroxide, an organic peroxide, Type F, 5.2, UN 3110, PG II, was utilized as the test material. It is a colorless to pale yellow solid at ambient conditions. The purity was 99.7% as dicumyl peroxide with a melting point (M.P = 101 F). The material will decompose before the boiling point is re a c h e d . Decomposition products are flammable and include methane. A total of 3,500 pounds of the material was involved in the test. Test Container. A D.O.T. Spec. 57 container, designed and custom fabricated for GEO Specialty Chemicals, was used for the test. It was identified as Serial No. 1188-7081-F6. It had a tare weight of 765 pounds, a container volume of 460 gallons, and a height of 66 inches. A protective 3/16 stainless steel rim of 4.5 inches fitted about the top of the container added to the total height. A 20 psi frangible disk (relief vent) measuring 22.5 inches in diameter was positioned and centered on the top head. It was secured with a clamping ring torqued to 20 ft lbs at the locking bolt. An integral part of the disk design incorporated an additional 3 inch relief device, designed to melt at 300 F to provide early relief before the primary disk was engaged. The total weight, including organic peroxide, was 4,265 pounds (see Figure 4).
90 June 2002

The container had two minor modifications to permit test telemetry to be transmitted: 1. Fifteen 1/4 inch diameter holes were drilled into the top of the container to permit access of the thermocouples. These were made pressure tight by the use of Swagelok fittings. 2. An existing dip pipe used to pump out the contents of the container was modified for use as the pressure measurement connection. The 1.25 inch diameter pipe reaches to the bottom of the container. At a point just under the top of the container, at a position clearly in the vapor space, a 2 inch diameter hole was cut into the pipe to facilitate the external connection of the pressure measuring devices, without making a significant change in the containers design. Test Apparatus Test Assembly. The package consisted of the test container, a test stand, pipes to support the pressure and temperature telemetry lines and a 12 foot diameter containment tank to contain the JP-4 fuel. A 6 inch layer of water was added to the bottom of the containment tank, then, the fuel was added to just below the top of the tanks lip. As it burned off, additional fuel was pumped into the bottom of the tank through the water layer to provide the needed media to sustain the fire.
Process Safety Progress (Vol.21, No.2)

Thermocouples. Fifteen K-Type thermocouples, 1/8 inch, with grounded junctions and 1/8 inch 304 stainless steel sheaths were installed by GEO Specialty Chemicals prior to shipping (Omega Engineering, Inc. CASS-188-84-NHX, CASS-188-60-NHX, and CASS-18848-NHX). The internal thermocouples were arranged in a vertical plane through the center of the container and at three heights: near the top, at the center, and near the bottom. Radially, they were placed at the centerline, at a point half-way from the wall to the centerline, and at a point 3 inches from the wall. To reduce the chance of failure, eight exited on the east side and seven on the west side. Five additional KType, stainless steel sheathed, ungrounded thermocouples (Omega Engineering, Inc. CASS-116-U-12) were used to measure the temperatures on the exterior of the package. One was located on the west skin, one on the east skin, one in the west side flame, one in the east side flame, and one under the container near the west side. The west skin thermocouple was at the mid-point near internal thermocouple #3. The east skin thermocouple was at the mid-point near internal thermocouple #12. Figure 4 identifies locations. P re s s u r e Devic es. Two Sentra 280E variable capacitance pre s s u re transducers (one 500 psi and one 250 psi) remotely monitored the pressure at the top of the container. Oil (Thermosol 60) was fed t h rough 1/4 inch stainless steel lines to each transducer to provide a non-compressible pressure medium to the devices. Te m p e r a t u re Pro t e c t i o n . All of the thermocouple wires from the top of the container were enclosed in 4 inch steel pipes with a wall thickness of 1 inch to provide support and protection. The transducer lines were encased in a 3.5 inch diameter pipe with 3/4 inch walls. Air was blown through the outer pipe at approximately 100 cu ft/sec on the outside of the oil-filled lines. The ends of the pipe that were exposed to direct flame were encased in calcium silicate insulation. Test Procedures After the container was placed on the test stand, the thermocouples (internal and external) were attached to the computer links. Next, the transducer lines were filled with oil and connected to the transducers and pipe tee, respectively. Six inches of water was added to the bottom of the fuel tank for safety purposes and both electric diaphragm pumps were cycled to ensure that they were functioning correctly. The following day, fuel was pumped to fill the fuel tank. Two puff bags, or delayed fuse bags, were used to ignite the fuel. The fuel level in the tank was monitored by video, and was added when necessary to maintain a high level of heat and flame on the test container. Wind direction and velocity was monitored throughout the test. Based on a rough use of approximately 2,000 gallons of fuel for the two-hour test, it was determined that the burn rate was approximately 17 gallons/min. P re s s u re Data Collection. The signal from each pressure transducer was transmitted to separate Strawberry Tree, Inc., Dynares 16-channel Analog/Digital I/O boards that were installed in Pentium computers.
Process Safety Progress (Vol.21, No.2)

