Effects of Radiant Heat Flux on Clean Agent Performance for Class-C Fires

By Romil Patel & James A. Milke, P.E., Ph.D., FSFPE

Data server room and telecommunication facility fires can have significant financial impact.1 Interruption of service can erase permanent and temporary memory due to the physical loss of memory storage devices, costing companies millions. Average downtime costs for computing infrastructures are estimated at $42,000 per hour.2

Current fire protection standards provide guidance for protection hazards based on class, e.g., Class A, B, C, D and K. However, for some classes of hazards, the guidance is not supported by performance data. The lack of performance data may lead to over or under-design of fire extinguishing systems.

Class C Fires

In standards such as NFPA 20013 and ISO 14520,4 concentrations of extinguishing agents for Class C fires are based on concentrations for other classes of fires. The use of the Class A or B concentrations for the protection of Class C fuels is done because of the lack of test data to provide the technical basis for minimum extinguishing concentrations to suppress energy-augmented electrical fires. However, applying a multiple of Class A or B extinguishing concentrations for Class C fires is questionable as there is no data to support the relationship between either of these two classes of fire with Class C fires.

Energy Augmentation

Energy augmentation refers to the effects of external heat flux on the behavior of a fire. Blade server power densities has been a concern in the fire protection industry because radiant heating can enhance energized equipment fires, as demonstrated from studies conducted by Linteris 5 and Steckler et al.6

The total heat dissipated by energized electrical equipment is related to the power consumption of the equipment.  Hence, an increase in power consumption, such as that associated with the trend of using higher performing, smaller blade servers to efficiently maximize today’s technology needs, will cause an increase in heat dissipation to nearby components. 

Blade server equipment found in data processing facilities are made of units (CPUs), chipsets, storage devices, memory, voltage regulators and power suppliers. Since CPUs  consume the majority of all power supplied, ASHRAE sought out to determine the trends in power consumption by CPUs in blade servers. Figure 1 shows the trends in CPU power consumption by year 2020 for four processor types; high power and low power high-performance computing (HPC) and two-socket and four-socket servers. 7
Figure 1: CPU Power Consumption Trends7

Similarly, in 2010, Ponemon Institute conducted a survey of 453 individuals who have responsibility for data center operations in the U.S. 8  One of the questions asked for predictions of power density (in kW) per rack in data centers in two years. Poll results from the respondents showed an average value of 11.4 kW per rack power density.  

Several assumptions were made to determine the radiant heat flux levels emitted from the blade servers. First, it is assumed all racks hold 42, 1U blade servers. The "U” term refers to 1.75 in (44.5 mm) vertical height within a rack between blades. Since most blade server fires originate in a single component, the radiant heat flux augmenting that fire is from the surrounding two blade servers. Two horizontally-stacked blade servers are shown in Figure 2 with a cable running between them.
Figure 2: Horizontally Stacked Blade Servers Emitting Thermal Radiation on Cable.

With power consumption per rack of blade servers expected to be 11.4 kW by year 2012, power rack densities of 10 kW and 20 kW are used in calculating radiant heat flux emitted by each blade server. Radiant heat flux emitted by blade servers is calculated from equation 1:

Where:
Q=   P/H  (heat output (kW) per blade server)
P= power density (kW) per blade server rack
H=number of blade servers in a 42U rack (42 blade servers)
q”= radiant heat flux emitted by blade server
A=surface area of blade server

The radiant heat flux emitted by blade servers of various surface areas and power densities is presented in Table 1. 

Table 1: Radiant Heat Flux Emitted by Blade Server in respect to Surface Area and Power Density

 

Blade Server Area (m²)

10 kW Rack

20 kW Rack

Radiant Heat Flux (kW/m²)

0.1

2.38

4.76

0.15

1.59

3.17

0.2

1.19

2.38

0.25

0.95

1.90

0.3

0.79

1.59

0.4

0.60

1.19

 

REED Testing:

The Radiant Enhanced Extinguishing Device (REED) apparatus was developed to determine the relationship between extinguishing concentrations and radiant heat flux. The design of the REED apparatus takes features from both the cone calorimeter and cup-burner tests. The apparatus provides a radiant heat flux from coils at the top of the device and has similar design characteristics to a cup-burner test, allowing clean agent to be introduced to the fire at a laminar flow rate.

