The experimental 30-MW(t), carbon dioxide—cooled, heavy water-moderated nuclear power station at Lucens, Switzerland, combined the fuel and coolant of the British Magnox reactors with a heavy-water moderator. The fuel element consisted of a graphite column with seven parallel longitudinal channels (Fig­ure 5.22a). Each channel contained fuel rods made from slightly enriched ura­nium metal clad in a finned magnesium alloy (Magnox) can (Figure 5.22 b). Each fuel element was placed in a Zircaloy pressure tube, closed at the bottom end so that the flow of high-pressure (60 bars) carbon dioxide was directed down the annulus between the graphite column and the pressure tube before passing upward to cool individual fuel rods. The heavy-water moderator was contained in an aluminum alloy calandria tank 3 m in diameter and 3 m high, through which the vertical pressure tubes passed (Figure 5.22c).

On January 21, 1969, an accident occurred that resulted in the destruction of one of the fuel elements and the rupturing of its pressure tube. The carbon dioxide expanded into the moderator tank and, after fracturing its rupture disks, entered the reactor containment, which in this case was an underground cav­ern, carrying with it fission products and a large fraction of the heavy-water moderator. The reactor was subsequently dismantled.

Postmortem. The investigations into the causes of the accident were com­plex and lasted about 10 years (Fritzsche, 1981). The initiating cause of the ac­cident was ingress of water into some of the fuel channels around the edge of the core. This was caused by water leaking from the shaft seals of the carbon dioxide gas circulators. Because the pressure tube was closed at the bottom end, a standing water level was formed in these edge fuel channels when the reactor was shut down. Corrosion at the water-air interface resulted in complete removal of the finning over a short length of the fuel rod.

When the reactor was started up on January 21, 1969, water and corrosion products were blown out of the fuel channel. However, due to the lack of any extended surface in the region of the corrosion damage, the magnesium alloy cladding started to melt (at 640°C). The molten cladding soon ran down the channel and solidified, causing a blockage that prevented coolant flow to that channel. The uranium metal soon reached its melting point (1130°C). The ura­nium and the magnesium alloy ignited in the carbon dioxide and the molten metals slumped down inside the graphite column. This column, however, was heated nonuniformly.

Figure 5.22: Layout of Lucens reactor and fuel element.

The column bowed and contacted the pressure tube, which in turn over­heated and burst open under the action of the coolant gas pressure. Only sec­onds earlier the reactor had been tripped because of the release of fission products into the coolant gas stream.

Immediately following the pressure tube rupture, the pressure in the moder­ator tank rose rapidly. At a pressure of 8 bars the bursting disks blew, 0.1 s after the pressure tube mpture, and the expanding C02 bubble forced about 1 ton of heavy water out of the moderator tank

When the pressure tube ruptured, the graphite column also burst apart and the superheated liquid uranium and magnesium metals contacted the pressure tube wall. The Zircaloy wall melted locally and the liquid metal was ejected into the moderator. About 2 kg of the finely dispersed material reacted explosively with the heavy-water moderator. The resulting jet of fire damaged an adjacent pressure tube, which, however, was quenched by returning heavy water before it ruptured. The pressure spike as a result of the chemical explosion reached 16-25 bars and expelled more D20 from the tank.

Perhaps the most significant aspect of this event was the fact that the ingress of water to the core was not identified. The susceptibility of Magnox cladding to corrosion by water is well known, but the very localized and extensive na­ture of the corrosion process in removing the finning from the fuel was crucial. The disadvantage of closed-end fuel channels and the separate parallel chan­nels is also to be noted. It was later determined that even if one of the seven flow channels in the graphite column was completely blocked, the flow to that fuel assembly decreased by only 2%.