Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
4 Processing of M55 Rockets at ANCDF DESCRIPTION OF THE ANNISTON STOCKPILE The Anniston Chemical Agent Disposal Facility (ANCDF) was constructed near the Anniston Chemi- cal Activity, where stockpiled chemical agent muni- tions and containers are stored at the Anniston Army Depot in Anniston, Calhoun County, Alabama. The stockpile is stored in standard concrete, earth-covered igloos that are monitored to maintain the munitions in a safe and secure condition. The Anniston stockpile con- tains approximately 7.1 percent of the total 31,495 tons of agent in the original U.S. stockpile of unitary chemi- cal weapons. As noted in Chapter 1, as of July 2002, approximately one-fourth of this original tonnage had been destroyed during demilitarization operations at the Johnston Atoll Chemical Agent Disposal System (JACADS) and the Tooele Chemical Agent Disposal Facility (TOCDF). As noted in Table 1-1, ~ the Anniston stockpile has mustard agent in mortars, projectiles, and ton contain- ers. The nerve agent GB is contained in cartridges, pro- jectiles, and rockets. VX nerve agent is contained in projectiles, rockets, and mines (NRC, 1994a). At Anniston, 42,738 M55 rockets are GB-filled and 35,636 are VX-filled. M55 rockets contain a total of 457,300 lb of GB and 356,360 lb of VX (U.S. Army, 1998b). The number of gelled GB rockets at ANCDF was first estimated at about 33 percent of the inventory (Thomas, 2002~. A more precise estimate based on iTnformation from Army answers to questions from the Stock- those agent lots that are suspected to be in a gelled pile Committee as a follow-up to the September 25, 2002, fact- condition predicts a minimum of 8,706 rounds tap- finding meeting with the Army. proximately 20 percent. As indicated in Chapter 2, GB-filled rockets at Anniston are more prone to leak- age than munitions at other sites. A contributing factor could be the higher ambient temperatures at Anniston, which may accelerate aluminum corrosion relative to the other U.S. sites. At the time this report was pre- pared, 888 GB-filled M55 rockets were known to have leaked at Anniston. Public Concerns The Army' s plans for chemical demilitarization ac- tivities at Anniston have been delayed because of troubled relations between the various stakeholders. These include personnel in charge of emergency man- agement as part of the Chemical Stockpile Emergency Preparedness Plan, spokesmen for the Program Man- ager for Chemical Demilitarization (PMCD) and their contractor representatives, the Alabama Citizens Advi- sory Commission, and local officials ranging from the Calhoun County commissioners to the governor of Ala- bama and some members of the U.S. Congress. The previous governor had filed suit to postpone com- mencement of agent destruction operations until cer- tain local government demands were met. That suit has since been dropped. County commissioners have re- peatedly accused federal officials of failing to provide maximum protection for the surrounding populace and 26
PROCESSING OF M55 ROCKETS AT ANCDF of failing to keep promises relating to protective equip- ment, overpressurization of school buildings, and other protective measures. Individuals and organizations opposed to incineration technology per se have also voiced opposition to plans for stockpile disposal at the ANCDF. Steps were recently taken to overpressurize the schools, and additional protective equipment has been provided, improving the situation to some extent. However, early in 2003, anti-incineration propo- nents filed suit on the grounds that proper permitting procedures were not followed. They claim that the Army should therefore be required to obtain a new Resource Conservation and Recovery Act (RCRA) per- mit. They also claim that neutralization technologies are now available that should be used at Anniston in- stead of the baseline incineration system.2 As the Army's stockpile disposal program has pro- ceeded over more than a decade, the NRC has consis- tently urged the Army to engage community stakehold- ersintheiractivities.Somepreviousfindingsand recommendations on this topic are reviewed in Appen- dix A. ORIGINAL DISPOSAL PLAN FOR THE ANNISTON STOCKPILE The general design arrangements and proposed op- erations for disposal of GB M55 rockets at ANCDF are nearly identical with those employed at JACADS and TOCDF. They include loading of rockets at the storage site; transport to and unpacking at the disposal facility; processing in the rocket handling system (RHS), in- cluding draining the agent and slicing the rocket into eight sections using the rocket shear machine (RSM). Ungelled agent that is drained from rockets is processed in the liquid incinerator (LIC), and undrained gelled agent is processed in the deactivation furnace system (DFS) along with rocket energetics and metallic com- ponents. This rocket disposal system was shown in Fig- ure 3- 1. The original disposal plan that was used as a basis for risk analyses in the Phase 2 quantitative risk assess- ment (QRA) for the Anniston site called for processing the GB M55 rockets first, then processing the stock of VX munitions, followed by the GB projectiles, and fi- 2Chemical Weapons Working Group et al. v. United States De- partment of Defense and United States Army. This is a lawsuit filed in 2003 under the National Environmental Policy Act (NEPA) in the United States District Court for the District of Columbia. 27 natty the HDIHT munitions (SAIC, 2002c). The plan assumed that rocket processing would be at the TOCDF design rate of 32 rockets per hour, implying that all the rockets could be drained to a 5 percent heel. The GB M55 rocket campaign would thus take 390 days, and the destruction of all the munitions at Anniston would take 6.9 years. The original plan did not take into ac- count an early estimate that up to 13,000 rockets might contain gelled agent and therefore could not be drained. Even with a revised estimate of 8,706 gelled rockets, the Army has had to substantially revise the processing plans for ANCDF to allow for reduced processing rates for gelled rockets. Rockets containing GB are the most hazardous mu- nition because GB is the most volatile of the agents. Another factor contributing to the hazard presented by GB M55 rockets in storage is the fact that the acids formed during agent decomposition are corrosive to the aluminum rocket casing. The result is that these muni- tions generally have