Adding Value to Nuclear Engineering Processes
with Realistic Modeling
Tomasz Kisielewicz1, Pascale Goussard1, Marek Slovacek2
1. ESI Group, Paris, France
2. Mecas ESI, Plzen, Czech Republic
Abstract: The engineering practices in nuclear engineering have been for most of it defined in the 1970’s and have not evolved since then. While at that time these methodologies matched the available computational means, the latter have encompassed significant progress in the last 30 years delivering significant benefits. Some projects based on the realistic approach have been carried out for nuclear energy, the results of which have been accepted by the licensing authorities. The paper presents a status of the situation and defines a strategy to deploy such approach in a more general way.
1. Introduction
In the 1970’s, tens of nuclear power plants were ordered, designed and set in line as an answer to the first oil shock. This situation boosted the practices and processes of nuclear engineering providing the first major field for industrial applications of the finite element method. At that time however, the limitations of the hardware available in terms of memory and processing speed together with the stringent design criteria for insuring the safety of the plants led engineers to define a series of modeling processes based on simplified modeling processes and a series of ‘conservative’ assumptions based on increase of load factors and decrease of allowable limits, regardless of the consistency of the resultant physics.
With the resumption of interest for nuclear energy to fight the global warming and the increasing oil prices, and regardless of the events of Spring 2011 in Japan that have focused the attention on the actual safety of nuclear installations, nuclear engineering is being revived. The modeling means however have greatly evolved meanwhile led by the automotive industry since the mid 1980’s. In the same time, a number of engineering projects have been carried out in the nuclear energy field to re-assess the safety status of existing plants, especially those designed against less stringent regulations that have taken advantage of these progresses.
This paper intends to differentiate the conservative and realistic modeling approaches, discussing some of the benefits provided by the latter and describing at which level of the modeling process, realistic assumptions can be introduced to get the benefits of such approach.
2. Simulations in Nuclear Engineering
2.1 Purposes
Nuclear engineering has been historically the first field of industry where numerical modeling has been institutionalized as the commonly accepted approach for designing plants and justifying their level of safety. This resulted from the impossibility to obtain full-scale test results to justify the assumptions and validate the results.
In order to compensate the lack of experimental validation and still justify the safety level required by regulations, regulators and engineers have relied on a series of conservative approach by selecting at each step what was deemed to be on the safe side.
Figure 1 - Purposes of Simulation for Nuclear Engineering
Modeling of processes and components in the nuclear engineering were and are carried out for a series of purposes starting from Design and Stress Reports components and Operating Processes. Figure 1 above lists the field of application of numerical modeling for nuclear engineering.
Beyond this list of destinations, it is worth mentioning that such modeling needs encompass all fields of physics from neutronics to crack propagation including thermal-hydraulics, chemistry, structural dynamics and thermo-mechanics, electromagnetism, as well as a series of multiple physics coupling several of them.
In the following sections, this paper focuses on the design of components for producing Design and Stress Reports. This does not imply that the realistic modeling is not applicable for other purposes.
Figure 2 - Simulation Process for Nuclear Engineering
2.2 Typical Engineering Process
Figure 2 above outlines a typical flow chart for the engineering processes with two specificities for nuclear engineering. When selecting the approach for carrying out the modeling tasks, the project manager has to determine whether the traditional ‘conservative’ approach of the new realistic approach will be used. This choice has to be done very early in the process.
The second specific point is related to the qualification of the modeling tools, data and users as per the nuclear safety regulations. While the ‘conservative’ approach has been validated and accepted by safety authorities since the early 1970’s and, provided the past methodologies are followed without any change, are accepted, the realistic approach requires some specific tasks for such validation.
On the other hand, as will be detailed later, the realistic approaches rely on first physics principles (generally accepted equations of physics) and hence are somewhat simpler to validate.
3. Realistic Approach
3.1 Definition
Realistic modeling is an approach that tends to bridge the abstract world of theoretical physics and its modeling means with the concrete world of reality.
This does not mean that the numerical models are exact representation of the reality. This would not be possible as too many uncertain and unknown data makes reality unreachable.
The realistic models will include all details and parameters needed to capture the major and significant secondary effects. The idealizations used traditionally in terms of boundary conditions, loading conditions, initial conditions, geometry and equations of state are not taken for granted and replaced by conditions closer to reality, provided they are deemed to have an influence on the results of the simulations.
In order to enhance the quality of the models in such way, commonly accepted equations of physics are extensively used in their available programming state. With the dramatic progress of computer hardware during the past thirty years, this provides nowadays almost unlimited means to obtain as accurate as needed results from simulation.
These developments have been thorough fully validated fro a variety of applications in the automotive industry since the mid 1980’s starting with car crash simulations, and then spread to aerospace and mostly all other industrial sectors.
In the field of nuclear engineering, a number of projects have been carried out in the 1990’s with similar realistic approaches to assess more accurately the level of safety of existing nuclear power plants that were not originally designed and erected in compliance with the international standards. Without such realistic approach, i.e. using the ‘conservative’ approach, these plants would have had to been shut down while their level of safety was proven to be sufficient. The next paragraph briefly describes tow such projects.
3.2 ‘Conservative’ vs. Realistic
In the conservative approach, the input data are selected to provide pessimistic results with respect to a given safety criterion. However, it is also based on a series of simplifications and idealizations matching the available analysis means. For example, geometry is accepted ‘as-designed’, i.e. ideal dimensions without any imperfection resulting from the manufacturing process.
