Shah 2011

INSTITUTE OF TECHNOLOGY, NIRMA UNIVERSITY, AHMEDABAD – 382 481, 08-10 DECEMBER, 2011

1

Fault Detection and Diagnosis in Nuclear Power Plant - A Brief Introduction Manit D. Shah Abstract--This paper reviews fault detection and diagnosis techniques and their application in nuclear power plant. Fault is an unexpected change or malfunction in the system.Fault occurring at a point in the plant may propagate further in the system. It is necessary to arrest fault propagation for economy and safety of plant. For this purpose quick detection of an unexpected deviation of process variableis necessary. This function is achieved with the help of Fault detection and diagnosis unit. Diagnostic unit also assists operators in retaining normal and safe state of the system. In this paper,I have mentionedthe characteristic requirements of Fault detection and diagnosis unit and the classification of diagnostic methods. With the help of Fault detection and diagnosis, it is possible to realize continuous monitoring of plant which enables diagnosis of incipient faults and thus prevents unexpected operational upsets.Here, I have done case study for two approaches, and I have observed that use of hybrid approaches for Fault detection and diagnosis yields quick and accurate assistant to the operators compared to conventional approaches. Index Terms—Fault detection, Fault identification, Fault isolation, Fault diagnosis, Nuclear power plant I.

INTRODUCTION

I

n recent times, regulatory control could be automated with the help of computers and thus, the continuous controller actionsrequired from the human operators are reduced to minimal. With this, a great progress in safety, product quality and economy of the plant was observed. In present era, current challenge for control engineers is the automation of abnormal event management (AEM) [1]. AEM deals with detection and diagnosis of abnormal events and suggesting supervisory actions to operators, for bringing system back in normal state. People in process industries view this development as next milestone in control systems. Due to wide scope of this diagnostic system, various techniques have been developed over the time. Fault detection and diagnosis (FDD) is the process to detect, isolate and identify faults in the system. Fault detection is the determination of faults, present in the system. Fault isolation is finding type and location of fault, followed by the faultdetection. Fault identification is analyzing the size and time variant behavior of a fault [2], followed by the fault isolation. FDD and AEM both have same functions but in complex process industry, FDD nomenclature is widely used. We will also use, in this paper, FDD terminology. Though early attempts were made for fault diagnosis using diagraphs, fault trees, hardware redundancy and analytical redundancy, these methods are still used in practical application. In recent times more elegant methods such as expert system, knowledge based method and neural networks, due to advancement in computer technology, are being used.

Before we begin further on discussing diagnostic approaches, let us see few of the terminologies which are widely used in literature: -Failure: Permanent disruption of the system in performing its desired function -Disturbance: An unknown input adversely affecting operation - Irregularity: Intermittent departure from normal response of the system -Abnormality: Fault or failure constitutes an abnormal event The rest of the paper is organized as follows: Section II describes fundamentals of FDD unit, Section III shows requirements of FDD in NPP and various diagnostic approaches employed in NPP and Section IV concludes the paper. II. FAULT DETECTION AND DIAGNOSIS A. Classification of faults We can classify types of faults based on their physical location or their effect on system performance. 1) Classification based on physical location of fault: - System faults - Sensor faults - Actuator faults 2) Classification based on fault characteristics: - Additive/Multiplicative faults - Abrupt/incipient faults - Permanent/Transient/Intermittent faults Fault occurring due to leakage in any one of the pipes, damage in shaft of turbine generator, etc. are considered under system faults. Such faults beings changes in dynamic input/output properties of the system. When system under observation faces malfunctioning of any one of the sensors in use, then these are called sensor faults. In this case, sensor reading has substantial errors. Due to actuator faults, influence of the controller on plant gets affected. Actuator faults may be jam of pump, stuck of control valve(s), etc. Multiplicative fault are observed when fault gets multiplied by system gain, while in additive fault system sees addition of faults. In abrupt faults, sudden change in behavior of the system occurs, e.g. a step function; while incipient is gradual change in fault, e.g. a ramp function. Permanent fault is the total failure of the equipment, transient is temporary malfunctioning while intermittent is the repeated occurrences of transient faults. Automated FDD heavily relies on measurements taken from sensors, since they act as inputs to diagnostic unit.So, two of the values which matters the most during system behavior analyses are: Reference value and dynamic reading

