Experimental and numerical methodology for the analysis of fireproofing materials

Journal of Loss Prevention in the Process Industries 28 (2014) 60e71

Contents lists available at SciVerse ScienceDirect

Journal of Loss Prevention in the Process Industries journal homepage: www.elsevier.com/locate/jlp

Experimental and numerical methodology for the analysis of fireproofing materials Francesca Argenti, Gabriele Landucci* Dipartimento di Ingegneria Civile e Industriale, Università di Pisa, Largo Lucio Lazzarino 1, 56126 Pisa, Italy

a r t i c l e i n f o

a b s t r a c t

Article history: Received 1 April 2013 Received in revised form 15 May 2013 Accepted 15 May 2013

In this study, a methodology for the assessment of fireproofing materials performance is presented. The methodology is based on a combined experimental and numerical approach. A modified version of the ASTM E162 standard fire test was used to expose specimens of steel board protected with different types of fireproofing materials to a steady radiation source. The temperature of the steel board was recorded with an infrared camera in order to evaluate the heat up due to the fire and characterize the protective performance. Experimental results were used to validate a simplified mono-dimensional model which allowed simulating more severe conditions and different protection configurations. A specific key performance indicator (KPI) was used for the quantitative assessment of fireproofing effectiveness. Finally, the professional career of Menso Molag, safety pioneer in the framework of hazardous materials transportation, was outlined. Ó 2013 Elsevier Ltd. All rights reserved.

Keywords: Passive fire protection Fireproofing Fire test Numerical modeling Hazardous materials transportation

1. Introduction Severe fires, mainly due to the ignition of accidental releases, may affect process equipment or transport vessels leading to a catastrophic loss of containment (Center for Chemical Process Safety [CCPS], 2000; Cotgreave, 1992; Cowley & Johnson, 1992; Khan & Abbasi, 1999; Lees, 1996; Roberts, Medonos, & Shirvill, 2000). In the case of flammable liquefied gases (such as Liquefied Petroleum Gas - LPG, propylene, ammonia, etc.), this type of rupture may be followed by a BLEVE (Boiling Liquid Expanding Vapour Explosion) and associated fireball with extremely severe consequences for workers and population (Abbasi & Abbasi, 2007, 2008; CCPS, 1996; Reid, 1979; Roberts, 1981, 1982; Manas, 1984). Hence, a key issue to enhance safety and to reduce the risks related to both fixed installations and hazardous materials transportation is the development and the application of specific protections, able to reduce the thermal weakening of the fired equipment. The adoption of passive fire protection materials (PFP), e.g., installation of protective coatings able to withstand severe fire exposure conditions, may represent a highly safe and effective solution (Di Padova, Tugnoli, Cozzani, Barbaresi, & Tallone, 2011; Roberts, Shirvill, Waterton, & Buckland, 2010; Tugnoli, Cozzani, Di

* Corresponding author. Tel.: þ39 050 2217907; fax: þ39 050 2217866. E-mail addresses: [email protected], [email protected] (G. Landucci). 0950-4230/$ e see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.jlp.2013.05.005

Padova, Barbaresi, & Tallone, 2012). It is worth mentioning that, depending on the fire exposure severity and applied PFP material layer, this type of measures may not totally avoid the occurrence of catastrophic failure and thus BLEVE, as remarked by Salzano, Picozzi, Vaccaro, and Ciambelli (2003). Nevertheless, since the presence of PFP reduces the temperature and pressure increase in the vessel, a stretch in the time to failure (Droste & Schoen, 1988; Molag & Kruithof, 2005; Salzano et al., 2003; Steel Construction Institute [SCI], 1992; Townsend, Anderson, Zook, & Cowgill, 1974) may be obtained leaving a safety margin for the external emergency teams’ intervention for equipment cooling and fire suppression (Hobert & Molag, 2006; Landucci, Gubinelli, Antonioni, & Cozzani, 2009), thus eventually preventing the accident escalation. PFP systems are widely applied in fixed installations (e.g., storage units, critical process units, etc.) and several standards rule the specific design and testing of materials (American Petroleum Institute [API], 2010; International Organization for Standardization [ISO], 2007; National Fire Protection Agency [NFPA], 1991; SCI, 1992; Underwriters Laboratories Inc. [UL], 1994). On the contrary, several issues are still open concerning the possible implementation of effective fire protections, based on thermal coatings, for road and rail tankers in the specific European context (European Commission, 2006a, b; Paltrinieri et al., 2009). In this case, severe exposure conditions and specific issues related to transportation (e.g., damage to coating following accidents or collisions, defective coating installation, deterioration due to coating erosion/corrosion, etc.) must be taken into account (Birk,1999; Birk, 2005; VanderSteen & Birk, 2003);

F. Argenti, G. Landucci / Journal of Loss Prevention in the Process Industries 28 (2014) 60e71

besides, emergency response may be not as effective as in the case of fixed installations, thus a higher strength in the protective performance is required (Molag & Kruithof, 2005). Hence, PFP testing in severe fire conditions is of fundamental importance for the development of robust fireproofing that could be suitable also for the application on road tankers or tank wagons. Several experimental configurations reproducing typical fire scenarios on real scale have been proposed in the last 30 years, in order to test materials performance (Birk, Poirier, & Davison, 2006; Droste, Probst, & Heller, 1999; Droste & Schoen 1988; Kielec & Birk, 1997; Landucci, Rossi, Nicolella, & Zanelli, 2009; Townsend et al., 1974). Those tests, carried out on large or pilot scale were aimed at the characterization of both materials and protected structures response to fire (Cowley and Johnson, 1992; SCI, 1992). The main advantage of such large scale tests is that design guidelines for PFP systems to be applied on industrial equipment can be directly derived, without requiring scale up protocols (Cowley & Johnson, 1992; NFPA, 1991; Roberts et al., 2010). On the other hand, large scale tests are not easily reproducible and require high financial efforts for experimental set up preparation and management. In addition, their realization may arise environmental and safety concerns. Therefore, in order to carry out a preliminary design and screening of technological solutions for PFP systems development, bench scale laboratory tests are an effective solution, with lower costs both for equipment and tested specimens (American Society for Testing Materials [ASTM], 1994a, b; Cowley & Johnson, 1992; Landucci, Rossi, et al., 2009). In the present study, a methodology for the fireproofing performance assessment is presented. The method is based on combined experimental and modeling activities. The experimental set-up for the evaluation of fireproofing materials performances is based on a modified version of the ASTM E162

