A Study of Intermittent Buoyancy Induced Flow Phenomena in CANDU Fuel Channels

A Study of Intermittent Buoyancy Induced Flow Phenomena in CANDU Fuel Channels by Zheko Petrov Karchev A thesis submitted in conformity with the requirements for the degree of Master of Applied Science Department of Chemical Engineering and Applied Chemistry University of Toronto Copyright by Zheko Petrov Karchev 2009

A Study of Intermittent Buoyancy Induced Flow Phenomena in CANDU Fuel Channels Zheko Petrov Karchev Abstract Master of Applied Science Department of Chemical Engineering and Applied Chemistry University of Toronto 2009 The present work focuses on the study of two-phase flow behavior called Intermittent Buoyancy Induced Flow (IBIF) resulting from the loss of coolant circulation in a CANDU nuclear reactor core. The main objectives are to study steam bubble formation and migration through the pressure tube and into the feeder tubes and headers, and to study the effect of pressure tube sagging on the two-phase flow behavior during IBIF. Experiments are conducted using air and water flow at atmospheric pressure to qualitatively examine the IBIF phenomena. The test showed oscillating periodic behavior in the void fraction as the air vents. In addition to this a mathematical model based on a simplified momentum balance for the liquid and gas phases was formulated. The model was further solved and compared to the experimental data. The model predictions showed a reasonable agreement within the investigated range of void fractions. ii

Acknowledgments I wish to express my appreciation to Professor Masahiro Kawaji for his support, professional guidance and excellent supervision throughout the course of this work. I would like to acknowledge the financial support of Ontario Power Generation, Bruce Power Inc. and Atomic Energy of Canada. I would also like to express my gratitude to Mr. Laurence Leung, Mr. Muhammad Ali and Mr. Marc Kwee for their valuable input on this work iii

Table of Contents Abstract ii Acknowledgements iii Table of Contents iv List of Figures vi List of Tables viii Nomenclature viii List of Appendices viii 1. Introduction 1 1.1 Nuclear Power Plant - Overall View 1 1.2 CANDU Nuclear Reactor Core 2 1.3 Project Objectives and Scope 3 2. Literature Review 7 2.1 Overview of Two Phase Flow Models 7 2.2 Bubble Formation and Propagation 9 2.3 Void Fraction Measurements 10 2.4 CANDU Reactor behaviour in case of circulation outage 11 3. Experimental Design 12 3.1 Facility Design 12 3.2 Experimental Set-Up 15 3.2.1 Overall View 15 3.2.2 Pressure Tube and Fuel Bundles Design 17 4. Experimental Results and Discussion 19 4.1. Preliminary Observations 19 4.2. Effect of Air Injection Nozzle Location 21 4.3. Combined Effect of Feeder Water Level and Air-Injection Rate on the Venting Time 25 4.4. Effect of Multiple Air Injection on the Venting Time 27 4.4.1. Air Injections through Two Air-Injection Nozzles 28 4.4.2. Air Injections through Three Air-Injection Nozzles 29 iv

4.5. Effect of Pressure Tube Sagging on the Venting Time 30 4.6. Air-Lift Effect 32 4.7. Oscillatory Behavior 35 4.8 Summary of Experimental Results 39 5. Mathematical Model 40 5.1. Mathematical Description 40 5.2. Numerical solution of the mathematical model 44 5.2.1. Calculating the basic geometric parameters 44 5.2.2. Numerical Solution of the model 51 5.3. Calculation Results 53 5.4 Summary of Model Development 57 6. Conclusions 58 References 60 Appendices 63 v

List of Figures Fig. 1-1 Nuclear power plant overall diagram Fig. 1-2 CANDU nuclear reactor core Fig. 1-3 Decay power 6 hours to 29 days after reactor shutdown Fig. 3-1 Experimental set-up block diagram Fig. 3-2 Experimental set-up overall view Fig. 3-3 CANDU nuclear reactor steam supply system Fig. 3-4 Photograph of a CANDU pressure tube with a fuel bundle placed inside Fig. 3-5 Photograph of the simulated fuel bundle used in the current design Fig. 4-1 Schematics of the experimental set-up Fig. 4-2 Photograph of the system behavior upon air injection Fig. 4-3 Photograph of the slug rising in the vertical feeder Fig. 4-4 Schematics of the experimental set-up effect of the air injection location Fig. 4-5 Consecutive photographs of the bubble propagation front (65 ms time interval between the frames) Fig. 4-6 Effect of air injection location at different simulated power level (SPL) Fig. 4-7 Schematic of the experimental set-up to investigate the combined effect of feeder water level and air-injection rate Fig. 4-8 Venting time as a function of water level in the feeder line at different simulated power levels Fig. 4-9 A typical CANDU reactor pressure tube axial heat flux distribution Fig. 4-10 Schematic of the experimental set-up for air injection through two air- nozzles Fig. 4-11 Schematic of the experimental set-up air injections through three air- nozzles Fig. 4-12 Schematic of the experimental set-up to study the effect of pressure tube sagging Fig. 4-13 Photograph of the inclined pressure tube (sagging of 5.08 cm (2 ) in the mid section Fig. 4-14 Effect of pressure tube sagging on the venting time Fig. 4-15 Schematic of the experimental set-up for studying the air-lift effect Fig. 4-16 Air-lift effect on flow velocity Fig. 4-17 Effect of water level on the frequency of oscillations (SPL 1.1 kw) vi

