On Saturday at about 13h00, a major leak occurred as a result of the catastrophic failure of the nuts on the studs connecting the bonnet and the body of a 2” manual drain valve in a blow down line. The bonnet was ejected from the valve body, creating a 2” opening in the system to atmosphere. Nobody was injured and no ignition of the flammable gas occurred. The system was stopped, isolated and decompressed to flare. At around 18h00, the pressure in the system has dropped sufficiently to start maintenance works to replace the damaged valve. In total about 45 tons of flammable gas was released to atmosphere. In total about 45 tons of flammable gas was released to atmosphere. A mixture of flammable gasses was released. The main components were methane and ethylene.

History of 2” drain valveLittle information is available about the history of the 2” drain valve. This valve is not used during normal operation of the unit, only during startup (connection for flushing with nitrogen).The following information is available: The line was originally installed in 1984;o Works were done on the system in 2003; The valve is installed at a height of about 4 meters above grade;o Works on this type of valves can only be done during turnarounds; The construction of the valve (bonnet kept in place with 4 bolts) does not allow works to be done in run; There are no records about repairs of the valve;o According to the original specifications, the 2” drain valves should be a carbon steel drain valve. However, the installed valve is a stainless steel valve. It is not clear whether the valve that failed on 25/04/2015 is the originally installed valve or whether the original valve has been replaced or worked on during one of the turnarounds between 1984 and 2015. Possibly, the carbon steel nuts were installed during one of these turnarounds. As a result of previous incidents with valve packing followers, a visual verification of isolation valves in stainless steel systems was launched prior to the incident on 25/04/2015. The 2” drain valve that failed on 25/04 was in the scope of this verification and some comments were reported about the status of the valve: No ice formation o Corrosion of bolts Incident investigation results Investigation of the failed valve shows clearly advanced corrosion of the nuts of the studs connecting the bonnet and body of the valve. The following construction materials were encountered:o The bonnet and body of the valve are constructed from austenitic stainless steel (316); The valve is installed on a line constructed of low temperature carbon steel; The 4 studs used to connect the bonnet and body of the valve are constructed of ferritic stainless steel (410); The 8 nuts on the studs connecting the bonnet and body of the valve are constructed of carbon steel;The following conclusions can be drawn upon investigation of the recovered valve: The severe corrosion of the 8 carbon steel nuts on the 4 stainless steel studs/valve (see Figure 8) is consistent with a combination of galvanic corrosion and atmospheric corrosion of the nuts. The 4 stainless steel studs of the valve bonnet/body connection are in good condition.o Some of the nuts are missing because they were torn off during the ejection of the valve bonnet.o It is possible that the severely corroded nuts (including the missing nuts) were all located at the same side of the valve bonnet/body connection (back side). According to the original specification, the 2” drain valve should have been a low temperature carbon steel valve instead of a stainless steel valve.o Besides carbon steel nuts on the studs of the valve bonnet/body connection, carbon steel nuts were also discovered on the studs of the valve flange/body connection (the 8 nuts on the blind flange are a mix of carbon steel nuts and stainless steel nuts. The carbon steel nuts are corroded.). Upon ejection of the bonnet of the 2” drain valve, the bonnet hit a nearby propylene pipeline. This propylene pipe was hit in an horizontal plane above the horizontal plane of the original location of the 2” drain valve, suggesting that the nuts that failed first were the ones located on the bottom half of the bonnet/body connection. Extensive corrosion of carbon steel nutsThe catastrophic failure of the carbon steel nuts of the valve bonnet/body connection occurred because of severe corrosion of the carbon steel nuts on the stainless steel studs.The extensive corrosion of the carbon steel nuts is a combination of galvanic corrosion and atmospheric corrosion. Installation of stainless steel valve with stainless steel studs in low temperature carbon steel line The line was originally installed in 1984. According to the original specification, the 2” drain valve should have been a low temperature carbon steel valve. However, the installed valve was made of austenitic stainless steel (316). It is possible that the valve that failed is the originally installed valve. Installation of carbon steel nuts on stainless steel studs of stainless steel 2” valveWorks on this type of valves can only be done during turnarounds since the construction of the valve (bonnet kept in place with 4 bolts) does not allow works to be done in run. There are no records about repairs of the valve (records were checked until 1998). It is not clear whether the valve that failed on 25/04/2015 is the originally installed valve or whether the original valve has been replaced or worked on during one of the turnarounds between 1984 and 2015. Possibly, the carbon steel nuts were installed during one of these turnarounds. Valve in humid environment The valve is constantly in a wet environment due to the condensation of water vapor in ambient air. The presence of this wet environment is a necessary condition for both galvanic corrosion and atmospheric corrosion to occur (presence of electrolyte). Nuts not protected against atmospheric corrosionThe carbon steel nuts were not protected against atmospheric corrosion (e.g. by paint or selection of corrosion resistant material). Nuts in carbon steelThe nuts used to tighten the stainless steel studs were made of carbon steel. Condensation of water vapor in airThere is a constant condensation of water vapor in air at the location of the valve due to the low temperature of the fluid in the low temperature carbon steel line and because of the lack of thermal insulation on the valve. Installation of valve not meeting specificationsIt is not unusual to install equipment parts made of more noble materials in systems made of less noble materials, provided that the compatibility of the materials is checked and that appropriate measures are taken to avoid galvanic corrosion. Valve not thermally insulatedThe valve is installed in a thermally insulated low temperature carbon steel line. The valve as such was not insulated. Process at -6°C. The valve is installed in an insulated low temperature carbon steel line. The operating temperature of the fluid in the line is about -6°C. Inspection of lineThe 2” drain valve was part of lines that are inspected inside the RBI scope. The line with the valve that failed was assessed to lie in the low risk area of the RBI matrix. The lines in this low risk area are inspected by visual verification of the state of the line during site visits (frequency of these visits is every 5 years).An inspection report of this particular line from a previous inspection cycle (2009) is available. This line was scheduled to be subjected to a more detailed inspection (with thickness measurements etc.) in the next inspection cycle (to be carried out in the turnaround of 2016). A request of this more detailed inspection was applied in a pre-inspection report of end 2014. The state of the valve that failed was insufficiently detected in the visual inspections of 2009.