These systems were programmed to collect data at 100 MHz. Te m p e r a t u re Data Collection. The signals fro m seven of the thermocouples were transmitted to a Strawberry Tree, Inc., Dynares 16-channel Analog/Digital I/O board that was installed in a Pentium computer (Computer #1). Also, data from the west skin, west air, and underside of container, and two additional thermocouples were recorded on this computer system. The additional thermocouples read fuel temperature at 6 and 12 inch depths in the fuel tank. The remaining eight internal therm o c o u p l e s transmitted data to a second Strawberry Tree, Inc., D y n a res 16-channel Analog/Digital I/O board that was installed in a Pentium computer (Computer #2). In addition to these signals, signals from the east skin and east air were also recorded. Both systems were set to record data at 100 MHz.

U.N. 6 (C) FULL-SCALE BURN TEST RESULTS

Burn Sequence of Events At five minutes before ignition of the JP-4 fuel (T-5 min), monitoring and data acquisition systems were turned on. Wind velocity was measured at 4 to 5 mph, which was within the allowable 11 mph set by U.N. 6(c) format. At time Zero (t = 0), the igniters were tripped and a fire was observed on the surface of the fuel tank. External skin temperatures rapidly rose to 450 F, while internal temperatures reached 112 to 125 F at (t+ 10 min) at the probes closest to the walls and bottom. Within minutes, the 3 inch relief device melted, indicating temperatures at the top of the container at > 300 F. When it gave way, a low-velocity jet of light decomposition gases (single phase) could be seen emanating from it. The gases, which were most likely methane, ignited. This would indicate decomposition at the walls and bottom of the container with early release of internal pressure as designed. The primary frangible disk was still intact. At (t+ 23 min), external skin temperatures had reached 500 to 800 F, while temperatures throughout the interior of the container were at least 170 F, with higher temperatures at the walls and, particularly, the bottom of the container. A high velocity jet of gas could be seen emanating from the 3 inch vent, but t h e re was no measurable internal pre s s u re and the primary frangible disk was still intact. Release of the primary relief device and maximum p re s s u re pulse (spike) came at (t+23.54min). The maximum pressure recorded was 15 and 25 psig on the two respective monitoring systems. The pressure spike was brief in duration, (< 0.15 seconds), and re q u i red the 1,000 points/second re c o rding rate to detect and measure accurately. There was a major two-phase release of material from the vent indicating significant amounts of non-decomposed pero x i d e leaving the container prior to its decomposition, energy release, resultant heat release, and re s u l t a n t gaseous pressure generation. There was a minor secondary pressure pulse at (t+ 23.65 min) or (0.11) secJune 2002 91

Table 3. Results of full-scale U.N. 6 (c) burn.

Thermocouples 1, 3, and 6 in the right (west) side of tank.

Figures 5a and b. Full-scale Spec. 57 burn.

onds after the primary pulse, and measured 1 to 2 psig with a duration of ~0.03 seconds. During this time there was a rapid surge of internal temperatures. At (t+ 35 to 45 min), the internal temperature ranged from 345 to 800 F. There was rapid fluctuation on individual probes indicating internal turbulence and/or mixing/roiling effects, with the probes being alternately exposed to vapor and liquid which varied considerably in temperature. Two-phase emissions from the relief device were still occurring but with rapidly diminishing volume. By (t+ 50 min) fires due to internal peroxides had died down and temperatures started to diminish. The fuel fire was allowed to continue for another 70 minutes or a total burn time of 120 minutes. No further surges of temperature and/or pressure was observed in that time period. After the fuel fire was extinguished and the apparatus had cooled down, a visual inspection was made of the container. It was essentially empty with only some evidence of char on the internal walls and heating coils. Table 3 and Figures 5a and b represent typical temperature and pressure profiles for the burn.