1. Selection of reference fireproofing materials

2. Experimental characterization by small scale fire tests

3. Development and validation of numerical model

61

standard test for materials surface flammability evaluation (ASTM, 1994a), which allows reproducing severe fire exposure conditions on small scale. Several commercial inorganic materials are selected and compared in the study. Next, a mono-dimensional model, developed following a simplified thermal nodes approach (Modest, 2003), is presented. The model, validated against the experimental results, allows both to extend the results obtained in the tests to more severe fire exposure conditions and to evaluate a specific Key Performance Indicator (KPI) to support the effective design of passive fire protections. Finally, the contribution of the safety pioneer Menso Molag in the field of hazardous materials transportation safety progress will be analyzed, evidencing the key aspects related to fire protection of road and rail tankers. 2. Materials and methods 2.1. Overview of the methodology The proposed methodology for the assessment of Passive Fire Protection (PFP) materials performances combines the results of experimental characterization with the modeling of the behavior of protected steelworks during the fire exposure. The methodology steps are summarized in Fig. 1. The first step consists in the selection of fireproofing materials to be analyzed (step 1 in Fig. 1) and the focus was on inorganic fireproofing solutions, considering commercially available products designed for insulation of industrial equipment, as well as for fireproofing of road/rail tankers and structural elements. Experimental characterization of materials behavior is performed through a small scale fire test, featuring the experimental set up described in ASTM standard E162 (ASTM, 1994a). This standard provides a laboratory test procedure for measuring and comparing the surface flammability of materials when exposed to a fixed level of radiant heat energy. It is intended for use in measurements of the surface flammability of materials exposed to fire. In this work, the test is modified in order to increase the severity of fire exposure (step 2 in Fig. 1). Panels made of the tested PFP material coupled with a steel board, aimed at reproducing the presence of a protected steel structure, are exposed to a steady radiative source and the heat up of the steel is monitored. The detailed description of experimental procedure is presented in Section 2.2. Due to the heterogeneity of tested products, the information obtained in the standard tests alone cannot provide sufficient elements for the assessment of PFP performances: a uniform screening criterion and a specific performance assessment tool are required. To the purpose, the experimental results are analyzed, integrated and extended through a numerical approach, based on the implementation of a simplified model for reproducing the temperature profiles in the specimen exposed to fire (step 3 of Fig. 1). The input data required for the development of a robust model are essentially the thermal properties of selected materials (namely density, heat capacity, thermal conductivity and emissivity). Details on model characteristics are discussed in Section 2.3.

Table 1 Geometry of the tested samples of fireproofing materials.

4. Case-studies analysis and key performance indicators evaluation Fig. 1. Overview of the methodology for the evaluation of fireproofing materials performance.

Dimensions

Height (mm) Length (mm) Thickness (mm)

Tested material Type 1 (Rock wool)

Type 2 (Fiber mineral wool)

Type 3 (Silica Aerogel)

150 460 20

150 460 12

150 460 6

62

F. Argenti, G. Landucci / Journal of Loss Prevention in the Process Industries 28 (2014) 60e71

The developed model is applied for the analysis of fire exposure conditions and duration relevant for real scale applications, in order to assess and compare the performances of reference materials and to obtain guidelines for the design of fireproofing layer thickness (step 4 in Fig. 1). 2.2. Experimental analysis 2.2.1. Tested materials Three commercial fireproofing and insulating materials having an inorganic formulation were selected as references for the present study:  Coating type 1: Marine Firebatt 100, supplied by Rockwool Italia S.p.A. (http://www.ais-group.com.au/product/uploaded_ files_download/pds_pds_marine_firebatt__33.pdf/29)  Coating type 2: Fibercon Silica Needled Blanket 1200, density 130 supplied by Insulcon BV (http://www.insulcon.com/page/ products/fibre-products/blankets.htm)  Coating type 3: Pyrogel 6650, supplied by Aspen Aerogels (http://www.aerogel.com/products/pdf/Pyrogel_6650_DS.pdf) Coating type 1 consists of a semi-rigid rock wool slab and is typically used in offshore and marine applications. Coating type 2 is made up of silica glass fibers that are mainly composed of silicon dioxide, aluminum dioxide, and some binders as minor components. This type of coating is usually applied as a blanket onto the shell of the protected vessels and then covered with a steel jacket. Specimens used in the current study were sections cut from commercial rolls. Finally, coating type 3 is a flexible blanket made up of light weight silica solids, obtained from a gel formulation where the liquid phase was substituted with a gaseous phase, typically air. The solid phase constitutes only the 3% of total volume and has a low value of thermal conductivity, thus heat conduction into the solid is extremely slow. The remaining 97% of the volume has a nanoporous structure filled with air, aimed at reducing heat transfer effectiveness. The specimens used in experimental runs were made up of a steel board covered by a layer of the fireproofing material to be tested. Table 1 summarizes the dimensions of selected fireproofing specimens. The dimensions are in accordance with the ASTM standard E162, while the thicknesses are the one commercially available for fireproofing purposes. The steel board features the same height and length of the fireproofing layer (see Table 1) and is 5 mm thick. This value is representative of industrial equipment or tankers minimum wall thickness. 2.2.2. Experimental set-up and instrumentation Fig. 2 shows the schematic representation of the experimental set-up, whose overview is reported in Fig. 3a. As mentioned in Section 2.1, a modified version of the ASTM E162 standard (ASTM, 1994a) was used to perform the fire tests in order to obtain more severe fire exposure conditions. The whole set-up is the RP-1A model, supplied, modified and certified by Govmark Inc. The system can be divided into three main sections, as depicted in Fig. 2: (1) the framework, that includes the specimen support frame; (2) the radiating panel and the flame generation system; (3) the data acquisition system. The framework consists of the supporting structure made of stainless steel covered by epoxy varnish. The support frame, made in aluminum (see the sample image in Fig. 3c and the schematization in Fig. 3d) is fixed to the bigger structure and holds material

samples of 150  460 mm, with variable thickness. The blocking system is made with screws able to fix the specimen against the edge of the support frame. The radiating panel (item 10 in Fig. 2) is made of porous refractory material; it is vertically mounted on a stainless steel framework and it features a radiating surface of 300  460 mm. The panel is equipped with a Venturi fan for fuel gas and air mixing, as a premixed flame is required for making the panel rapidly incandescent providing a constant radiative heat flux. The modifications introduced in the standard set up basically concern the relative orientation between the specimen and the radiating panel. In the ASTM test, the specimen has a 30 orientation respect to the radiating source. In this work, the specimen is instead positioned parallel to the radiant panel, hence shortening the separating distance between the specimen and the radiating panel and increasing the severity of the fire exposure (see scheme in Fig. 3b). Besides, the selection of LPG instead of methane as fuel gas also increases the thermal potentiality of the apparatus, which can provide heat radiation in the range 50O100 kW/m2. The flame generation system consists of the fuel gas storage and the supplying system connected to the flame generator. LPG is stored in a pressurized cylinder. On the fuel line, appropriate devices allow pressure and flow regulation. In the tests, the LPG flow is set to 10 L/s, while air flow is set to 50 L/s. The ignition of the mixture fed to the panel is guaranteed by an electrical spark manually generated by the operator, even if a pilot flame, generated through an oxy-acetylene welding tip, is provided (item 9 in Fig. 2). The data acquisition system consists of a set of thermocouples, an optical pyrometer and an infrared camera. An eight thermocouples Chromel/Alumel K type stack is provided in the standard ASTM (1994a) configuration and located inside the hood on top of radiant panel (see Fig. 3a). This is aimed at measuring the temperature of exhaust gases exiting from the radiant panel. The optical pyrometer, provided by Govmark Inc in accordance to ASTM E162, allows monitoring the temperature of the radiant panel in the start-up of the panel (see Section 2.2.3). The pyrometer is positioned in front of the panel at a distance of 1.3 m. An analog acquisition unit (data logger) Agilent 34970A, interfacing the thermocouples and pyrometer with a computer, is provided. The temperature-time profile on the non-exposed side of the specimen (e.g., steel wall) is monitored using an IR camera (Thermovision A40M) from FLIR systems, which is directly connected to the computer with a firewire connection. The IR camera is located orthogonally to the panel at a distance of 1.3 m, e.g., the same pyrometer distance (see Fig. 2, item 11). 2.2.3. Experimental procedure The test begins with the radiating panel start-up. After positioning the pilot flame, the radiant panel is fed with the fuel-air mixture and ignited by the operator. Next, the panel takes about 20 min to heat up to steady state conditions. During this phase, radiant panel surface is monitored with the optical pyrometer, while the hot fumes temperature is measured by the eight thermocouples stack. Based on ASTM (1994a) indications, steady state conditions are reached when for more than 20 min a constant temperature is kept both by the radiant panel and fumes. In this work, a steady state temperature of 685  5  C is set for the radiant panel, while 190  10  C is set for the thermocouple stack. During the heating phase, the support frame holding the specimen is located on the bearing structure but not directly exposed to the heat source (see Fig. 3a). This allows a uniform pre-conditioning of all the specimens.

