by John Nedderman

The importance of flow patterns

30 May 2000

A coolant leak at Tsuruga revealed limitations in the design of heat exchangers, and has prompted further investigation of thermal stratification.

On 12 July 1999, a reactor coolant leak occurred inside the containment of Japan Atomic Power Company’s Tsuruga 2 reactor (1160MWe PWR). A total of 51m3 of coolant leaked over a period of 14 hours. The leak was traced to cracks in an elbow joint in the regenerative heat exchanger of the charging and volume control system (CVCS). Investigation showed that the cracks were caused by high cycle fatigue due to temperature variations resulting from thermal stratification in the outlet of one stage of the heat exchanger. The plant had been in commercial operation for about 13 years.

The regenerative heat exchanger transfers heat from the letdown flow (a flow of reactor coolant taken from the reactor to the CVCS system for purification) to the charging flow, which returns water to the reactor. The heat exchanger comprises three horizontal stages (see diagram). The shells of the heat exchanger are exposed to full reactor pressure. The letdown flow enters the shell–side of the top stage at about 290°C and is discharged from the bottom unit at about 140°C. The charging flow enters the bottom stage at about 55°C, passes through the U-tubes in each stage and leaves at about 240°C.

As the pressure is high, the designers had adopted a conservative design with a large radius at the point where the tube plate joins the shell to avoid stress concentrations, and this meant there was a large gap between the tubes and the outer shell. An inner cylinder was then placed around the tube bundle to keep the coolant in close contact with the tubes and a small bypass flow was allowed to pass through the space between the cylinder and the shell. According to the design, this flow would not be cooled by the tube bundle but would mix with the cooled water discharged from the inner cylinder before leaving through the exit nozzle.

The inner cylinder was manufactured with seven support rings, which help to locate the cylinder within the shell, and the gaps between the rings and the shell (approximately 1.5mm on each side) controlled the amount of bypass flow. In the case of the middle stage, the outlet nozzle is located at the opposite end from the tube sheet and this contributed to the failure. The water entered the inner cylinder through a hole on the side of the cylinder diametrically opposite to the inlet nozzle so that it had to pass around the outside of the cylinder before entering the tube space. Similarly, the exit hole in the cylinder was on the side opposite the exit nozzle.

Failure investigation

The leak occurred in an elbow joint immediately downstream of the exit nozzle of the middle stage. The crack was longitudinal and the opening on the outside of the elbow was 47mm long.

Investigation showed that the crack was caused by high cycle fatigue and had propagated from the inner surface. Detailed examination in a laboratory showed 11 circumferential and longitudinal cracks on the inner surface, in addition to the through–wall crack. A large number of cracks were also found on the inner surface of the shell of the middle stage of the heat exchanger. There were no cracks in other parts of the heat exchanger.


Measuring the actual gaps between the support rings and the shell and by analysing heat balances, it became clear that the bypass flow was much larger than the design value. About 40% of the total flow was bypassing the tubes and passing, essentially uncooled, to the exit nozzle. Therefore the part of the middle stage close to the exit nozzle was exposed to two flow streams: the bypass flow at about 250°C and the main flow at 170°C. The temperature difference could cause high thermal stresses. To investigate the flow paths in detail a full scale model was constructed.

The model had an outer shell of acrylic resin so that the flow paths could be seen by injecting dyes and the tests were carried out at a reduced temperature of 70°C for the bypass flow and 30°C for the main flow. This combination was expected to give the same flow patterns at atmospheric pressure as in the actual heat exchanger under operating conditions. The behaviour of the heat exchanger was also studied, using a finite element model in a computer simulation.

The inner cylinder and the tube bundle are deflected downwards due to gravity. In the case of the middle stage, this means that the gaps between the support rings and the shell in the vicinity of the exit nozzle are smaller at the bottom than at the top. The investigation showed that when the heat exchanger is put into service, all the bypass flow initially passes along the top of the shell and the water in the annular space at the bottom, stagnates. The top of the shell is therefore at bypass flow temperature and the bottom is at the temperature of the tube space (because the water is stagnant and heat is conducted through the inner cylinder). The resulting differential expansion distorts the shell which bows upwards at the centre.