Post-Burn Container Examination Post-examination of the container indicated there had been a slight deflection of the top and bottom flat heads. There was no evidence of loss of structural integrity, breach, or fragmentation. The container had relieved the peroxide decomposition and overpressurization successfully as designed. American Petroleum Institutes API 520/521Calculated Heat-up Rate vs. Full Scale Results The United Nations approved calculation for heatup rate in a fuel fire for a non-insulated tank (API 520/521) predicted 6.0 C/min/min. Test work in the 10 liter bench-scale vessel was conducted at 6.0 C/min. Post full-scale burn calculations of actual heat-up rate in the Spec. 57 container indicated 6.0+/-0.3 C/min. The API 520/521 prediction was confirmed and the 10 liter studies had been conducted therefore at the correct heat-up rate to predict the heat-up rate peak pressure pulse for the full-scale burn. 10 Liter Pressure Pulse Prediction vs. Full Scale Results Test runs conducted on the 10 liter vessel with equivalent A/V venting ratio of 1.00 sq.in./working-heated
Process Safety Progress (Vol.21, No.2)

92 June 2002

Figure 6. Results of an earlier test.

gallon with a heat-up rate of 6.0 C/min predicted a maximum pressure pulse of ~230 psig. During the fullscale burn test, the maximum pressure pulse was 15 to 25 psig, an order of magnitude lower than expected. The larger venting orifice size, 22.5 inch diameter vs. 1.75 inch diameter, permitted a more efficient nozzleeffect, and two-phase transport with minimized walleffect, as a result. This was not unexpected. Other works indicated i m p roved efficiency and reduced pre s s u re with increasing scale. Scale difference is regarded as the reason for lower peak pressures as was demonstrated by earlier proprietary work conducted by Herc u l e s Incorporated wherein an electrically heated 65 gallon D . O . T. Spec. 57 container was tested with a peak pressure of 170 psig for a similar design ratio of A/V = 1.00 (see Figure 6). The 10 liter studies therefore represent an extremely conservative method for safely predicting peak pressure pulse for a larger, full-scale, commercial container.
CONCLUSIONS

The nozzle efficiency for venting a two-phase mixture increases with increasing scale and, thereby, reduces the maximum pressure pulse significantly. Reduced maximum pressure pulses observed during the full-scale burn are attributable to improved nozzle efficiency and two-phase transport with increasing scale. The OPPSD/SPI Emergency Venting Methodology for establishing the safety of a specific peroxide for an approved container has been demonstrated. Likewise, the methodology can be applied to design a new container for an existing or new organic peroxide. The OPPSD/SPI Methodology is a suitable alternative method to U.N. Appendix 5 Sample Method. It a ff o rds the advantage of more recent technical advances in the determination of container ultimate strength. The OPPSD/SPI Methodology follows, and is presented in its entirety as submitted to the D.O.T. for consideration in September 2000. Identification assignment: P-1404.
EXAMPLE OF A TEST METHOD FOR VENT SIZING: OPPSD/SPI METHODOLOGY

The 10 liter OPPSD/SPI designed test vessel is suitable for organic peroxide decomposition studies. R e p roducibility has been demonstrated and is comparable in results to the U.N. Appendix 5, Test Method, test vessel. There is a near-linear relationship between heat-up rate and maximum pressure pulse for a constant A/V; venting area to working-heated volume ratio. H y d rostatic testing is an acceptable method for d e t e rmining the ultimate structural strength of a container through destructive testing. Linear and curvilinear finite element analysis (FEA) is an alternative method to hydrostatic testing and acceptable as a non-destructive method for identifying the ultimate, structural strength of a container. The U.N. approved API 520/521 method for calculating the heat-up rate in a fuel fire was verified as accurate in full-scale testing. The 10 liter test is a very conservative predictor of the full-scale maximum pressure pulse during an organic peroxide decomposition/overpressurization.
Process Safety Progress (Vol.21, No.2)

