Belle R. Upadhyaya

Application of Neural Networks for Sensor Validation and Plant Monitoring

Sensor and process monitoring in power plants requires the estimation of one or more process variables. Neural network paradigms are suitable for establishing general nonlinear relationships among a set of plant variables. Multiple-input/multiple-output autoassociative networks can follow changes in plantwide behavior. The backpropagation (BPN) algorithm has been applied for training multilayer feedforward networks. A new and enhanced BPN algorithm for training neural networks has been developed and implemented in a VAX workstation. Operational data from the Experimental Breeder Reactor II (EBR-II) have been used to study the performance of the BPN algorithm. In this paper several results of application to the EBR-II are presented.

Sensor validation for power plants using adaptive backpropagation neural network

Signal validation and process monitoring problems in many cases require the prediction of one or more process variables in a system. The feasibility of using neural networks to characterize one variable as a function of other related variables is studied. The backpropagation network (BPN) is used to develop models of signals from both a commercial power plant and the Experimental Breeder Reactor-II (EBR-II). Several innovations are made in the algorithm, the most significant of which is the progressive adjustment of the sigmoidal threshold function and weight updating terms. >

Software-based fault-tolerant control design for improved power plant operation

The use of fault-tolerance has increasingly become an important requirement for power plant control because of safety and availability considerations. Typical applications in process control employ redundant hardware for data acquisition and processing. The primary objective is to maintain the operation of the plant, even in the presence of anomalies in instrument channels. The performance of a system with existing hardware can be improved by the additional design of diverse, digital control systems in the form of software redundancy. This design has been integrated with signal and command validation with option for manual and automatic control through a graphical user interface. The fault-tolerant design has been demonstrated with application to the steam generator system control in a pressurized water reactor plant.<<ETX>>

Monitoring Feedwater Flow Rate and Component Thermal Performance of Pressurized Water Reactors by Means of Artificial Neural Networks

The fouling of venturi meters, used for steam generator feedwater flow rate measurement in pressurized water reactors (PWRs), may result in unnecessary plant power derating. On-line monitoring of these important instrument channels and the thermal efficiencies of the balance-of-plant components are addressed. The steam generator feedwater flow rate and thermal efficiencies of critical components in a PWR are estimated by means of artificial neural networks. The physics of these systems and appropriate plant measurements are combined to establish robust neural network models for on-line prediction of feedwater flow rate and thermal efficiency of feedwater heaters in PWRs. A statistical sensitivity analysis technique was developed to establish the performance of this methodology.

Chemometric Data Analysis Using Artificial Neural Networks

The on-line measurement of chemical composition under different operating conditions is an important problem in many industries. An approach based on hybrid signal preprocessing and artificial neural network paradigms for estimating composition from chemometric data has been developed. The performance of this methodology was tested with the use of near-infrared (NIR) and Raman spectra from both laboratory and industrial samples. The sensitivity-of-composition estimation as a function of spectral errors, spectral preprocessing, and choice of parameter vector was studied. The optimal architecture of multilayer neural networks and the guidelines for achieving them were also studied. The results of applications to FT-Raman data and NIR data demonstrate that this methodology is highly effective in establishing a generalized mapping between spectral information and sample composition, and that the parameters can be estimated with high accuracy.

Multivariate signal analysis algorithms for process monitoring and parameter estimation in nuclear reactors

Abstract The random fluctuations of the process variables during the normal operation of a nuclear reactor may be processed to yield information about changing parameters, vibration of reactor components, sensor response characteristics, stability margin and noise source distribution. Multivariate time domain algorithms are developed to study the relationship between reactor dynamic variables, both in pressurized water reactors (PWR) and boiling water reactors (BWR). Evaluation of dynamic system descriptors in PWRs using the multivariate analysis is presented for the first time in this paper. The analysis is performed at low frequency range to study the long term behavior of PWR dynamics. Interpretation of driving noise sources and cause-and-effect relationships between the signals are derived using spectral decomposition obtained from the multivariate models. The assumptions in the analysis are tested using an orthonormal projection of the source noise vector into independent components. Applications of mini computer oriented algorithms are evaluated using test data from operating power reactors.

