Fire on a flange in a refinery

30 pounds of gaseous hydrogen (14 kg) were accidentally released from a heat exchanger at a refinery. The location was the flange between the heat exchanger channel and shell and occurred during unit start-up. The released hydrogen ignited, leading to an emergency shutdown, and caused over $1.5 million in property damage.

The heat exchanger was assembled in 2013 (ten year before) and had undergone 43 thermal cycles. During that time, the bolts had relaxed. This is an expected event, but the bolts had relaxed to the point that the flange could no longer contain the hydrogen within the heat exchanger. The investigation found that the bolt torque value used to assemble the heat exchanger in 2013 was too low and should have been much higher to prevent leakage.
The company found that the assembly instructions for both the affected heat exchanger and a nearby similar heat exchanger listed incorrect torque values that were too low, because the instructions listed the wrong bolt sizes. In addition, CITGO suspected that since it is common practice at the site to assemble similar flanges to similar torque values, the incorrect torque instructions from the nearby heat exchanger may have been applied to the incident heat exchanger.
The company determined that the hydrogen may have ignited from contacting an adjacent hot heat exchanger, friction from the release, or a spark.

The INTITATING CAUSE was the loss of the confinement capacity of the flange due to the relaxation of the bolts.
The ROOT CAUSE was the application of the wrong torque, due to shortcoming in the procedures.

flange

The unit was restarting.

DESCRIPTION of the UNIT
The unit affected was a shell-and-tube heat exchanger, which is the most common type of heat exchanger in oil refineries and other large chemical processes, and is suited for higher-pressure applications. As its name implies, this type of heat exchanger consists of a shell (a large pressure vessel) with a bundle of tubes inside it. One fluid runs through the tubes, and another fluid flows over the tubes (through the shell) to transfer heat between the two fluids.

From the photo in the report it appears that the flange is surrounded by bridges and other structural elements, so that the space in which the hydrogen release occurred has to considered at least semi-confined.

A mistake during assembling of a flange had an effect ten years later. A too low torque had been applied to the bolds of a flange, because of wrong or wrongly applied instructions. The incident highlights the need of regular reviews of inspection and maintenance procedures, for an early detection and elimination of malfunctions which could bring to loss of confinement. Ten years improvements in technical knowledge and diagnostic tools should also be used to refine the management of operation. At the same time, attention has to be paid to the loss of historical expertise and memory that may occur due to the considerable turnover of operative and management personnel expected happening in in ten years.