2

INTERNATIONAL CONFERENCE ON CURRENT TRENDS IN TECHNOLOGY, „NUiCONE – 2011‟

of the same variable. Actions of diagnosis unit are based on reference value and reading taken in real-time. Errors in any of above values may result in catastrophic consequences. A great care has to be taken while selecting above values. B. Characteristics requirements of FDD For operator to rely on FDD result FDD should suffice few of the qualities, i.e. few of the characteristic requirements [1], a FDD unit should fulfill are: Quick detection and diagnosis: Early diagnosis of a fault, as early as possible should be done. Isolation-ability:Ability to distinguish between different types of fault. Novel Identifiability: System should identify unknown, novel, fault occurring in the operation. Robustness: System should be less prune to noise and system uncertainty. Adaptability: Diagnostic model should be adaptive to dynamic behavior of the system. Multiple fault identification: System should identify multiple faults. Explanation facility: Decisions and actions of a diagnostic unit should be justified. Diagnostic method should be able to distinguish model uncertainties, process disturbances and real faults. Diagnostic unit in the case of abnormalities come up with set of faults that explains an abnormality. Completeness of a diagnostic unit would require the actual fault to be a subset of the proposed fault set. Resolution of a diagnostic unit would demand fault set to be as minimal as possible. So, a trade-off between completeness and resolution exist. These two concepts would originate whenever modeling and designing of a diagnostic unit is performed. Earlier we have mentioned classification of diagnostic system, now we understand them in more details. Diagnostic methodscan be classified in number of ways due to similarity in approaches in [1], [3]. We classify diagnostic methods as model based and process history (data-driven) based methods. Model based methods are developed based on some fundamental understanding of the physics of the system. This priori knowledge based approach can be realized either by quantitative or qualitative models. Quantitative models can be developed from detailed physics models or simplified physics models, see fig. 1.1. While qualitative methods use either rule based models or qualitative physics based models.

In process history based method, diagnostic systems use available large amount of historical data. These data are transformed and presented to the diagnostic system as a priori knowledge. The process of data transformation is known as feature extraction process. This extraction process can be achievedeither as quantitative or qualitative extraction process(Black box or Gray box). Quantitative model based approach can be further subdivided into detailed and simplified physical models. For modeling transient behavior of a dynamic system, detailed physics model is highly useful.In simplified physicalmodels lumped parameter approach is generally employed, which is computationally simpler because a space-time coupled equations in partial differential form are transformed into ordinary differential and algebraic equations [3].Simplified physical models generally use explicit mathematical models. This enables detection of faulty behavior with an ease. However, quantitative models are complex and difficult to develop[1], [3], [4]. Unlike quantitative model based approach which uses quantitative mathematical relationsto represent knowledge of the system, qualitative models use qualitative relations or knowledge bases to represent the state of the system. Qualitative model based method can be further classified as rule based and qualitative physics based model approaches. Rule based method uses a priori knowledge to derive set of if-then-else rules. They are easy to develop and apply. Thus, this technique enables transparent reasoning[3]. Qualitative model involves deriving qualitative behavior from physical behavior of the system[5]. This methodis useful in noncritical processes [3]. Qualitative model approach is specific to a system and we require complete list of rules to identify faults accurately. Process history based method develops a relationship between measured inputs and measured outputs. Quantitative model develops a mathematical relation from available historical data.When model shows physical relationship between available historical data, then such models are called qualitative models[4]. In other words, when the model parameters or features have no physical significance, they are referred to as black-box models[3].Modelparameters in a model, when designed based on first principle shows physical significance of the system. These models are known as gray box models[3]. Process history based method find its significance where training data are easy to create orcollectand are implemented where no other methods exist[3]. We have seen so far various standard methods for the realization of a diagnostics unit. Now, we see general details about NPP operation and requirements of FDD for NPP. III. REQUIREMENT OF FDD FOR NPP A. Description of NPP