Fig. 4-18 Effect of the air injection rate (SPL) on the frequency of oscillations (Feeder Line Water Level 230 cm) Fig. 5-1 Diagram of the modeled two-phase system Fig. 5-2 Diagram of the gas liquid interface Frig. 5-3 Schematics of the simulated pressure tube with 37 acrylic rods Fig. 5-4 Schematics of the gas-liquid interface Fig. 5-5 Schematics of the gas-liquid interface with the rods placed inside Fig. 5-6 Wall wetted perimeter schematics Fig. 5-7 Schematics of the wetted perimeter calculation with rods placed inside Fig. 5-8 Newton-Raphson method - graphical representation Fig. 5-9 Variations of the liquid-solid and gas-solid interface lengths as a function of void fraction Fig. 5-10 Variation in the gas-liquid interface length as a function of the void fraction Fig. 5-11 Variations of the liquid and gas-solid cross sectional areas as a function of the void fraction Fig. 5-12 Schematic of the experimental set-up for model validation Fig. 5-13 Variation of the liquid phase velocity with the void fraction in the pressure tube Fig. 5-14 Comparison between the predicted and the calculated liquid phase velocity vii

List of Tables Table 3-1 Comparison of test section and CANDU reactor component dimensions Table 4-1 Effect of the air injection location on the venting time Table 4-2 Venting time data for air injections through two air- nozzles Table 4-3 Venting time for air injections through three air- nozzles Table 4-4 Effect of the pressure tube sagging Table 4-5 Air-lift effect on flow velocity (m/s) Table 4-6 Air-lift effect on header water level Table 4-7 Effect of water level on the frequency of oscillations (SPL 1.1 kw) Table 4-8 Effect of air injection rate (SPL) on the frequency of oscillations Table 4-9 Effect of air injection rate (SPL) on the frequency of oscillations (Ten fold decrease in the header tank volume) Nomenclature IBIF SPL Intermittent Buoyancy Induced Flow Simulated Power Level List of Appendices Appendix 1: Numerical Code for Interfacial Area Calculation Appendix 2: Numerical Code for Liquid Velocity Calculation Appendix 3: User Input Function viii

1 1. Introduction 1.1. Nuclear Power Plant - Overall View CANDU PHWR is an essential part of Ontario s Power system. Twenty reactors of this type generate about 15,000 MW of electricity. The CANDU nuclear technology combines cost efficiency, low capital costs and a design which has proven its reliability and safety for more than 30 years. A nuclear power plant usually includes three basic compartments illustrated in Fig 1-1: Vacuum building, reactor building and a building housing the steam turbine and the electrical generator. The heart of the plant is the nuclear reactor in which the heat generated as a result of the fission reaction is used to produce high pressure steam which is transferred to the steam turbine. The turbine transforms the energy carried by the steam into mechanical energy which is subsequently transferred to the electrical generator and converted into electrical energy. As a safety measure the reactor building is connected through a large diameter pipe to the Vacuum Building. The purpose of this structure is to ensure fast seam condensation in case of Loss of Coolant Accidents (LOCA) in the reactor building. In order to prevent even small releases of radioactive materials from the nuclear power plant, all the air released into the atmosphere is filtered through a filter system. Reactor Building Secondary Loop Boiler Steam Turbine Primary Loop Vacuum Building Filter System Reactor Generator Fig 1-1. Nuclear power plant overall diagram Cooling Water from Lake

2 1.2 CANDU Nuclear Reactor Core The CANDU Reactor uses two independent water loops for removing the heat from the nuclear core. The primary loop is indicated in Fig. 1-2 with yellow color and uses heavy water as a heat transfer fluid. The heavy water flows through the pressure tubes (#10) in which the fuel bundles (#1) are placed. The heat released as a result of nuclear fission reactions is transferred to the water. The water in the primary loop is kept under high pressure which allows it to be heated to higher temperature and more heat to be removed from the core. The water temperature in the primary loop is lower than the boiling temperature which means that there is a single-phase flow through the core. The heavy water flows through the pressure tubes, gets heated and reaches the steam generator (#5). In the steam generator the hot heavy water is used to heat up light water in the secondary loop (#12) and generate steam (#11) which is fed to the steam turbine. 14 1. Fuel Bundles; 2. Calandria; 3. Control Rods; 4. Pressurizer; 5. Steam Generator 6. Light Water Pump; 7. Heavy Water Pump; 8. Fuel Loading Machine; 9. Moderator 10. Pressure Tube; 11.High Pressure Steam (to Steam Turbine); 12. Water Condensate (from Condenser); 13.Reactor Containment Building; 14 Primary Loop Fig 1-2. CANDU nuclear reactor core (source: http://en.wikipedia.org/wiki/candu)

3 All of the pressure tubes are placed inside the Calandria (#2) containing heavy water which is used as a moderator. The control rods (#3) are 28 cadmium rods which serve as an emergency shutdown system. In case of an accident they are submerged into to the heavy water moderator in the Calandria by a gravitational drop. This is possible since the moderator is kept at low pressure. As a backup to the primary shutdown system the CANDU reactor uses a second system, which injects gadolinium through 6 nozzles into the Calandria. The gadolinium with a large neutron absorption cross section acts as a neutron poison and can rapidly terminate the fission reaction. One of the main design characteristics of the CANDU reactor is the use of a horizontal core containing many small diameter pressure tubes (about 4 ID) in which the uranium fuel bundles are placed. This allows on-line re-fueling of the reactor at full power. A refueling machine (#8) attached to both ends of the pressure tube push in a new fuel bundle at one end and removes an old bundle at the other end. In contrast, light water reactors which are more popular in other countries, must be shut down for re-fueling purposes. The horizontal pressure tubes are typically 9 m long and can sag in the middle after many years of service. The horizontal orientation of the pressure tubes, fuel bundles and sagging phenomenon are closely related to the objectives and scope of the present project as discussed below. 1.3 Project Objectives and Scope The complete understanding of the thermal-hydraulic phenomena taking place in the nuclear reactor core is a requirement for the safe operation of all nuclear reactors. The current investigation is focused on the thermal-hydraulics of a CANDU reactor under loss of coolant circulation conditions. In such an event, the reactor is shut down and the pumps circulating the heavy water coolant through the primary loop will cease their operation. Consequently, the coolant will become stagnant inside the pressure tubes for an extended period of time, while the fuel rods continue generating a varying amount of heat due to decay heat.

