You are on page 1of 4

0260-9576/05/$17.63 + 0.

00 Institution of Chemical Engineers 2005

Explosion during a welding operation

Introduction
An explosion occurred on an ammonia plant, where welding repairs were being carried out on defective pipework. A welder, who was carrying out the repair, was injured, sustaining burns to the face and shoulder. No damage to plant or injury to other personnel resulted, and the injured man was released from the hospital later that week.

Description of the plant


The section of plant in which the work was to be carried out is shown in Figure 1. This section of plant transfers gas from the low temperature shift converter (LTS) to the carbon dioxide removal section of the plant. Heat exchangers in the section provide boiler feedwater pre-heat (C307) and heat for Benfield solution regeneration (C401 and C402). In doing so, the gas is cooled to below its dew point, and condensate is collected and removed in catchpot G401.

point, it was decided to shut down the front end of the plant to save gas and to stop the problem of condensation affecting the area to be welded. In an attempt to prevent water reaching this area Klingerite (compressed asbestos joint material) disks were fitted in the flange exit C401 (condensate reboiler) and below the bypass butterfly valve (V4001), which was left closed. By the afternoon of the incident all the cracked welds had been identified, and significant sections of pipework had been removed (see Figure 2). This involved significant amounts of welding and grinding (grinding had taken place on the end of the open pipe only 15 minutes prior to the explosion). At this point, just prior to the explosion, gas testing was carried out on all drains on the line and at the open end where the welding was to take place. At approximately 15.30 hours, a welder struck an arc at one of the open ends of the pipework to repair a minor defect. An explosion occurred within minutes, injuring the welder.

Events leading up to the incident


A leak had been discovered (as indicated in Figure 1) and it was decided that the plant should be partially shut down to repair these cracks by grinding out the defective parts and welding from the outside. The plant was to be vented, isolated and purged by pressuring and depressurizing with nitrogen and then tested to ensure that no flammable gas was present. This job had initially been expected to take eight hours, but two problems significantly extended this period. Firstly, there was difficulty in getting the line dry enough to carry out the welding. Condensate draining slowly from the tubes in the condensate reboiler C401 (see Figure 1) had to be blown away from the work area. Secondly, radiography checks on welds adjacent to the one that had been found to be leaking indicated that they also had cracks in them, which had to be ground out. At this

The investigation
It was initially thought that the explosion had taken place in the piece of pipe under repair and that possible fuels could be: (a) Residual hydrogen in the system due to inadequate isolation or purging. (b) Acetone left behind following dye penetrant crack detection. (Acetone is used as a cleaning solvent and as a solvent in the developer). (c) Ingress of acetylene from a burning torch, present some 2 m from the work piece. Item (c) was eliminated when immediate checking revealed isolation of the torch and bottles at source and adequate separation from the work piece. Item (b) was eliminated when calculations showed that there could not be sufficient quantities of acetone vapour present to give the force of the explosion.

LOSS PREVENTION BULLETIN 186

FIGURE 1: P & I DIAGRAM OF THE PLANT The initiation point of the explosion was just above the butterfly valve V4001 (see Figure 3). The flame front and pressure wave travelled through C401, bursting the disc at the outlet, ejecting the pieces, and causing injury to the welder.

The ignition source


FIGURE 2: KLINGERITE FITTED AREAS Inspection of the electrical circuit in place for the welding operation revealed that the return cable was clamped to a boss on the base of catchpot G401 (as shown on Figure 1). As part of the pipework between the work piece and G401 had been removed, the electrical return circuit was through the plant. As both joints connecting the workpiece to the plant had been disturbed to insert the Klingerite discs and only partly

Calculations showed that the explosion must have taken place within the system enclosed by the Klingerite discs (Figure 2). The power needed to burst the disc under C401 could not have been produced by the volume of gas between the discs and the point of welding.