significantly higher leakage rates than other munitions in the stockpile. M55 rockets are packed in fiberglass storage tubes, 15 to a bundle, which are stacked on pallets in storage igloos. The concern that accidental ignition of a single stored rocket might trigger a large conflagration and a release of agent was the main reason for congressional authoriza- tion for the Chemical Stockpile Disposal Program (CSDP) in 1985, as noted earlier (NRC, 1994a). Compounding this concern has been the fact that stabilizers added to the rocket propellant are gradually depleting as they react with propellant degradation products. Agent leakage into pro- pellant may also hasten stabilizer degradation and gener- ate internal heat. Several times over the last two decades the Anny and its contractors reviewed the likelihood that these conditions could result in accidental ignition (U.S. Army, 2002a). These reviews are summarized briefly in Chapter 2 of this report and will be covered more thor- oughly in a forthcoming National Research Council (NRC) report on the status of stockpile degradation under storage conditions that is being prepared by the Stockpile Committee. MODIFIED DISPOSAL PLAN FOR THE ANNISTON STOCKPILE Description of the Modified Plan An initial challenge faced by the Army as it readies the ANCDF for commencement of disposal operations is to determine a safe rate for processing gelled GB M55
28 rockets. A related challenge is to obtain regulatory ap- proval to process rockets at a higher rate than the low rate allowed as a regulatory compromise and proven at TOCDF, and then to establish a schedule compatible with both regulatory approval and system capability. Over the last decade, the NRC Committee on Re- view and Evaluation of the Army Chemical Stockpile Disposal Program (the Stockpile Committee) has re- peatedly concluded and reported that the greatest risk to the public presented by the chemical weapons stock- pile is its continued storage (NRC, 1994a). The pres- ence of gelled agent in as many as 20 percent of the 42,738 M55 GB rockets stored at Anniston, and the TOCDF precedent of reducing production rates for these rockets to limit agent loading to the DFS kiln, means it will probably be necessary to extend the dis- posal processing schedule beyond that originally planned. This would extend the storage period and, hence, the period of increased risk to the public, work- ers, and the environment. To minimize the total time for stockpile destruction at the ANCDF and to deal with the need to process gelled rockets at a reduced rate, a modified schedule of operations has been proposed. The Army commissioned Continental Research and Engineering (CR&E) formerly the Denver office of Maumee Research & Engineering (MR&E) to con- duct a study to determine the maximum number of gelled M55 rockets each containing 10.7 lb of gelled agent that could be safely processed per hour in the DFS. In May 2000, CR&E reported that the DFS could handle up to 34 gelled rockets per hour according to the modeling analysis that it had performed (CR&E, 2000~. CR&E addressed a number of uncertainties and recommended that a staged ramp-up in the rate of dis- posal processing during the agent trial burn be followed for gelled rockets to show safe operation at a given rate before proceeding to a higher rate. This method allows demonstration of the maximum safe rate, which might be below the CR&E estimate owing to uncertainties in the analysis. A description and critique of this modi- fied process follows later in this chapter. Based on the CR&E report, the Army decided to seek approval for an increase in the disposal production rates for gelled GB M55 rockets (CR&E, 2000~. A RCRA permit modification request was submitted in June 2002 that redefined the plan for an agent trial burn (U.S. Army, 2002g). One section of the new plan proposed that 34 gelled rockets be processed per hour. The complementary processing (described in Chap- ter 3 under "Coprocessing") of GB rockets and GB pro- ASSESSMENT OF PROCESSING GELLED GB M55 ROCKETS AT ANNISTON jectiles is another approach for schedule improvement at Anniston (U.S. Army, 1999b). According to this con- cept, gelled rockets would be processed for a maxi- mum of 96 hours per week to provide 24 hours weekly for maintenance of the RHS. The rest of the time (48 hours) would be used to process GB projectiles inter- mittently. This complementary processing regimen could reduce the total time required for processing the rockets and the projectiles in separate campaigns. It would also reduce the number of agent changeovers required. Coprocessing, also described in Chapter 3, will also be utilized at ANCDF, where rockets and reconfigured munitions will be coprocessed. The ANCDF process- ing sequence envisions the reconfiguration of many projectiles in the inventory. All energetics are removed in the reconfiguration processing step and disposed of at the depot. At the plant, while one explosion contain- ment room (ECR) is processing rockets, the other can process the reconfigured, de-energized munitions. Agent extracted from these can be handled in the LIC, and metal parts can be processed in the metal parts fur- nace (MPF), all separately from the rocket process and its associated DFS. This complete separation of pro- cess equipment and flow permits simultaneous process- ing of gelled rockets and de-energized munitions con- taining the same agent (GB). Based on the TOCDF experience, which had only one gelled GB M55 rocket in the DFS kiln at any given time, it requires 6.5 min of residence time for rocket segments to traverse the length of the DFS kiln at a kiln rotation rate of 1.85 rpm, yielding a maximum charg- ing rate of 9.2 rockets per hour (WDC, 2002~. For an assumed system availability rate of 70 percent, as was used in preparing the life-cycle cost estimates, the con- tinuous production rate for scheduling purposes be- comes 6.4 rockets per hour.3 A summary of the various disposal campaign plan schedules for ANCDF is given in Appendix B (SAIC, 2002c). While the estimated time required to dispose of all GB munitions is 687 days (plus 268 days for the two agent changeovers) according to the original schedule in the second column of Table A-1, it drops to 594 days without any additional changeover time if complementary processing is undertaken (fifth col- 3From the notes of a meeting between a fact-finding group from the Stockpile Committee and the Army, September 25, 2002.
