Abstract
Owner/operators of chemical processing and petroleum refining sites often ask whether unconfined hydrogen vapor cloud explosions (VCEs) can actually occur. This question normally arises during the course of a consequence-based facility siting study (FSS) or a quantitative risk assessment (QRA). While it is generally recognized that a hydrogen release within a process enclosure could lead to an explosion, the potential for an external hydrogen release to cause a VCE is not as widely recognized and is often questioned. This uncertainty appears to stem from the impression that a hydrogen release always ignites quickly and near the point of release such that a flammable cloud does not have time to develop prior to ignition and/or that a hydrogen release never produces a flammable cloud of any significant volume due to its positive buoyancy.
Unfortunately, neither impression is correct. Hydrogen releases are actually susceptible to delayed ignition, and hydrogen releases can form significant flammable gas clouds near grade level. Unconfined hydrogen VCEs can and do occur. Furthermore, given the potential for rapid flame acceleration associated with hydrogen, the consequences of a hydrogen VCE can be severe. Consideration of such events in FSS and QRAs is, therefore, warranted.
Prior accidental hydrogen VCEs are reviewed to establish that such events do occur. Selected hydrogen VCE tests are also discussed to establish the potential severity of such events. Moosemiller and Galindo [10th Global Congress on Process Safety, 2014 Annual AIChE Meeting, New Orleans, LA, March 30–April 2, 2014] reviewed the ignition characteristics of hydrogen relative to the potential for a delayed ignition, and only the conclusions from that article are presented here. Example dispersions, using both simplified dispersion and computational fluid dynamics methods, are presented to illustrate the flammable gas volumes that can be created by hydrogen release scenarios. Blast load predictions are presented to illustrate the range of loads that could result from a hydrogen VCE due to such a release. © 2014 American Institute of Chemical Engineers Process Saf Prog 34: 36–43, 2015
Owner/operators of chemical processing and petroleum refining sites often ask whether unconfined hydrogen vapor cloud explosions (VCEs) can actually occur. This question normally arises during the course of a consequence-based facility siting study or a quantitative risk assessment (QRA). While it is generally recognized that a hydrogen release within a process enclosure (e.g., a building) could lead to an explosion, the potential for an external hydrogen release to cause a VCE is not as widely recognized and is often questioned. This uncertainty appears to stem from the impression that a hydrogen release always ignites quickly and near the point of release such that a flammable cloud does not have time to develop prior to ignition and/or that a hydrogen release never produces a flammable cloud of any significant volume due to its positive buoyancy.
While the practices and policies of specific petroleum refining and chemical processing facility owner/operators are confidential and hence cannot be discussed within the context of this article, it can be stated that some facility owner/operators do not consider unconfined H2-air VCEs as credible events. A number of other owner/operators do consider such events, but they limit the associated release sizes to small values compared to those normally considered for typical hydrocarbon releases. Others treat hydrogen releases essentially the same as typical hydrocarbon releases. A similar range of treatments exists in the recommendations provided by consultancies engaged in explosion hazard analyses, facility siting studies, and QRAs of such facilities.
It is noted that the Factory Mutual (FM) Global data sheet for the evaluation of VCEs [1] specifically excludes consideration of VCEs due to gaseous hydrogen releases, irrespective of the hydrogen pressure or temperature, although it does call for the evaluation of liquid hydrogen releases. The FM Global data sheet does note that 3% of the VCEs reported in a VCE incident database were due to hydrogen or synthesis gas. It is further noted that FM Global’s goals with respect to VCE evaluation may be different than those associated with facility siting studies and QRAs performed by facility owner/operators, but such considerations are beyond the scope of this article.
The focus of this article is on unconfined hydrogen VCEs. Since it is generally accepted that hydrogen released within an enclosure (i.e., a structure with roof and full walls) can result in an explosion, accidental explosions of that type are not addressed in this article. Of course, a hydrogen explosion within an enclosure still requires delayed ignition. It is also assumed that hydrogen-hydrocarbon mixtures are widely viewed as credible unconfined VCE scenarios, and hence are not addressed in this article.
The terms “congestion” and “confinement” are used within this article relative to the flame speed achieved in a VCE. Congestion is typically present within refining or processing areas in the form of piping, structural supports, instrumentation, conduit, and other similar items. Congestion induces turbulence in the flow field ahead of the flame, and hence increases the combustion rate and accelerates the flame front. Confinement is typically present in the form of limited solid decking or roofing, larger vessels, and/or small enclosures within the region of interest. Confinement restricts the free expansion of the product gas and hence accelerates the flame front. The presence of this type of limited confinement does not denote “fully confined” (e.g., as with a VCE within an enclosure), and a VCE is still “unconfined” even when such limited confinement is present.
Zalosh and Short [2] reviewed over 400 accidents involving hydrogen from a variety of sources covering the time period from 1965 to 1977. The referenced report provides details only for selected accidents. Zalosh and Short concluded that the data “indicate that hydrogen explosions have been a more serious problem than other types of hydrogen accidents in terms of the number of incidents, casualties, and reported property damage.” Their analysis showed that slightly more than half the hydrogen incidents were explosions, and that explosions accounted for three-quarters of the injuries and fatalities. Three-quarters of the incidents reviewed involved hydrogen gas, with most of the remaining incidents involving liquid. Hence, the analysis of Zalosh and Short [2] would indicate that a significant fraction of the incidents involved explosions of hydrogen gas, and that explosions caused a disproportionally large fraction of the casualties.
Perhaps the two most well publicized accidental unconfined H2 VCEs are the Polysar Sarnia [3] and Jackass Flats [4, 5] incidents. These two events were also reported in the summary provided by Lenoir and Davenport [6]. Lenoir and Davenport also reported an unconfined H2 VCE in 1975 at a hydrogen unit in Watson, CA involving 300 kg of H2, but because the source cited by Lenoir and Davenport is private communication, no further information is available; this event is also included in the incident collection provided by Gugan [7]. More recent accidental unconfined H2 VCEs occurred at the Silver Eagle refinery in Woods Cross, UT [8], although few details related to that event have been published, and at the Muskingum River power plant [9]. A number of additional incidents are summarized in the “H2 Incidents” database [10], although the information sources for these incidents are not reported and not a great deal of information is provided for any specific incident. The Manufacturing Chemists’ Association (MCA) Case History collection [11] also describes an accidental unconfined H2 VCE. Ordin’s review of accidents involving hydrogen in NASA operations [12] identifies a number of unconfined H2 VCEs. Limited information is also available on the unconfined H2 VCEs which occurred in Hanau, Germany [13] and Sodegaura, Japan [14]. Each of these incidents is discussed separately below. Table 1 summarizes the congestion and confinement present in each of these incidents; note that most of the congestion and/or confinement levels assigned in this table are judgments based on the description provided in the referenced source, as the authors did not personally investigate many of these incidents.
Fuente: http://onlinelibrary.wiley.com
