Soon-Heung Chang

Power prediction in nuclear power plants using a back-propagation learning neural network

This paper proposes an artificial neural network - a data processing system with a number of simple highly interconnected processing elements in an architecture inspired by the structure of the human brain for the prediction of thermal power in nuclear power plants (NPPs). The back-propagation network (BPN) algorithm is applied to develop models of signal processing. A number of case studies are performed with emphasis on the applicability of the network in a steady-state high power level. The studies reveal that the BON algorithm can precisely predict the thermal power of an NPP. It also shows that the defected signals resulting from instrumentation problems, even when the signals comprising various patterns are noisy or incomplete, can be properly handled.

Neural Network Model for Estimating Departure from Nucleate Boiling Performance of a Pressurized Water Reactor Core

A new approach for estimating the departure from nucleate boiling (DNB) performance of a pressurized water reactor core is proposed in which a neural network model is introduced to predict the DNB ratios (DNBRs) for given reactor operating conditions. This model is trained against the detailed simulation results of DNBRs obtained from optimized random input vectors that are generated by Latin hypercube sampling on a wide range of parameters. The trained network is examined to verify the generalized prediction capability of the model. The test results show that a higher level of accuracy in predicting the DNBR can be achieved with the neural network model for both steady-state and transient operating conditions. The neural network model can be developed as a viable tool for on-line DNBR estimation in a nuclear plant.

Development of an uncertainty quantification method of the best estimate large LOCA analysis

Received 1 November 1991 A statistical uncertainty quantification methodology for evaluation of the emergency core cooling system (ECCS) performance is proposed and applied in assessing the best-estimate peak cladding temperature (PCT). In the proposed methodology, the Latin hypercube sampling method is adopted, and separate model uncertainties are used as input variables. The independency of the input variables is verified through a correlation coefficient test for statistical treatment of their uncertainties. Next, the PCT response distribution is determined through a goodness-of-fit test. Finally, the PCT with a one-sided 95% probability and a confidence level of 0.95 is estimated. This methodology is applied to the large-break loss-of-coolant accident (LBLOCA) of Kori Nuclear Units 3 and 4. This study shows that the proposed methodology is a useful one.

The manipulation of time-varying dynamic variables using the rule modification method and performance-index in NPP accident diagnostic expert systems

When an expert system is being developed for nuclear power plant (NPP) accident diagnosis, the most difficult problem is the manipulation of the time-varying dynamic variables. To meet this need, the authors propose modification of the rules for accident diagnosis when plant parameters and conditions are varying. In addition to the rule-modification method, a pattern-matching method using the performance index is suggested. This system also uses the results of transient analysis and accident analysis codes as a database. To simulate this expert system, PROLOG was used to construct the knowledge base. The inherent backtracking inference strategy was used for the inference engine. Feedwater line piping failure was selected for system verification. >

Performance Test of the Quenching Meshes for Hydrogen Control

The quenching distance of hydrogen gas was experimentally investigated by considering the effects of the initial pressure and steam addition. The quenching distance decreases with the initial pressure and there is a little increase with the addition of steam. Performance tests have been carried out to check the applicability of quenching mesh for the purpose of arresting hydrogen flame propagation during a severe accident in nuclear power plants. The experimental facility for the performance test of the quenching mesh consisted of a model compartment, a visualization system and an ignition system. Dimensions of the single model compartment were 300x300x300 mm. Three-compartments are connected in parallel. The quenching mesh is located between the first and second compartments. It was observed that the flame from the first compartment where the ignition starts does not propagate to the second compartment. The quenching mesh played a role of preventing flame propagation.

Uncertainty analysis of safety assessment for high-level radioactive waste repository

Abstract For the treatment of uncertainties arising in the safety assessment of high-level radioactive waste repository, a new probabilistic safety assessment methodology is developed. The present methodology describes a radioactive waste repository as a stand-by redundant system in a continuous operation using the system reliability analysis method in order to consider the model uncertainties. The risk of waste repository is presented as a product of the failure probability density of repository and the radionuclide inventory in the repository when the failure of repository occurs. Simultaneously, to evaluate how the safety assessment results are affected by the uncertainties involved in input parameters, the Monte Carlo simulation with Latin Hypercube sampling and a statistical estimation are used.

Development of an expert system for failure diagnosis of primary side systems

In this paper a prototype expert system, the NSSS-DS, is described. It is used for the diagnosis of three main systems-the rod control system, the reactor coolant pumps (RCPs), and the pressurizer-in the primary system of the Kori-2 nuclear power plant. This system diagnoses system malfunctions quickly and offers appropriate guidance to operators. The method of inference applied to this system is rule-based deduction with certainty factor operation. The diagnostic symptoms include alarms, indication lamps, parameter values, and valve lineup that can be acquired at the main control room. The application of this system to the rod control system, mainly consisting of electrical components, and the RCPs, mainly consisting of mechanical components, is described from the point of view of diagnostic strategies.

Transient-Effects modeling of critical heat flux

Abstract A method of classifying the transient critical heat flux (CHF) mechanisms has been proposed. The method is based on a novel transient CHF map which considers upstream and local effects. The prediction approach developed in this study is based on a correction factor which relates transient and steady-state CHF for both flow and power transients. The equation for the correction factor depends on the location on the transient CHF map. The predicted transient CHF provides reasonable agreement with experimental data.

APPLICATION OF A DUAL-ENERGY MONOCHROMATIC X- RAY CT ALGORITHM TO POLYCHROMATIC X-RAY CT: A FEASIBILITY STUDY

In this study, a simple post-reconstruction dual-energy computed tomography (CT) method is proposed. A dual-energy CT algorithm for monochromatic x-rays was adopted and applied to the dual-energy CT of polychromatic x-rays by assigning a representative mono-energy. The accuracy of algorithm implementation was tested with mathematical phantoms. To test the sensitivity of this algorithm to the inaccuracy of representative energy value in energy values, a simulation study was performed with mathematical phantom. To represent a polychromatic x-ray energy spectrum with a single-energy, mean energy and equivalent energy were used, and the results were compared. The feasibility of the proposed method was experimentally tested with two different micro-CTs and a test phantom made of polymethyl methacrylate (PMMA), water, and graphite. The dual-energy calculations were carried out with CT images of all possible energy pairs among 40, 50, 60, 70, and 80 kVp. The effective atomic number and the electron density values obtained from the proposed method were compared with theoretical values. The results showed that, except the errors in the effective atomic number of graphite, most of the errors were less than 10 % for both CT scanners, and for the combination of 60 kVp and 70 kVp, errors less than 6.0 % could be achieved with a Polaris 90 CT. The proposed method shows simplicity of calibration, demonstrating its practicality and feasibility for use with a general polychromatic CT.

DEVELOPMENT OF STATISTICAL CORE THERMAL DESIGN METHODOLOGY USING A MODIFIED LATIN HYPERCUBE SAMPLING METHOD

This paper presents a statistical core thermal design methodology for generating the limit departure from nucleate boiling ratio (DNBR) and is used in assessing the best-estimate thermal margin in a reactor core. This new methodology adopts a modified Latin hypercube sampling method. In this method, the independencies of the input variables are verified through a correlation coefficient test for statistical treatment of their uncertainties. Next the DNBR response distribution is determined through a goodness-of-fit test. Finally, a limit DNBR with a one-sided 95% probability and a confidence level of 0.95 is estimated. This methodology is simpler than the conventional statistical method using the response surface and Monte Carlo simulation technique, but it maintains the same level of confidence in the limit DNBR result. From this study, it is deduced that the proposed methodology is useful for design application.