PROCESSING OF M55 ROCKETS AT ANCDF umn). Overall, the total number of agent changeovers would be reduced from three to two according to the modified plan, and the elapsed time would drop from 7.2 years to 6.3 years. Rationale for Implementing the Modified Plan The key features of the modified plan are designed to minimize the total time necessary for stockpile de- struction while conforming to regulatory limitations and safety requirements. Recognizing that gelled rock- ets must be destroyed in the DFS and that this is going to delay the processing rate, complementary process- ing of GB rockets and projectiles, along with coprocessing, has been embraced as a technique for reducing the overall disposal schedule at ANCDF. Ex- perience at JACADS and TOCDF has shown that agent changeover activity is hazardous to workers, time con- suming, and expensive (SAIC, 2002c).4 The elimination of one changeover by implementa- tion of the modified plan is viewed as desirable, pro- vided the overall risk remains acceptable, both in an absolute sense and as perceived by the public. A simi- lar decision to complete GB processing before process- ing VX M55 rockets was made at TOCDF because the 20-week planned changeover to VX activities could be made to coincide with the Olympic Games that would be taking place in Salt Lake City early in 2002. EXPERIMENTAL AND MODELING RESULTS Personnel at CR&E have broad experience in the operation of chemical demilitarization furnaces, includ- ing over 6 years with the four furnaces at JACADS. The furnaces and related equipment were field tested and proved successful, and all the furnaces (except the dunnage incinerators DUNs which are no longer used) have performed well, destroying about 8,000 tons of chemical munitions to date. In 2000, CR&E examined the ability of the DFS fur- nace planned for ANCDF to destroy M55 rocket segments containing gelled GB. Simplified computational fluid dy- 4During agent changeover operations, all areas exposed to agent are decontaminated by workers in DPE suits. Because automatic continuous air monitoring system (ACAMS) monitors are agent specific, monitors calibrated for the previously processed agent are replaced with monitors calibrated for the next agent to be processed. Changeover operations typically take about 4 months. 29 namics (CFD) models were constructed of the DFS burn- ing gelled rockets. Non-CFD modeling studies had been used by MR&E since at least 1989 in the design of the MPF of the Army's baseline incineration system. More recently, CR&E has been collaborating with Reaction Engineering International (REI) to produce and refine the CFD and reaction kinetic models for the baseline furnaces. The Anny asked CR&E to estimate the number of gelled rockets that could be safely destroyed per hour. The mod- eling study showed that up to 34 gelled rockets per hour could be processed safely, compared with 38 ungelled rockets per hour for rockets that have been at least 95 percent drained (CR&E, 2000~. CR&E recommended that "standard ramp up procedures be utilized for the shake- down period prior to the trial burn," starting at 10 rockets per hour for 2 hours (CR&E, 2000~. The first simplified model (heat transfer only) envisioned the kiln as a cylinder 30 in. in radius and 10 ft long. It contained a second, smaller cyl- inder 4.5 in. in diameter and 10 in. long, simulat- ing a rocket segment. The small cylinder contained a Composition B burster charge, GB agent, and M28 propellant, all of which were assumed to be open at the ends to the ambient conditions in the kiln. The kiln gas temperature was set at 1000°F, which resulted in a steady-state, no-load wall tem- perature of 800°F. Under these conditions, the model predicts the propellant face temperature will reach a steady state of about 950°F in about 0.1 min. This value is above the ignition tempera- ture, and CR&E predicted that the burster charge, agent, and propellant would "ignite within a few seconds after entering the furnace" (CR&E, 2000~. The heating of fiberglass tube sections was also modeled. In this case, the combustion of the tube section should be complete after a couple of min- utes (CR&E, 2000~. The feed rate was calculated from this steady-state heat transfer model, using the distance between flights (approximately 2.74 feet) to establish a separation se- quence for dropping the rocket segments through the feed chute into the inlet section of the furnace. The residence time in this section was about 1 min at 1 rpm. CR&E assumed that all the heat from combustion of the rocket components would be released in this sec- tion in 1 min. The maximum 1-min average heat re- lease the DFS could sustain was calculated from the performance experience at JACADS, where as many as 38 ungelled rockets per hour were processed. The heat of combustion from the burster charge, the propel-
30 lant, and the resin in the shipping/firing tube of each drained rocket had been estimated in 1989 as 177,345 Btu (CR&E, 2000~. This number does not include any Btu contribution from the agent. The next calculation was to estimate the percentage of the GB that would burn in the inlet section (1 min at an assumed kiln rotation rate of 1 rpm). CR&E assumed that about 20 percent of the gelled GB would burn in the first minute. Combustion of the 10.7 lb of GB con- tained in a rocket would release 93,197 Btu, with 20 percent of it, or 18,640 Btu, released in the first minute. If this is added to the 177,345 Btu per minute heat re- lease for the drained ungelled rocket, the total heat re- leased per rocket in the first minute is 195,985 Btu. Then, CR&E made a simplified calculation for the rate at (177,345/195,985) x 38, or 34 rockets per hour. Another 20 percent of the GB was estimated to burn in the second minute, and the final 60 percent was as- sumed to burn in the third minute. A second model was constructed of the DFS furnace itself; it was used to test the feed rate of 34 rockets per hour with gelled agent. This could have been a very com- plex model; however, several simplifying assumptions were made to enable results to be calculated using a rea- sonable amount of computer time. One major assumption was that a steady-state condition existed with respect to heat release and the resulting kiln temperatures. The flights in the kiln were assumed to be perpendicular to the axis of flow, not spirally located as they really are. Also, hot air was assumed to be the heating medium rather than the combustion process, and no spray water was included. It was further assumed that 20 percent of the GB (2.1 lb), all of the Composition B (3.2 lb) in the burster, and all of the M28 propellant (19.1 lb) burned in the inlet section. The rest of the GB (8.6 lb) was added between the first and second kiln flights. Field data were used to set the exterior kiln wall temperature at 300°F, the interior wall temperature at 900°F, and the infiltration air at 5.2 lb/s. The rotation rate was 0.1 rad/s (about 1 rpm), which gives a residence time of 1 min in the inlet section. The output from this model showed a maximum gas temperature of about 2600°F to 3000°F 3 ft from the charge end of the kiln, decreasing to about 2200°F at the gas exhaust duct (without cooling water sprays). Since these temperatures were not deemed by CR&E to be excessive, the rate of 34 rockets per hour was recommended with the qualification that this rate "be approached