New paper discussing Large-scale vented deflagration tests on predictions from NFPA 68 published in Journal of Loss Prevention in the Process Safety Industry

An article titled “Large-scale vented deflagration tests” by James Kelly Thomas, Philip Parsons, and Peter Diakow, Blast Effects Engineers for BakerRisk, was recently published in the Journal of Loss Prevention in the Process Industries.

The Abstract is below. Click here for full access to the paper.



This paper presents results from a test program carried out to determine the peak deflagration pressure achieved within a congested enclosure vented through one wall of the enclosure. The industry standard in the United States for predicting the peak pressure developed in a vented deflagration is the National Fire Protection Association’s Standard on Explosion Protection by Deflagration Venting (NFPA 68). The NFPA Explosion Protection Committee has compiled a database of published and unpublished explosion venting test data. This data was summarized in a 2008 report (Zalosh) that served as the foundation of the development for the vented deflagration correlations in the latest (2013) edition of NFPA 68. In this latest edition, NFPA 68 (2013), the vent area correlation accounts for varying degrees of congestion if the ratio of the obstacle surface area (Aobs) to that of the enclosure internal surface area (As) is greater than 0.4 (i.e., Ar = Aobs/As > 0.4). Congestion is accounted for within the correlation at all values of Ar, however when Ar is < 0.4, variations in the level of congestion are not accounted for. The tests described in this paper were performed using an obstacle array with an Ar ratio of less than 0.4.

These tests were conducted in a rig with a 48-foot width, 24-foot depth, and 12-foot height. The rig is enclosed with solid walls, roof, and floor, allowing for venting through one of the long walls (i.e., 48-foot by 12-foot). The venting face of the rig was sealed with a 6 mil (0.15 mm) thick plastic vapor barrier to allow for the formation of a near-stoichiometric propane-air mixture. The flammable gas cloud was ignited near the center of the rear wall. Steel vent panels (20-gauge, 2 lbm/ft2) were installed over the plastic vapor barrier using explosion relief fasteners. The vent panels were configured to release at 0.3 psig; vent panel restraint devices were not utilized. The congestion inside the rig was provided by a regular array of vertical cylinders (2-inch schedule 40 pipe and 2-inch outer diameter cylinders) giving area and volume blockage ratios (ABR and VBR) within the congestion array of 4.9% and 0.5%, respectively. The obstacle-to-enclosure surface area ratio (Ar) is 0.3 with the array extended throughout the rig and vent panels installed, which is less than the critical value to account for congestion in the NFPA 68 correlation.

Four series of tests were conducted with varying vent parameters, flammable gas cloud sizes, and congestion levels. Baseline tests were performed with the congestion array and flammable gas cloud extending throughout the entire rig without vent panels present (i.e., vapor barrier only). The second test series included the addition of vent panels for the same congestion pattern as that employed for the baseline tests. The third test series utilized a flammable gas cloud that filled only the back half of the rig. For the fourth test series, the congestion array occupied only ¼ of the rig. The peak pressures and impulses for each test series are provided, along with pressure histories internal and external to the rig for selected tests. The steel vent panel throw distance is also provided as a function of internal peak pressure.

The test data were compared with the predictions of the vent area correlations provided in NFPA 68. For all but the fourth test series (i.e., congestion array occupying ¼ of the rig), the average internal peak pressures were approximately a factor of 2 larger than those predicted by NFPA 68. Adjustments to the NFPA 68 correlation were investigated to improve the agreement with the current test data.