Introduction This example of a method for vent sizing is used to determine the required emergency vent capacity to be fitted to a specific IBC or portable tank for a particular organic peroxide, Type F, or self-reactive substance Type F, or any formulation thereof. The method is based on experimental data which indicates that, for organic peroxides or self-reactive substance formulations, the ratio of the minimum emergency vent area to the capacity of the IBC or tank is constant and can be determined using a reduced-scale tank with a 10 liter capacity. In the tests, the reduced scale tank is heated at rates equivalent to that given by complete f i re engulfment or, in the case of insulated IBCs or portable tanks, the heat transfer through the insulation with the assumption that 1% of the insulation is missing (see 4.2.1.13.8 and 4.2.1.13.9 of Recommendations on the Transport of Dangerous Goods, Model Regulations, United Nations, Eleventh Revised Edition). Other methods may be used provided that they adequately size the emergency relief device(s) on an IBC or portable tank to vent all the material evolved during self-accelerating decomposition or a period of not less than one hour of complete fire engulfment. Warning: The method does not take into account the possibility of initiation of deflagration. If this is a possibility, particularly if initiation in the vapor phase can propagate to the liquid phase, then tests should be performed which take this into account. Apparatus and materials The reduced-scale tank consists of a stainless steel test vessel with a gross volume of 10 liters. The top of the vessel, optionally, is provided with either a 1 mm opening, which simulates the pressure relief valve (PRV) of the IBC, or portable tank, or a real PRV of a diameter that is scaled using the vent area to vessel volume ratio.
June 2002 93

A second opening simulates the emergency vent opening and is closed by a bursting disk. The diameter of this vent opening can be varied by using orifice plates with different apertures. The bursting pressure of the disk fixed to the 10 liter vessel should be equal to the maximum rupture pressure of the bursting disks to be fitted to the IBC or portable tank. This pressure should be lower than the test pressure of the IBC or portable tank involved. Usually, the bursting pressure is set at a level that can cope with the pressures encountered during normal transport conditions, such as hydrostatic pressure from the liquid due to turnover of the IBC or portable tank, slopping of the contents, etc. The 10 liter vessel should be provided with a bursting disk with a set pressure in the range of the disk(s) fitted on the tank or IBC as to be used in transport. For safety, it is recommended to provide the test vessel with an extra bursting disk (bursting pressure approximately 80% of the design pressure of the 10 liter test vessel) with a large opening for additional emergency venting of the test vessel in the event that the chosen orifice diameter is too small. The outer surface of the test vessel, below the liquid level, is provided with an electrical heating coil or cartridge heaters connected to a power supply. The 10 liter test vessel contents should be heated at a constant rate independent of the heat being generated by the organic peroxide. The resistance of the heating coil or cartridge heaters should be such that, with the power available, the calculated heating rate (see section 3) can be achieved. The whole vessel is insulated with rock wool, cellular glass, or ceramic fiber. The temperature inside the vessel is measured by means of three thermocouples, two located in the liquid phase (near the top and bottom) and one in the gas phase. Two thermocouples are used in the liquid phase in order to check the homogeneity of the heating. The pressure is recorded by a pressure transducer(s) capable of recording slow and fast (at least 1,000 point/sec.) changes in pressure. Examples of test vessels are illustrated in Figure 1. Additional information may be obtained if the vessel is mounted in a tray designed to collect any solids or liquids ejected and/or on a scale to measure weight loss vs. time. The tests should be performed at a test site with suitable safety distances. Alternatively, the test can be performed in a bunker provided with sufficient ventilation and vent openings to prevent pressure build-up. Explosion-proof electrical equipment should be used in such a bunker to minimize the risk of ignition. However, the tests should be performed on the assumption that the decomposition products will ignite. Calculation of the heating rate to be used in the test If an IBC or portable tank is non-insulated, the heat load to the shell as given in 4.2.1.13.8 of the Model Regulations is required. For an insulated IBC or tank, the U.N. Model Regulations require that the heat load to the shell be equivalent to the heat transfer through the insulation, plus the heat load to the shell, on the assumption that 1% of the insulation is missing. The following information on the IBC or portable tank, and organic peroxide or self-reactive substance, is needed for the heating rate calculation:
94 June 2002

