Fault diagnostics of dynamic system operation using a fault tree based method

For conventional systems, their availability can be considerably improved by reducing the time taken to restore the system to the working state when faults occur. Fault identification can be a significant proportion of the time taken in the repair process. Having diagnosed the problem the restoration of the system back to its fully functioning condition can then take place. This paper expands the capability of previous approaches to fault detection and identification using fault trees for application to dynamically changing systems. The technique has two phases. The first phase is modelling and preparation carried out offline. This gathers information on the effects that sub-system failure will have on the system performance. Causes of the sub-system failures are developed in the form of fault trees. The second phase is application. Sensors are installed on the system to provide information about current system performance from which the potential causes can be deduced. A simple system example is used to demonstrate the features of the method. To illustrate the potential for the method to deal with additional system complexity and redundancy, a section from an aircraft fuel system is used. A discussion of the results is provided.     

Mathematical foundations of event trees

A mathematical foundation from first principles of event trees is presented. The main objective of this formulation is to offer a formal basis for developing automated computer assisted construction techniques for event trees. The mathematical theory of event trees is based on the correspondence between the paths of the tree and the elements of the outcome space of a joint event. The concept of a basic cylinder set is introduced to describe joint event outcomes conditional on specific outcomes of basic events or unconditional on the outcome of basic events. The concept of outcome space partition is used to describe the minimum amount of information intended to be preserved by the event tree representation. These concepts form the basis for an algorithm for systematic search for and generation of the most compact (reduced) form of an event tree consistent with the minimum amount of information the tree should preserve. This mathematical foundation allows for the development of techniques for automated generation of event trees corresponding to joint events which are formally described through other types of graphical models. Such a technique has been developed for complex systems described by functional blocks and it is reported elsewhere. On the quantification issue of event trees, a formal definition of a probability space corresponding to the event tree outcomes is provided. Finally, a short discussion is offered on the relationship of the presented mathematical theory with the more general use of event trees in reliability analysis of dynamic systems.     

Human communication, mutual awareness and system dependability. Lessons learnt from air-traffic control field studies

The dependability of many complex and critical systems strongly relies on human operators, both through human reliability and human ability to handle adequately the unexpected events. This paper focuses on ergonomics field studies of air traffic control activities, and more specifically on the analyses of communication within teams of controllers. We show how operators use spontaneously the natural redundancy and diversity of human communication (multimodality, addressing features,…), so as to successfully maintain mutual awareness. This is the key for reliable cooperation, for the sake of global system dependability that rests on mechanisms such as error detection, recovery, and prevention (by anticipation and regulation). This study helps in providing specifications for the design of systems efficiently supporting both human cooperation and human ability to contribute to dependability.     

Production reliability in flexible manufacturing systems

A typical flexible manufacturing system, Westland Helicopters' sheet metal detail manufacturing complex, has been analysed for reliability. The techniques of fault tree analysis and event tree analysis are presented and their applicability to this study investigated. Event tree analysis has been found to be a more effective method for analysing manufacturing systems. The failure states of the system have been identified from the construction of an event tree which considers random hardware faults that influence production. Failure rate data have been used to quantify the critical production failure states in terms of machine failures. Estimates are made of the system's MTTF and percentage availability using typical MTTR figures. The probability that a selected production route fails to complete the manufacture of a set of parts is also evaluated. A dependency of systems reliability on the production demand has been discovered, and a possible method for modelling and assessing the reliability of systems capable of producing several products is proposed.     

An ordering heuristic to develop the binary decision diagram based on structural importance

Fault tree analysis is often used to assess risks within industrial systems. The technique is commonly used although there are associated limitations in terms of accuracy and efficiency when dealing with large fault tree structures. The most recent approach to aid the analysis of the fault tree diagram is the Binary Decision Diagram (BDD) methodology. To utilise the technique the fault tree structure needs to be converted into the BDD format. Converting the fault tree requires the basic events of the tree to be placed in an ordering. The ordering of the basic events is critical to the resulting size of the BDD, and ultimately affects the performance and benefits of this technique. A number of heuristic approaches have been developed to produce an optimal ordering permutation for a specific tree. These heuristic approaches do not always yield a minimal BDD structure for all trees. This paper looks at a heuristic that is based on the structural importance measure of each basic event. Comparing the resulting size of the BDD with the smallest generated from a set of six alternative ordering heuristics, this new structural heuristic produced a BDD of smaller or equal dimension on 77% of trials.     