4 The reactor has several safety systems which are designed to terminate the fission reaction shortly after any accident occurs. This leads to a drastic decay in the power level inside the core. For example, the decay power 6 hours after the CANDU reactor with 37- element fuel assemblies is shut down is just 0.794% of the full power. The heat generation continues inside the fuel long after the shut-down although the power level is significantly low (0.105% of full power 29 days after the shut-down) as shown in Fig. 1-3. Power Level per Chanel, kw 70.00 60.00 50.00 40.00 30.00 20.00 10.00 0.00 0.00 5.00 10.00 15.00 20.00 25.00 30.00 35.00 Days after shutdown Fig. 1-3 Decay power from 6 hours to 29 days after reactor shutdown (Ref. Report # N- REP-03500.2-10002, page 61, Darlington NGS) In case of a loss of coolant circulation event, the pressure in the primary loop is rapidly reduced so that an over-pressurization can be avoided. Under these conditions it is expected that the stagnant water boils in the pressure tube. The main safety concern in this case is the uncovery of fuel rods due to the formation of large vapor bubbles, and the time interval between the vapor bubbles formation and their venting through the feeder tubes located at both ends of the pressure tube. Venting is important since the bubbles rising in feeder tubes induce two phase flow inside the feeder tubes and consequently the

5 pressure tube which facilitates the heat removal from the fuel bundles. This two phase flow resulting from a loss of coolant circulation accident is referred to as Intermittent Buoyancy Induced Flow (IBIF). The formation of steam bubbles inside the pressure tubes under normal operation is not desirable since it can lead to an increase in the nuclear fission rate. This phenomenon is referred to as positive void reactivity and it results in an increase in the reactor power level. This property of the CANDU reactor is significantly different from the Light Water Reactors which have a negative void reactivity and the fission rate decreases with the void volume. This positive void coefficient of CANDU reactors is not important as long as the emergency shutdown systems (Cadmium control rods and Gadolinium Injection) are activated. After long term operation, the horizontal pressure tubes inside the CANDU reactor core tend to sag slightly due to the thermal and mechanical stresses to which they are subjected. The degree of sagging is at maximum about 5.0 cm (2 inches) measured at the centre of the pressure tube. This phenomenon is expected to enhance the venting of the steam bubbles formed in the pressure tubes as a result of boiling induced by the loss of coolant circulation event. This assumption is based on experimental studies of two phase flow behavior according to which the bubbles remain stagnant in horizontal tubes and also the bubble rise velocity in inclined pipes generally exceeds the value of this parameter in vertical pipes [10]. Despite all this, the effect of pressure tube sagging has not been experimentally studied. The present objectives and scope of the current project can be summarized as follows: Design and construct an experimental facility which nearly exactly duplicates the geometry of a CANDU reactor pressure tube together with the thirteen replicas of 37- element fuel bundles placed inside. Study of bubble formation and migration throughout the system by injecting air bubbles into the rod bundles. Study of the effect of pressure tube sagging on the IBIF phenomena and two-phase flow behavior in feeder tubes.

6 Study on the possible existence of buoyancy-induced two phase flow circulation inside the system.

7 2. Literature Review The two phase flow behavior has been an object of extensive theoretical and experimental studies in energy systems. The reason behind this is the fact that this type of flow occurs in wide industrial applications including nuclear reactors, boilers, oil wells and pipelines, etc. Despite the large amount of work done on studying the behavior of the two- and multi phase flow systems there is still some uncertainty associated with them. 2.1 Overview of Two-Phase Flow Models There are different mathematical descriptions proposed for two-phase flow systems. Some of them are based on theoretical considerations and others are derived from experimental investigations. Due to a high level of complexity arising from the presence of the two phases the practical models apply different simplifications which result in certain models applicable only to specific flow patterns. The simplest model used for describing a two-phase flow is the so called homogeneous flow model. It treats the two-phase system as a homogeneous mixture with averaged properties [1]. This model leads to relatively simple equations but it is applicable to a limited number of systems in which the homogeneous equilibrium flow assumption can be justified. The model predictions tend to strongly deviate from the experimental data as the void fraction increases [2]. A little more sophisticated two-phase flow model is the drift flux model based on the concept of a drift flux which uses a reference plane moving with a given velocity. The flow of each phase is referred to this plane. This significantly simplifies the analysis and allows a much wider range of flow patterns to be covered. The drift flux model shows a good agreement with the experimental data for co-current and counter-current flow as well as for stratified flow [2]. Taitel and Duler [3] combined the momentum balance equations for the liquid and the gas phases into a single expression. The model accounts for the interfacial and fluid to

8 wall momentum interactions as well as for the effect of flow channel inclination. It is especially well suited for the stratified flow pattern. As a further extension of the stratified flow model, Sadatomi et al. [4] included the interfacial level gradient into the momentum equations for both phases. The model showed good agreement with the experimental data for some specific applications involving stratified flow in large diameter horizontal pipes. In addition to the more fundamental mathematical descriptions discussed above, some simplified one-dimensional two-phase flow models applicable to a specific flow pattern have also been formulated. One such model applicable to the bubbly flow was proposed by Wallis [1]. The model can be applied to a wide range of void fractions. Another model of this type is the one proposed by Nicklin et al. [5] and applicable to the slug flow. The major limitation of the models discussed above is the fact that their application is restricted to a given flow pattern. In order to cover a wider range of two-phase flow conditions, Lockhart and Martinelli [6] proposed an empirical correlation relating the void fraction to a parameter called the Lockhart-Martinelli parameter. This parameter is defined as the square root of the ratio of the friction pressure drop of liquid to that of the gas under the assumption that each phase flows alone in the channel. The Lockhart- Martinelli approach has been widely applied to a large spectrum of two-phase systems such as air-lift pumps [7, 8], two-phase heat exchangers, etc.