Fr = fraction of tank directly heated (1 if non-insulated, 0.01 if insulated) [-] M = total mass of organic peroxide and diluent [kg] K = heat conductivity of the insulation layer [W.m-1.K-1] L = thickness of insulation layer [m] U = K/L= heat transfer coefficient [W.m-2.K-1] A = wetted area of IBC or portable tank [m2] Cp = specific heat of the organic peroxide or selfreactive substance formulation [J.kg-1.K-1] Tpo = temperature of organic peroxide or self-reactive substance formulation at relieving conditions [K] Heat input, qi (W), via indirectly exposed surface (insulated part), is calculated by Equations 1 and 2: qi = 70961 * F * (1 - Fr) * A0.82 where F = insulation factor; F = 1 for non-insulated shells, or (1)

F = 2

U (923 Tpo ) for insulated shells. 47032

(2)

In the calculation of F, a multiplication factor of 2 is introduced to take into account a 50% loss in insulation efficiency in an incident. Heat input, qd (W), via the directly exposed surface (non-insulated part), is calculated by Equation 3: qd = 70961 * F * Fr * A0.82 (3)

where F = insulation factor = 1 (non-insulated) The overall heating rate, dT/dt (K/min), due to fire engulfment, is calculated by Equation 4: (4)

dT / dt =

(q i + q d ) 60 Mt C p

Example 1. Insulated portable tank

For a typical 20 m3 insulated portable tank: Fr = fraction of tank directly heated = 0.01 Mt = total mass of organic peroxide and diluent = 16268 kg K = heat conductivity of the insulation layer = 0.031 W.m-1.K-1 L = thickness of the insulation layer = 0.075 m U = heat transfer coefficient = 0.4 W.m-2.K-1 A = wetted area of portable tank = 40 m2 Cp = specific heat of the organic peroxide form = 2000 J .kg-1. K-1 Tm = temperature of peroxide at relieving conditions = 100 C and

Process Safety Progress (Vol.21, No.2)

0.4 (923 373) 47032 (1 0.01) 400.82 = 13533W q i = 70961 2

qd = 70961 * 1 * 0.01 * 400.82 = 14611 W

dT / dt =

(13533 + 14611)
16268 2000

-1 60 = 0.052K.min

Example 2. Non-insulated IBC

For a typical 1.2 m3 non-insulated stainless steel IBC (only direct heat input, qd): Fr = fraction of tank directly heated = 1 Mt = total mass of organic peroxide and diluent = 1,012 kg A = wetted area of IBC = 5.04 m2 Cp = specific heat of the organic peroxide form = 2190 J .kg-1. K-1 and qd = 70961 * 1 * 1 * 5.040.82 = 266 kW

dT / dt =

(0 + 266000)
1012 2190

60 = 7.2K.min -1

Procedure Fill the test vessel shell with the amount of organic peroxide or self-reactive substance required to give the same degree of fill (by volume of the shell) as to be used in the portable tank (maximum degree of fill 90%, by volume) and then install the required orifice plate* and bursting disk. For example, it is common practice to fit four 250 mm diameter bursting disks to a 20 ton portable tank. This corresponds to a test vessel orifice diameter of about 11 mm. * It is recommended that either small-scale vent experiments (100 to 200 ml scale) or experiments using a very strong vessel (> 100 bar) be perf o rm e d prior to the perf o rmance of the 10 liter vent test in order to obtain information on pressure effects from the test substance and on the required orifice diameter to be used in the first 10 liter scale vent test. For safety, start with the largest orifice. The vessel is heated at the desired rate by applying power to the heating coil or cartridge heater. A higher than calculated heating rate may be applied initially until a temperature 5 C above the self-accelerating decomposition temperature (for a 50 kg package) of the organic peroxide or self-reactive substance is reached. The calculated heating rate should be applied once this temperature is reached. The temperature and pressure in the test vessel are recorded during the entire experiment. After rupture of the bursting disk, the temperature should be maintained for a p p roximately 30 minutes more to be sure that all
Process Safety Progress (Vol.21, No.2)