13.00 Alarm by gas detectors in the area. A major leak is being reported that cannot be approached. Intervention team is installing a water screen for dilution of the vapor cloud Activation of emergency stops of different parts of the installation by Central Control Room operators due to multiple gas alarms and after confirmation of internal fire brigade Evacuation alarm is given Phase 2 of the internal emergency plan is activated. 13.55 Monitoring of pressure profiles allows diagnosing were the leak is situated. This diagnose is confirmed by visual observations outside by maintenance and production people. The exact location of the leak is fine-tuned to be on a 2” drain valve on a blow down line. The lead is isolated. Since there is no isolation valve between the section with the leak and the next section both have to be emptied to stop the leak. 15.00 The pressure in the system has fallen from 32 bar to 2.4 bar 18.00 The pressure drops quickly to ± 0.1 bar after opening of a blow down. Scaffolding is erected and the leaking 2” valve is replaced. The replaced 2” valve with studs and nuts is recovered for further investigation.

Development of PMI program (Positive Material Identification) Quality control of delivered materials and equipment parts. Quality control of installed materials in the field by additional controls (e.g. supervision, Positive Material Identification). Organization of campaign to raise awareness about galvanic corrosion. During the incident investigation, it became clear that the risks associated to galvanic corrosion are not always clearly understood by all layers of the organization. To raise understanding and awareness about galvanic corrosion hazards, a sensibilisation campaign needs to be organized by the site. Development of formal protocol for visual inspection of valves There is no formal procedure describing the visual inspection of equipment. To improve the quality of visual inspections during site visits, a formal protocol needs to be developed to describe the requirements related to visual inspection. This formal protocol needs to include specific a measurable acceptance criteria for each of the inspected equipment. Gas detection The gas detection system in the installation gave a quick warning of the leak allowing prompt action by the unit operators. It took a long time to depressurise the section of the installation with the leak. Partly because of the white cloud formed by the leak that didn’t allow visual verification of the leak origin. Partly because there was no block valve on the top of the destillation colomn to which the leaking valve was connected. Therefore also the following equipement needed to be depressurised. The need for an additional block valve is still under investigation.