A Comparison of the DDIFF-1 and GOTHIC Computer Codes for Analyzing Sub Compartment Pressurization Transients by Felix Meissner An Engineering Project Submitted to the Graduate Faculty of Rensselaer Polytechnic Institute in Partial Fulfillment of the Requirements for the degree of MASTER OF ENGINEERING IN MECHANICAL ENGINEERING Approved: _________________________________________ Ernesto Gutierrez-Miravete, Project Adviser Rensselaer Polytechnic Institute Hartford, CT April, 2012 (For Graduation May 2012) © Copyright 2012 by Felix Meissner All Rights Reserved ii CONTENTS LIST OF FIGURES .......................................................................................................... iv GLOSSARY ...................................................................................................................... v ACKNOWLEDGMENT .................................................................................................. vi ABSTRACT .................................................................................................................... vii 1. Introduction.................................................................................................................. 1 1.1 Problem Statement ............................................................................................. 2 2. Code Methodologies .................................................................................................... 3 2.1 DDIFF-1 Methodology ...................................................................................... 3 2.2 GOTHIC Methodology ...................................................................................... 6 3. Model Setup ............................................................................................................... 10 3.1 SGSC Model .................................................................................................... 10 3.2 DDIFF-1 Model ............................................................................................... 12 3.3 GOTHIC Model ............................................................................................... 12 4. Results........................................................................................................................ 15 4.1 DDIFF-1 Noise Reduction ............................................................................... 15 4.2 Comparison of Results ..................................................................................... 15 5. Conclusions and Recommendations .......................................................................... 21 REFERENCES ................................................................................................................ 22 iii LIST OF FIGURES Figure 1: Typical PWR Containment Rector Building……………………………...……. 1 Figure 2: Typical DDIFF-1 Input Deck…………………………………………………... 5 Figure 3: Typical DDIFF-1 Output Deck………………………………………………… 6 Figure 4: …………………………………………………………. 9 Figure 5: Cross Section of the Analyzed SGSC………………………………………… 11 Figure 6: General Outline of a SG………………………………………………………..12 Figure 7: GOTHIC Model Diagram…………………………………………………...... 14 Figure 8: Maximum Pressure Difference Across the SG in Level 1……………………. 17 Figure 9: Maximum Pressure Difference Across the SG in Level 2……………………. 18 Figure 10: Maximum Pressure Difference Across the SG in Level 3…………………... 18 Figure 11: Maximum Pressure Difference Across the SG in Level 4…………………... 19 Figure 12: Maximum Pressure Difference Across the SG in Level 5…………………... 19 Figure 13: Maximum Pressure Difference Across the SG in Level 6…………………... 20 iv GLOSSARY BWR - boiling water reactor CAD - computer aided design CFD - computational fluid dynamics EPRI - Electric Power Research Institute GOTHIC - Generation of Thermal-Hydraulic Information for Containments GUI - graphical users interface NAI - Numerical Applications Inc. NRC - Nuclear Regulatory Commission PWR - pressurized water reactor RCS - reactor coolant system RV - reactor vessel RVSC - reactor vessel sub-compartment SC - sub-compartment SG - steam generator SGSC - steam generator sub-compartment v ACKNOWLEDGMENT I’d like to acknowledge and thank my advisor, Ernesto Gutierrez-Miravete, for providing direction and guidance during this project and my master’s degree. I’d also like to acknowledge and thank the other RPI faculty, colleagues, friends and family; all of who provided tremendous support and assistance which was essential for the successful completion of my degree. vi ABSTRACT This paper successfully benchmarks the GOTHIC computer code to the DDIFF-1 computer code and recommends replacement of DDIFF-1 with GOTHIC. Transient response of a SGSC following a feedwater pipe break was analyzed with both codes. The main result of interest, pressure differences across the SG, was calculated and found to be comparable. DDIFF-1 is an old code (written in the 1960s) using elementary solution techniques and input methods (no GUI). GOTHIC, developed by NAI, is a modern, industry standardized, and accepted computer code that uses sophisticated solution methods with increased precision and accuracy. Faster computing speed combined with a GUI