Application of artificial neural networks to nuclear power plant transient diagnosis

A study on various artificial neural network (ANN) algorithms for selecting a best suitable algorithm for diagnosing the transients of a typical nuclear power plant (NPP) is presented. NPP experiences a number of transients during its operations. These transients may be due to equipment failure, malfunctioning of process systems, etc. In case of any undesired plant condition generally known as initiating event (IE), the operator has to carry out diagnostic and corrective actions. The objective of this study is to develop a neural network based framework that will assist the operator to identify such initiating events quickly and to take corrective actions. Optimization study on several neural network algorithms has been carried out. These algorithms have been trained and tested for several initiating events of a typical nuclear power plant. The study shows that the resilient-back propagation algorithm is best suitable for this application. This algorithm has been adopted in the development of operator support system. The performance of ANN for several IEs is also presented.     

Safety of nuclear power plants

The safety of operating nuclear power plants of the CANDU type is described in this paper. The need for a systematic study on these types of heavy water reactors similar to the safety studies done on light water reactors is brought out in this paper. Some of the work done on station blackout, operational transients, small and large break loss of coolant accidents is reviewed. Recent nuclear power plant accidents, namely Three-Mile Island-2 and Chernobyl, seem to indicate that an understanding of man-machine interaction and human behaviour under stress is important for the safety aspects and more work needs to be done in these areas.     

Symptom based diagnostic system for nuclear power plant operations using artificial neural networks

Nuclear power plant experiences a number of transients during its operations. These transients may be due to equipment failure, malfunctioning of process systems and unavailability of safety systems. In such a situation, the plant may result into an abnormal state which is undesired. In case of an undesired plant condition generally known as an initiating event (IE), the operator has to carry out diagnostic and corrective actions. The operator's response may be too late to mitigate or minimize the negative consequences in such scenarios. The objective of this work is to develop an operator support system based on artificial neural networks that will assist the operator to identify the IEs at the earliest stages of their developments. These abnormal plant conditions must be diagnosed and identified through the process instrument readings. A symptom based diagnostic system has been developed to investigate the IEs. The event identification is carried out by using resilient back propagation neural network algorithm. Whenever an event is detected, the system will display the necessary operator actions in addition to the type of IE. The system will also show the graphical trend of relevant parameters. The developed system is able to identify the eight IEs of Narora Atomic Power Station. This paper describes the features of the diagnostic system taking one of the IEs as a case study.     

A data-driven fuzzy approach for predicting the remaining useful life in dynamic failure scenarios of a nuclear system

This paper presents a similarity-based approach for prognostics of the Remaining Useful Life (RUL) of a system, i.e. the lifetime remaining between the present and the instance when the system can no longer perform its function. Data from failure dynamic scenarios of the system are used to create a library of reference trajectory patterns to failure. Given a failure scenario developing in the system, the remaining time before failure is predicted by comparing by fuzzy similarity analysis its evolution data to the reference trajectory patterns and aggregating their times to failure in a weighted sum which accounts for their similarity to the developing pattern. The prediction on the failure time is dynamically updated as time goes by and measurements of signals representative of the system state are collected. The approach allows for the on-line estimation of the RUL. For illustration, a case study is considered regarding the estimation of RUL in failure scenarios of the Lead Bismuth Eutectic eXperimental Accelerator Driven System (LBE-XADS).     

Efficiency parameters estimation in gemstones cut design using artificial neural networks

This paper deals with the problem of estimating cut results for faceted gemstones. The proposed approach applies artificial neural networks for a faceted gemstones analysis tool that could be further developed for incorporation in a computer-aided-design (CAD) context. Basic concepts concerning gemstone processing are introduced and the design of computational tools using neural networks is discussed. The model presented proposes two criteria to assess the efficiency of lapidary designs for rock crystal quartz: brilliance and yield. Closing the article, 62 different lapidary models were used to train and test the neural network tool.     

