Moving Fuel Load Surveys to Digital Platforms

Europe_Digital_Magazine.png

View the PDF here

By Thomas Gernay and Negar Elhami Khorasani

Evaluating fire action in a building compartment relies on codes and standards that provide design values for fuel load. These values are based on statistics of data collected from surveys. However, the data are scarce and rarely updated because no efficient on-site collection method exists. With support from the NFPA Research Foundation, we proposed a new fuel load survey method that harnesses recent technological advances to modernize the inventory method with a digitized approach using machine vision. We also applied the new method to three university office buildings and measured a mean movable fuel load density of 1115 MJ/m² with a standard deviation of 614 MJ/m², which is considerably larger than values found in older surveys and most code provisions. This article is the compressed version of two articles published in Fire Technology describing the methodology[1] and its application[2] to collect fuel load data in buildings.

Introduction

The fuel load quantifies the amount of energy available to fuel the fire. Surveying fuel load density (in MJ/m²) entails quantifying the amount of combustible materials (in kg) and the nature of said materials (quantified by their calorific value in MJ/kg) in a compartment of given floor area (in m²). Combustible materials include movable contents such as furniture items as well as fixed content such as doors, carpets, and interior finishes.

Despite being a crucial input for the fire-safe design of the built environment, fuel load data are scarce because current surveying methods remain relatively impractical and/or inaccurate. A review of fuel load surveys[1] shows that the data underpinning the design value recommendations in codes is limited and often dated. For surveying fuel load in buildings, the NFPA 557[3] recommends adopting a combination of inventory and weighing method, where the inventory method obtains mass of common furniture items by visually matching the furniture items to items from manufactorers’ catalogs. In this research, we aim to disrupt this surveying method by proposing the use of a new interactive electronic surveying form with cloud storage and machine vision, which speeds up the on-site survey, enables the item matching process through online retail search engines, and automates the database construction.

Digitized Methodology Using Electronic Survey Form and Machine Vision Post-Processing

The fuel load survey is completed through an electronic form accessed from a mobile device (such as a tablet). The method, illustrated in Figure 1, can be divided into four steps:

1.   Digital inventory. Onsite, the surveyor collects data through the digital application, including snapshots and digital measurements of the room content, general characteristics about the surveyed room, and possibly building drawings.

2.   Data organization. The developed application stores the snapshots and introduced data into a structured online database (SQL database in formatted tables).

3.   Item matching. Off-site, the surveyor performs an image search on the snapshots to find matching furniture and corresponding specifications online, using image recognition and data mining techniques from existing online inventory database. Matching with the online catalog item informs the recording of mass and material composition of the surveyed item.

4. Fuel load calculation. The application converts furniture specification (mass and material composition) into fuel load using material calorific values. The fuel load is stored in the database and is available for analysis and statistics (e.g. classification per occupancy, material type, etc.).

 

Figure 1. Overview of Digitized Fuel Load Survey Method

Application and Results

The method is validated for sample items by comparing dimensions and mass obtained by physical measurements to those obtained through image search in online database, showing good agreement.[1] Compared with the former inventory method, the proposed machine vision approach improves accuracy and efficiency because it gives immediate access to a greater and continuously updated repository of items. Yet, the accuracy of the procedure depends on the availability of the item online and on the quality of the image. At this stage, user input is still needed to review the options pre-selected by the AI and select an appropriate matching item and material composition. Future improvements of the method could leverage machine learning algorithms to complete the matching process on the developed application.

We applied the survey method to 34 closed offices and 161 cubicles within 12 large, open-plan office spaces in three buildings in Buffalo, New York. The measured fuel load density for movable content had a mean of 1115 MJ/m² with a standard deviation of 614 MJ/m² (Figure 2). Paper represented 54% of this movable fuel load, followed by 36% furniture, 8% electronics, and 2% other content. When including the fixed content, the measured total fuel load density had a mean of 1486 MJ/m² and standard deviation of 726 MJ/m².

​

 

Figure 2. Movable Fuel Load Density in Surveyed Office Buildings: Data and Distribution Fit[2]

The mean surveyed movable fuel load density (1,115 MJ/m²) significantly exceeds the values provided for office buildings in the Eurocode 1 part 1-2 (420 MJ/m²) and NFPA 557 (600 MJ/m²). Several reasons can explain this result, including the small sample size surveyed in this proof-of-concept application, the large quantity of stored paper in university offices, and the fact that no derating factor was applied to the stored content. However, a few surveys conducted in the 1990s and 2000s indicated comparably large values,[2] which stresses the need to collect more data to provide robust design recommendations. Furthermore, new definitions of occupancy sub-categories may be needed to determine fuel load density, at least to distinguish between office occupancies with lower versus higher fuel loads.

Conclusion

The new method to collect fuel load data in buildings reduces disruption for room occupants through the use of an efficient mobile application, speeds up the process of quantifying item properties by leveraging machine vision, and standardizes fuel load surveying owing to interactive digitized form and data storage. Moving forward, the digitized application can be used broadly to survey buildings and populate the database with comprehensive fuel load statistics to support the establishment of design values in codes and standards.

The authors gratefully acknowledge the Fire Protection Research Foundation (FPRF) and the National Fire Protection Association (NFPA) for funding, as well as Amanda Kimball and the members of the Project Technical Panel for their support.

References

Thomas Gernay is with Johns Hopkins University, USA.

Negar Elhami Khorasani is with the State University of New York at Buffalo, USA.