makes the standard GOTHIC code a better tool for SC analyses. vii 1. Introduction In the unlikely event that a high energy piping system ruptures (either a main or tributary line break) inside a sub-compartment (SC) in the reactor containment building of a pressurized water reactor (PWR, see Figure 1 for a typical example), asymmetric pressurization (local pressure differences) may occur around components or walls, resulting in loads that could damage structures within the SC. Typically these pressurization transients only last on the order of seconds. Figure 1: Typical PWR Containment Rector Building The Nuclear Regulatory Commission (NRC) has stringent guidelines on accident management and analysis. It ranks postulated scenarios according to severity and dictates that a postulated accident cannot lead to a higher order accident. For example, a feedwater line break inside the steam generator (SG, component #8 in Figure 1) SC cannot lead to a failure of the SG which would breach the reactor coolant system (RCS) boundary. Or in other words, a small pipe break cannot lead to the failure of a huge component. Therefore, pressurization of the steam generator sub-compartment (SGSC, and the resulting loads on the components within the SC) must be analyzed to demonstrate that the NRC guidelines on accident management are adequately met. 1 DDIFF-1 [1] is designed to analyze the asymmetric pressurization of reactor containment building SCs, either a SGSC or a reactor vessel sub-compartment (RVSC). DDIFF-1 is based on the original DDIFF code which was developed by Kraftwerk Union of the Federal Republic of Germany. Once calculated, the pressure time histories of all the nodes in the SC are forwarded to a structural group, which uses them to determine the loadings on the components of interest. GOTHIC [2] is a modern computer code developed by Numerical Applications Inc. (NAI) for transient thermal hydraulic analyses of multiphase systems in complex geometries. The latest version (version 8) was just released. GOTHIC was developed for the nuclear industry, specifically to analyze reactor containment buildings. References [3] and [4] document research using the GOTHIC code for reactor containment building analyses. This paper investigates the feasibility of replacing the DDIFF-1 code with GOTHIC for SC asymmetric pressurization transient analyses. The investigation process is completed in two steps; first gain more insight into each code by studying the methodology and solution methods, second perform a benchmark analysis of a GOTCHIC SGSC model to a similar DDIFF-1 model. Compare the results and determine the feasibility of using the GOTHIC computer code for applications which DDIFF-1 has been the historical solution vehicle choice. 1.1 Problem Statement The DDIFF-1 code was written in FORTRAN code in the 1960s. It is an old code that uses basic correlations and approximation schemes. Working with DDIFF-1 is not easy; it does not have a graphical users interface (GUI); inputs are provided via an input deck and outputs are reported in a similar fashion. Resources familiar with and able to debug older programs are becoming harder to find. DDIFF-1 needs to be replaced by a more modern code. 2 2. Code Methodologies 2.1 DDIFF-1 Methodology DDIFF-1 is a code which was specifically written to calculate the transient response of a SC in the unlikely event of a loss-of fluid incident in a high energy fluid system. The DDIFF-1 computer code is primarily used to predict SC conditions for the first few seconds following incident initiation during which the maximum pressure differentials on structures or components would occur. The transient calculations include determination of mass flow rates, mass and energy inventories, absolute and differential pressures, and temperatures in the SC system. The SC system is a control volume-flow path network created based upon the geometry of the SC being analyzed. Blowdown mass and energy release rates from the postulated pipe rupture are determined by an independent computer program and these histories are input to DDIFF-1. The output of interest is the differential (asymmetric) pressurization of the SC, as this will induce dynamic loads on components within the SC. The calculated pressurization history is used as input in a downstream structural analysis which will transform the differential pressures into stresses on the components. Methodology employed by DDIFF-1 is documented in Reference [1]. DDIFF-1 solves the mass, momentum and energy balance equations, as well as the equation of state for each control volume (or node) assuming lumped parameters. Inertia terms are included in the momentum balance equations. Homogenous equilibrium two phase flow correlations and Moody’s critical flow models are used to determine the flow between the control volumes. Numerical computations are performed using a simple forward integration method in a finite difference scheme. Time steps are specified in the input and can be dependent on time in transient, but are not dynamic based on the quality of the computations. Homogeneity and equilibrium are assumed for each control volume, therefore each component (state) occupies