Spark Box Case Study


At a local Recycling Steel Mill, a "Mini-Mill", two explosions occurred within a two week period in the Spark Arrestor Box causing severe damage and shutting down the mill for several days each time while repairs were made.


The 1000 °F hot gases from the Electric Arc Furnaces, which melt the scrap steel, are pulled from the furnaces through 10 ft diameter ducts, then through the Spark Arrestor Box where any larger pieces of slag entrained in the airstream are dropped out before the gasses pass through the exhaust fans and then are pushed out through the bag house to the atmosphere.

  • The client knew that the problem had never occurred in the previous 8 years of operation since the plant was commissioned.
  • The constituents of the gas stream at the time of the events were not known.
  • There are many impurities on the scrap steel which produce combustible gases.
  • Two possible candidates for the combustible gas responsible for the explosions were carbon monoxide (CO) and hydrogen, both of which are by-products of the steel making process.
  • The composition of the gases varies depending on the ingredients used to produce the desired grade of steel. The ingredients for a given grade can also vary depending on the current cost of the raw material.
  • There was a possibility that the ingredient responsible for producing the explosive gases was Direct Reduced Iron.


The proposed solution, which was acceptable to the client, was to design a new Spark Arrestor Box with doors which would open to relieve the overpressure before the box would be damaged. The methodology to be used was the procedures contained in NFPA 68, "Guide for Venting Deflagrations". The phenomena that was occurring was a "deflagration", which is a rapidly burning flame front passing through a mixture of a combustible material, either gas or dust, and air. This is a different phenomenon than an "explosion", which is the ignition of an explosive material, which in turn is a material that already contains both the combustible substance and the oxygenating agent within itself.


Due to the time constraints, the design had to proceed without definitive input information. The client purchased a new gas sampling system that could sample fast enough that it might capture the fluctuations in the gas content when a deflagration occurred, but it would not be operational for several months. The installation of the new Spark Box was to be made during the regular August shut down. In order to meet the schedule, the permissible time for each phase was worked backwards to allow time for fabrication and shop drawing preparation and design drawings. The decision was made to base the design on the most likely culprit, which was CO, as mentioned above. There was full understanding with the client that the system would not work if the combustible gas was hydrogen. Because of the large uncertainties associated with the design parameters, it was agreed that the box would be a structurally robust design with strength in excess of the calculated demand and that all available surface area on the walls of the Spark Box would be used to locate pressure relief doors, thus providing a relief area in excess of the NFPA requirements and providing the maximum opportunity for successful performance.

For functional considerations, the new design was to duplicate the original Spark Box in size, 60 ft wide x 20 ft deep x 40 ft tall with a triple hopper bottom and internal baffle wall. To facilitate the speed of reconstruction it was decided to repair and reuse the existing hoopers which were not severely damaged. To maintain appropriate door sizes, the box was designed in 4 tiers of about 10 ft height with the walls panelized in approximately 20 ft in lengths. Door size is important because the larger the size, the greater the rotational moment of inertia resisting the door opening rapidly when a deflagration event occurs and consequently the higher the corresponding release pressure, Pred, that had to be sustained by the roof and walls of the box as well as the doors and their hinges. The erector later decided to preassemble the wall panels into boxes for each of the 4 tiers, to be erected like the layers of a wedding cake during the shutdown. See the attached drawing. Among the custom made items that had to be selected were a gasketing material for the doors and the seal material for the joints in the walls. These materials had to withstand the operating temperatures of 1000 °F, provide an airtight seal for the negative 12 inch water gage static pressure during normal operating conditions and not be blown off the door during a deflagration event. After some discussions about the requirements with the gasket manufacturer, he made up a sample consisting of woven stainless steel cording wrapped in an asbestos fabric which was hand coated with high temperature silicone on the inside. This proved to be a very effective gasket.

Selecting hinges and latches for the doors that would be able to withstand the service and be relatively maintenance free was a challenge. The hinges are stainless steel with high temperature lubricant and can carry 6,000 lbs each. There are 3 hinges per door, the heaviest of which weighs 1,300 lbs. The required 18,000 lb hinge capacity is controlled by blast loading. The latches are also stainless steel and can be calibrated to keep the doors from being sucked open by negative wind pressure while at the same time having the ability to release during a deflagration event and re-latch after the event for continued operation of the plant.

Once the design was completed the client solicited fabrication and erection bids from the fabricators who usually serviced the plant. No one would bid the job because they could not meet the schedule due to the shop drawing preparation time required. One fabricator offered to bid on fabrication and erection if shop drawings were supplied by the owner. In order to meet the client's needs, AMG, Inc. agreed to prepare shop drawings, which is not normally a service AMG, Inc. chooses to provide. The software, "Mechanical Desk Top, R.4", which is typically used for mechanical engineering machine design, was selected for this project. Every piece is modeled 3-D in its assembled configuration and given a piece mark. Reports can then generated for each piece for tracking purposes and a detailed 2-D shop drawing is generated for each piece from the 3-D Model. There were over 54,000 pieces required for the Spark Box, including bolts gaskets and seals, hinges and latches, etc.

The new Spark Box was tested by a deflagration event within the first week of service. It performed as intended and has continued to do so for 7 years since. It has done its job so problem free that the maintenance catwalks that were designed to provide service access to the door hardware and permit resetting the Door latches have never been purchased or installed.

Since the box was built, observation a deflagration event at night by an employee gives a better understanding of the mechanics of the phenomena than was available at the time of design. Based on the sound, the deflagration begins at the furnace and propagates through the duct until it reaches the Spark Box. The duct has tremendous reserve strength for sustaining a deflagration because it is loaded in hoop tension by the overpressure. After reaching the Spark Box the flame front bursts into it from the 10 ft diameter duct. This creates a huge ignition source, much greater in volume than the point ignition source on which the NFPA equations are based. The result is much higher over pressures than predicted by the NFPA equations.

Nevertheless, the most severe load that the box sees, based on observations of yielding patterns of the metal deck panels in some of the larger doors is the vacuum pressure created inside the box when the air is blasted out and then the doors slam closed again after an event.

The employee who witnessed the event said that when the deflagration hit the Spark Box, "sixteen doors instantly flew open and each shot out a twenty foot long fireball that looked like dragon's breath, and then it was over, with no evidence that it had ever happened".