F. Argenti, G. Landucci / Journal of Loss Prevention in the Process Industries 28 (2014) 60e71

63

Fig. 2. Schematic representation of experimental set-up.

When the radiant panel is steady state conditions, the support frame holding the specimen is displaced and positioned in front of the radiating panel, the IR camera is switched on and fire test starts.

Each test is carried out for a maximum time of 45 min, which allows the safe operation of the radiant panel. Then the specimen is displaced to the initial position (see Fig. 3a) and the cooling phase is observed and registered.

Fig. 3. Details of experimental set-up: position of radiant panel and specimen, a) photograph and b) scheme; specimen within the framework, c) photograph and d) scheme.

64

F. Argenti, G. Landucci / Journal of Loss Prevention in the Process Industries 28 (2014) 60e71

Examples of the results obtained by the described test procedure are discussed in Section 3.1. 2.3. Numerical modeling approach A simplified mono-dimensional physical model was developed for simulation of the fire tests. The model is aimed at estimating an averaged value of the temperature of the non-exposed side of the specimen during the tests (e.g., the steel board) and at reproducing the average temperature profile in time. The base assumption for model development is that in the fire test the heat propagation through the sample exposed to radiant heat takes place mainly in the direction perpendicular to the exposed face. Therefore, the occurring heat exchange phenomena are described as transient mono-dimensional thermal conduction in two adjacent layers made up of two different materials. The governing equations are represented by of a system of two Partial Differential Equations (PDEs) of second order in space and first order in time, one for each layer. Hence, two initial conditions, assuming uniform initial temperature for both layers, together with four boundary conditions, describing heat flux on the exposed side, heat flux on the rear side and temperature continuity and heat flux continuity at the interface, are required for resolution. The transient heat balance is reported in Table 2 with the assumed initial and boundary conditions. In order to consider also the properties of coating material as variable with temperature, the above described system is solved through numerical integration (see Table 3). More specifically, a spatial discretization is adopted following the “thermal nodes modeling” approach (Modest, 2003): the specimen (layer of insulating material coupled with the steel board) is schematized through a one-dimensional grid in the direction of thickness. The grid consists of a finite number of nodes having the same width and height of the sample board; each node is described by a simple set of parameters representing physical quantities (e.g., temperature, thermal conductivity, etc.) averaged over each node.

Fig. 4 shows the mesh adopted for the definition of calculation nodes. As it can be seen, a non-uniform mesh is considered. The total thickness of the coating is subdivided into n thermal nodes having the same thickness ds, plus a single smaller node of thickness ds/2, which is located just before the interface between fireproofing layer and steel board. The parameter n can be increased to guarantee mesh-independent results, while the value of node thickness is automatically calculated, having defined the condition of uniform cells in the coating layer and finer cell next to the interface. The total thickness of the steel board is subdivided into two nodes: the thickness of ds/2 was selected for the first one, thus creating a symmetric node respect to the interface; while the final node covers the whole residual thickness of the steel board. This latter choice is based on the assumption that the heat transfer by conduction within the steel board is far more rapid than the convective heat transport in air. This assumption may be considered quite accurate due to the small thickness of the steel board (5 mm) and the high thermal conductivity of steel. Through the spatial discretization, a system of (n_ins_nodes þ 3) ordinary differential equations in time is obtained and implemented in the model. The considered model equations are reported in Table 3. Nomenclature of equations is reported in Table 4. Each equation represents the thermal balance on the corresponding node: it includes a term representing the variation of the node temperature with time, which is set equal to the difference between incoming heat flux and dispersed or released heat flux. In this way, the thermal interactions between nodes and between a node and the surrounding ambient air are taken into account. According to the presented equations, the first node receives a net radiative heat flux from the radiant panel and transfer heat to the next node by conduction. Within the solid (Eqs. (c), (e), (g), in Table 3), heat is transferred only by conduction. The last node, on the rear side of the sample board, receives conductive heat by the previous node (Eq. (i) in Table 3) and disperses heat by convection in air (Eq. (j) in Table 3).

Table 2 Transient heat balance equations and correspondent nomenclature. ID (a)

Equation

rins cins

vTins vt

  v vTins ¼ k vx ins vx

(b)

vTsteel v2 Tsteel ¼ asteel vt vx2

(c)

Tins(t ¼ 0) ¼ Tair

(d)

Tsteel(t ¼ 0) ¼ Tair vT 4  ε T4 Þ kins ins exposed ¼ sb $ðεrad Trad ins ins vx side

(e)

(f)

(g)

(h)

a b

vTsteel ¼ hair $ðTsteel  Tair Þ vx rear side vT vT ¼ kins ins ksteel steel vx interface vx interface

ksteel

Tsteel(t) ¼ Tins(t) (interface)

Average constant value is considered. Variable value is considered.