9 2.2 Bubble Formation and Propagation Bubble formation and propagation problems have been investigated extensively. A pioneering study in this field was performed by Bretherthon [9] who investigated the propagation of a long axisymmetric isothermal bubble of inviscid gas through a viscous liquid. A solution was obtained under the assumption of strong surface tension effects and it showed that the bubble dynamics is dominated by two capillary-statics regions located at both ends of the bubble. The analysis subsequently translates into a transitional region in which the surface tension forces are balanced by the viscous ones. Finally a third region is defined in which the viscous forces are dominant and this results in a thin liquid film surrounding the gas bubble [9]. Wilson et al. [10] expanded Bretherton s analysis focusing on the unsteady mass-transfer expansion and contraction of a twodimensional vapor bubble located in-between subcooled or superheated plates. In a subsequent study Kenning et al. [11] investigated the growth of a bubble in a capillary tube under superheated conditions and proposed a mathematical model describing the bubble generation and propagation under these conditions. A quite comprehensive study on the growth and departure of bubbles from a submerged needle was performed by Hassan et al. [12]. The study employed a simplified model for the bubble growth based on the Rayleigh-Plesset equation. The results showed the existence of two separate bubble growth regimes which depend on the rate at which the gas flow is injected into the bubble. In the case of high velocity gas injection the bubbles formed at the injection nozzle tend to elongate in the direction of the gas flow due to the axial momentum of the gas. Under these conditions the bubble growth can be modeled well enough by an ellipsoid expanding upwards [13]. Yuan et al. [14] defined a model for the vapor bubble growth and collapse in a small channel connecting two reservoirs. The model was based on the potential flow theory and did not account for the initial variations in the internal bubble pressure assuming that the bubble dynamics is inertia controlled. The bubble growth in a confined space causes a displacement of the surrounding liquid phase. This behavior can result in unidirectional liquid transport upon careful timing of the bubble generation and collapse. Yuan at al.

10 [15] further investigated the possibility of existence of a net pumping action under these conditions. The study showed that under certain system parameters it is possible for the liquid transport to be initiated. Ory et al. [16] performed a similar study defining an analytical model for the bubble growth and collapse in a narrow tube which showed a good agreement with the experimental result. As a further investigation Alira et al. [17] proposed a one-dimensional model incorporating the coupled heat transfer and phase change phenomena in a narrow channel. The mathematical description was divided into several time periods, covering the bubble generation, growth, collapse and channel refill and provided an insight into the effect of the fluid viscosity on the system behavior. 2.3 Void Fraction Measurements The complex nature of the two phase flows introduces significant difficulties in the experimental studies. The main parameters under investigation in these systems are pressure drop, heat transfer coefficient, mass-transfer coefficient and void fraction. The accurate measurement of the void fraction is especially important. The fraction of the flow channel occupied by the gas or vapor phase is related to the flow regime of the twophase system and enters directly into the gravitational and acceleration terms of the pressure drop calculations. In nuclear reactor engineering calculations the void fraction is a major design parameter due to the fact that it affects the neutron absorption rate. The most widely used method for void fraction measurements is based on the attenuation of high energy electromagnetic waves (gamma or X-rays) passing through the flow channel. The main difficulty associated with this method arises from the need to handle radiation. In addition to this there is a significant error resulting from the gas-liquid interface orientation, the effect of the tube wall, the effect of the temporal fluctuations, etc. [18]. Another class of techniques employed in the void fraction measurements is based on the fact that the electrical conductance and capacitance of the fluid depend on the concentration of the phases. These methods are known as impedance methods. In order for the readings to be accurate it is required for the electrical field between the electrodes

11 to be homogeneous, the electrodes to be located close enough and there should be no discontinuity in the channel cross section [18]. The void fraction can also be measured by mounting fast acting valves at the entrance and the exit of the flow channel of interest. By simultaneously closing both valves one can easily measure the fraction of the volume occupied by each phase. The method is precise but it is not applicable to online measurements [18]. In addition to the above mentioned techniques there are other methods such as acoustic techniques, electro-magnetic measurements, optical methods, etc. [17]. 2.4 CANDU Reactor Behaviour in Case of Circulation Outage There is not much information available in the literature regarding CANDU reactors under a loss of the forced circulation of the coolant through the core. Due to the complexities of the system involved and the phenomena to be studied as well as the fact that they are specific to the CANDU design only a limited number of papers have been focused on this problem. Previous experimental studies and analytical investigations showed that under a loss of coolant circulation there are three modes of core cooling that can be induced singlephase thermo-siphoning, two-phase thermo-siphoning and intermittent buoyancy induced flow (IBIF) [19, 20]. Feyginberg et al. proposed a lumped parameter model based on a transient energy balance for a single channel in a CANDU reactor core. The model divided the heat transfer in each channel into five time periods including preheating, local saturation, venting and channel refill. Subsequently each stage was described in terms of a transient heat balance. In order to validate the model the researchers used an experimental setup in which a 6 m long channel was heated up and IBIF was induced. The model showed good agreement with the basic trend observed during the