the entire control volume and temperatures for these components are equal within the control volume (node). In Reference [1], results predicted by DDIFF-1 are benchmarked to test results. DDIFF1 over predicts the initial pressure spike following blowdown (especially in the “break room”, the control volume where blowdown is introduced). This over-prediction results 3 in conservative loadings on the components calculated by the downstream structural group. For setting up the control volumes, node volume is the only input needed. For setting up the flow paths, flow area, inertia length, geometric loss coefficient and frictional loss coefficients and elevation differences are needed. These inputs must be user calculated with information from drawings. With all the components within the SC, it is time consuming and difficult to accurately calculate the nodal interface and flow areas. Inputs are specified via a user generated input deck (typical example shown in Figure 2). The structure of this input deck is based on the card system employed by early FORTRAN codes. Outputs are provided in a similar fashion (Figure 3), via output decks that must be post processed using Excel or a similar data manipulation program. Performing a full SGSC analysis using the DDIFF-1 code involves gathering and calculating the required inputs, setting up the physical model, applying the boundary conditions, running the model and analyzing the results. It takes about 400 man-hours to conduct a typical SC analysis using DDIFF-1. 4 Figure 2: Typical DDIFF-1 Input Deck 5 Figure 3: Typical DDIFF-1 Output Deck 2.2 GOTHIC Methodology The GOTHIC thermal-hydraulics computer code, described in Reference [2] is becoming the industry standard for performing containment analyses, as well as analyses for auxiliary buildings outside containment. The code was been developed by NAI with funding by the electric power research institute (EPRI). GOTHIC has been reviewed and approved by the NRC for containment analysis applications for several U.S. utilities. GOTHIC was specifically designed for use in design, licensing, safety and operating analysis of nuclear power plant containments and SCs. The prime application of GOTHIC is the evaluation of containment and containment SC response after a high energy line break. GOTHIC solves the mass, momentum and energy balance equations, as well as the equation of state for multi-phase and multi-component flow. Models for mass, energy 6 and momentum transfer at the interfaces between phases and components span from single phase flow to complex boiling. Interface models make it possible for GOTHIC to analyze non thermal equilibrium states between phases and unequal phase velocities. This is one major advantage of GOTHIC over DDIFF-1, the ability to more accurately predict phase interactions and analyze non-thermal equilibrium between different components within a control volume. Phases that GOTHIC can model and analyze include gas, liquid, droplet and mist. The gas phase includes water steam and various (up to 8) non-condensing gases. Inertia terms are included in the momentum balance equations, applied in the junctions/flowpaths that connect nodes. GOTHIC transient results have been successfully compared with results from other containment analysis codes (COCO, CONTEMPT, CONRANS, CONTAIN and COPATTA). Differences between the GOTHIC results and the results from other codes are attributed to the ability of GOTHIC to better model droplet phase interface heat and mass transfer. GOTHIC can perform its calculations using two different methods, in either a lumped parameter (similar to DDIFF-1) or domain model. By using the domain model, basic computational fluid dynamics (CFD) capabilities are introduced into the model. Nodes can be subdivided by simply drawing x, y and z grid lines. Subdividing a node is similar to meshing a domain/model in a conventional CFD program. In GOTHIC however, this meshing/subdividing is limited to rectangular schemes (unlike more complex geometry which is available in conventional CFD programs) and must be drawn via continuous grid lines. The GOTHIC terminology will be used throughout this paper. A node is a control volume with the lumped parameter model applied; a subdivided node is a domain that is “meshed” into individual cells. Using the subdivided method considers each cell as an individual control volume. Blockages of different geometric shapes can be used in a subdivided node to “carve-out” the geometric shape of the model. Coordinates and dimensions of these blockages are specified so that they can be placed appropriately. Internal blockages represent volume where there is no fluid flow and are also used to model the components within the SC. For instance, pump motors (which are located within the analyzed SGSC) can be 7 approximated as rectangular prisms. In order to model that flow blockage in GOTHIC, a rectangular prism blockage is placed at the coordinates of the pump motors. In a subdivided scheme, ends of flowpaths are placed into an explicit cell and the orientation of each flowpath is specified. A SC can be modeled as one subdivided node by making use of blockages to model the components