Water distribution under trickle irrigation predicted using artificial neural networks

An artificial neural network (ANN) technology is presented as an alternative to physical-based modeling of subsurface water distribution from trickle emitters. Three options are explored to prepare input–output functional relations from a database created using a numerical model (HYDRUS-2D). From the database the feasibility and advantages of the three alternative options are evaluated: water-content at defined coordinates, moment analysis describing the shape of the plume, and coordinates of individual water-content contours. The best option is determined in a way by the application objectives, but results suggest that prediction using moment analyses is probably the most versatile and robust and gives an adequate picture of the subsurface distribution. Of the other two options, the direct determination of the individual water contours was subjectively judged to be more successful than predicting the water content at given coordinates, at least in terms of describing the subsurface distribution. The results can be used to estimate subsurface water distribution for essentially any soil properties, initial conditions or flow rates for trickle sources.     

Prediction of FRP-confined compressive strength of concrete using artificial neural networks

Strengthening and retrofitting of concrete columns by wrapping and bonding FRP sheets has become an efficient technique in recent years. Considerable investigations have been carried out in the field of FRP-confined concrete and there are many proposed models that predict the compressive strength which are developed empirically by either doing regression analysis using existing test data or by a development based on the theory of plasticity. In the present study, a new approach is developed to obtain the FRP-confined compressive strength of concrete using a large number of experimental data by applying artificial neural networks. Having parameters used as input nodes in ANN modeling such as characteristics of concrete and FRP, the output node was FRP-confined compressive strength of concrete. The idealized neural network was employed to generate empirical charts and equations for use in design. The comparison of the new approach with existing empirical and experimental data shows good precision and accuracy of the developed ANN-based model in predicting the FRP-confined compressive strength of concrete.     

Design of artificial neural networks using a genetic algorithm to predict collection efficiency in venturi scrubbers

In this study, a new approach for the auto-design of neural networks, based on a genetic algorithm (GA), has been used to predict collection efficiency in venturi scrubbers. The experimental input data, including particle diameter, throat gas velocity, liquid to gas flow rate ratio, throat hydraulic diameter, pressure drop across the venturi scrubber and collection efficiency as an output, have been used to create a GA-artificial neural network (ANN) model. The testing results from the model are in good agreement with the experimental data. Comparison of the results of the GA optimized ANN model with the results from the trial-and-error calibrated ANN model indicates that the GA-ANN model is more efficient. Finally, the effects of operating parameters such as liquid to gas flow rate ratio, throat gas velocity, and particle diameter on collection efficiency were determined.     

Identifying significant predictors of injury severity in traffic accidents using a series of artificial neural networks

Understanding the circumstances under which drivers and passengers are more likely to be killed or more severely injured in an automobile accident can help improve the overall driving safety situation. Factors that affect the risk of increased injury of occupants in the event of an automotive accident include demographic or behavioral characteristics of the person, environmental factors and roadway conditions at the time of the accident occurrence, technical characteristics of the vehicle itself, among others. This study uses a series of artificial neural networks to model the potentially non-linear relationships between the injury severity levels and crash-related factors. It then conducts sensitivity analysis on the trained neural network models to identify the prioritized importance of crash-related factors as they apply to different injury severity levels. In the process, the problem of five-class prediction is decomposed into a set of binary prediction models (using a nationally representative sample of 30358 police-recorded crash reports) in order to obtain the granularity of information needed to identify the "true" cause and effect relationships between the crash-related factors and different levels of injury severity. The results, mostly validated by the findings of previous studies, provide insight into the changing importance of crash factors with the changing injury severity levels.     

Cyber-based design for additive manufacturing using artificial neural networks for Industry 4.0

Additive Manufacturing (AM) requires integrated networking, embedded controls and cloud computing technologies to increase their efficiency and resource utilisation. However, currently there is no readily applicable system that can be used for cloud-based AM. The objective of this research is to develop a framework for designing a cyber additive manufacturing system that integrates an expert system with Internet of Things (IoT). An Artificial Neural Network (ANN) based expert system was implemented to classify input part designs based on CAD data and user inputs. Three ANN algorithms were trained on a knowledge base to identify optimal AM processes for different part designs. A two-stage model was used to enhance the prediction accuracy above 90% by increasing the number of input factors and datasets. A cyber interface was developed to query AM machine availability and resource capability using a Node-RED IoT device simulator. The dynamic AM machine identification system developed using an application programme interface (API) that integrates inputs from the smart algorithm and IoT interface for real-time predictions. This research establishes a foundation for the development of a cyber additive design for manufacturing system which can dynamically allocate digital designs to different AM techniques over the cyber network.     

