Explosion in Cone-Roof Gas Oil Tank

An internal explosion occurred in a cone-roof gas oil tank. The explosion completely ruptured the weak roof-to-shell weld seam and projected the roof into the diked area beside the tank. The ensuing fire was relatively minor and involved only a small quantity of product which spilled over into the diked area.
The explosion happened while a refinery contractor was taking a sample through the gauging hatch. It was followed by a fire in two adjacent tanks located in the same dike and caused one fatality.

The tank was in a refined gas and oil product rundown service, and the explosion is believed to have resulted from ignition of a flammable mixture of hydrogen-rich treat gas and air in the vapour space. The vapour space above gas oil (flash point of 66 C), is normally non-flammable. However, at the time of the accident a flammable condition had been generated as a result of non-complete gas-oil stripping at the exit of a hydro treating unit.

The IGNITION SOURCE The source of ignition was an electrostatic spark initiated by synthetic rope (instead of a cotton one) used with a device to obtain samples from the tank.

The ROOT CAUSE(S) are related to the shortcoming of the risk assessment and of the safety design. Management of changes was not in place, and the preventive systems inadequate for the possible presence of hydrogen in the tank.

gas oil tank

The tank had been filled up to the maximal capacity with various type of oil streams, and was waiting for dispatching.
The tank inventory was 15,000 m3 of gas-oil. At ful storage capacity, the vapour space left at the top of the tank, under the roof, was 880 m3. Assuming that the whole space was filled with air plus hydrogen and a minimal hydrogen volumetric concentration of at least 10% to achieve explosive conditions, it is possible to calculate a minimal hydrogen inventory of approximately 8 kg. In case of hydrogen concentration near stoichiometry, the hydrogen mass would have been 24 kg.

The environmental temperature was 48 C.

DESCRIPTION OF HYDROGENROLE IN THE PLANT
The refinery was making use of hydrogen in the two hydrogen desulphurisation units, each include a stripping section. These sections had been modified during the lifetime of the plant, and the last modification consisted in replacing the steam injection stripping
with hydrogen as the stripping material.
The gas-oil products from the two desulphurisation units, and the streams that bypass the HDS units, were routed and blended to specific products in the tank farm.

Both internal and external emergency operations of the refinery and outside services were very effective, and fires in both tanks and the dike were put out without any additional damage to other equipment in the tank farm.
Foam was used on the on the roof and surface cooling systems enabled protection of the tanks nearby.

Based on the findings and conclusions of the paper of Riziel (see References), the following lessons can be learnt.

(1) A risk assessment considering the plant / facility in its entirety is required. The main gas-oil streams flowing from the bottoms of the two HDS strippers operated many years with hydrogen stripping, but a HAZOP (hazard and operability) review study was never performed on the entrance to the storage tanks, the risk assessment had only focussed the stripping tower itself. The question: “Is there any possibility of hydrogen sweeping or solution with the product gas oil?” was never asked. Because of the high diffusivity of hydrogen, its possible presence in the stripped oil stream was not considered relevant.

(2) Preventing measures must be based on effective and correct measurements of the critical operative parameters, to detect early enough deviations from their ‘safety’ values. The flash tests were performed on a final product consisting of streams with different boiling points and flash points, some of them with a flash-point temperature below the standard 66 C. However, when arriving in the storage tank at atmospheric pressure, the hydrogen in these streams was leaving the solution and concentrate in the tank vapour space under the roof. A fuel sample taken from the tank to measure the oil flashpoint would not have detected any presence of hydrogen. The detection/analysis of the vapour space under the roof would have delivered the required safety parameter assessment.

(3) The definition of the worst-case accidental scenarios is often based on pre-knowledge of the experts, which may exclude some accidental path. The operation of the storage tank could not exclude under all conditions the formation of a flammable/ explosive atmosphere in the gas phase, since air was always present at the top of the tank. An air vent was installed, through which air was sucked into the tank while lowering the oil level and vented out when the level was rising. Therefore, effective preventing measures were adopted to avoid ignition of the fuel. The worker which was extracting the fuel sample wore anti-static clothing and anti-static shoes. However, in presence of hydrogen and a full tank which was minimising the vapour space and facilitating the formation of air-hydrogen mixture above the low flammability limit, the antistatic measures were not enough. The most probable ignition source was an electrostatic charge on the sampling device rope, made of synthetic nylon and not cotton as it should have been. This detail had been disregarded, but had not have any consequence till the moment of the explosion.

(4) On the positive side, the fact that the roof was only weakly welded to the walls provided to the explosion an easy path to release energy, without endangering the containment of the oil inside the tank. The full release of the tank content would have had much higher negative consequences and raised the possibility of an escalation.