within the SC. When using a subdivided node model, GOTHIC automatically determines the interface areas and flow information between cells. Blockages are included in this determination. This eliminates the manual and time intensive flow path parameter calculations that are needed when setting up a DDIFF-1 model, saving time and effort in the development of the SC model. Numerical computations within subdivided nodes are performed using bounded second order upwind methods in a finite volume scheme. For large and complex problems, Gauss-Seidel, conjugate gradient and multigrid iterative techniques are also available in GOTHIC. Time steps are automatically calculated and adjusted so that the solution remains stable, however minimum and maximum values must be specified in the input. Inputs are specified via an interactive GUI (Figure 4). This screenshot shows a typical GUI used in the process of subdividing (meshing) the SGSC using the gridlines. On the right hand side is a graphic of the modeled SC with the blockages in place. The gray triangles at the top form the trapezoidal shape of the SC, the circles represent the SG and the pumps within the SC. Note that the pressurizer is not shown in this view. In the center is a dialog box used for specifying the inputs, in the upper left is a table showing the actual inputs that will be used. Outputs are available in graph form, as well as user specified data files. Only the data files will be used for this analysis, as some data manipulation/post-processing will be done in Excel for convenience. Performing a full SGSC analysis using the GOTHIC code anywhere between 100 to 200 man-hours (25-50% of the time it takes to complete a similar DDIFF-1 analysis). This time saving is due to the ability of GOTHIC to automatically calculate the parameters for flow inside the SC (by using a subdivided node), which has to be done manually for the DIFF-1 model. 8 Figure 4: This typical GOTHIC GUI screenshot shows the process of subdividing (meshing) the SGSC in the vertical direction via gridlines. The table in the upper left hand corner displays the selected gridlines. The table near the center of the screen is used to specify the inputs. The figure at the right of the screen shows the model with the blockages (gray) in place. 9 3. Model Setup 3.1 SGSC Model The analyzed SGSC in a PWR (Figure 5) has the shape of a trapezoidal prism (about 60 feet by 30 feet and 80 feet high) and is located within containment right next to the RVSC. Major components within the SGSC are a pressurizer, two pumps and the SG (Figure 6). The pressurizer is a large cylinder with dome ends, approximately 35 feet high with a diameter of 10 feet. Its function is to ensure the RCS (not shown) remains in a sub-cooled state. Water within the pressurizer is kept at saturation at the desired RCS pressure. A small surge pipe connects the pressurizer to the RCS and this ensures that the coolant within the RCS remains at the right pressure and does not boil (like in a boiling water reactor (BWR)). The pumps within the SGSC are the large coolant pumps (diameter of about 6 feet). They each move about 100,000 gpm of coolant which transfers the thermal energy form the reactor to the SG. The SG is a U-tube heat exchanger which removes heat from the coolant and transfers it to the secondary side. Secondary side water in the SG is at a lower pressure (less than half of RCS pressure) and gets boiled. The resulting steam is piped through the turbine which turns the generators. A feedwater pipe (approximately 1 foot diameter) supplies the SG with “cold” water. Its nozzle is located near the top of the SG. A break of the feedwater pipe at this nozzle will be analyzed. Such a pipe break would allow the SG (initially at saturated conditions, pressure of around 1000 psi) and the feedwater pipe to blow down into the SGSC. In the analyzed SGSC nodal model, the SG nozzle is located in the fifth (out of seven) vertical level, split into the north-west and north-east (of the SG) nodes. At the time when the postulated pipe break occurs, the north side of the SGSC pressurizes quicker than the south side, resulting in differential pressure and loads on the SG. Initially the differential pressure spikes as the inertia of the stagnant air in the SC is overcome and then it settles to a quasi steady-state value. This initial pressurization (magnitude of and time) is of interest to the downstream structural analyses. During the short duration of the transient (on the order of seconds), level changes in the feedwater 10 pipe and SG will not impact the blowdown, so the mass and energy release into the SC is assumed constant. Figure 5: Cross Section of the Analyzed SGSC 11 Figure 6: General Outline of a SG 3.2 DDIFF-1 Model The DDIFF-1 model for the SGSC being analyzed has 65 nodes (64 for the SGSC plus 1 for containment), arranged in 7 equally spaced vertical levels. In each of the bottom 4 levels, there are 12 nodes, 4 for each pump area and 4 for the SG area. In levels 5 through 6 (above the pumps but still within the SC walls) there are 6 nodes per level, 4 for the SG area and 1 per pump area. The SC walls housing the pumps do not reach the 7th level and only 