Description

Nomenclature

Heat balance equation for the PFP layer

Tins ¼ PFP layer temperature (K) t ¼ time (s) x ¼ spatial mono-dimensional coordinate rins ¼ average density of the PFP layera (kg/m3) cins ¼ average heat capacity of the PFP layera (J/(kgK)) kins ¼ thermal conductivity of the PFP layerb (W/(mK)) Tsteel ¼ steel layer temperature (K) asteel ¼ effective thermal diffusivity of the steel layera (m2/s) which takes into account also the heat dispersion from the side of the steel panel Tair ¼ initial temperature (environmental temperature)

Heat balance equation for the steel layer

Initial condition 1 (for the PFP layer) Initial condition 2 (for the steel layer) Boundary condition 1 (for the PFP layer): heat flux received by the fire exposed side Boundary condition 2 (for the steel layer): heat flux dispersed from the rear side Boundary condition 3 (for both layers): heat flux at the interface between the layers Boundary condition 4 (for both layers): temperature continuity between the two layers

(see above) Trad ¼ temperature of the radiating source kins ¼ PFP layer thermal conductivityb (W/(mK)) sb ¼ 5.670373  108 W/(m2K4) Stefan Boltzmann’s constant εrad ¼ emissivity of the radiating sourcea εins ¼ emissivity of the PFP layera ksteel ¼ steel layer thermal conductivitya (W/(mK)) hair ¼ average heat transfer coefficientb (W/(m2K)) (see above)

(see above)

F. Argenti, G. Landucci / Journal of Loss Prevention in the Process Industries 28 (2014) 60e71

65

Table 3 Equations implemented in the numerical “thermal nodes” model. Energy balance equation

Node

External surface of coating, directly exposed to radiant panel :   vTj kins ðTj Þ 4 $ðTj  Tjþ1 Þ $Ax  Qdisp;j ¼ s$ðεrad Trad  εins Tj4 Þ  (a) rins $cpins $ds$Ax $ ds vt

j¼1

(b) Qdisp;j ¼ ðTj  Tair Þ$½hair

vert ðTj ; Tair Þ$2$Ay ds

þ ðhair

hor down ðTj ; Tair Þ

þ hair

hor up ðTj ; Tair ÞÞ$Az ds 

j ¼ 2 to nins_nodes

Internal nodes within the coating layer:   vTj kins ðTj1 Þ kins ðTj Þ $ðTj1  Ti Þ  $ðTj  Tjþ1 Þ $Ax  Qdisp;j ¼ (c) rins $cpins $ds$Ax $ ds ds vt (d) Qdisp;j ¼ ðTj  Tair Þ$½hair

vert ðTj ; Tair Þ$2$Ay ds

þ ðhair

hor down ðTj ; Tair Þ

þ hair

hor up ðTj ; Tair ÞÞ$Az ds 

j ¼ nins_nodes þ 1

Coating node before solidesolid interface:   vTj kins ðTj Þ k $ðTj1  Tj Þ  2$ ins $ðTj  Tjþ1 Þ $Ax  Qdisp;j ¼ (e) rins $cpins $ds=2$Ax $ ds vt ds (f) Qdisp;j ¼ ðTj  Tair Þ$½hair

vert ðTj ; Tair Þ$Ay ds

þ ðhair

hor down ðTj ; Tair Þ

þ hair

hor up ðTj ; Tair ÞÞ$Az ds =2

  First steel node: vTj kins ðTj1 Þ ks $ðTj1  Tj Þ  2$ $ðTj  Tjþ1 Þ $Ax  Qdisp;j ¼ 2$ (g) rs $cps $ds=2$Ax $ ds vt ds (h) Qdisp;j ¼ ðTj  Tair Þ$½hair

vert ðTj ; Tair Þ$Ay ds

þ ðhair

hor down ðTj ; Tair Þ

þ hair

j ¼ nins_nodes þ 2

hor up ðTj ; Tair ÞÞ$Az ds =2

j ¼ nins_nodes þ 3

Second steel node, rear of the specimen: vTj ks ¼ 2$ $ðTj1  Tj Þ$Ax  Qdisp;j (i) rs $cps $dss $Ax $ vt ds  h ðT ; T Þ$ðAy steel þ 4$Ay fr Þ (j) Qdisp;j ¼ ðTj  Tair Þ$ air vert j air þðhair hor down ðTj ; Tair Þ þ hair hor up ðTjþ ; Tair ÞÞ$ðAz

Given the specimen dimensions, an infinite plate model was not suitable for describing the heat exchange phenomena intervening during the experiments. More specifically, boundary effects were considered: particularly, convective heat dispersion from the side (not normal to the thickness direction) surfaces was modeled for all nodes and the presence of a steel made support for the specimen was taken into account by appropriately increasing the surface area available for heat dispersion in the case of the steel node. The dispersion equations are reported in Table 3 for each group of nodes (see Eqs. (b), (d), (f), (h)). The definition of each dispersion term and correspondent heat exchange surface is reported in Table 3 together with the nomenclature for Table 4. A pure radiative model is proposed as the heat flux received by the surface of the fireproofing layer that is directly exposed to the radiating panel is considered constant and calculated by applying Stefan Boltzmann’s law. More specifically, the radiating panel is considered as a black body (emissivity equal to 1) and a constant value of “effective radiant temperature” is assumed to account for the convective heat transfer contribution of hot gases flowing over the exposed surface of the specimen (Eq. (a) in Table 3). Effective radiant temperature of the black body is selected as the arithmetic mean of the steady state temperature of the panel and the peak temperature registered by the pyrometer when the specimen is removed from in front of the panel. Considering that the three experimental trials were conducted at a steady state panel temperature of 685  5  C, which corresponded to a peak temperature higher than 790  C at the end of the test, the adopted value of effective radiant temperature was 740  C. The geometry of the sample board and the physical properties relevant for predicting thermal behavior of steel and coating material are required inputs for the model. For each insulating material constant values of density, heat capacity and emissivity are adopted. As far as thermal conductivity is concerned, a more accurate description based on temperature dependent correlations is

 steel

þ 2$Az

fr Þ

proposed and derived from manufacturer’s material data sheets (see Table 5). Only in the case of rock wool (coating type 1) an average value is selected due to missing information. In the case of silica blanket (coating type 2), an average value between the solid and air thermal conductivity (respectively ksolid and kair), weighted on the pore fraction (εp) is considered (see Table 5), as suggested in previous studies (Gomez-Mares, Tugnoli, Landucci, Barontini, & Cozzani, 2012; Gomez-Mares, Tugnoli, Landucci, & Cozzani, 2012). Section 3.2 reports the model validation, while the model application to specific case studies is discussed in Section 3.3 in order to test the potentiality of the method. 3. Results and discussion 3.1. Experimental analysis Fig. 5, 6 and 7 report the thermal images of the back side of the specimen registered by the IR camera after 15, 30 and 45 min for material type 1, material type 2 and material type 3 respectively. As it can be seen from the thermal images, temperature shows an increasing trend moving from the bottom to the top of the specimen surface, i.e. higher temperature values were observed in the upper region, due to the convective contribution to heat transfer by buoyant hot gases. Fig. 8 shows averaged temperature profiles representing the thermal conditions existing in the upper, central and lower region of the steel board. The reported profiles were derived from an analysis of the data registered by the IR camera during the whole duration of the test. Particularly, three couples of “reference spot areas” were identified on the non-exposed surface of the specimen and used as reference points for the evaluation of averaged temperature profiles. Fig. 8 highlights that the temperature profiles are quite similar for the three materials. This is mainly due to the thermal behavior of the protected steel board, which features a

66

F. Argenti, G. Landucci / Journal of Loss Prevention in the Process Industries 28 (2014) 60e71

Az_up_ins

silica blanket are respectively available in tripled and doubled thickness respect to aerogel (see Table 1). Therefore, effective quantitative considerations and PFP performance ranking is not effective on the sole basis of experimental trials: in the second phase of this study, data on the thermal behavior of the three reference materials, for fixed values of thickness, are obtained through the simulation of fire test conditions with the simplified model, and then, compared to assess the fireproofing performances.