Figure 1.1Classification of Diagnostic methods

Based on the statistics from World nuclear association,

INSTITUTE OF TECHNOLOGY, NIRMA UNIVERSITY, AHMEDABAD – 382 481, 08-10 DECEMBER, 2011

nuclear power plants are delivering about 6% of world energy and about 15% of world electricity. NPPcan be seen as a thermal power station in which the heat source is one or more nuclear reactors. As in a conventional thermal power station the heat is used to generate steam which drivessteam turbine connected to a generator which produces electricity. The heart of a NPP is the Nuclear reactor. In its central part, the reactor core‟s heat is generated by controlled nuclear fission. Coolant circuit removes heat energy from the reactor. Since nuclear fission creates radioactivity, the reactor core is surrounded by a protective shield. This containment absorbs radiation and prevents radioactive material from being released into the environment. In addition, many reactors are equipped with a dome of concrete to protect the reactor against external impacts. Apart from NPP and its facilities, health care facilities, research institutes, natural sources like civil aircrew, mines, etc. also suffers from radiation exposure. Data from International atomic energy agency shows that, for nuclear reactor annual average effective dose for an individual is 1.4mSv; while in civil aircrew effective dose is 3mSv and in metal mines effective dose is 2.7mSv. Still in NPP, safety focuses on unintended conditions or events which may lead to radiological release from authorized activities. In other words, operational safety is a prime concern for those working in nuclear plants. These are supported by continuous monitoring of individual doses and of the work environment to ensure very low radiation exposure compared with other industries.So early detection and quick diagnosis of a fault is of paramount importance for NPP. The complexity and voluminous nature of NPP make it a very difficult task to ascertain whether the plant is operating within acceptable limits. Furthermore, when plant is not operating within acceptable limits, it is difficult to determine what has gone wrong. With use of FDD determination of, part of the system which is failing and which kind of fault it is facing would be accurate.In addition, operator‟s task of handling plant becomes easy. B. FDD in NPP All model based diagnostic methods rely on an explicit model of the plant, under observation. Model based method consists of two steps: 1) Residual generation: Generating inconsistencies between the actual and expected behavior 2) Selection of decision rule Quantitative model based approaches can be realized by analytical redundancy, i.e. observer,extended kalman filters, parity relation, etc., [6]-[9].In analytical redundancy unlike physical redundancy (measurements from multiple sensors are compared to each other), sensor measurements are compared to values computed analytically. Observer based method develops a set observers, each of which is sensitive to a subset of faults and insensitive to the remaining faults.Kalman filters, a recursive algorithm for state estimation, in state space model is equivalent to an optimal predictor for input output model. Extended kalman filters are widely used in non-linear systems by employing two step

3

processes. Firstly linearization of nonlinear models is done. Secondly kalman filters are used on these linear models. General problem with model based methodology is the modeling uncertainties due to deviations in the parameters,linear approximations achieved for nonlinear models can be poor over the wide range andwe need to model faults specifically for novelty identifiability. Diagraphs, fault tree, structural and functional methods are used forqualitative model based approaches[10]-[15]. Signed diagraphs represents the cause-effect relations. Arc is directed from cause node to effect node[1]. In case of fault tree, layers of nodes exist. Fault tree is constructed by using logic gate symbols[16]. This method is top down, deductive method aimed at analyzing the effects of initiating faults and events on a complex process system. Structural approach specifies the connectivity information of a system and subsystem while functional approach specifies the output of a unit as a function of its inputs. Disadvantage with use of qualitative models is the generation of spurious solutions. Considerable work has been done so for reducing spurious solutions like generation of latent constraints in signed directed graphs. Process history method develops model in terms of quantitative or qualitative nature from available data[4], [8], [16]-[19]. Qualitative models are realized using expert systems and Qualitative trend analysis (QTA) while quantitative models are realized using principle component analysis (PCA)/partial least square, neural networks, etc.[10],[20]. Indevelopment of expert system, we have ease in development process, transparent reasoning and the ability to provide explanations for the solutions provided. With QTA, we can achieve process monitoring, malfunction diagnosis and prediction of future states. This method provides valuable information that explains the process behavior. PCA is based on an orthogonal decomposition of the covariance matrix, i.e. data set are reduced to lower dimensions which can still describe the major trends in the original data set.NN refers to the network of biological neurons. Modern usage refers to artificial neural networks (ANN). ANN can be realized using supervised learning strategy or unsupervised learning strategy. They mainly consist of three layers. First layer is input, then hidden layer and finally an output layer. ANN being process historybased approach has minimal modeling requirements. Once network is trained, on-line computation complexity is minimal. ANN does not possess an explanation facility for its actions and also its underlying structure is difficult to understand. We have seen in this sectionvarious approaches used for diagnosis. These methods individually satisfy few of the above listed characteristics requirements of FDD unit while still lacks in few other qualities. Different approaches employed for diagnosis purpose has relative strengths and weaknesses. Also some of these methods can complement one another[1]-[3].Hybrid techniques combine two or more standard techniques with the aim to provide accurate diagnosis to the system over the wide range of operational conditions. In next section, we look upon few of the hybridtechniques proposed or used in NPP for diagnostic purpose.