LOSS PREVENTION BULLETIN 186

and loosely rebolted, there was inadequate electrical continuity. It was therefore concluded that a spark was generated by strong currents above the butterfly valve, causing the explosion of a hydrogen/air mixture that was present in the closed system.

During purging
Considering the volume being purged, if there was perfect mixing with nitrogen, then the hydrogen concentration could be about 0.4% maximum. A sample of purge taken at the end of the blowdown may have been as low as 0.1% in hydrogen and, therefore, would not have registered on the gas testing meter. This could have led to the erroneous conclusion that the system was purged and that the isolations applied to the shift converter were tight and acceptable.

Presence of process gas


It could be proved that the system had initially been completely purged of process gas, so it could only be concluded that one of the isolation valves was passing hydrogen from the LT Shift (see Figure 1). Helium leak tests showed that both the LT Shift exit and bypass isolation valves were leaking, and a study to explain why these leaks had been missed during initial preparation was carried out. The leak tests showed that hydrogen from the LT Shift would clearly have provided sufficient fuel for the explosion. However, gas testing conducted during purging and testing (see figure 1) did not reveal the presence of any such hydrogen.

With no purge
If the vent to F stack had been open and valves V3058 and 3059 (Figure 1) opened to take a sample at the end of the blow down, with no nitrogen purge then no reading would have been shown on a gas testing meter, as air would have been inspirating into the system via the drains. Figure 4 shows the relative height of all the vessels. A glass model of the pipework confirmed, by experiment, that a leak at (A) would result in air inspiration at (B), (C) and (D).

The welding and grinding operation


Two days prior to the explosion, the decision was made to remove a section of pipe. The presence of hydrogen was not detected during this operation. It is thought that this was because regular nitrogen purging was being carried out during this period, preventing the build-up of an explosive mixture. Nitrogen purging ceased on the evening prior to the incident and the Klingerite joints were inserted. The leak presumably persisted. Although the drains downstream of C307 were open, hydrogen would gradually progress towards C401 by buoyancy migration mixing with the

THE

FIGURE 3: BUTTERFLY VALVE V4001. INITIATION POINT OF EXPLOSION WAS JUST ABOVE THE VALVE

FIGURE 4: AIR INSPIRITING INTO THE SYSTEM VIA THE DRAINS

LOSS PREVENTION BULLETIN 186

air that had been introduced by breaking the joints at V4001 and C401 exit to fit the Klingerite discs. In addition, there would be inspiration of air through open drains. Thus, the mixture would slowly move into the explosive range. When welding took place earlier that day, better electrical circuitry presumably existed. It is possible that the gasket at V4001, which showed distinct signs of having been wetted with condensate, dried out increasing the circuit resistance. Whatever the mechanism, the implication is that a stray unintended arcing at a position remote from the welding work did not develop immediately.

Conclusions
The direct cause of the incident was the failure to adequately isolate hydrogen within the plant while maintenance work was being carried out. The incident cause was the inadequate management system that was based on the principle of firstly identifying the hazard, then taking the necessary precautions. The very complex mechanisms that led to this incident were not foreseen, and even after the event the explanation was only found after two full days of intensive enquiry. The instruction on process isolations was amended to insist on positive (slip plates) isolations for hazardous materials.

Incident summary
The explosion was initiated by the ignition of a mixture of hydrogen and air by a spark, generated by stray electrical welding currents. This was due to inadequate welding return arrangements and exacerbated by the disturbance of bolted flanges on the work piece. The hydrogen entered the system by a leaking valve, undetected at the time of the initial job preparation and during subsequent checks due to the purging sequence and the inspiration of air at sample points.

Editors note
It appears in this case study that gas tests were carried out after nitrogen purging had taken place (i.e. no oxygen would be present in the samples). Standard gas detectors give false readings in oxygen deficient atmospheres and are therefore not recommended for testing inert atmospheres. For more information on gas testing, see the new toolbox talk Gas testing not as straightforward as it seems on this issues LPB Online.

LOSS PREVENTION BULLETIN 186

You might also like