gradually (standard shake- down procedure), monitoring conditions closely over a period of several hours." ASSESSMENT OF PROCESSING GELLED GB M55 ROCKETS AT ANNISTON CRITIQUE OF MODELING AND AREAS FOR FURTHER INVESTIGATION While the committee respects the technical skill and accomplishment represented by the May 2000 CR&E modeling work, it believes that processing 34 rockets per hour may be unreasonably optimistic and that the actual maximum safe operating rate may be substan- tially lower. For one thing, if more than 20 percent of the GB is volatilized in the inlet section of the kiln by the heat released from the burster charge, the maxi- mum heat release in the inlet section of the kiln may be higher than the 195,985 Btu/min projected by CR&E. If the heat release rate is higher, it will probably be necessary to reduce the rocket feed rate to fewer than 34 per hour to avoid excessive temperatures in the inlet section. It is possible that the peak instantaneous heat release rate could be two or more times the 195,985 Btu/min maximum heat release averaged over the first minute because the heat will not be released uniformly over this 1-min period. The CR&E assumption of a 1 rpm kiln rotation rate, rather than the planned rate of 1.85 rpm, also allows more residence time in the model system as the rocket moves through the furnace. A counterbalancing factor is the plan to utilize an air in- filtration flow of up to 10 lb/s versus the 5.2 lb/s used for the model (U.S. Army, 2002h). This significantly higher air flow will help to reduce the maximum gas temperature for a given amount of combustibles in the inlet section and reduce the possibility that the DFS kiln will emit puffs of hot gases. The DFS kiln system at TOCDF processing 1 rocket per hour had 53.5 min (60 min less 6.5 min) to cool before another rocket was charged into the kiln. If the charge rate is increased to 9.2 rockets per hour, there will be no additional time for cooling beyond the 6.5- min charge intervals. By using a low start-up rate and gradual ramping up during the agent trial burn, DFS kiln gas and metal temperatures can be monitored be- fore deciding whether to continue increasing the rocket charging rate. Another area of uncertainty concerns the DFS feed chute. It is probable that a significant portion of the gelled agent will melt in the feed chute, vaporize, and be thermally destroyed (thermal decomposition and/or thermal oxidation). Thermal oxidation could heat the chute to excessive temperatures if rocket sections were fed too quickly. Adequate time must be allowed for the burning rocket components to clear the chute and for the chute to cool before introducing additional rocket
PROCESSING OF M55 ROCKETS AT ANCDF segments that might raise the chute temperature exces- sively. Another concern directly related to the peak instan- taneous heat release rate is transient pressure puffs oc- curring when large quantities of highly volatile materi- als burn within a short time. If the pressure in the kiln shroud (through which the combustion air from the DFS kiln is aspirated) becomes higher than the ambi- ent external pressure, there could be a short-term re- lease (puff) of gases into the DFS furnace room and/or into the ECR. Still another concern is the possibility of cracks in the charge chute and DFS kiln shell, as happened at TOCDF, where they were probably the result of ther- mal stresses caused by intermittent (unsteady-state) charging of rocket pieces, especially when gelled GB rockets were being charged only once per hour. The committee believes these thermal stresses at ANCDF might be less at a charging rate of 9.2 rockets or more per hour and might produce less cracking than oc- curred at TOCDF because the chute and kiln metal temperatures would remain more uniform (i.e., the heat released over a given 1-h period would be more uniform at 9.2 rockets per hour or more than at 1.0 or 1.6 rockets per hour. DETERMINING THE MAXIMUM SAFE OPERATING RATE The experience gained at TOCDF from having one gelled rocket in the DFS at a time is clear. As discussed elsewhere, this experience would support an assump- tion of a maximum processing rate of 9.2 rockets per hour. At an availability of 70 percent, this yields a full rate of 6.4 gelled GB rockets per hour. However, be- cause of the concerns cited above, the committee does not believe the processing goal of 9.2 gelled rockets per hour or any rate above the proven 1.0 or 1.6 rockets per hour has been confirmed by the modeling work so far. The actual maximum safe operating rate may be more or less than the 9.2 per hour goal. The only way to establish a maximum safe rate is to test for it during the agent trial burn. Therefore, the Army plans to start at a rate of two gelled rockets per hour and to demon- strate over a period of at least 2 weeks that observed pressure and temperature fluctuations are within de- sign limits at this rate. Once this is achieved, a ramp-up to four rockets per hour seems prudent. Again, safe performance must be demonstrated for 2 weeks before a further ramp-up to six rockets per hour is attempted. 31 This process should be continued until the maximum safe operating rate has been determined (DePew, 2000; Thomas, 2002~. With regard to ramp-up during the agent trial burn to establish an optimum safe performance rate in the throughput tests, all relevant existing process measure- ments acquired from the Process Data and Recording System of the baseline incineration system during pre- vious operations will need to be evaluated. In addition to the use of five infrared pyrometers along the length of the kiln to measure the kiln shell temperature pro- file, the following continuous, real-time measurements should be made at least every 2 s and recorded on high- speed continuous recorders for the time periods of in- terest to determine the maximum feed rate: Differential pressure between the lower end of the feed chute and the surrounding room using at least two differential pressure transmitters with low draft range. · Gas temperatures in the lower feed chute and the gas exhaust duct using at least three fast-response thermocouples. · Metal temperatures with at least two thermo- couples located close to the section of the chute nearest the DFS kiln. Even with the added measurement and monitoring ca- pabilities mentioned above, some further modeling may be desirable in order for operators to understand and inter- pret field measurements.5 This may bolster the under- standing of combustion efficiency issues. Graphic visuals from the models might prove useful in this regard. SCHEDULE IMPLICATIONS The processing of GB M55 rockets, agent change- over, and the processing of VX M55 rockets are typi- sThe committee has learned that a more comprehensive model of DFS operations on GB M55 rockets has recently been constructed by Reaction Engineering International and CR&E. This computa- tional fluid dynamics (CFD) model apparently reproduces quite ac- curately the actual operating results of processing GB M55 rockets at JACADS and TOCDF for example, average DFS exit gas tem- peratures and exit gas oxygen contents. The work is described in "Computational Modeling of a Chemical Demilitarization Deacti- vation Furnace System" (Denison et al., 2003~. The paper was pre- sented in May 2003 at the 22nd Annual Conference on Incineration and Thermal Treatment Technologies in Orlando, Fla.