4

INTERNATIONAL CONFERENCE ON CURRENT TRENDS IN TECHNOLOGY, „NUiCONE – 2011‟

C. Case study: Diagnosis techniques used in NPP 1)Identification of transients in NPP Identification of transients for timely monitoring of nuclear plants operation is of fundamental importance. A fuzzy-logic based method for the identification of transients is being proposed[17]. This approach uses if-then rules, which constitutes the heart of the approach, developed from the available input-output signal data. Marzio uses Quandry based reactor kinetics code (QUARK) for simulation, to obtain necessary data for the identification. QUARK is being used to model the operations of the Westinghouse Advanced Pressurized water reactor, AP600. Code allows simulating the neutronic and thermal-hydraulic transient behavior of a 3-D Light water reactor. In the QUARK, analytical nodal method is used to solve two-group neutron diffusion equation and the upgraded version of the COBRA code is used to develop the core thermal-hydraulic model.The output of the code is taken as the correct (pseudo) experimental values of the process parameters, which are used in the fuzzy model training and identification tasks. Possible four forcing functions considered by Marzio which can originate transients are: - Primary pressure in the core (Pp) - Primary core inlet temperature (Tin) - Core inlet flow rate (Γin) - Boron concentration (CB) Selection of forcing functions allows simulation of various transients listed as: - Leakage for the pressurizer (PP) - Pressure increase (PP) - Increase of coolant inlet temperature (Tin) - Decrease of coolant inlet temperature (Tin) - Loss of coolant accident (Tin) - Rise in boron concentration (CB) - Decrease in boron concentration (CB) The proposed approach by Marzio begins with automatically generating the Fuzzy rule base (FRB) starting from plant input/output data pairs. Set of rules designed maps measured data to system behavior required for the system identification. Each signal is characterized by a classificability threshold, given as:

Where, smax, smin and savg are the maximum, minimum and average value of the signal in the transients. FRB algorithm takes care of possible ambiguities in the classification. During simulation, the responsible forcing function is readily classified with small percentage of errors from signals measured within first 7s after the beginning of the transients. Result by Marzio shows that by increasing the signal variation threshold classification accuracy can be increased.