As the distortion increases, the support rings near the exit nozzle move upwards so that the shell–to–support ring gaps become larger at the bottom. At some point the gap becomes large enough to allow the relatively cool water in the annular space at the bottom to be flushed out and replaced by hot bypass-flow water. When this happens the distortion of the shell reverses and the gaps decrease again and, after some time, the flow in the bottom stagnates once more. The computer simulation showed that this pattern of thermal expansion would continue in a cyclical patternwith a period of about 500 seconds. Consequently the pattern of the bypass flow also changed cyclically.

Full–scale model tests showed that when the gap between the second support ring and the shell was small, the bypass flow was in the top part of the shell. The cooler water leaving the tube space through the hole in the top of the cylinder did not mix with the bypass flow. Instead, the main flow passed down one side of the inner cylinder and the bypass flow passed down the other side. The separation of the flows continued into the nozzle and downstream elbow. As a result one side was hotter than the other.

On the other hand when the gap between the second support ring and the shell was large, the bypass flow approached the nozzle area from the bottom side. Again, the bypass flow did not mix with the main flow but in this case the flow pattern in the nozzle and downstream pipework was different, and the hotter zone was rotated through 90°.

In this way a pattern of differential thermal expansions and resultant stresses was set up which changed cyclically according to the distortion cycle of the shell. In addition, fluctuations in the flow patterns caused random temperature fluctuations of about 14°C in the elbow with a period of several seconds. It was the continuously changing stresses created by these differential temperatures that caused the growth of high-cycle fatigue cracks.


The damaged heat exchanger was replaced with a new exchanger of a different design and the plant is now back in commercial operation. Heat exchangers manufactured to this design were first used on two of Kansai Electric’s PWRs, Ohi 3 and 4, which went into service in 1991 and 1992,


The new design is much simpler as it does not have an inner cylinder around the tube bundle. Using modern computer programs it was possible to make more accurate stress analyses and this allowed the radius of the shell-to-tube sheet joint to be reduced, therefore eliminating the need for an inner cylinder.

As well as replacing the heat exchanger with a new design, JAPC will implement the following changes:

•In the past, the in-service inspections which have to be carried out on the reactor coolant system (non-destructive tests every 10 years) have not been applied to the letdown system because there are isolation valves between the two systems. In future all piping and equipment inside the containment vessel which is subject to reactor coolant system operating conditions will undergo ultrasonic crack detection tests at the same time intervals as for the reactor coolant system.

•JAPC will install an additional nine TV cameras inside the containment so that leaks can be located more easily (it took 14 hours to locate the source of the leak in this case).

•Procedures for cooling down and depressurising the reactor will be revised so that future leaks can be stopped more quickly.


The test facilities used to investigate the Tsuruga 2 leak were also used to see if similar heat exchangers on other reactors would be likely to fail in the same way. The heat exchangers on 3-loop and 2-loop PWRs are somewhat shorter in length and the designs differ in small ways. The model tests showed that these heat exchangers would not experience the same problem.

The tests also showed that cyclic flow pattern changes would not have occurred on Tsuruga 2 if the bypass flow had been less than 30%.

Thermal stratification problems are unusual and difficult to predict. The Japan Society of Mechanical Engineers (JSME), the reactor vendors and the utilities, together with CRIEPI (the Central Research Institute of the Electric Power Industry), will work under the leadership of Professor Madarame of Tokyo University to investigate other cases by collecting data from similar failures and carrying out model tests. Similar thermal stratification problems have been encountered elsewhere in Japan (due to valve seat leakage at Genkai, for example). Overseas the problem has been observed in several places, including on the RHR system of a French reactor at Civaux and at the Farley reactor in the US.

The group hopes to report back to MITI and draw up design engineering standards in three to five years.

Privacy Policy
We have updated our privacy policy. In the latest update it explains what cookies are and how we use them on our site. To learn more about cookies and their benefits, please view our privacy policy. Please be aware that parts of this site will not function correctly if you disable cookies. By continuing to use this site, you consent to our use of cookies in accordance with our privacy policy unless you have disabled them.