32 catty the first tasks scheduled at chemical demilitariza- tion facilities in order to minimize overall risk to work- ers and the public. However, in planning ANCDF op- erations, such a schedule would prolong the overall time required because of the need for an additional agent changeover to complete the disposal of GB pro- jectiles after completing the disposal of all VX muni- tions. The additional changeover would probably con- tribute additional worker risk. Complementary processing would reduce the time and cost of the overall ANCDF operations and contrib- ute to eliminating the risk from the stored stockpile at Anniston sooner. However, the risk management analy- sis of August 2002 indicates that if complementary pro- cessing is undertaken, the GB M55 rocket campaign would be extended and the VX M55 rocket campaign delayed by some 120 days by interspersing GB projec- tile processing (SAIC, 2002c). Additional concerns associated with munitions de- livery during complementary processing periods at the Anniston site are these: · Deliveries of two types of munitions via on-site containers (ONCs) on trucks and trailers from storage igloos to the unpack area of the container handling building would require careful planning and scheduling. There would need to be adequate quantities of munitions on hand for processing in the two individual lines, and deliveries of one type of munition would need to be kept from interfer- ing with delivery and storage of the other type of munition in the unpack area of the container han- dling building. The risks associated with muni- tions delivery and storage at ANCDF have in gen- eral been estimated to be minimal (SAIC,2002b). There could be long-term effects of agent and de- contamination solution splashes in and around the RSM and the chute. However, there was no indi- cation in the end-of-campaign reports on rocket processing at TOCDF that the processing of gelled M55 GB rockets decreased the availability of the RHS, including the RSM. . The TOCDF experience with gelled rockets shows that a destruction and removal efficiency (DRE) of 99.9999 percent is unfailingly achieved with a kiln resi- dence time of 6.5 min. Based on this residence time, the committee believes that a DFS processing rate of 9.2 rockets per hour (with a 70 percent availability rate) ASSESSMENT OF PROCESSING GELLED GB M55 ROCKETS AT ANNISTON will prove feasible and safe. Such a rate would allow only one rocket in the kiln at any given time. In a seminal 1994 report, Recommendations for the Disposal of Chemical Agent and Munitions, the Stock- pile Committee considered the options for configuring a chemical agent and munitions disposal facility based on incineration technology (NRC, 1994a). The report confirmed the Army's earlier decision that four sepa- rate furnace systems should be installed to handle agent, energetics, metal parts, and dunnage. The rea- son for this was that the separation of these streams was considered an important safety feature of the baseline incineration system. The report states that the separation into streams ". . . provides the designer the freedom to tailor the design of each disposal system to the properties of the separate (and quite different) ma- terials to ensure safe, controllable operations" (NRC, 1994a). The separation of streams idea was followed in the original designs for JACADS and TOCDF. However, as operating experience was gained and specific prob- lems and opportunities arose, changes were made in the original design concept. Four specific examples of process evolution follow: The dunnage incinerators at JACADS and TOCDF were not operated because it was deter- mined that the dunnage (pallets, containers, DPE suits, etc.) could be stored safely and then pro- cessed through the MPF periodically or shipped off-site. Accordingly, the dunnage incinerator was omitted from the ANCDF design. The Army also determined that the brine reduc- tion area (BRA) was not needed since the brine could be processed safely off-site and at lower cost. Although the BRA was included in the ANCDF design, its use is not currently intended (personal communication between Col. Christo- pher Lesniak, PMCD, and the M55 Committee on February 4, 2003). The third example of process evolution, PAS fil- ter systems (PFS), was not in the JACADS or TOCDF designs but was installed at ANCDF and the other new baseline facilities. The PFS uses high-efficiency particulate air filters and beds of activated carbon to treat flue gas at these facili- ties. Its purpose is to further reduce the already low traces of various products of incomplete com- bustion and metals and virtually eliminate the
PROCESSING OF M55 ROCKETS AT ANCDF possibility of an accidental release of chemical agent through the stack. This modification should serve also to enhance public confidence in the safety of the disposal operation (NRC, 1999~. · The fourth example is the development of a modi- fied baseline process concept for processing mus- tard agent munitions at the Pueblo, Colorado, site (NRC, 2001~. This process utilized a single MPF to process agent and metal parts simultaneously. The energetics would be handled in a DFS or sent off-site.6 These four examples show that the baseline incin- eration system is continually being modified by evolu- tionary development and operational improvements to enhance safety or to increase efficiency while main- taining a high standard of safety. Improvements may stem from lessons learned during the course of opera- tion or the need to meet newly identified, specific pro- cessing requirements. The proposal put forth by the Army for processing gelled GB M55 rockets at ANCDF is the most recent example of attempting to meet new processing needs. As pointed out in Chapter 3, crystallized or gelled agent was unexpectedly encountered in some of the GB rock- ets processed at TOCDF. Special permit modifications had to be obtained to allow their processing, but these modified conditions slowed the disposal. At Anniston, which has roughly twice as many gelled rockets as Tooele, the delay could be significant and would in- crease storage risk. The Army's experience at TOCDF in processing gelled GB rockets, coupled with suffi- cient modeling and agent trial burn testing for safely increasing the throughput rates during DFS operations at ANCDF, could lead to an improved processing se- quence that saves time and reduces storage risk. RISK IMPLICATIONS OF ACCELERATED PROCESSING The Chemical Stockpile Disposal Program (CSDP) was established to destroy the U.S. stockpile of unitary chemical weapons while ensuring maximum protection 6A separate NRC committee examined alternative technology op- tions for the Pueblo disposal facility while the Stockpile Committee prepared its report on a modified baseline process. The modified baseline process and an alternative neu~alization-based process were separately determined to be technologically feasible. The neu~aliza- tion process was selected after all factors, including public reaction and preferences, were considered. 