12 experimental studies but the predicted results were generally higher than the values obtained from the experiment [20]. 3. Experimental Design 3.1 Facility Design The phenomenon under investigation is referred to as the Intermittent Buoyancy Induced Flow (IBIF). In order to investigate this phenomenon resulting from a loss of coolant circulation in a CANDU reactor core, an experimental set up was designed and constructed. The experimental set-up simulates a CANDU reactor pressure tube as schematically shown in Fig. 3-1. The boiling phenomenon was simulated by injecting air bubbles into the stagnant water inside the pressure tube. The air was distributed into three parallel lines and injected into the pressure tube through 12 nozzles mounted throughout the length of the pressure tube. In order to measure the air flow rate, one air velocity meter was mounted on each air line (#4). The velocity meters used in the design were high precision ones with integrated temperature and pressure correction. This meter allows the air flow rate to be measured precisely. In addition, there were 12 measuring needle valves (#8) with a Vernier handle mounted on each air injection nozzle. The power density in the actual CANDU pressure tube is not uniform throughout the whole length. By using the needle valves the air injection at each nozzle location could be precisely adjusted to closely simulate the real power distribution. A differential pressure transducer (DP) with a range of 6.87 kpa (1.0-psid) was installed on each of the two vertical feeder pipes to measure the collapsed water level in each pipe. Based on that, the instantaneous void fraction in the feeder pipes could be calculated. In addition to this, two pressure transducers were added to both feeders, allowing the changes in the water level inside the water tanks at the top of the feeder pipes to be detected. This would allow the amount of water dragged by the air flow during its venting to be measured.

13 Other instrumentation included three 0-25.4 cm (0-10 inch) water level differential pressure (DP) transducers mounted at different locations throughout the length of the pressure tube. They allowed the horizontal liquid level distribution and respectively the void fraction inside the pressure tube to be measured. In addition, by comparing the signals of two adjacent differential pressure transducers the propagation velocity of the waves formed inside the pressure tube could be measured. There were 8 pairs of ball valves mounted on the pressure tube which allowed the differential pressure transducers to be used at different locations. The two water tanks were connected through a pipe connection (#5). This allowed the possible existence of a continuous IBIF to be studied. Through the valve (#6), the connection between the two tanks could be interrupted which permitted the investigation of different system configurations.

14 1 5 6 1 2 2 D D P 3 7 9 Fe D D D 4 4 8 4 F F F 1. Water Tank 2. Feeder Tube 3. Pressure Tube 4. Air Distribution Line 5. Water Tank Pipe Connection 6. Valve T DP Differential Pressure Transducer PT Pressure Transducer F Air Flow-Meter T - Thermocouple Air Flow Fig. 3-1 Experimental set-up block diagram

15 3.2 Experimental Set-Up 3.2.1 Overall View The test section was a close simulation of a real CANDU pressure tube. It was 9.0 meters long made of 10.0 cm (4-inch) ID acrylic resin tubing. It consisted of six sections connected to each other by flanges. Two 2.2-m long and 5.08 cm (2.0-inch) ID vertical pipes were attached at the end sections of the pressure tube to simulate the feeder pipes. Each feeder pipe was connected to a rectangular water tank with a 0.3 m 3 volume, simulating the header tank in the CANDU reactor. Ball Valve Header Tanks Feeder Pipe Connection Air Injection Feeder Fig. 3-2 Experimental set-up overall view In the CANDU reactor two pressure tubes work in parallel forming a closed loop for the heavy water coolant circulating between the reactor core and the steam generators. Under these conditions there is a possibility for the existence of a continuous flow throughout the loop in case of boiling inside the reactor core resulting from the loss of coolant

16 circulation. In order to investigate this both tanks were connected with a pipe with a valve in-between. This connection represented a simplified model of the second pressure tube working in parallel with the one being studied. This would allow us to study the possible existence of the above mentioned phenomenon. Light Water Condensate Steam Generators Primary Pumps Light Water Steam Heavy Water Coolant Heavy Water Moderator Fuel Channel Assembly Calandria Fig. 3-3 CANDU nuclear reactor steam supply system [21]

17 3.2.2 Pressure Tube and Fuel Bundles Design The dimensions of the simulated pressure tube and the fuel bundles were selected as close as possible to those of the real pressure tube. An acrylic resin was selected as the main material of construction since its transparency would allow us to directly view and record the two-phase flow phenomena taking place inside the system. A comparison between the actual pressure tube used in a typical CANDU reactor and the one used in the current design is presented in Figure 3-4. Fuel Rod Zircaloy end plate a) Real pressure tube b) Simulated pressure tube Fig. 3-4 Photograph of a CANDU pressure tube with a fuel bundle placed inside (Ref. Report # N-REP-03500.2-10002, page 61, Darlington NGS) Thirteen simulated fuel bundles were fabricated from acrylic rods and placed inside the pressure tube. Each bundle was composed of 37 acrylic rods with 12.7-mm OD simulating a 37-Fuel Rod Assembly. The acrylic rods were attached to actual zircaloy end plates provided by Bruce Power Inc. A photograph of the fabricated fuel bundle is shown in Fig. 3-5.