dangerous effects are measured. After the test, the vessel should not be approached until the contents have cooled. The diameter of the orifice should be varied (if necessary) until a suitable opening is determined at which the maximum recorded pressure is: for portable tanks tested according to 4.2.1.13.4, not more than 85% of the maximum pre s s u re before failure as determined for the portable tank based on the hydrostatic limit or as determined by Finite Element Analysis, but not to exceed 25 barg. for IBCs tested according to 6.5.4.8.4 of the Model Regulations, not more than 85% of the maximum pressure before failure as determined for the IBC based on the hydrostatic limit, or as determined by Finite Element Analysis, but not to exceed 25 barg. The minimum test pressure of the IBC should be at least as specified in 6.5.4.8.4 of the Model Regulations (e.g., 2.0 bar for metal IBCs) or higher as tested by h y d rostatic pre s s u re testing or by Finite Element Analysis (FEA). The minimal test pre s s u re of the portable tank should be 4.0 barg (see 4.2.1.13.4 of the Model Regulations). Since these portable tanks and IBCs are used to transport organic peroxides (and other self-reactive substances) and are not pro c e s s vessels, permanent deformation is permitted during hydrostatic peak pressure testing so long as there is no fragmentation. A shipping vessel, should there be an incident resulting in permanent deformation, will be discarded without exception (no further use). The step size used should be related to the options available in practice for the IBC or portable tank, that is larger vent sizes or more vents. The test should be performed in duplicate, with the minimum total vent a rea having sufficient capacity. An altern a t i v e approach is to perform a series of six runs at varying orifice sizes so as to generate a continuous plot of peak pressure vs. the venting area to working volume ratio. This permits improved working latitudes with respect to the approval of additional peroxides and/or the design of additional containers. Test criteria and method of assessing the results The minimum or suitable (if it is acceptable to use a vent size larger than the minimum vent size) IBC or portable tank vent area, AIBC or Aportable tank (m2), can be calculated using the minimum orifice vent are a , Atest vessel (m2), determined in the 10 liter test at which the maximum pressure is not more than specified in the previous section of this method, and the volumes of the test vessel (Vtest vessel, m3) and IBC or portable tank (VIBC or Vportable tank, m3). The minimum total IBC or portable tank vent area is given by: For IBCs: AIBC = VIBC * (Atest vessel/Vtest vessel) For portable tanks: A portable (Atest vessel/Vtest vessel)
tank

= V portable

tank

June 2002 95

Example

For a typical organic peroxide in a 20 m 3 insulated portable tank: Atest vessel = Minimum suitable orifice area found in test = 9.5 10-5 m2 Vportable tank = Volume of portable tank = 20 m3 Vtest vessel = Volume of test vessel = 0.01 m3 Aportable tank = 20 * (9.5 10-5/0.01) = 0.19 m2 Utilizing the peak pressure curve for a given standard peroxide generated in a 10 liter study in the previous section on pro c e d u re, another pero x i d e exhibiting a lower peak pressure for the same venting ratio may use the same full-scale container without further testing of the container or the peroxide in that container (full-scale). Likewise, another pero x i d e which exhibits a higher pressure for a given venting ratio in the 10 liter 6 point study may be used in the full-scale container so long as the containers venting ratio is increased so as to produce a lower peak press u re than the standard organic peroxide pre v i o u s l y tested. Note: It is recognized in developing this method for vent sizing that dicumyl peroxide behaves differently than most other organic peroxides in vent testing. T h e re f o re dicumyl peroxide can continue to be shipped in the portable tanks and IBCs curre n t l y approved by the D.O.T., based on extensive testing and proven performance.

LITERATURE CITED

1. Banks, M., et al. , Safe Emergency Venting of Organic Peroxide Containers, EMRTC Report FR98-07, Research conducted for the Organic Peroxides Producers Safety Division of the Society of the Plastics Industry (OPPSD/SPI), Energetic Materials Research & Testing Center, New Mexico Institute of Mining & Technology, Socorro, NM 87801, February 1998. 2. Banks, M., et al. , Effects of Heating Rate on Organic Peroxide Reactions, E M RTC Report FR9 9 - 0 3 , Research conducted for the OPPSD/SPI, E n e rgetic Materials Research & Testing Center, New Mexico Institute of Mining & Te c h n o l o g y , Socorro, NM 87801, March 1999. 3. Code of Federal Regulations, Title 49 CFR, Subtitle B (Other Regulations Relating to Transportation), Chapter 1 (Research and Special Programs Administration, D.O.T.), Subchapter C Haza rdous Materials Regulation Part 178 (Specifications for Packagings), Subpart N Interm e d i a t e Bulk Container PerformanceOriented Standards and Subpart O Testing of Intermediate Bulk Containers. 4. C O S M O S / M , Finite Element Analysis System, Structural Research and Analysis Corporation, 12121 Wi l s h i re Blvd., Los Angeles, CA 90025, October 1996.

96 June 2002

Process Safety Progress (Vol.21, No.2)