Az_up_s Az_up_fr

Ay_fr Ax Ay_ins

Ay_s

Az_down_fr

z

y x x (thickness direction)

3.2. Model validation Model validation is carried out on the basis of experimental data obtained with the small scale fire tests and discussed in Section 3.1. In particular, the temperature profile predicted by the model in the thermal node corresponding to the external side of the steel board is compared with the measured temperature profiles that were registered in the spot areas numbered as 03 and 04, located in the central region of the specimen (see Fig. 8). The results of the validation procedure are reported in a graphical form in Fig. 9. It can be seen that a good agreement between experimental data and model simulation is reached, with a maximum relative error lower than 10%. It is worth mentioning that the model, due to the strong simplifications, allows the determination of temperature values at different time steps and different positions only among the specimen thickness. In other words, the model has the main limitation of not quantifying temperature variation along the sample height, which requires a deeper characterization of convective contribution to heat transfer by the buoyant hot gases. Nevertheless, due to easiness of use and low computational costs, the model can be used to generate an extended data set on the coating response to fire, thus supporting the selection and design of passive fire protection systems, as discussed in Section 3.3. 3.3. Application to case-studies

ds/2 ds/2

ds

sins

ss

Fig. 4. Geometry features of the specimen and mesh implemented in the numerical model.

high thermal inertia showing a relatively “slow” thermal response to the incoming heat fluxes: a slow temperature increase is registered in the transient phase until an almost constant value of the specimen temperature is reached. The maximum value of temperature reached by the steel is mainly influenced by the thermal conductivity of the fireproofing material (Cotgreave, 1992; Droste & Schoen 1988; Landucci, Molag, & Cozzani, 2009; Malloy, 1969; Roberts et al., 2010; Townsend et al. 1974) and represents a significant parameter for the effectiveness assessment of fireproofing. It can be observed that the maximum temperature measured on the steel board is lower than 250  C in all cases, which indicates that thermal performances can be considered satisfactory for all specimens. As a matter of fact, no critical weakness caused by the heat up is expected for the steel at this temperature (SCI, 1992). However, it should be remarked that the thickness of the fireproofing material layer is different in the three tests; rock wool and

As previously outlined in Section 2.1, a critical issue in the design of fireproofing materials is providing suitable tools for assessing their effective performances while subject to fire conditions that have specific relevance for the final application. In the present work, the experimental characterization through small scale fire tests presented in Section 3.1 is a significant and cost effective benchmark, but the severity of the fire exposure is limited, especially if compared with the severe scenarios that can take place during road and rail accidents involving hazardous materials tankers. Therefore, the model developed is applied for extrapolating the behavior of the protected steelwork in presence of large scale fires. In particular, the engulfment in a hydrocarbon pool fire is selected as reference heat exposure scenario and implemented for the analysis of specific case studies. Pool fire exposure is simulated considering a radiation source at 1098  C. This value is the steady radiation temperature of a pool fire according to the standard hydrocarbon fire curve in UL 1709 Standard (UL, 1994). The model allows reproducing the thermal response of specimen protected by layers of different thickness of the three reference materials. Three sets of case studies are considered, one for each inorganic coating (type 1, type 2 and type 3) in order to analyze the effect of increasing coating layer thickness (3, 6, 12, 20 and 40 mm). For all case studies, a uniform temperature condition (25  C) for both specimen and surrounding environment is considered. Simulation run time is extended to 100 min, allowing a higher fire exposure time with respect to the test (which maximum

F. Argenti, G. Landucci / Journal of Loss Prevention in the Process Industries 28 (2014) 60e71

67

Table 4 Summary of input parameters implemented in the model; ED: parameter value is fixed on the basis of experimental measures before test starts; PD: parameter value is fixed on the basis of manufacturer data; CALC: parameter calculated as a function of node temperature on the basis of input data. Operation

ID

Item

Description/Definition

Value

Units

1) Selection of input parameters: characterization of test conditions

1.1 1.2 1.3

Trad Tair Tin

1013 ED ED

K K K

1.4 1.5 1.6

sB

CALC 5.670373  108 CALC

K W/(m2K4) kW/m2K

(Kern, 1950)

1.7

hair_hor_up

CALC

kW/m2K

(Kern, 1950)

1.8

hair_hor_down

CALC

kW/m2K

(Kern, 1950)

cpins kins εins εrad W H sins

Effective radiative temperature Surrounding air temperature Temperature of the specimen at the beginning of test Temperature of the j-th node Stefan Boltzmann’s constant Heat transfer coefficient for convective dispersion toward air for vertical walls hair vert ¼ 0:0017034$ð1:8$jTJ  Tair jÞ0:25 Heat transfer coefficient for convective dispersion toward air for horizontal walls facing upwards hair hor up ¼ 0:00215764$ð1:8$jTJ  Tair jÞ0:25 Heat transfer coefficient for convective dispersion toward air for horizontal walls facing downwards hair hor down ¼ 0:0011356$ð1:8$jTJ  Tair jÞ0:25 Steel density Steel heat capacity Steel thermal conductivity Fireproofing material density Fireproofing material heat capacity Fireproofing material thermal conductivity Fireproofing material emissivity Radiant panel emissivity Specimen total length Specimen total height Fireproofing layer thickness

PD PD PD PD PD CALC 0.85e0.95 1 0.15 0.46 PD

kg/m3 kJ/kg K kW/m K kg/m3 kJ/kg K kW/m K

3.4

ss

Steel plate thickness

0.005

m

3.5 3.6 3.7

stot nnodes_ins nnodes

CALC User defineda CALC

m e e

3.8

ds

CALC

e

3.9

dss

CALC

m

3.10

Ax

CALC

m2

Fig. 4

3.11

Ay_ds

CALC

m2

Fig. 4

3.12

Az_ds

CALC

m2

Fig. 4

3.13

Ay_steel

CALC

m2

Fig. 4

3.14

Az_steel

CALC

m2

Fig. 4

3.15 3.16

sfr Ay_fr

0.03175 CALC

m m2

Fig. 3 Fig. 4

3.17

Az_fr

Total specimen thickness stot ¼ sins þ ss Number of nodes within fireproofing layer-1 Total number of nodes of the grid nnodes ¼ nnodes ins þ 3 Spatial discretization step within fireproofing layer ds ¼ 2$sins =ð2$nnodesins þ 1Þ Thickness of external steel node dss ¼ ss  0:5$ds Area of the external surface normal to x axis for every node Ax ¼ H$W Area of the external surface normal to y axis, for nodes having thickness ds Ay ds ¼ H$ds Area of the external surface normal to z axis, for nodes having thickness ds Az ds ¼ W$ds Area of the external surface normal to y axis, for nodes having thickness ds Ay steel ¼ H$ds Area of the external surface normal to z axis, for nodes having thickness ds Az steel ¼ W$ds Thickness of steel supporting framework Area of the external surface normal to y axis of the framework Ay fr ¼ H$sfr Area of the external surface normal to z axis of the framework Az fr ¼ W$sfr