33 of the general public, personnel involved in the destruc- tion effort, and the environment. To attain this goal, PMCD established effective risk management systems, outlined in its Guide to Risk Management Policy and Activities (U.S. Army, 1997b). The guide defines the process and a series of assessments to be used to evalu- ate CSDP project risk and discusses how the assess- ment results are used in decision making to ensure that any changes to a project will not be made unless the changes continue to provide maximum protection to the public, workers, and the environment. In accor- dance with this guidance, the storage risk, the worker risk, and the general public risk from agent exposure were extensively investigated in developing the Phase 1 and Phase 2 QRAs7 for the Anniston site. However, because the original schedule plan for disposal (see the second column in Table B-1) was used for the analy- ses, the QRAs did not provide for schedule extensions associated with having to process gelled GB M55 rock- ets at a reduced rate in the DFS. Consequently, the QRAs also assumed that the GB M55 rocket campaign would be followed by the VX campaign (with all M55 rockets being destroyed first) and that the remainder of the GB munitions would be destroyed next, followed by the HD/HT stockpile. Public Risk The public risks calculated in all the QRAs per- formed to date show that the risk associated with con- tinued storage is larger than the risk associated with processing (SAIC, 2002~.8 These values were estimated using a comprehensive QRA methodology, which was reviewed earlier by the NRC in a risk report (NRC, 1997~. Thus, the public risk is essentially controlled by the duration of potential exposures from stored stock- pile components. Table 4-1 shows the revised QRA public risk estimates for Anniston for the four sched- ules shown in Table B- 1 in Appendix B. The risk com- pared in Table 4-1 is the Public Acute Fatality Risk, which is the estimated expected number of fatalities 7A Phase 1 QRA evaluates public risks from a proposed facility before it is constructed. A Phase 2 QRA is a detailed evaluation of the risks and consequences of accidental releases of agent to work- ers and the community based on the site-specific design and opera- tions. 8In general, risk estimates of fatalities in QRAs are arrived at by considering the probability of an accident occurring in combination with the likely consequences of such an accident.
34 ASSESSMENT OF PROCESSING GELLED GB M55 ROCKETS AT ANNISTON TABLE 4-1 Comparison of Storage Risk for the Anniston Public Under Four Different Rates of Rocket Disposal Schedule Option Total Public Risk of Storage from Start to Completion of Disposal Processing Public Acute Fatality Risk (Meanja Original schedule 32 ungelled rockets per hour Original schedule 9.2 gelled rockets per hour Modified schedule 9.2 gelled rockets per hour Modified schedule 1.6 gelled rockets per hour 5.1 x 10-2 5.8 x 10-2 6.5 x 10-2 9.5 x 10-2 aThese numbers range from 0.051 to 0.095 fatalities. Source: Adapted from SAIC (2002c). Over the planned duration of the disposal process for each of the four options. The original schedule risk is not applicable unless gelled rockets can be processed through the DFS at the same rate as drained ungelled rockets. This seems un- likely, and even if it were possible, the delays in dem- onstrating its safety would probably more than offset the time that would be saved by processing at this rate. Table 4-1 shows that for the projected rate of process- ing gelled GB M55 rockets, 9.2 rockets per hour, mov- ing from the original schedule to the modified schedule entails some increase in public risk. This increase is associated with delaying the destruction of the VX M55 rockets by about 120 days while the other GB muni- tions are destroyed. Storage of VX M55 rockets poses a higher risk than storage of the other GB munitions. However, it is useful to note that the differences in these risk estimates are less than the uncertainty ranges asso- ciated with the absolute numeric estimates of risk, and that other sources of 4-month delays can cause equiva- lent increases in public risk. Worker Risk Industrial accidents include all manner of non-agent- related injuries, such as cuts and falls. The work de- mands at a chemical agent disposal facility may be more complex than at a typical industrial facility. The committee believes that continuing improvement in training and attention to rules are essential to the safety of CMA operations. The need for this continuing im- provement was a consistent theme of earlier NRC re- ports, as exemplified by Recommendation 13 from Evaluation of Chemical Events at Army Chemical Agent Disposal Facilities (NRC, 2002b): Recommendation 13. A generous allotment of time should be given to training and retraining chemical demilitarization plant operating personnel to ensure their total familiarity with the system and its engineering limitations. All plant personnel should receive some education on the total plant operation, not just the area of their own special responsibility. The extent of this overall training will be a matter of judgment for plant management, but the training needs to focus on how an individual's activities affect the integrated plant and its operational risk. Each facility should develop training pro- grams using the newly-designed in-plant simulators to simulate chal- lenges that require knowledge based thinking. The training programs should include a process for judging the effectiveness of the train- ing. Including "design" experts in the start-up crew for new plants could be helpful in identifying latent failures in process and facility design. The likely number of disposal worker fatalities from agent contact during processing was estimated to be 0.5 over the anticipated 7-year disposal period. This is the calculated result of a very detailed fault tree analy- sis covering a wide variety of conceivable accidents and mishaps. It is higher than average industrial expe- rience, which would predict about 0.1 fatality for the same work period (SAIC, 2002b). It should be recog- nized that these probabilities are the best estimates from risk experts working in the field. The fact that there has not been an agent-related fatality in the 20 years of combined experience at JACADS and TOCDF sug- gests that these computed values may be high. The one worker fatality that has occurred happened during a maintenance operation, while JACADS was shut down. It was not related to the presence of agent. The QRAs do include some worker risks associ- ated with agent changeover periods, but these risks have not been developed in detail because of the per- cention that risks are reduced when agent processing is stopped between disposal campaigns. The committee notes that there have been several agent events while facilities are in shutdown condition (NRC, 2002b). A