2) Monitoring system for NPP An on-line monitoring system for the NPP has been developed using the neural networks and the expert system [21]. With conventional monitoring methods in NPP, detecting anomalies can be achieved. But detecting symptoms of anomalies is difficult because of wide error boundary covering power level operation from zero to full[21]. Neural network can detect deviation from the normal state of operation, while interpretation of the deviation by an expert system is performed to diagnose the cause. Neural networks lacks in interpreting the cause of deviation while use of expert systems alone for the plant monitoring is too much computational complexity. Compact simulator of Surry-I, U.S.A. helps in simulating many kinds of malfunctions caused by equipment failure during steady state and transient operation for the testing purpose. Time interval of the simulation is 2 seconds. Nabeshima has considered the Borssele NPP, a two loop pressurized water reactor, with normal electric power output of 477 MWe. On-line data acquisition system sends 72 plant signals to the neuro-expert system every two seconds. Of these 21 most significant signals are selected for the inputs of neural network. Others signals are considered to be unchanged during controlled operation of power rise or fall, so that conventional monitoring method can detect anomaly of these signals. Anomaly detection patterns by NNs are created using PWR simulator. In anomaly detection principle, measured sensor value is compared with predicted value for checking fault level. Software of NN and expert system are programmed in FORTRAN and executed on the PC. The advisory displays show the status of NPP diagnosed by neuro-expert system. Java language based program is used in the graphical advisory displays [21]. NNs are trained by the current and past system input and output for predicting next output of the system. This process of one-step-ahead prediction can be implemented by ANN, and applied for dynamic tracking. Expert system use sensory signals and outputs of the NNS along with information provided from human operators as the input. Few of malfunction cases considered by Nabeshima are listed below: - Small reactor coolant system leak - Leakage of atmospheric steam dump valve - Partial loss of feed water - Turbine governor valves fails - Volume control tank level control fails - Steam generator level control fails Last two malfunctions mentioned above are some kind of controller failures. With conventional alarm system detecting those failures is possible. But with these systems identifying the cause of these anomalies is difficult. The neuro-expert system in next time step can identify the failures for the failed signals because the failed signals show deviations which are much larger than the others. The simple expert system called DISKET is compared with neuro-expert system by Nabeshima. DISKET diagnosis process utilizes information obtained from the conventional alarm system for knowing plant status. Neuro-fuzzy expert

INSTITUTE OF TECHNOLOGY, NIRMA UNIVERSITY, AHMEDABAD – 382 481, 08-10 DECEMBER, 2011

system shows better result in identifying anomaly than DISKET [21]. Nabeshima applied neuro-expert system on Borssele NPP and the on-line PWR simulator. Neural network successfully detected the symptoms of anomalies. The expert system correctly recognized the plant operation mode and diagnosed the plant status. IV. CONCLUSION In this paper, we have shown types of faults which may occur in system. For early detection of such faults, we have explained the need of fault detection and diagnostic unit. To rely on such automatic process system we need fulfillment of some necessary qualities, i.e. characteristic requirements of FDD unit are expressed. Then classifications of diagnostic approach into model and process history based methods are explained in brief. Ihave observed that diagnostic methods realized with help of quantitative models are difficult to develop. Qualitative models have transparent reasoning and are well suited in data-rich environment. While process history based method are applicable for virtually any kind of pattern recognition problem. I have presented here two practical application of FDD in NPP. One is conventional approach of fuzzy logic while second approach isneuro-expert system, where we integrate their individual strengths and compensates their weaknesses.Our observation on conventional and hybrid approach enables, me to decide our future course of action i.e. future work would be aimed to elaborate more on Expert systemand ANN approaches along with redundancy method, so that we can prepare a better framework of diagnostic unit which accommodate all of these methods to yield wide operation range of diagnostic unit and still, retaining accuracy of our result. V. HELPFUL HINTS A. Abbreviations and Acronyms AEM: Abnormal event management ANN: Artificial neural network FDD: Fault detection and diagnosis FRB: Fuzzy rule base NN: Neural network NPP: Nuclear power plant PCA: Principle component analysis PWR: Pressurized water reactor QTA: Qualitative trend analysis QUARK: Quandry based reactor kinetics code VI. REFERENCES [1]

[2]

Venkat Venkatasubramanian, Raghunathan Rengaswamy, Kewen Yin and Surya N. Kavuri, "A review of process fault detection and diagnosis Part I: Quantitative model-based methods", Computers & Chemical Engineering 27, pp. 293-311, Apr. 2002. Jianping Ma and Jin Jiang, “Application of Fault detection and diagnosis methods in Nuclear power plants: A review”, Progress in nuclear energy 53, pp. 255-266, Dec. 2010.