A fault tree analysis strategy using binary decision diagrams

The use of binary decision diagrams (BDDs) in fault tree analysis provides both an accurate and efficient means of analysing a system. There is a problem, however, with the conversion process of the fault tree to the BDD. The variable ordering scheme chosen for the construction of the BDD has a crucial effect on its resulting size and previous research has failed to identify any scheme that is capable of producing BDDs for all fault trees. This paper proposes an analysis strategy aimed at increasing the likelihood of obtaining a BDD for any given fault tree, by ensuring the associated calculations are as efficient as possible. The method implements simplification techniques, which are applied to the fault tree to obtain a set of ‘minimal’ subtrees, equivalent to the original fault tree structure. BDDs are constructed for each, using ordering schemes most suited to their particular characteristics. Quantitative analysis is performed simultaneously on the set of BDDs to obtain the top event probability, the system unconditional failure intensity and the criticality of the basic events.     

A stochastic approach to consequence tree analysis

In this paper a stochastic approach to consequence tree analysis is considered. A consequence tree is a set of events logically combined by OR and AND connections that occur in sequence, some being prerequisites for others. Consequence trees are applicable to failure propagation in plants. Facilitating paths and inhibiting cuts are defined and considered. The distribution of the time the system needs to reach a certain top event is obtained. Probability weights are defined that can be used to obtain the weakest link in the consequence tree.     

Criteria for evaluating protection from single points of failure for partially expanded fault trees

Fault tree analysis (FTA) is a technique that describes the combinations of events in a system which result in an undesirable outcome. FTA is used as a tool to quantitatively assess a system's probability for an undesirable outcome. Time constraints from concept to production in modern engineering often limit the opportunity for a thorough statistical analysis of a system. Furthermore, when undesirable outcomes are considered such as hazard to human(s), it becomes difficult to identify strict statistical targets for what is acceptable. Consequently, when hazard to human(s) is concerned a common design target is to protect the system from single points of failure (SPOF) which means that no failure mode caused by a single event, concern, or error has a critical consequence on the system. Such a design target is common with “by-wire” systems. FTA can be used to verify if a system is protected from SPOF. In this paper, sufficient criteria for evaluating protection from SPOF for partially expanded fault trees are proposed along with proof. The proposed criteria consider potential interactions between the lowest drawn events of a partial fault tree expansion which otherwise easily leads to an overly optimistic analysis of protection from SPOF. The analysis is limited to fault trees that are coherent and static.     

Anomaly Detection

The paper presents a revolutionary framework for the modeling, detection, characterization, identification, and machine-learning of anomalous behavior in observed phenomena arising from a large class of unknown and uncertain dynamical systems. An evolved behavior would in general be very difficult to correct unless the specific anomalous event that caused such behavior can be detected early, and any consequence attributed to the specific anomaly following its detection. Substantial investigative time and effort is required to back-track the cause for abnormal behavior and to recreate the event sequence leading to such abnormal behavior. The need to automatically detect anomalous behavior is therefore critical using principles of state motion, and to do so with a human operator in the loop. Human-machine interaction results in a capability for machine self-learning and in producing a robust decision-support mechanism. This is the fundamental concept of intelligent control wherein machine-learning is enhanced by interaction with human operators.     