CALC

m2

Fig. 4

2) Selection of input parameters: materials thermal properties

3) Selection of input parameters: characterization of the specimen geometry

a

2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 3.1 3.2 3.3

Tj

hair_vert

rs cps ks

rins

m m m

Reference

Fig. 3 Fig. 3 Figs. 3 and 4 Figs. 3 and 4

Fig. 4

The parameter is increased up to mesh-independent predictions.

duration is 45 min). This time lapse represents an estimate of the maximum time required for emergency teams intervention in the case of fire scenarios involving road or rail accidents as discussed in a previous study (Hobert & Molag, 2006; Landucci, Molag, & Cozzani, 2009; Molag & Kruithof, 2005). Following the approach suggested by Landucci, Molag, and Cozzani (2009), a key performance indicator for fireproofing

materials (namely temperature index TI) is calculated for each given thickness of the coating and its behavior during fire exposure is monitored. In this work, TI is defined as the ratio between the maximum temperature reached by the non-exposed surface of the steel board and the temperature of the external coating surface, directly facing the radiating source. More specifically, the index is calculated by using the above-defined values of temperature as

68

F. Argenti, G. Landucci / Journal of Loss Prevention in the Process Industries 28 (2014) 60e71

Table 5 Thermal properties of fireproofing materials implemented in the model and derived from manufacturer data. Material type 1a Density (kg/m3) Heat capacity (kJ/kgK) Thermal conductivity (kW/mK) Emissivity Material type 2b Density (kg/m3) Heat capacity (kJ/kgK) Thermal conductivity (kW/mK)

Emissivity Material type 3c Density (kg/m3) Heat capacity (kJ/kgK) Thermal conductivity (kW/mK)

Emissivity

Material type 2

100 0.84 1.12$104 0.95 130 1.2 k ¼ (1ε)  ksolid þ ε  kair where ε ¼ 0.95 is the pore fraction and ksolid ¼ 3$1010T21.2578$107$T þ 5.37288$105; kair ¼ 1$103$(3.1417$104$T0.7786)/ (17.116$101/T2121.7/T2); 0.95 110 1.046 if T < 436.1 K k ¼ 1.2$1010$T2  3.668$108$T þ 1.304475$105; if T  436.1 K k ¼ 4.719$108$T  7.3536$107; 0.95

a Data were derived from http://www.ais-group.com.au/product/uploaded_files_ download/pds_pds_marine_firebatt__33.pdf/29. b Data were derived from http://www.insulcon.com/page/products/fibreproducts/blankets.htm. c Data were derived from http://www.aerogel.com/products/pdf/Pyrogel_6650_ DS.pdf.

derived from model simulations. Therefore, average values of temperature are used instead of considering the maximum reached temperatures, which can be registered at the top of the specimen during tests but not predicted through the sole use of the model. Even if this choice leads to underestimation of the index and nonconservative representation of the heat up of the exposed board,

(a)

(b)

(c)

Temperature (°C) 20

250

Fig. 6. Thermal images of the back side of the specimen registered by the IR-camera after 15 min of exposure (a), after 30 min of exposure (b) and at the end of tests (45 min) (c) for material type 2.

the application of a normalized index to average temperature values can as well provide an effective and rapid estimation of the temperature gradient within the specimen thickness and, hence, of the thermal insulation capability of the analyzed coating material.

Material type 3 Material type 1

(a) (a)

(b)

(c)

Temperature (°C) 20

(b) Temperature (°C)

20 250

Fig. 5. Thermal images of the back side of the specimen registered by the IR-camera after 15 min of exposure (a), after 30 min of exposure (b) and at the end of tests (45 min) (c) for material type 1.

(c)

250 > 250°C

Fig. 7. Thermal images of the back side of the specimen registered by the IR-camera after 15 min of exposure (a), after 30 min of exposure (b) and at the end of tests (45 min) (c) for material type 3.

F. Argenti, G. Landucci / Journal of Loss Prevention in the Process Industries 28 (2014) 60e71

a)

01

69

250

Tmax Temperature (°C)

02 03

T

04

av

05

T

06

200 150 100

T mod

T exp 03 50

T exp 04

min

0 0

5

10

250 200

b) 250

150

200

Temperature (°C)

Temperature (°C)

a)

100 Tmax Tav Tmin

50

0 0

5

10

15

20

25

30

35

40

45

15

20

25

30

35

40

45

Time (minutes)

150

100

T mod T exp 03 T exp 04

50

Time (minutes) 0 0

200

10

15

20

25

30

35

40

45

Time (minutes)

150 100

Tmax Tav Tmin

50 0 0

5

10

15

20

25

30

35

40

45

Time (minutes)

c)

5

c) 250 Temperature (°C)

Temperature (°C)

b) 250

200 150 100

T mod Texp 03

50

Texp 04

250

Temperature (°C)

0 0

200

5

10

15

20

25

30

35

40

45

Time (minutes) 150

Fig. 9. Graphical results of model validation for material type 1 (a), type 2 (b) and type 3 (c). Tmod: temperature predicted by the model; Texp03: average temperature in the spot 03 (see Fig. 8); Texp04: average temperature in the spot 04 (see Fig. 8).

100 Tmax Tav Tmin

50 0 0

5

10

15

20

25

30

35

40

45

Time (minutes) Fig. 8. Averaged temperature profiles representing the thermal conditions existing in the upper, central and lower region of the steel plate for material type 1 (a), type 2 (b) and type 3 (c).

Therefore, this approach is suitable considering the final aim of the study, which is to show the potential applications of the simplified model as a tool for the preliminary screening and assessment of PFP performances, without the need of carrying out experimental trials. It is worth to remark that pilot or large scale bonfire tests on process equipment, on one side, and more detailed three-dimensional models, on the other, are needed to detail and finalize the assessment of PFP performance and complicating factors, such as inner fluid temperature behavior, pressure build up, thermal stress

increase on the equipment, possible damages to the PFP layer, etc. (Droste & Schoen, 1988; Landucci, Molag, Reinders, & Cozzani, 2009; Landucci, Molag, & Cozzani, 2009; Molag & Kruithof, 2005; Salzano et al., 2003; SCI, 1992; Townsend et al., 1974). Fig. 10 reports the values of the performance index TI for 100 min simulation time of all defined case studies. A threshold value of TI (namely TI* ¼ 0.36) is also reported in Fig. 10 as a horizontal bold line. According to Landucci, Molag, and Cozzani (2009), TI* is calculated as the ratio between the reference temperature value for the thermal weakening of the steel, fixed at 400  C according to references in the literature (Cotgreave, 1992; Lees, 1996), and the standard hydrocarbon fire temperature (e.g., 1098  C) (UL, 1994). The region that lies over the horizontal line identifies potentially critical conditions for the structural integrity of the protected steelwork, caused by insufficient insulation performances of the applied fireproofing layer.