[3] [4]

[5]

[6] [7] [8] [9]

[10]

[11]

[12] [13] [14] [15]

[16]

[17] [18] [19] [20]

[21]

5

Srinivas Katipamula and Michael R. Brambley, “Methods for Fault detection, diagnosis and prognostics for building systems- A review, Part I”, HVAC&R Research, vol. 11, No. 1., July 2004. Venkat Venkatasubramanian, Raghunathan Rengaswamy, Surya N. Kavuri and Kewen Yin, “A review of process fault detection and diagnosis Part III: Process history based methods”, Computers & Chemical Engineering 27, pp. 327-346, Apr. 2002. Venkat Venkatasubramanian, Raghunathan Rengaswamy and Surya N. Kavuri, “A review of process fault detection and diagnosis Part II: Qualitative models and search strategies”, Computers & Chemical Engineering 27, pp. 313-326, Apr. 2002. A. Ashokan and D. Sivakumar, “Fault detection and diagnosis for a three-tank system using structured residual approach”, ICGST-ACSE Journal, Vol.7, Issue 2, Nov. 2007. Edward Y. Cho and Alan S. Willsky, “Analytical redundancy and the design of robust failure detection systems”, Automatic Control, Vol. AC-29, No. 7, pp. 603-614, July 1984. C. Angeli, “On-line fault detection techniques for technical systems: A survey”, International Journal of Computer Science & Applications Vol. I, No. 1, pp. 12-30, 2004. Richard Dorr, Frederic Kratz, Jose Ragot, Francois Loisy and Jean-Luc Germain, “Detection, Isolation and Identification of sensor faults in Nuclear Power Plants”, Control Systems Technology Vol. 5, No. 1, Jan. 1997. Chun-ling Xie, Jen-Yuan (James) Chang, Xiao-cheng Shi and Jing-min Dai, “Fault diagnosis of nuclear power plant based on genetic-RBF neural network”, 15th International conference on Mechatronics and Machine vision in Practice, Dec. 2008. Irina Obreja, “Diagnosis of power plant faults using qualitative models and heuristic rules”, IEA/AIE ’90 Proceedings of the 3 rd International conference on Industrial and engineering applications of artificial intelligence and expert systems Vol. 1, pp. 41-46, 1990. Hiranmayee Vedam, “Signed Diagraph based multiple fault diagnosis”, Computers & Chemical Engineering Vol. 21, Issue 1-2, pp. 8655-8660, 1997. H. Chung, Z. Bien, J. Park, and P. Seong, “Incipient multiple fault diagnosis in real time with application to large-scale systems”,Nuclear Science, vol. 33, no. 8, pp. 1943–1954, Aug. 1994. Joon On Yang and Soon Heung Chang, “A Diagnostic expert system for the nuclear power plant based on the Hybrid Knowledge Approach”, Nuclear Science Vol. 36, No. 6, pp. 2450-2458, Dec. 1989. Chun-ling Xie, Jen-Yuan (James) Chang, Xiao-cheng Shi and Jing-min Dai, “Fault diagnosis of nuclear power plant based on genetic-RBF neural network”, International Journal of Computer Applications in Technology Vol. 39, No. 1/2/3, pp. 159-165, 2010. Enrico Zio, Piero Baraldi and Irina Crenguta Popescu, “A fuzzy decision tree method for fault classification in the steam generator of a pressurized water reactor”, Annals of Nuclear Energy 36, pp. 11591169, May 2009. Marzio Marseguerra, Enrico Zio, Andrea Oldrini and Enrico Brega, “Fuzzy identification of transients in nuclear power plants”, Nuclear Engineering and Design 225, pp. 285-294, Jun 2003. Dieter Wach, “Experience with knowledge based systems for maintenance diagnosis”, Control systems, Robotics and Automation Vol. XVI, No. 1/2/3, pp. 159-165, 2010. C. Angeli, “Diagnostic expert systems: From expert‟s knowledge to real-time systems”,Advanced knowledge based systems: Model, Application & Research Vol. 1, pp. 50-73, 2010. Shengwei Wang and Jingtan Cui, “Sensor- fault detection, diagnosis and estimation for centrifugal chiller systems using principal component analysis method”, Applied Engineering 82, pp. 197-213, Feb. 2005. K. Nabeshima, T. Suzudo, S. Seker, E. Ayaz, B. Barutcu, E. Turkcan, T. Ohno and K. Kudo, “On-line Neuro-Expert monitoring system for Borssele Nuclear Power Plant”, Progress in Nuclear Energy Vol. 43, No. 1-4, pp. 397-404, 2003.