An enhanced component connection method for conversion of fault trees to binary decision diagrams

Fault tree analysis (FTA) is widely applied to assess the failure probability of industrial systems. Many computer packages are available, which are based on conventional kinetic tree theory methods. When dealing with large (possibly non-coherent) fault trees, the limitations of the technique in terms of accuracy of the solutions and the efficiency of the processing time become apparent. Over recent years, the binary decision diagram (BDD) method has been developed that solves fault trees and overcomes the disadvantages of the conventional FTA approach. First of all, a fault tree for a particular system failure mode is constructed and then converted to a BDD for analysis. This paper analyses alternative methods for the fault tree to BDD conversion process.For most fault tree to BDD conversion approaches, the basic events of the fault tree are placed in an ordering. This can dramatically affect the size of the final BDD and the success of qualitative and quantitative analyses of the system. A set of rules is then applied to each gate in the fault tree to generate the BDD. An alternative approach can also be used, where BDD constructs for each of the gate types are first built and then merged to represent a parent gate. A powerful and efficient property, sub-node sharing, is also incorporated in the enhanced method proposed in this paper. Finally, a combined approach is developed taking the best features of the alternative methods. The efficiency of the techniques is analysed and discussed.     

Automatic Detection and Analysis of Voltage Events in Power Systems

This paper presents an automatic system for real-time detection and analysis of voltage events in power systems. The voltage events considered are those that cause a temporary increase or decrease in the rms voltage magnitude over the limits recommended in the international standards. The system proposed uses three Kalman filters to detect when a voltage event begins and to estimate the three-phase voltage supply during the event. The results obtained from real measurements in a low-voltage distribution network show that the system accurately detects and analyzes in real time different types of voltage events in power systems.     

中国沙尘暴及其监测预警

利用1954—2001年中国大陆681个站的气象资料，分析了中国沙尘暴的时空分布特征和分级标准，给出了沙尘暴若干观测事实。通过分析沙尘暴的发生发展机制，讨论了沙尘暴的定性预报和定量预报技术，指出发展包含大气运动、风蚀物理过程和下垫面地表信息的集合数值预报模式是沙尘暴定量预警技术的发展方向。同时，指出传统的监测网尚不能全面监测沙尘暴的发生和发展，建立包含下垫面生态环境要素、大气物理化学要素的专业化沙尘监测网络是深化对沙尘暴特性认识和发展沙尘暴定量预警技术的基础。     

Spatial generalized linear mixed models of electric power outages due to hurricanes and ice storms

This paper presents new statistical models that predict the number of hurricane- and ice storm-related electric power outages likely to occur in each 3 km×3 km grid cell in a region. The models are based on a large database of recent outages experienced by three major East Coast power companies in six hurricanes and eight ice storms. A spatial generalized linear mixed modeling (GLMM) approach was used in which spatial correlation is incorporated through random effects. Models were fitted using a composite likelihood approach and the covariance matrix was estimated empirically. A simulation study was conducted to test the model estimation procedure, and model training, validation, and testing were done to select the best models and assess their predictive power. The final hurricane model includes number of protective devices, maximum gust wind speed, hurricane indicator, and company indicator covariates. The final ice storm model includes number of protective devices, ice thickness, and ice storm indicator covariates. The models should be useful for power companies as they plan for future storms. The statistical modeling approach offers a new way to assess the reliability of electric power and other infrastructure systems in extreme events.     

Posbist fault tree analysis of coherent systems

When the failure probability of a system is extremely small or necessary statistical data from the system is scarce, it is very difficult or impossible to evaluate its reliability and safety with conventional fault tree analysis (FTA) techniques. New techniques are needed to predict and diagnose such a system's failures and evaluate its reliability and safety. In this paper, we first provide a concise overview of FTA. Then, based on the posbist reliability theory, event failure behavior is characterized in the context of possibility measures and the structure function of the posbist fault tree of a coherent system is defined. In addition, we define the AND operator and the OR operator based on the minimal cut of a posbist fault tree. Finally, a model of posbist fault tree analysis (posbist FTA) of coherent systems is presented. The use of the model for quantitative analysis is demonstrated with a real-life safety system.     