4 nodes for the SG area are needed. Node volume is the only major input needed. Nodal connections and interfaces are modeled with flowpaths. Required information for the flowpaths include: flow area, inertia length, geometric loss coefficient, frictional loss coefficient, and elevation difference. Nodal and flowpath information is typically obtained from a customer provided computer aided design (CAD) model of the SGSC. The model is imported into ANSYS, inversed, stripped of all minor components (like ladders, stairs etc.) and divided into nodes. An ANSYS -routine is written to output nodal volume and to calculate the required information to model the flowpaths. Mass and energy blowdown is input by release rate and node number. The time interval for calculations is specified as 0.00005 seconds. 3.3 GOTHIC Model The GOTHIC model consists of 2 nodes, 1 subdivided node (consisting of 84 cells) to model the SGSC and 1 node for the containment which contacts the top of the SGSC. The SGSC node is subdivided into 7 vertical layers (levels), each having 12 cells (4 in each pump area and 4 in the SG area, for a total of 84 cells). The levels are numbered consecutively with elevation, with level 1 being the bottom level. Blockages are used to shape the trapezoidal shape of the SGSC and to account for the pumps, SG, pressurizer, walls between the SG and pump areas of the SC at the higher elevations and various piping near the bottom of the SC. Blockages are also used to cap the height of the pump areas of the SC, as that portion of the SC does not extend as high as the SG portion of 12 the SC. Twelve flowpaths model the openings at the top of the SGSC to containment. They connect each subdivided node at the top of the SGSC to the containment node. Mass and energy blowdown is introduced into the model via flow boundary conditions, and connected to the appropriate subdivided node via flowpaths. A diagram of the nodes, boundary conditions and flowpaths is shown in Figure 7. The dashed-line boxes (labeled 1 and 2s) are the atmospheric (lumped parameter) and SC (subdivided into cells) nodes. The solid line boxes (labeled 1F and 2F) are the flow boundary conditions which add the mass and energy blowdown into the model. The blowdown is transferred into the SC node via flowpaths #1 and #2 (solid lines). The remaining flowpaths (solid lines #3 #14) connect the cells at the top of the SC to the atmosphere node (1). Input information consists of dimensions of the SGSC and of all the components within the SC. This is obtained from plant drawings and was confirmed to match the information used in the DDIFF-1 model. GOTHIC is able to dynamically adapt the time step size based on the accuracy of the computation (by analyzing the computational residuals). Time step intervals are limited to be between 1E-10 and 1E-05 seconds for this SC model. GOTHIC was able to successfully calculate the solution using the maximum time step interval (i.e. the residuals remained below the threshold even with the maximum specified time step interval). By using the blockages feature to create a SGSC model in GOTHIC, it makes the process much easier and more effective than manually dividing the SGSC into nodes and calculating the info for the interfacing flowpaths. This is a major reason for the reduced time required for a GOTHIC analysis. 13 Figure 7: GOTHIC Model Diagram 14 4. Results 4.1 DDIFF-1 Noise Reduction The output data obtained from DDIFF-1 reports pressures for each node at 227 different time steps during the relatively short transient (on the order of seconds). These data points contain a lot of noise, due to the simplistic schemes employed by DDIFF-1 and make data comparisons impossible. In order to better compare the DDIFF-1 results to those obtained by GOTHIC; noise was filtered out by averaging every 5 consecutive data points. Without this noise reduction routine, it would be impossible to make a deterministic comparison between the results of the two codes. 4.2 Comparison of Results The most heavily loaded (and most critical) component during a SGSC transient is the SG. Therefore, the most important result of a SGSC analysis is the pressure difference and loading across the SG. Pressure differentials across the pump and pressurizer are of secondary importance. Since the break is located near the SG, a comparison of the peak pressures across the SG at each level is a good method to compare the two codes. Maximum pressure differences predicted by each code across the SG (maximum of the absolute value of either the north-east minus south-west or north-west minus south-east nodes) at each time step and level in the vertical direction were calculated. Results were normalized to the maximum pressure differential (from all levels, on the order of tens of psi) and to the time in transient. Comparisons between the results obtained from each code are plotted and shown in Figure 8 through Figure 13 for the first 6 levels of the SC. Results from GOTHIC are somewhat noisy (especially in the lower levels), but dampen to more consistent values and have lower amplitude low-frequency noise than those from DDIFF-1, even after the noise reduction scheme was applied to the