18 Fig. 3-5 Photographs of the simulated fuel bundle used in the current design The dimensions of the simulated pressure tube and fuel rods are summarized in Table 3-1 and compared to those of the CANDU reactor. As it can be seen from this Table the dimensions of the simulated pressure tube and fuel bundles were chosen as close as possible to the real ones in order to more accurately simulate the IBIF phenomena in the real system. Table 3-1. Comparison of test section and CANDU reactor component dimensions Simulated Actual Pressure Tube ID, mm 101.6 102.4 Fuel Bundle OD, mm 87.9 99.7 Fuel Rod OD, mm 12.7 13.08

19 4. Experimental Results and Discussion 4.1. Preliminary Observations A schematic of the flow phenomena studied is shown in Fig. 4-1. The approach employed in the current experimental study was to investigate the effect of different process variables on the two-phase flow behavior. The main parameters that were varied included the air-injection rate which corresponded to the simulated reactor power level, the water level inside the feeder pipes and the depth of sagging of the pressure tube in the middle from a horizontal position. A system of pressure transducers and video cameras were used to collect the data. This information was subsequently processed and analyzed to present the results. Water Tank Water Tank Air Injection Fig. 4-1 Schematics of the experimental set-up Pressure Tube Preliminary studies showed that the air bubbles injected into the pressure tube rapidly rose towards the top, merged together and formed a continuous layer rather than remaining as discrete bubbles. The observed behavior is illustrated in Fig. 4-1 and shown in Fig. 4-2. Air then flowed along the pressure tube towards the end section and vented into the feeder pipes forming a slug flow in the feeder pipes (Fig. 4-3).

20 a) bubble merging b) air-layer formation Fig. 4-2 Photographs of the system behavior upon air injection Fig. 4-3 Photograph of a gas slug rising in the vertical feeder

21 4.2. Effect of Air Injection Nozzle Location The goal of this series of experiments was to investigate the effect of air injection location on the gas venting time. Air was injected at different locations throughout the length of the pressure tube (Fig. 4-4) and the time interval between the start of the air injection and air venting was measured. In Figure 4-5 are shown two photographs of the bubble propagation front taken at a 65 ms time interval between the frames. In order for the results to be consistent the air injection rate as well as the water levels inside the water tanks were kept constant for each run. The only parameter which was varied in different runs was the location of the air injection. Each measurement presented is an average of five separate runs which allows the error to be minimized. The experiments were performed at three different simulated power levels and the results are summarized in Table 4-1 and Fig. 4-6. Water Tank Venting Distance Feeder Tube Air Injection Pressure Tube Fig. 4-4 Schematics of the experimental set-up effect of the air injection location

22 Bubble Front Bubble Front Fig. 4-5 Consecutive photographs of the bubble propagation front (65 ms time interval between the frames) Simulated Power Level(SPL), kw 1.1 1.5 2.0 Table 4-1. Effect of the air injection location on the venting time Average Venting Time, s Venting Distance, cm Bubble Expansion Velocity, cm/s 5.55 345 62.16 4.06 258 63.5 2.46 150 60.97 5.37 345 64.25 4.03 258 64.02 2.38 150 63.03 5.11 345 67.51 3.92 258 65.82 2.24 150 66.96 Average Bubble Expansion Velocity, cm/s 62.2 63.8 66.8

23 Bubble Expansion Velocity, cm/s 80 70 60 50 40 100 150 200 250 300 350 400 Venting Distance, cm Average Velocity Velocity Measured Series1 Velocity Series3 Series4 a) SPL 1.1 kw Bubble Expansion Velocity, cm/s 80 70 60 50 40 100 150 200 250 300 350 400 Venting Distance, cm b) SPL 1.5 kw Average Velocity Measured Series1 Velocity Series3 Series4

24 Bubble Expansion Velocity, cm/s 80 70 60 50 40 100 150 200 250 300 350 400 Average Velocity AverageVelocity Measured Series1 Velocity Series3 Series4 Venting Distance, cm c) SPL 2.0 kw Fig. 4-6 Effect of air injection location on bubble expansion velocity for different Simulated Power Levels (SPL) The observed system behavior showed that the air bubbles introduced into the pressure tube rapidly rose towards the top and formed a continuous layer rather than remaining as discrete bubbles. Air then flowed along the pressure tube towards the end section and vented into the feeder pipes forming a slug flow in the feeder pipes. The current results showed only slight variations of the bubble expansion velocity as a function of the airinjection location. The experiments were performed at three different injection rates and the results were consistent. Therefore, we can consider the bubble front to move with a constant velocity.

25 4.3. Combined Effect of Feeder Water Level and Air- Injection Rate on the Venting Time The goal of this series of experiments was to investigate how the water level inside the feeder pipes and the rate of air injection would affect the venting time. The experiments were performed at three different simulated power levels and at nine different water levels in the feeder pipes. Following the results from the previous experiments which showed that the bubble would not accelerate along the pressure tube, all the tests were performed by injecting air through a single injection point. The venting distance was kept constant and equal to 345 cm (Fig. 4-7). The total number of runs was 135 and each data point is an average of five tests performed under the same conditions. The results are presented graphically in Fig. 4-8. Water Tank Feeder Tube 345 cm Water Level Air Injection Pressure Tube Fig. 4-7 Schematic of the experimental set-up to investigate the combined effect of feeder water level and air-injection rate

26 Average Venting Time, s 6.5 6 5.5 5 4.5 4 3.5 45 95 145 195 245 Feeder Line Water Level, cm SPL - 1.1 kw SPL - 2 kw SPL - 2.85 kw Fig. 4-8 Venting Time as a function of water level in the feeder line at different simulated power levels As expected, the experimental results showed an increase in the venting time with an increase in the water level inside the feeder pipes and with the decreasing simulated power level, i.e. the air injection rate. The experimental data were subjected to a multivariable regression analysis in order to derive a correlation between the parameters under investigation. T = 3.54-0.177*SPL + 0.011*WL (4-1) where T = venting time [s] SPL = simulated power level at the point of injection [kw] WL = water level in the feeder line [cm] The maximum absolute deviation of the data from the correlation was MaxErr = 0.68 [s]. By taking into account the venting distance we can express the bubble expansion velocity (BEV) as a function of the SPL and WL as follows. BEV = 0.10-1.95*SPL + 30.4*WL (4-2) where BEV = bubble expansion velocity [cm/s]