Integrated system fault diagnostics utilising digraph and fault tree-based approaches

With the growing intolerance to failures within systems, the issue of fault diagnosis has become ever prevalent. Information concerning these possible failures can help to minimise the disruption to the functionality of the system by allowing quick rectification. Traditional approaches to fault diagnosis within engineering systems have focused on sequential testing procedures and real-time mechanisms. Both methods have been predominantly limited to single fault causes. Latest approaches also consider the issue of multiple faults in reflection to the characteristics of modern day systems designed for high reliability. In addition, a diagnostic capability is required in real time and for changeable system functionality. This paper focuses on two approaches which have been developed to cater for the demands of diagnosis within current engineering systems, namely application of the fault tree analysis technique and the method of digraphs. Both use a comparative approach to consider differences between actual system behaviour and that expected. The procedural guidelines are discussed for each method, with an experimental aircraft fuel system used to test and demonstrate the features of the techniques. The effectiveness of the approaches is compared and their future potential highlighted.     

Fault detection using parameter estimation

This paper presents a brief summary of fault-detection methods using parameter estimation techniques. An overview of the fault-detection system design methodology is first presented, followed by the principles of parameter-estimation fault-detection techniques. Applications from the field of industrial processes are given and finally a case study is described which applies the general techniques to the fault detection of D.C. motors using multiprocessor systems.     

An Integrated mining approach to discover business process models with parallel structures: towards fitness improvement

Process mining (PM) is a technique to extract a process model from an event log to represent the process behaviour recorded in that event log. A mined process model with high fitness means that it can reflect most of the process behaviour recorded in the event log. Previous studies have shown that the mined model with high fitness can be used in process improvement, such as fraud detection, continuous process improvement and benchmarking. Genetic process mining (GPM) is a famous PM approach, which can simultaneously identify several process structures from event logs. However, GPM cannot effectively discover parallel structures from event logs. This study proposes a PM approach based on integration of GPM, particle swarm optimisation and differential evolution to find process models with high fitness for event logs involving multiple parallel structures. The results show that the proposed approach does indeed lead to improvement in gaining process models with high fitness for event logs involving multiple parallel structures.     

Attack–norm separation for detecting attack‐induced quality problems on computers and networks

Cyber attacks on computer and network systems induce system quality and reliability problems, and present a significant threat to the computer and network systems that we are heavily dependent on. Cyber attack detection involves monitoring system data and detecting the attack‐induced quality and reliability problems of computer and network systems caused by cyber attacks. Usually there are ongoing normal user activities on computer and network systems when an attack occurs. As a result, the observed system data may be a mixture of attack data and normal use data (norm data). We have established a novel attack–norm separation approach to cyber attack detection that includes norm data cancelation to improve the data quality as an important part of this approach. Aiming at demonstrating the importance of norm data cancelation, this paper presents a set of data modeling and analysis techniques developed to perform norm data cancelation before applying an existing technique of anomaly detection, the chi‐square distance monitoring (CSDM), to residual data obtained after norm data cancelation for cyber attack detection. Specifically, a Markov chain model of norm data and an artificial neural network (ANN) of norm data cancelation are developed and tested. This set of techniques is compared with using CSDM alone for cyber attack detection. The results show a significant improvement of detection performance by CSDM with norm data cancelation over CSDM alone. Copyright © 2006 John Wiley & Sons, Ltd.     

A New Database Intrusion Detection Approach Based on Hybrid Meta-Heuristics

A new secured database management system architecture using intrusion detection systems (IDS) is proposed in this paper for organizations with no previous role mapping for users. A simple representation of Structured Query Language queries is proposed to easily permit the use of the worked clustering algorithm. A new clustering algorithm that uses a tube search with adaptive memory is applied to database log files to create users’ profiles. Then, queries issued for each user are checked against the related user profile using a classifier to determine whether or not each query is malicious. The IDS will stop query execution or report the threat to the responsible person if the query is malicious. A simple classifier based on the Euclidean distance is used and the issued query is transformed to the proposed simple representation using a classifier, where the Euclidean distance between the centers and the profile’s issued query is calculated. A synthetic data set is used for our experimental evaluations. Normal user access behavior in relation to the database is modelled using the data set. The false negative (FN) and false positive (FP) rates are used to compare our proposed algorithm with other methods. The experimental results indicate that our proposed method results in very small FN and FP rates.