DDIFF-1 results. This is expected, since GOTHIC uses a second order upwind solution method, which is more numerically stable than the forward integration method used by DDIFF-1. The feedwater pipe (analyzed break location) is located in the fifth vertical level. Pressure differentials in levels 4 and above are the most important; since they exert forces on the SG at a higher elevation, they cause the highest momentum and stress on 15 the SG. Since blowdown is introduced on level 5, they also have the highest pressure differential. As expected (and described in Section 2.1), DDIFF-1 over-predicts the initial pressure spike in the break room on level 5 (shown in Figure 12). Initially, the DDIFF-1 pressure differential reaches its transient peak and then dampens down to about 20%, whereas the GOTHIC result spikes to 35% and then dampens down to about 20%. Examining the DDIFF-1 initial pressure spike over-prediction data documented on Reference [1], the GOTHIC predicted peak of 35% is much closer to the test data and is as expected. Level 6 (directly above level 5) connects to level 7, which vents to the atmosphere. Pressure differentials for level 6 are shown in Figure 13. Despite the DDIFF-1 over-predicted pressure spike in level 5, for level 6, the GOTHIC and DDIFF-1 values are very comparable, both in oscillation shapes and magnitude. Level 4 pressure differentials predicted by DDIFF-1 (Figure 11) are affected by the overpredicted initial pressure spike in level 5, peaking at 26% versus 10% and then dampening down to about 2%. The pressure differentials are very similar (both in magnitude and shape) at the most important elevations (levels 4 and above). For conservatism, only the top openings/vents in the SGSC were modeled. This leads to greater numerical oscillations in levels 4 and below, as the blowdown gets pushed downward, then travels outward to the pump areas and must then rise to the top to vent to containment (the flow must pass through more control volumes before it vents). The oscillation continues into levels 3 through 1 with decreasing amplitude. Pressure differentials in levels 3 and below are of secondary importance; because they are at a lower elevation, they exert less momentum and stress on the SG and they are lower in magnitude. In level 3 (Figure 10), the pressure differential oscillations are evident and the magnitudes between DDIFF-1 and GOTHIC vary by up to 6%. In level 2 (Figure 9), the differentials trend reasonably well but vary by up to 4%. Differentials in level 1 (Figure 8) trend similarly to those in level 3 but vary by up to 4%. The pressure differentials in levels 4 and above (the most significant ones) are very similar and compare well. The differentials in the first 3 levels are not as similar, but as those differentials are of secondary importance, the variations have an insignificant 16 impact on the result of interest, the loading on the SG. Therefore, the results show that the DDIFF-1 and GOTHIC computer code calculate somewhat similar results of interest. Sensitivity studies in time space were run with the GOTHIC model. The maximum time step interval of 1E-05 sec could have been relaxed to at least 1E-04 without any impact on the results. Sensitivities on the number of nodes/cells (node/cell size) were not run, since the objective was to recreate (as best as possible) the DDIFF-1 SGSC model in GOTHIC. Running sensitivity cases on the number of nodes/cells would be a requirement if the GOTHIC computer code is used for runs in official calculations. Sensitivity studies were not done for the DDIFF-1 model, since it served as the basis of the comparison and the methodology employed is NRC approved. Figure 8: Maximum Pressure Difference Across the SG in Level 1 17 Figure 9: Maximum Pressure Difference Across the SG in Level 2 Figure 10: Maximum Pressure Difference Across the SG in Level 3 18 Figure 11: Maximum Pressure Difference Across the SG in Level 4 Figure 12: Maximum Pressure Difference Across the SG in Level 5 19 Figure 13: Maximum Pressure Difference Across the SG in Level 6 20 5. Conclusions and Recommendations It is concluded that for a SGSC analysis, the GOTHIC computer code produces comparable results to those predicted by the DDIFF-1 computer code. GOTHIC uses more advanced modeling and solution techniques. It has a user friendly GUI. Performing SGSC analyses in GOTHIC is more cost-effective; using GOTHIC it only takes 25% to 50% of the time required for performing a similar analysis using DDIFF-1. It is recommended that the GOTHIC computer code be considered as a tool to replace the DDIFF-1 computer code to perform SGSC analyses. 21 REFERENCES 1. CENPD-141, Revision 2, A Description of the DDIFF-1 Digital Computer Code for Reactor Plant Subcompartment Analysis, March, 1978. 2. GOTHIC Datasheet, published by NAI, http://www.numerical.com/pdf/gothic_datasheet.pdf 3. M. Gavrilas, GOTHIC Code Evaluation of Alternative Passive Containment Cooling Features, Nuclear Engineering and Design, 166 (1996): 427-42. 4. M. Gavrilas, The Design and Evaluation of a Passively Cooled Containment for a High-Rating Pressurized Water Reactor, Nuclear Engineering and Design, 200 (2000): 233-49. 22