A new method to evaluate alternate AC power source effects in multi-unit nuclear power plants

In order to evaluate accurately a station blackout (SBO) event frequency of a multi-unit nuclear power plant that has a shared alternate AC (AAC) power source, an approach has been developed which accommodates the complex inter-unit behavior of the shared AAC power source under multi-unit loss of offsite power conditions. The SBO frequency at a target unit of probabilistic safety assessment could be underestimated if the inter-unit dependency of the shared AAC power source is not properly modeled.The approach is illustrated for two cases, 2 units and 4 units at a single site, and generalized for a multi-unit site. Furthermore, the SBO frequency of the first unit of the 2-unit site is quantified. The methodology suggested in the present paper is believed to be very useful in evaluating the SBO frequency and the core damage frequency resulting from the SBO event. This approach is also applicable to the probabilistic evaluation of the other shared systems in a multi-unit nuclear power plant.     

基于神经网络的长距离油气管道安全预警系统

提出了一种基于分布式光纤传感和人工神经网络判别的长距离输油管道安全预警系统.该系统利用光纤传感器收集管道周围土壤的振动信号,通过神经网络判断是否存在针对管道的破坏性行为和判别破坏性行为的类别,实现对油气管道的长距离安全预警.系统在预处理阶段对信号大幅度降维,降低数据处理的时间复杂度,以满足实时性的要求.在识别阶段则采用人工神经网络模型,包括反向传播(BP)网络和支持向量机(SVM).试验结果表明,这两种神经网络模型对打夯、镐刨、电钻三类破坏行为的识别率分别达到96.5和97.1%,均优于以往文献中的报道.     

Crushing behavior of laterally compressed composite elliptical tubes: Experiments and predictions using artificial neural networks

Composite materials have been increasingly used in the automobile industry for weight saving and part integration purposes. In this regard, composite elliptical tubes have been effectively employed as energy absorber devices. This increases the need for accurate and simple prediction techniques to optimize these structures.

The present work deals with the implementation of artificial neural networks (ANN) technique in the prediction of the crushing behavior and energy absorption characteristics of laterally loaded glass fiber/epoxy composite elliptical tubes. Predicted results are compared with actual experimental data in terms of load carrying capacity and energy absorption capability showing good agreement. This shows that ANN techniques could effectively be used to predict the response of collapsible composite energy absorber devices subjected to different loading conditions. As is the case for experimental findings, the predictions obtained using ANN also show the significant effect of the ellipticity ratio on the crushing behavior of laterally loaded tubes.     

Reliability analysis of laminated composite structures using finite elements and neural networks

Saving of computer processing time on the reliability analysis of laminated composite structures using artificial neural networks is the main objective of this work. This subject is particularly important when the reliability index is a constraint in the optimization of structural performance, because the task of looking for an optimum structural design demands also a very high processing time. Reliability methods, such as Standard Monte Carlo (SMC), Monte Carlo with Importance Sampling (MC–IS), First Order Reliability Method (FORM) and FORM with Multiple Check Points (FORM–MCPs) are used to compare the solution and the processing time when the Finite Element Method (FEM) is employed and when the finite element analysis (FEA) is substituted by trained artificial neural networks (ANNs). Two ANN are used here: the Multilayer Perceptron Network (MPN) and the Radial Basis Network (RBN). Several examples are presented, including a shell with geometrically non-linear behavior, which shows the advantages using this methodology.     

基于神经网络和功率谱的水中目标信号检测方法

将功率谱和神经网络相结合，应用于高海况、低信噪比条件下，水中目标信号的特征提取中．文中首先对信号进行功率谱估计，利用目标信号功率主要集中在低频部分的特点，提取低频信号的能量作为特征，然后利用人工神经网络对目标信号进行检测．利用不同浪级情况下海洋水压场的仿真信号数据，对某型目标舰船的水压信号进行了检测计算，验证了该方法的有效性，尤其是达到了在高海况、低信噪比条件下，对目标信号检测率比较高、虚警率比较低的效果．     

Fatigue life prediction of composite materials using polynomial classifiers and recurrent neural networks

Due to their massively parallel structure and ability to learn by example, artificial neural networks can deal with nonlinear problems for which an accurate analytical solution is difficult to obtain. These networks have been used in modeling the mechanical behavior of fiber-reinforced composite materials. Although promising results were obtained using such networks, more investigation on the appropriate choice of their structure and their performance in the presence of limited and noisy data is needed. On the other hand, polynomials networks have been known to have excellent properties as classifiers and are universal approximators to the optimal Bayes classifier. Not being dependant on various user defined parameters, having less computational requirements makes their use over other methods, such as neural networks, an advantage.

