Graphs of DREG reactor parameter data from Skala computer

 (back)

The most detailed, publicly available reactor data is reproduced in graph form below, based on this page from the renowned site by Viktor Dmitriev of VNIIAES: http://accidont.ru/datable.html

Some observations from the graph above:

• Remember that the reactor is split into two separate coolant loops, each with its own set of Main Circulating Pumps and Drum Separators. Therefore each parameter is listed twice, for left-hand side and right-hand side.
• The Skala's DREG program stopped recording reactor parameters three times that night. Just after midnight the system crashed due to a power supply problem. The system was also manually reset twice because the control staff intended to either delay or cancel the rundown test. One of these Skala restarts is presumably indicated by the period of no data ending around 1:18am. This reset could also mark a transition between the turbine vibration test and the rundown test.
  
• Feedwater is additional coolant added to the drum separators to make up for losses to evaporation. In the case of the two large influxes of feedwater depicted above, SIUB Stolyarchuk was likely trying to refill the left-hand side drum separators in accordance with safety parameters. The addition of relatively cool feedwater could depress boiling in the core, requiring the removal of control rods. Conversely, note the increase in feedwater temperature just before the rundown test begins, which reflects the feedwater flow rate returning to normal. This likely introduced positive reactivity into the core in the final minute.
  
• The green lines indicate that the left-hand side drum separators suffered from low water levels, which was the cause of periodic warning signals during the shift. Around 1:05am the water level actually dropped below the -700mm lower limit, hence the SIUB's urgent insertion of feedwater. Water in the drum separators is an important reservoir for emergency cooling in the result of an accident, but did not play a direct role in this one.
• A key parameter in the lower graph segment is the high temperature of the coolant (after mixing with the cooler feedwater) as it enters the reactor. The RBMK is intended to induce boiling in coolant with a temperature of 270 degrees Celsius, increasing its temperature by about 10 degrees during the trip through the active zone. But after 1:00am the temperature is generally above 280 degrees, which is referred to as 'low subcooling' in INSAG-7. This was a key element in the reactor's instability, given that boiling could suddenly accelerate very low in the core, precisely where the tip effect made itself felt. INSAG-7 and other commentators emphasize the connection of the additional main circulating pumps to explain this low subcooling. But the graph makes it clear that subcooling was low or nonexistent before either of the additional pumps was engaged. Dyatlov seems to have been justified in describing (in an article published in NEI) low subcooling as a fact of life during low power operation.

The final minutes from the previous graph are produced above, right after the reset of the DREG program. The Main Circulating Pumps inlet temperature parameter remained constant in this period, which reminds us of the fact that the level of boiling in the core is determined by pressure. Drum separators pressure can be seen to decline and then increase by a few kg/cm^2 during the rundown test. It then skyrockets as the accident sequence begins.

The graph above depicts the only direct and unambiguous violation of safety rules that occurred that night. Most of the main circulating pumps were periodically exceeding the flow rate limit of 7000 cubic meters per hour. Flow rates were supposed to be lower during low power operation to prevent cavitation of the pumps. Studies cited in INSAG-7 found that the pumps came closest to cavitation a few minutes before the rundown test began, although it never actually occurred. Nevertheless, cavitation is the preferred culprit for the accident in papers published in the 1990s by the RBMK's design bureau (NIKIET).

Why were flow rates so high? Maintaining control of these parameters was especially difficult at low power, and the MCPs often exhibited roving flow rates, with one pump 'robbing' the output from another. A former SIUB of my acquaintance speculated that Stolyarchuk set the flow rate window to 6500-6800 m^3/hr, which made fluctuations above 7000 inevitable. With a lower setpoint of 6000, occasional fluctuations below the automatic pump shutdown threshold of 5000 m^3/hr could have occurred (see the next graph below).

At 1:23:04, four of the MCPs begin 'running down' as they receive their electrical power from the coasting turbine. The steady decline of output is apparent in the graph, and there is some evidence that the four 'rundown' pumps were shut down by their emergency protection systems just as AZ-5 was triggered. However, there is no sign of interruption in coolant supply until a sudden drop around timestamp 1:23:45. Given that this is at least two seconds after the beginning of the power surge, any abrupt cessation of MCP flow would be a result of the reactor's destruction, rather than its cause.