PROCESSING OF M55 ROCKETS AT ANCDF recent incident involved a release of VX outside engi- neering controls during processing of waste materials at JACADS; another recent incident involved the re- lease of GB during agent changeover operations at TOCDF. The latter occurred inside engineering con- trols but did involve exposure of two workers to a non- lethal concentration of GB. The Army has observed a higher frequency of unplanned events when nonroutine operations, including maintenance and changeovers, are being performed.9 Therefore, the worker risks as- sociated with changeovers may not be fully considered in current QRA estimates. The net impacts on worker risk of the modified plan scenarios for ANCDF are difficult to assess. For ex- ample: · Worker risks are increased due to complementary processing activities, but the total time of worker exposure to risk is decreased. · Risks to workers are decreased by the elimination of one changeover period if all GB munitions are processed together. Although detailed worker risk analyses for the sce- narios in Appendix B that consider processing of gelled GB rockets have not been done, it is likely that worker risk is somewhat reduced for the modified schedule with processing at 9.2 (sustained rate of 6.4) gelled rockets per hour. Health and Environmental Risks Chemical agent disposal operations can pose other, more general risks to the public. These may be health risks posed by exposure to hazardous materials other than chemical agents for example, from stack emis- sions of metals or organic materials. Estimates of the extent of this risk are contained in a health risk assess- ment (HRA). The development of the HRA for Anniston required data from yet-to-be-performed agent trial burns; consequently, an HRA could not be made available to the committee when this report was being prepared. Emissions over the course of destroying the stockpile at 9Personal communication between Conrad Whyne, PMCD, and the Committee on Review of Army Planning for the Disposal of MSS Rockets at the Anniston Chemical Weapons Disposal Facil- ity, March 26, 2003. 35 Anniston will produce small receptor exposures to nonagent emissions within regulatory guidelines, and it does not appear that shifts in schedule of the type envi- sioned in the modified schedule options given in Table B-1 will have much, if any, effect on the Anniston HRA. Although data on the security risk from extended stockpile storage were not provided to the committee, it is intuitively obvious that the longer weapons of mass destruction are maintained intact, the greater the risk of misappropriation and misapplication by an unautho- rized person or persons. Overall Risk to the Public, Workers, and the Environment The minimization of risk to the public, workers, and the environment is a guiding precept of the CSDP. With regard to the options for processing rockets containing gelled agent at Anniston, it appears that maintaining the original schedule probably entails increased worker risk because of the additional agent changeover needed to expedite processing of VX M55 rockets before com- pleting the processing of other GB munitions. On the other hand, the original schedule reduces public risk by a small amount by eliminating the VX M55 rockets about 4 months earlier. However, neither the original nor the modified schedules will minimize both public and worker risk. Thus, the Army will have to make a decision that is based on judgment and proactive con- sultation with regulators and other concerned parties. SURROGATE TRIAL BURN IN THE DFS From May 29 through June 4, 2002, ANCDF con- ducted a surrogate trial burn in the DFS using an agent surrogate. The surrogate was a mixture of 67 weight percent liquid monochlorobenzene (MCB) and 33 weight percent solid hexachloroethane (HCE) (U.S. Army, 2002i). MCB was chosen because it has ther- mally stable bonds and is harder to destroy than agent. HCE was selected because it simulates a solid gel and has a high chlorine content. This mixture challenges the combustion efficiency of the DFS and the ability of the PAS to remove the acid hydrogen chloride (HC1) gases. Three sets of test conditions were employed: · a low-temperature test (LTT) without the PFS · a high-temperature test (HIT) without the PFS · an HTT in which the PFS was operating (HTT- PFS)
36 The LTT was designed to demonstrate the permit lim- its for maximum hourly rolling average feed rates and minimum combustion temperatures. The DFS inlet was kept at more than 950°F (i.e., a nominal operating tem- perature of 1100°F). The afterburner operated in a tem- perature range between 1700°F and 2200°F (i.e., a nomi- nal operating temperature of 1850°F). The heated discharge conveyor ran at 1000°F or hotter. A plastic bottle containing approximately 7.6 lb of the surrogate was put into a wet burlap bag and fed to the DFS through the feed chute every 106 s. The total surrogate feed rate was 257.1 lb/in. This was intended to approximate the heat release from feeding the GB in 34 gelled rockets per hour, or 363.8 lb/in. However, each gelled rocket also contains 3.2 lb Composition B burster charge, 19.1 lb M28 propellant, and 7.2 lb epoxy resin in the fiberglass shell, a total of 29.8 lb of other combustible material. This was not compensated for in the LTT but was to some extent in the HTT and HTT-PFS tests. The HTT and HTT-PFS test conditions were designed to test the maximum hourly rolling average feed rates for metals and the maximum hourly average kiln tem- peratures (U.S. Army, 2002i). A high chlorine injection rate was employed because the volatilization rates of metals increase in a high chlorine