27 4.4. Effect of Multiple Air Injection on the Venting Time As it was noted earlier, the power density in the real pressure tube is not uniformly distributed throughout its length. The heat generation rate is higher in the central sections of the pressure tube and lower at the ends. A typical axial heat flux distribution in a CANDU Reactor pressure tube is presented in Fig. 4-9. Relative Heat Flux, local/average Relative Location, x/l Fig. 4-9 A typical CANDU reactor pressure tube axial heat flux distribution [21] Under these conditions it is expected that in the event of a loss of coolant circulation, the steam generation would initially occur at the centre of the pressure tube and subsequently in the other sections. The goal of the current experiment was to investigate how the venting time would be influenced if the steam generation occurs simultaneously at several different locations. The experiments were conducted with two or three simultaneous air injections. In each case the total air injection rate was kept the same and equal to SPL = 1.1 kw and the water level in the feeder pipes equal to WL = 234 cm. The results could be compared with those of the single injection tests. For each case twenty runs were conducted at three different locations of the air injection points. The data were analyzed to determine how the relative distance between the injection points would affect the venting time.

28 4.4.1. Air Injection through Two Air-Injection Nozzles The two air-injection locations used are shown in Fig. 4-10. The first air-injection location was fixed at the middle of the pressure tube and the second air-injection was varied. Water Tank Feeder Tube 345cm WL= 234cm Pressure Tube Air Injection 1 Variable Distance Air Injection 2 Fig. 4-10 Schematic of the experimental set-up for air injection through two air- nozzles The results shown in Table 4-2 indicate that the simultaneous air injection leads to a decrease in the venting time. The decrease is more significant when air is injected through nozzles which are located close to each other. One explanation for the observed behavior can be the fact that the bubbles formed above the two injection points interacted with each other. This interaction restricted the bubble expansion towards each other and the bubbles predominantly expanded towards the feeder tubes. This behavior is expected to enhance the pressure tube venting since in the real reactor the steam bubbles will be first formed predominantly in the centre of the pressure tube.

29 Table 4-2 Venting time data for air injections through two air- nozzles Venting Distance, cm Venting Time (dual injection), s Venting Time (single injection), s 307 4.6 5.41-14.9 258 3.95 4.55-13.2 200 3.06 3.52-13.1 150 2.39 2.64-9.5 Change in the Venting Time, % 4.4.2. Air injections through Three Air-Injection Nozzles An experimental schematic is shown in Fig. 4-11 and the results are summarized in Table 4-3. The same venting distance from the feeder pipes was used for the second and third air-injection locations. Water Tank Feeder Tube 345cm WL= 234cm Pressure Tube Variable Distance Variable Distance Air Injection 2 Air Injection 3 Air Injection 1 Fig. 4-11 Schematic of the experimental set-up air injections through three air- nozzles Table 4-3 Venting time for air injections through three air- nozzles Venting Distance, cm Venting Time (triple injection), s Venting Time (single injection), s 307 4.37 5.41-19.2 258 3.86 4.55-15.2 200 3.01 3.52-14.5 150 2.31 2.64-12.5 Decrease in the Venting Time, %

30 The experimental results show that the simultaneous injection through three nozzles further reduces the venting time. Similar to the two air-injection location experiments, with an increase in the relative distance between the injection points the effect of the multiple air injection gets weaker. It can be expected that the addition of more injection points will lead to a further decrease in the venting time which is expected to enhance the pressure tube venting in the CANDU reactor core due to the significant number of nucleation sites existing on the fuel rods inside the pressure tube during the loss of coolant circulation event. 4.5. Effect of Pressure Tube Sagging on the Venting Time The goal of this experiment series was to investigate how the pressure tube sagging would affect the venting time. In order to study this phenomenon we gradually lowered the supports of the pressure tube at the centre as shown in Figs. 4-12 and 4-13. As a result, it was possible to achieve smooth sagging of the tube in the middle between the two ends. This configuration is considered to be very close to the one in the CANDU reactor. The experiments were conducted by injecting air through a single nozzle keeping the water level in both feeder pipes constant and equal to 234 cm. The air injection rate was also kept constant and equal to an equivalent of 1.1 kw of Simulated Power Level. The experimental results are summarized in Table 4-4. Each data point is an average of five runs performed under the same conditions. Water Tank Sagging Distance Feeder Tube Air Injection 345cm Inclination Angle WL= 234cm Pressure Tube Fig. 4-12 Schematic of the experimental set-up to study the effect of pressure tube sagging

31 Horizontal Line Inclined Pressure Tube Fig. 4-13 Photograph of the inclined pressure tube (sagging of 5.08 cm (2 ) in the mid section) The experimental results are summarized in Table 4-4 and Fig. 4-14. The average venting time data showed that even small sagging of the pressure tube in the middle by 12.52 mm (0.5 inch) could cause a significant decrease in the venting time by 8% as compared to the horizontal pressure tube. Upon further increases in the pressure tube sagging the venting time was further reduced, however, the effect became less significant as the depth of sagging was increased to 1.0 and 2.0-inches (25.4 and 50.8 mm). Although it was not possible to perform additional experiments at greater depths of sagging due to the risk of fracturing the pressure tube at the midpoint, the venting time is expected to continue to be reduced as shown by the trend seen in Fig. 4-14. Sagging Distance, Inch Table 4-4. Effect of the pressure tube sagging Inclination Angle, deg Venting Time (sagged tube), s Venting Time (horizontal tube), s Decrease in the Venting Time, % 0.5 0.16 5.57 8 1.0 0.32 5.40 11 6.08 1.5 0.43 5.28 13 2.0 0.65 5.23 14