Incipient Fault Detection and Isolation of Field Devices in Nuclear Power Systems Using Principal Component Analysis

Abstract An integrated method for the detection and isolation of incipient faults in common field devices, such as sensors and actuators, using plant operational data is presented. The approach is based on the premise that data for normal operation lie on a surface and abnormal situations lead to deviations from the surface in a particular way. Statistically significant deviations from the surface result in the detection of faults, and the characteristic directions of deviations are used for isolation of one or more faults from the set of typical faults. Principal component analysis (PCA), a multivariate data-driven technique, is used to capture the relationships in the data and fit a hyperplane to the data. The fault direction for each of the scenarios is obtained using the singular value decomposition on the state and control function prediction errors, and fault isolation is then accomplished from projections on the fault directions. This approach is demonstrated for a simulated pressurized water reactor steam generator system and for a laboratory process control system under single device fault conditions. Enhanced fault isolation capability is also illustrated by incorporating realistic nonlinear terms in the PCA data matrix.

Fault monitoring of nuclear power plant sensors and field devices

Existing and new generation of nuclear power plants have economic and reliability concerns as addressed by overall plant performance, unscheduled downtime and the long-term management of critical assets. The key to achieving these needs is to develop an integrated approach for monitoring, control, fault detection and diagnosis of plant components such as sensors, actuators, control devices and other equipment. Both single and multiple fault cases have been considered. This paper presents the following approach for achieving this goal: 1. Development of data-driven system models using Group Method of Data Handling (GMDH), Principal Component Analysis (PCA) and Adaptive Network-based Fuzzy Inference System (ANFIS), 2. Fault detection by tracking model residuals of selected process variables and control functions, and 3. Fault isolation using a rule-based technique and/or a pattern classification technique. This approach is illustrated for a nuclear plant steam generator. Fault detection and isolation (FDI) of sensors and field devices is an important step towards the implementation of an automated and intelligent process control strategy, especially for Generation-IV reactors.

Model Based Approach for Fault Detection and Isolation of Helical Coil Steam Generator Systems Using Principal Component Analysis

A robust Principal Component Analysis (PCA) model-based approach to Fault Detection and Isolation (FDI) was developed for the Helical Coil Steam Generator (HCSG) systems of the International Reactor Innovative and Secure (IRIS) nuclear reactor. The PCA model was developed using the data generated from a simulation of the system dynamics. Because all the operation modes can be excited through well-designed simulations, a PCA model developed from simulation data is more robust to changes in operation conditions than a PCA model developed from historical data collected during routine operation conditions. In order to deal with model uncertainty that exists in a simulation model, the mismatch between PCA model predictions and online plant measurements was analyzed to characterize the model uncertainty. Based on the assumption that model uncertainty has a structured property, a complete algorithm for the PCA-based FDI was derived such that the FDI results are robust to model uncertainty. When this new method was applied to the IRIS HCSG systems, the results showed that this approach could avoid false alarms and fault misdiagnosis due to changes in operation conditions and model certainty. The approach would overcome the limitations of traditional data-driven model based FDI when routine operational data do not contain sufficient information, and the limitations of physical model based FDI when these models are too complicated for a direct use in FDI design and contain model uncertainty

Fault Diagnosis of Helical Coil Steam Generator Systems of an Integral Pressurized Water Reactor Using Optimal Sensor Selection

Fault diagnosis is an important area in nuclear power industry for effective and continuous operation of power plants. Fault diagnosis approaches depend critically on the sensors that measure important process variables. Allocation of these sensors determines the effectiveness of fault diagnostic methods. However, the emphasis of most approaches is primarily on the procedure to perform fault detection and isolation (FDI) given a set of sensors. Little attention has been given to actual allocation of the sensors for achieving efficient FDI performance. This paper presents a graph-based approach as a solution for optimization of sensor selection to ensure fault observability, as well as fault resolution to a maximum possible extent. Principal component analysis (PCA), a multivariate data-driven technique, is used to capture the relationships among the measurements and to characterize by a data hyper-plane. Fault directions for the different fault scenarios are obtained using singular value decomposition of the prediction errors, and fault isolation is then accomplished from new projections on these fault directions. Results of the helical coil steam generator (HCSG) system of the International Reactor Innovative and Secure (IRIS) nuclear reactor demonstrate the proposed FDI approach with optimized sensor selection, and its future application to large industrial systems.