atmosphere. The DFS temperature was held at the same condition as in the LTT (a nominal operating temperature of 1100°F), but the afterburner ran much hotter, at a nominal operating set point of 2150°F. Plastic bottles in this case contained 19.4 lb of a mixture of metal oxides (11 weight percent), ethylene glycol (56 weight percent), MCB (17 weight percent), and HCE (16 weight percent). The ethylene glycol was added to boost the thermal loading of the surrogate feed to that of a gelled GB M55 rocket, includ- ing energetics. A bottle was charged every 112 seconds for a total feed rate of 621.6 lb/in. Three runs were made under each test condition, so there were nine runs altogether. Selected data from these nine runs are summarized in Table 4-2 (U.S. Army, 2002j). The highest measured emission rates or concen- trations for each test condition are listed. Values that are above allowable standards are in bolt/face type. The regulatory 99.9999 percent DRE targets for agent surrogate destructions were met. There were 21 metals injected in the HTT trials, and all except cad- mium, lead, and mercury tested within permit limits in the exhaust gas. However, the PFS was able to reduce these three emissions to well below the permit levels. Vinyl chloride was not detected, but the detection limit of the analytical equipment was so high it was not pos- ASSESSMENT OF PROCESSING GELLED GB M55 ROCKETS AT ANNISTON sible to judge if the standard was met or not. Only 1 of the 17 substances of potential concern (SOPCs), tetrachlorodibenzofuran, was above the limit; the rest were not, and the total ITEMS levels for polychlori- nated dibenzo-p-dioxins and dibenzofurans were very much lower than allowed. Carbon monoxide levels were also very low, which indicates that products of incomplete combustion were minimal. Chlorine, HC1, and particulate emissions were also very low. On September 1, 2002, the Alabama Department of Environmental Management (ADEM) approved the surrogate trial burn report and asked ANCDF to make it available for public review (ADEM, 2002~. A surro- gate trial burn was also undertaken at the LIC furnace from March 16 to March 23, 2002. The results are available but are not germane to the processing of gelled rockets and so are not covered in this report (U.S. Army, 2002k). On December 20, 2002, ADEM also approved that report. The DFS surrogate trial burn enhances the prospects that more than 1.6 gelled GB M55 rockets per hour can safely be processed through the DFS. However, the conditions in the furnace when rocket segments are charged may be much different from conditions when bags of surrogate, each representing either the agent content alone or the agent and energetics content of a rocket, are charged. Proof for the ability of the DFS to process more than 1.6 gelled GB M55 rockets per hour awaits an agent trial burn. In June 2000, the Army sub- mitted a RCRA revision to test processing of 34 gelled rockets per hour. However, in 2002, the Army revisited its plans for processing 34 rockets per hour. A new procedure calls for processing one gelled GB rocket in the DFS kiln at a time, with 6.5 min residence time, for a total theoretical maximum processing rate of 9.2 gelled GB rockets per hour (U.S. Army, 2002m). Tak- ing into account a 70 percent DFS availability yields a maximum full rate of 6.4 gelled GB rockets per hour. l°International toxic equivalency quotient (ITEQ) dioxin is the amount of 2,3,7,8-TCDD (2,3,7,8-tetrachlorodibenzo-p-dioxin) with toxicity equivalent to the complex mixture of 210 dioxin and furan isomers with between 4 and 8 chlorine atoms found in flue gases. This equivalency is based on the International Toxic Equiva- lence Factor scheme adopted by EPA and most countries to sim- plify the reporting of dioxin emissions.
PROCESSING OF M55 ROCKETS AT ANCDF TABLE 4-2 Results of ANCDF Surrogate Trial Burn Runs for the DFSa Condition 1 Condition 2 Condition 3 RCRA or CAAb Parameter LAY: HEY: H1Y7-PFS Permit Limit DREforMCB, % >99.99994 NAC NA 99.9999 DRE for HCE, % >99.999997 NA NA 99.9999 HC1, g/s <6.40E-04 <6.81E-04 <6.48E-04 1.66E-02 C12 g/s <2.71E-03 <2.00E-03 <2.14E-03 4.03E-03 HC1 / C12 ppm at 7% O2 <0.60 <0.48 <0.50 21 HE, g/s <6.52E-04 <6.93E-04 <6.60E-04 1.718E-02 Particulates, lb/in <0.049 <0.086 £0.053 1.18 Particulates, g/dscf at 7% O2 <0.00084 <0.00126 £0.00076 0.015 NOX, lb/in <3.0 <4.5 <4.75 112 SO2 lb/in <6.25 <6.5 <5.0 14.5 CO, ppmv <0.02 <0.51 <0.10 100 Cadmium, g/s NA <4.00E-04 <2.60E-06 1.36E-05 Lead, g/s NA <1.98E-03 <1.43E-05 3.49E-04 Mercury, g/s NA <6.82E-05 <2.90E-06 5.42E-06 Vinyl chloride, g/s <1.00E-05 [ND]d NA NA 1.67E-06 Benzene, g/s <1.56E-05 NA NA 1.14E-04 Total PCDD/PCDF, ng TEQ/dscm at 7% O2 <0.023 <0.041 <0.028 0.20 aThe values reported for Conditions 1, 2, and 3 are the highest values measured during three test runs. Values above allowable standards are in boldface type. bClean Air Act. CNot applicable. ~Not detected. SOURCE: Adapted from U.S. Army (2002j). APPLICABILITY OF THE PROPOSED PROCESS FOR ANNISTON TO OTHER SITES The applicability of the proposed modified Anniston process to other sites depends on the presence of gelled GB agent in M55 rockets at those sites, as well as on the regulatory and public relations climate. The Army has estimated that gelled rockets are present at Umatilla but amount to only 3 percent of the total (2,791 of 91,375 GB M55 rockets in storage).~ Processing this small number at the TOCDF rate would extend the rocket campaign by 2,791/~1.6 x 24) = 73 days. If the Anniston modified system could be employed, the time required to destroy gelled rockets would be extended by 2,791/~6.4 x 24) = 18 days. The campaign extension llInformation from Army answers to questions from the Stock- pile Committee as a follow-up to the September 25, 2002, fact- finding meeting with the Army. 37 could therefore be reduced by 55 days if the Anniston modified process can be employed at Umatilla. This modest improvement and reduction of storage risk might not warrant the problems encountered in seeking regulatory approval and public understanding of the schedule change. Because the stockpile at Pine Bluff is not believed to contain any of the agent lots known to exhibit gelling,~2 there would be no perceivable benefit to employing the modified disposal plan for ANCDF at Pine Bluff. The stockpile at Blue Grass is to be destroyed by neutral- ization-based technology. Since a DFS is not a compo- nent of this technology, the Anniston modified disposal plan is not applicable to the Blue Grass site. 12Ibid.