Experiments were conducted on clean agents, IG-100, IG-55, IG-541, HFC-125, HFC-227ea and FK-5-1-12. Testing results for all the agents are presented in Figures 3 and 4.

Figure 3 shows results for clean agents tested from the 0-40 kW/m2 range.  In the results, the agent concentration increased at a more rapid rate with increasing heat fluxes. Beginning at 10 kW/m2, the curves start to plateau. This trend is similar to results seen from Steckler et al.6
Figure 3: FK-5-1-12, HFC-125 and IG-55 Clean Agent Extinguishing Concentration for 0-40 kW/m2

 Figure 4 shows results for clean agents, IG-100, IG-55 and IG-541. Results were conducted at the 0-5 kW/m2 heat flux levels. At these lower heat flux levels, the plateau effect seen from higher heat flux levels are not seen.
Figure 4: IG-541, IG-100 and HFC-227ea Clean Agent Extinguishing Concentrations for 0-5 kW/m2

Due to the actual method clean agents are released into the fire hazard and the effects of flame being weakened by the REED testing method, it was important to develop an alternate testing protocol.

Dubbed the "Impulsive Agent Release" (IAR) method, the testing procedures were modified. The proposed method reduces the 100 second time delay between increases in agent concentration used in the REED apparatus to two seconds. This method was conceived to eliminate the process of the flame being weakened before it succumbs to extinguishment.

Two tests were conducted using the reduced time between agent increases at each specified heat flux level. The resulting extinguishing concentrations were averaged. The average extinguishing concentration was then verified by running a test with the IAR technique. With the IAR method, clean agent is immediately supplied to the REED apparatus at the predefined flow rate based on extinguishing concentrations determined from the two second delay test. The idea was to introduce the clean agent to the flame at a known concentration.  
Figure 5: Original versus IAR Testing Method Extinguishing Concentrations using REED Apparatus, 0-5 kW/m2

A comparison of the IAR testing results versus the original method  is shown in Figure 5. It is evident that extinguishing concentrations in the new IAR testing method exceed the original method by approximately 30 to 45 percent. Also, the curves for the inert gas clean agents have a more curved behavior which plateaus at 2 kW/m2. In the original testing method, the curves didn’t start to plateau until roughly 10 kW/m2. It was also noticed that extinguishing concentrations of clean agent HFC-227ea were greater than HFC-125 in the IAR testing method. In contrast, the original testing method resulted in HFC-125 having greater extinguishing concentrations than HFC-227ea.

Romil Patel is with Siemens Industry, James A. Milke is with the University of Maryland

The authors would like to acknowledge the generous support of 3M in supporting the work described in this article.

References

  1. Robin, M. and McKenna, L., "Fire Protection Considerations for Telecommunication Central Offices," International Fire Protection, Issue 22, May 2005.
  2. Robin, M., et al., "Development of a Standard Procedure for the Evaluation of the Performance of Clean Agents in the Suppression of Class C Fires," Proceeding of the 2007 SUPDET Conference, Fire Protection Research Foundation, Quincy, MA, 2007.
  3. NFPA 2001, Standard on Clean Agent Fire Extinguishing Systems, National Fire Protection Association, Quincy, MA, 2012.
  4. ISO 14520, "Gaseous Fire-Extinguishing Systems -- Physical Properties and System Design," International Organization for Standardization, Geneva, 2006.
  5. Linteris, G. "Clean Agent Suppression of Energized Electrical Equipment Fires”, Fire Protection Research Foundation, Quincy, MA, 2009.
  6. Steckler, K., et al.  "Clean Agent Performance on Fires Exposed to an External Energy Source," National Institute of Standards and Technology, Gaithersburg, MD, 1998.
  7. Mark Owen et al. "Datacom Equipment Power Trneds and Cooling Applications”, ASHRAE Datacom Series, ASHRAE, Atlanta, GA, 2012.
  8. ”National Survey on Data Center Outages”.  Ponemon Institute, September 30, 2010, Ponemon Institute, Traverse City, MI, 2010.
Copyright 2013, Penton Media, Inc. All rights reserved. This eNewsletter is protected by United States copyright and other intellectual property laws and may not be reproduced, rewritten, distributed, re-disseminated, transmitted, displayed, published or broadcast, directly or indirectly, in any medium without the prior written permission of SFPE and Penton Media, Inc.