The main advantage of these methods if that they have been accepted by the licensing authorities. The main drawback is that the apparent high level of safety results from a series of inconsistent assumptions and the actual level of safety is impossible to assess. The intentional conservatism may not lead to conservative results (e.g. simplified simulated high power in LOCA leads to higher swell level providing better core cooling which is not conservative) Furthermore the original rational for their development (limitation of modeling and computational capabilities in the 1970’s) is not valid anymore.
By difference, the realistic approach tends to explore the parameters that will enable to get as close as necessary to the real case. In particular, as will be discussed later, ‘as-built’ and ‘as-operated’ conditions are defined (calculated). It also is very well adapted to integrating probabilistic ranges on parameters, hence making it the best choice for Probabilistic Safety Assessment.
These new approach is accepted by the design codes and increasingly by the licensing authorities. However, they require specific validation that, while not overly complex, creates a specific additional task, at least in the first instances of usage.
Their novelty also requires specific training and support to engineers for a proper usage and delivery of full benefits. The origin of realistic modeling stands in other sectors than the nuclear engineering. A transfer of know-how is required. The process has readily started with some visionary companies and has been applied in a series of projects that have convinced the utility company and the licensing authorities of their value.
3.3 Benefits
Typical benefits of the realistic approach are illustrated with two examples of projects carried out in the 1990’s on existing power plant to deliver more accurate safety assessment.
Medium leak size in cold leg [1]
The simulation of a medium leak of 160 cm2 in the cold leg of the KONVOI 1300 MWe NPP provided the following result to be compared with the result from the ‘conservative’ method.
The significant difference in temperature loadings leads to very different engineering decision on the detailing of the component. In particular the conservative approach ended up in predicting the full uncovering of the core while the realistic one predicted uncovering of the upper part only. These situations lead to drastically different operating reactions.
The available margin resulting from realistic prediction provides the means for improving the performance of the plant and its ultimate safety.
Figure 3 – Typical benefits of realistic modeling [1]
Large primary to secondary bleak in VVER [1]
Analyses were conducted for the case without intervention of operator using the conservative and the realistic approach. The results are sown on figure 4.
While maintaining the radiological releases acceptable, the realistic case provides much more time for proper operation and preserving the installation.
3.4 Examples
The influences of realistic modeling in nuclear reengineering practices are discussed hereunder for two extreme cases related to extreme external aggression (aircraft impact) and for accidental event on the system (pipe rupture and consequent pipe whip). For each situation, the traditional ‘conservative’ approach used in the 1970’s is described and the possible enhancements discussed.
Figure 4 – Typical benefits of realistic modeling [1]
Aircraft Impact
Aircraft impact is a severe event that must be considered in designing nuclear power plants based on statistical analyses of flight paths nearby nuclear plants and rate of airplane crashes for different categories of airplanes. This regulatory principle has led to a divergence of loading conditions depending on the countries and local statistics. For example, while in France the plants were designed against the impact of a rather light Lear Jet, in Germany much more severe impacts of fighter-bombers of type Phantom F4 were considered as a result of series of crashes of the predecessor of the Phantom in the German Air Force, the F104, and the habit of the Air Force to use NPP as virtual targets for their trainings, hence increasing the flight density nearby.
The ‘conservative’ approach defined to account for such event was drafted with the means available in the late 1960’s and early 1970’s. As detailed structural modeling of the airplanes for such crash was not possible (neither the software nor the hardware could handle such complex and large models), simplified models with lumped masses and non-linear springs were developed as illustrated on figure 5.1.
The modeling process was tedious. It relied on data provided not by the airplane manufacturers but by the local operators in the form of blue prints. A large number of engineering judgments were needed to define the lumping process. The non-linear behaviors were estimated based on assumptions on the collapse mode of the structure.
Figure 5.2 – Aircraft impact load generation Figure 5.1 – Aircraft structure modeling
These loads were then applied on different locations of the building models. 3D or shell models or beam models were used for such impact analyses, the results of which were used to define the rebars away form the impact points as well as to generate the floor response spectra for equipment qualification.
These uncertainties did lead some safety authorities at carrying out a series of experimental projects to validate these modeling processes. However, these were limited except for one case of impacting a full scale plane on a sliding concrete block and they were carried out much after the plants were designed using the ‘conservative’ approach.
The simplified non-linear models were not impacted locally to the models of the buildings (the impact analyses were carried out using linear elastic assumption). They were used instead to generate impulse loads as depicted on figure 5.2.
Complementary analyses were carried out for local detailing of the buildings always with the same target of justifying that the structure would resist without cracking. Military perforation formulas were used to define the transverse reinforcement needed to avoid that event. This led to very different wall thicknesses depending on the impinging aircraft. German reactor buildings have typically a heavily reinforced 2.1 meter thick concrete shell to resist the impact of Phantom F4 while the French reactor building have a much thinner wall sufficient to resist the impact of lighter Lear Jets.
Figure 5.3 – Impact analysis
Whether the French NPP could actually resist or not to the impact of a fighter-bomber cannot be deducted from these analyses. If the ‘conservative’ method is applied without adaptation, then the conclusion would be that they cannot resist it. However, the ‘conservative’ approach neglects a number of energy absorption capabilities of the building.
First, simplifying the structural modeling of the airplane structure for crash event can be unconservative and certainly is unrealistic. The collapse modes of structures as complex as airplanes can hardly be simplified into a single axial one.
This uncertainty can nowadays be easily overcome. Crash of aeronautics structures is being carried out by manufacturers for a number of severe loads such as had take-off or landing during which part of the structures can hit the tarmac [Fig. 6.1] or during ditching when the pressure of the water can crash the structure.
These modeling methodologies can be applied immediately for airplane crash onto NPP buildings either on rigid walls to follow the ‘conservative’ approach or onto deformable models of the buildings [Fig. 6.2].