70

F. Argenti, G. Landucci / Journal of Loss Prevention in the Process Industries 28 (2014) 60e71

a)

0.7 0.6

3mm 6mm 12mm 20mm 40mm TI*

TI

0.5

0.4 0.3 0.2 0.1

Table 6 Time (in minutes) at which the temperature of the steel plate reaches the critical value of 400  C according to model simulations; NR: the critical temperature of 400  C is not reached within 100 min. Fireproofing material thickness (mm)

Time to reach 400  C (minutes) Coating type 1

Coating type 2

Coating type 3

3 6 12

5 11 NR

7 21 NR

14 NR NR

0

0

20

40

60

80

100

Time (minutes)

b)

0.7 0.6

3mm 6mm 12mm 20mm 40mm TI*

TI

0.5 0.4 0.3 0.2

0.1 0 0

20

40

60

80

100

Time (minutes)

c)

0.7 0.6

3mm 6mm 12mm 20mm 40mm TI*

TI

0.5

0.4 0.3 0.2 0.1 0

0

20

40

60

80

100

Time (minutes) Fig. 10. Dynamic behavior of TI index during model simulations for material type 1 (a), type 2 (b) and type 3 (c) as a function of the increasing fireproofing thickness. TI*: threshold value for TI index (Landucci et al., 2009d).

behavior, increasing dilatation stress or stress intensification due to the support framework, etc.), the simulation results, through the use of Fig. 10, allow identifying the safety margins available for thermal protection design. Next, on the basis of simulations results, the chart in Fig. 11 can be derived in order to represent the maximum temperature reached in the steel board after 100 min of exposure, as a function of fireproofing materials and layer thickness. In Table 6, the time at which the temperature of the steel board reaches the critical value of 400  C is reported as a function of fireproofing materials and layer thickness. The graph presented in Fig. 11 may be used as a simplified screening tool to establish whether the applied thickness of PFP is sufficient to prevent critical conditions. Furthermore, these results can be adopted as reference for a preliminary design of PFP thickness based on simplified modeling. Comparing the results obtained for the three reference materials for a fixed value of thickness, it can be seen that rock wool (type 1) has the worst performances, while aerogel (type 3) can effectively reduce the temperature increase in the protected structure. These findings are in accordance with preliminary considerations that can be made on the basis of experimental results. Fig. 11 shows that every analyzed fireproofing material can limit the heat up of the specimen well below 400  C if the applied thickness is higher than 20 mm. Nevertheless, the feasibility of such a solution has to be examined also considering the actual costs for the application of a very thick layer of coating, and the resulting weight of the protected equipment, especially in the case of transport unit due to payload limitations. 4. The pioneer Menso Molag

Even if the model does not take into account some aspects which affect vessels resistance to fires, which would require advanced modeling (Birk, 1999, 2005; Moodie, 1988; Venart, 1986, 1999) (e.g., increasing vessel pressure due to heat load, inner fluid

Maximum temperature (°C)

800 Type 1

700

600

Type 2

500

Type 3

400 300

200 100

Dr. Menso Molag (1950) is a professor in risk management at Saxion University for Applied Science. From 1976 to 1986 he worked at University of Groningen, staff of the Chemistry Department, scientist for the occupational and environmental safety unit. Since 1987 he has worked in the department of industrial safety of the TNO (Netherlands Organisation for Applied Scientific Research). Dr. Molag led numerous project teams on the development and application of probabilistic and deterministic risk assessment models for the transport of persons and dangerous goods, and has been the project manager of TNO’s quantitative risk assessment of all activities with hazardous materials in the Haifa Bay area. He also coordinated the EC thematic network on tunnel safety, SafeT. In the last years he conducted large scale experimental studies aimed at determining the effectiveness of fireproofing materials in the protection of fired equipment, in order to reach a pioneering safety enhancement of tankers for road and rail shipment of hazardous materials in Europe.

0 0

5

10

15 20 25 30 Thickness (mm)

35

40

45

Fig. 11. Maximum temperature reached in the steel plate after 100 min of exposure, as a function of fireproofing material type and thickness.

5. Conclusions In the present work, a methodology for the assessment of passive fire protection effectiveness was presented. The methodology

F. Argenti, G. Landucci / Journal of Loss Prevention in the Process Industries 28 (2014) 60e71

is based on experimental and numerical modeling aimed at determining the behavior of coated steel components (e.g., vessels, pipelines or structures) exposed to the fire heat radiation. The experimental set up, derived from a modified ASTM standard, allowed comparing the protective performance of several coating materials by measuring the dynamic temperature evolution on a steel board covered by the tested PFP layer. The numerical model, validated against the obtained experimental results, allowed to simulate several case studies and to determine the optimum thickness of fireproofing. Although several simplifications were introduced in the model, the present methodology may be used as a preliminary design tool, to be further integrated with real scale experiments and models. Finally, due to the contribution in the risk studies and safety enhancement in the framework of hazardous materials, Menso Molag was recognized as a safety pioneer in Europe. Acknowledgment Authors gratefully acknowledge the “Consorzio Polo Tecnologico Magona e CPTM” (Cecina, Italy) for the assistance in the experimental activity. References Abbasi, T., & Abbasi, S. A. (2007). The boiling liquid expanding vapour explosion (BLEVE): mechanism, consequence assessment, management. Journal of Hazardous Materials, 141(3), 489e519. Abbasi, T., & Abbasi, S. A. (2008). The boiling liquid expanding vapour explosion (BLEVE) is fifty and lives on! Journal of Loss Prevention in the Process Industries, 21(4), 485e487. American Petroleum Institute. (2010). API recommended practice 2218-fireproofing practices in petroleum and petrochemical processing plants (3rd ed.).. American Society for Testing Materials. (1994a). ASTM E162-94-standard test method for surface flammability of materials using a radiant heat energy source. West Conshohocken: ASTM International. American Society for Testing Materials. (1994b). ASTM E84-94-standard test method for surface burning characteristics of building materials. West Conshohocken: ASTM International. Birk, A. M. (1999). Tank-car insulation defect assessment criteria: Thermal analysis of Defects (TP 13518E report). Transport Canada. Birk, A. M. (2005). Thermal model upgrade for the analysis of defective thermal protection systems. Transportation development centre, Transport Canada. Birk, A. M., Poirier, D., & Davison, C. (2006). On the response of 500 gal propane tanks to a 25% engulfing fire. Journal of Loss Prevention in the Process Industries, 19, 527e541. Center for Chemical Process Safety. (1996). Guidelines for evaluating the characteristics of vapour cloud explosions, flash fires and BLEVEs. Centre for Chemical Process Safety. New York (USA): AIChE. Center for Chemical Process Safety. (2000). Guidelines for chemical process quantitative risk analysis (II ed.). New York: American Institute of Chemical Engineers. Cotgreave, T. (1992). Passive fire protection: Performance requirements and test methods. OTI Report 92606. London: SCI HSE. Cowley, L. T., & Johnson, A. D. (1992). Oil and gas fires e characteristics and impact. OTI 92596N. UK: HSE. Di Padova, A., Tugnoli, A., Cozzani, V., Barbaresi, T., & Tallone, F. (2011). Identification of fireproofing zones in oil & gas facilities by a risk-based procedure. Journal of Hazardous Materials, 191(1e3), 83e93. Droste, B., Probst, U., & Heller, W. (1999). Impact of an Exploding LPG rail tank car onto a castor spent fuel cask. RAMTRANS Nuclear Technology Publishing, 10(4), 231e240. Droste, B., & Schoen, W. (1988). Full scale fire tests with unprotected and thermal insulated LPG storage tanks. Journal of Hazardous Materials, 20, 41e53. European Commission. (2006a). Directive 2006/89/EC of 3 November 2006, adapting for the sixth time to technical progress council directive 94/55/EC on the approximation of the laws of the member states with regard to the transport of dangerous goods by road. European Commission. (2006b). Directive 2006/90/EC of 3 November 2006 adapting for the seventh time to technical progress council directive 96/49/EC on the approximation of the laws of the member states with regard to the transport of dangerous goods by rail.