In this work, the fatigue behavior of unidirectional glass fiber/epoxy composite laminae under tension–tension and tension–compression loading is predicted using feedforward and recurrent neural networks. These predictions are compared to those obtained using polynomial classifiers. Experimental data obtained for fiber orientation angles of 0°, 19°, 45°, 71° and 90° under stress ratios of 0.5, 0 and –1 is used.

It is shown that, even when a small number of experimental data points is used to train both polynomial classifiers and neural networks, the predictions obtained are comparable to other current fatigue life-prediction methods. Also, polynomial classifiers are shown to provide accurate modeling between the input parameters (maximum stress, R-ratio, fiber orientation angle) and the number of cycles to failure when compared to neural networks.     

Nonlinear constitutive models for FRP composites using artificial neural networks

This paper presents a new approach to generate nonlinear and multi-axial constitutive models for fiber reinforced polymeric (FRP) composites using artificial neural networks (ANNs). The new nonlinear ANN constitutive models are complete and have been integrated with displacement-based FE software for the nonlinear analysis of composite structures. The proposed ANN constitutive models are trained with experimental data obtained from off-axis tension/compression and pure shear (Arcan) tests. The proposed ANN constitutive model is generated for plane–stress states with assumed functional response in some parts of the multi-axial stress space with no experimental data. The ability of the trained ANN models to predict material response is examined directly and through FE analysis of a notched composite plate. The experimental part of this study involved coupon testing of thick-section pultruded FRP E-glass/polyester material. Nonlinear response was pronounced including in the fiber direction due to the relatively low overall fiber volume fraction (FVF). Notched composite plates were also tested to verify the FE, with ANN material models, to predict general non-homogeneous responses at the structural level.     

A study on the development of a task complexity measure for emergency operating procedures of nuclear power plants

In this study, a measure called task complexity (TACOM) that can quantify the complexity of tasks stipulated in emergency operating procedures of nuclear power plants is developed. The TACOM measure consists of five sub-measures that can cover remarkable complexity factors: (1) amount of information to be managed by operators, (2) logical entanglement due to the logical sequence of the required actions, (3) amount of actions to be accomplished by operators, (4) amount of system knowledge in recognizing the problem space, and (5) amount of cognitive resources in establishing an appropriate decision criterion. The appropriateness of the TACOM measure is investigated by comparing task performance time data with the associated TACOM scores. As a result, it is observed that there is a significant correlation between TACOM scores and task performance time data. Therefore, it is reasonable to expect that the TACOM measure can be used as a meaningful tool to quantify the complexity of tasks.     

An artificial neural network-based expert system for the appraisal of two-car crash accidents

This paper employs artificial neural network (ANN) to develop an accident appraisal expert system. Two ANN models-- party-based and case-based-- with different hidden neurons are trained and validated by k-fold (k=3) cross validation method. A total of 537 two-car crash accidents (1074 parties involved) are randomly and equally divided into three subsets. For the comparison, a discrimination analysis (DA) model is also calibrated. The results show that the ANN model can achieve a high correctness rate of 85.72% in training and 77.91% in validation and a low Schwarz's Bayesian information criterion (SBC) of -0.82 in training and 0.13 in validation, which indicates that the ANN model is suitable for accident appraisal. Furthermore, in order to measure the importance of each explanatory variable, a general influence (GI) index is computed based on the trained weights of ANN. It is found that the most influential variable is right-of-way, followed by location and alcoholic use. This finding concurs with the prior knowledge in accident appraisal. Thus, for the fair assessment of accident liabilities the correctness of these three key variables is of critical importance to police investigation reports.