32 6.1 Average Venting Time, s 6 5.9 5.8 5.7 5.6 5.5 5.4 5.3 5.2 0 0.5 1 1.5 2 Sagging Distance, inch Fig. 4-14 Effect of pressure tube sagging on the venting time 4.6 Air-Lift Effect The goal of this set of experiments was to investigate the possibility of inducing a continuous IBIF when the two header tanks are connected as shown in Fig. 4-15. In order to perform the experiments, air was injected asymmetrically through 5 nozzles so that the injection was dominant in the left half of the pressure tube. This way the venting would occur preferentially through the left feeder pipe. Water Tank L 1 L 1 Feeder Tube UF Asymmetrical Air Injection Pressure Tube Fig. 4-15 Schematic of the experimental set-up for studying the air-lift effect

33 Under these conditions the air venting caused an increase in the water level inside the left header tank due to the air lift effect caused by the venting air on the stagnant water inside the system. It was noticed that this behavior was similar to the one of an air-lift pump [7, 8]. As the total amount of water remained constant, an increase in the water level in one of the headers resulted in a decrease in the water level in the other header. The difference in the water levels between the two headers caused a continuous flow from the left to the right header (or vice versa) through the pipe connecting the two header tanks. This system configuration resulted in a continuous flow of water circulating throughout the system, so an ultrasonic flow meter (UF) was used to measure the flow rate of water in the connecting tube. It was observed that when the initial water level inside the header tanks was low, the results of the experiments were altered. When the water level is low the water in the nonventing tank gets completely depleted which does not allow for the continuous flow of water to be sustained since the venting starts occurring simultaneously through both feeders. The experiments were performed at 6 different initial water levels and 5 different air-injection rates, i.e. simulated power levels. The results are presented in Table 4-5 and Fig. 4-16, where each data point represents an average of 3 repeated experiments performed under exactly the same conditions. A total of 90 experiments were conducted in this experiment. Table 4-5 Air-Lift Effect on Flow Velocity (m/s) Simulated Power Level kw Header Tank 0.7 1.6 1.8 2.5 4.5 Initial Water Level, cm 34 0.53 0.61 0.67 0.73 0.95 30 0.56 0.64 0.68 0.69 0.87 26 0.51 0.59 0.66 0.71 0.88 22 0.52 0.62 0.64 0.75 0.92 18 0.59 0.61 n.a. n.a. n.a. 16 n.a. n.a. n.a. n.a. n.a.

34 The results showed that when the initial level of water inside the headers is high enough the induced flow velocity depends linearly on the air injection rate. 0.95 0.9 Flow Velocity, m/s 0.85 0.8 0.75 0.7 0.65 0.6 0.55 0.5 0.5 1 1.5 2 2.5 3 3.5 4 4.5 Simulated Power Level, kw Initial Water Level - 34 cm Initial Water Level - 30 cm Initial Water Level - 26 cm Initial Water Level - 22 cm Initial Water Level - 18 cm Fig. 4-16 Air-lift effect on flow velocity With an increase in the air injection rate, more water was transported to one of the header tanks resulting in a higher water level (L 1 ) inside that tank and respectively a lower level (L 2 ) in the other tank. This caused a higher static pressure difference between the two header tanks which resulted in a higher flow velocity. Simulated Power Level, kw Table 4-6 Air-lift effect on header water level Flow Velocity (cm/s) L 1 (cm) L 2 (cm) L (cm) 0.7 0.53 32 22 10 1.6 0.61 35 23 12 1.8 0.67 36 22 14 2.5 0.73 37 20 17 4.5 0.95 40 13 27

35 When the initial amount of water inside the water tanks was reduced to lower than 18 cm from the tank bottom the venting started occurring simultaneously through both feeders. Under these conditions the continuous flow could not be sustained between the two header tanks any longer. 4.7 Oscillatory behavior Upon symmetrical air-injection throughout the pressure tube, the air layer that had formed on top of the stagnant water in the pressure tube propagated simultaneously to both feeders. Once the air front reached the feeder pipes the venting of air occurred. Due to the fact that complete symmetry could not be achieved the venting actually occurred through one of the feeders slightly earlier than through the opposite feeder pipe. Once the air injection was initiated the venting would occur through one of the feeders on a random basis. The air venting caused water accumulation in the header tank due to the air-lift effect. As a result, the water level in this header tank would increase and the level in the opposite tank would decrease. This means that a net amount of water is being transported between the header tanks. As the water level in the header tank increases the hydrostatic pressure in the corresponding part of the pressure tube increases, and the pressure in the opposite header tank decreases. This process continues until a critical water level is reached in the header tank, and the venting process switches directions and starts occurring through the opposite feeder. Afterwards the switching process repeats itself. This oscillatory behavior can be characterized by a certain frequency of oscillations. The goal of this series of experiments was to investigate how the air injection rate and the initial water level inside the header tanks would affect the oscillation frequency. The experiments were performed by injecting air at five different air injection rates for five different initial water levels inside the feeder line. The system behavior was observed and the frequencies of oscillations were recorded by measuring the time interval between switching. The experimental results are presented in Table 4-7 and Fig. 4-17. A total of 75 experiments were performed and each data point represents an average of five runs performed under the same conditions.