71

Gomez-Mares, M., Tugnoli, A., Landucci, G., Barontini, F., & Cozzani, V. (2012). Behavior of intumescent epoxy resins in fireproofing applications. Journal of Analytical and Applied Pyrolysis, 97, 99e108. Gomez-Mares, M., Tugnoli, A., Landucci, G., & Cozzani, V. (2012). Performance assessment of passive fire protection materials. Industrial and Engineering Chemical Research, 51, 7679e7689. dx.doi.org/10.1021/ie201867v. Hobert, J. F. A., & Molag, M. (2006). Fire brigade response and deployment scenarios to avoid a hot BLEVE of a LPG tank vehicle or LPG tank wagon. Report 2006-AR0069/B. Apeldoorn: Netherlands Organisation for Applied Scientific Research TNO. International Organization for Standardization [ISO]. (2007). International standard ISO 22899-1, determination of the resistance to jet fires of passive fire protection materialsdpart 1: general requirements (1st ed.).. Kern, D. Q. (1950). Process heat transfer. New York, CA: McGraw Hill. Khan, F. I., & Abbasi, S. A. (1999). The world’s worst industrial accident of the 1990s e what happened and what might have been: a quantitative study. Process Safety Progress, 18(3), 135e145. Kielec, D., & Birk, A. M. (1997). Analysis of tank deformation from fire induced ruptures and BLEVEs of 400 L propane tanks. ASME Journal of Pressure Technology, 119(3), 365e373. Landucci, G., Gubinelli, G., Antonioni, G., & Cozzani, V. (2009). The assessment of the damage probability of storage tanks in domino events triggered by fire. Accident Analysis Prevention, 41(6), 1206e1215. Landucci, G., Molag, M., Reinders, J., & Cozzani, V. (2009). Experimental and analytical investigation of thermal coating effectiveness for 3 m3 LPG tanks engulfed by fire. Journal of Hazardous Materials, 161(2e3), 1182e1192. Landucci, G., Rossi, F., Nicolella, C., & Zanelli, S. (2009). Design and testing of innovative materials for passive fire protection. Fire Safety Journal, 44, 1103e 1109. Landucci, G., Molag, M., & Cozzani, V. (2009). Modeling the performance of coated LPG tanks engulfed in fires. Journal of Hazardous Materials, 172(2009), 447e456. Lees, F. P. (1996). Loss prevention in the process industries (2nd ed.). Oxford: Butterworth- Heinemann. Malloy, F. (1969). Thermal insulation. New York, NY: Van Nostrand Reinhold Company. Manas, J. L. (1984). BLEVEs: their nature and prevention. Fire International Journal, 87(8), 27e39. Modest, M. F. (2003). Radiative heat transfer (2nd ed.). New York: Academic Press. Molag, M., & Kruithof, A. (2005). BLEVE prevention of a LPG tank vehicle or a LPG tank wagon. Report R2005/364. Apeldoorn (NL): Netherlands Organisation for Applied Scientific Research e TNO. Moodie, K. (1988). Experiments and modelling: an overview with particular reference to fire engulfment. Journal of Hazardous Materials, 20, 149e175. National Fire Protection Agency. (1991). Fire protection handbook (17th ed.).. Quincy. Paltrinieri, N., Landucci, G., Molag, M., Bonvicini, S., Spadoni, G., & Cozzani, V. (2009). Risk reduction in road and rail LPG transportation by passive fire protection. Journal of Hazardous Materials, 167, 332e344. Reid, R. C. (1979). Possible mechanism for pressurized-liquid tank explosions or BLEVE’s. Science, 203, 1263e1265. Roberts, A. F. (1981). Thermal radiation hazards from releases of LPG from pressurized storage. Fire Safety Journal, 4(3), 197e212. Roberts, A. F. (1982). The effect of conditions prior to loss of containment on fireball behaviour. The assessment of major hazards. Rugby (UK): Institute of Chemical Engineers. Roberts, A., Medonos, S., & Shirvill, L. C. (2000). Review of the response of pressurised process vessels and equipment to fire attack (OFFSHORE TECHNOLOGY REPORT e OTO 2000 051. HSL 2000. Manchester, UK: Health and Safety Laboratory. Roberts, T. A., Shirvill, L. C., Waterton, K., & Buckland, I. (2010). Fire resistance of passive fire protection coatings after long-term weathering. Process Safety and Environmental Protection, 88, 1e19. Salzano, E., Picozzi, B., Vaccaro, S., & Ciambelli, P. (2003). Hazard of pressurised tanks involved in fires. Industrial Engineering and Chemical Research, 42, 1804e 1812. Steel Construction Institute. (1992). Availability and properties of passive and active protection systems. OTI 92 607. UK: HSE. Townsend, W., Anderson, C. E., Zook, J., & Cowgill, G. (1974). Comparison of thermally coated and uninsulated rail tank-cars filled with LPG subjected to a fire environment. Report FRA-OR&D 75-32. US Department of Transportation e DOT. Tugnoli, A., Cozzani, V., Di Padova, A., Barbaresi, T., & Tallone, F. (2012). Mitigation of fire damage and escalation by fireproofing: a risk-based strategy. Reliability Engineering & System Safety, 105, 25e35. Underwriters Laboratories Inc. (1994). UL Standard for safety for rapid rise fire tests of protection materials for structural steel UL 1709 (2nd ed.).. VanderSteen, J. D. J., & Birk, A. M. (2003). Fire tests on defective tank-car thermal protection system. Journal of Loss Prevention in the Process Industries, 16, 417e425. Venart, J. E. S. (1986). Tank car thermal response analysis e phase II. Tank car safety research and test project. Chicago. Venart, J. E. S. (1999). Boiling liquid expanding vapor explosions (BLEVE): possible failure mechanisms. ASTM Special Technical Publication, 1336, 112e134.