Neural Network Expert System Research Articles

Reliable Star Pattern Identification Technique by Using Neural Networks

An approach to systematically structuring star patterns and an associated identification technique using a neural network and rule-based expert system is addressed. In stellar-inertial navigation systems based upon modern star trackers, star pattern identification has been characterized primarily as a database search problem. The proposed algorithm herein provides a methodology on how to efficiently construct a mission catalog that reduces the size of original data. Grouping neighboring stars for identification makes it possible to decrease data without noticeable performance degradation. The Radial Basis Function (RBF) well known for generic pattern recognitions is employed as a classifier in the simulation. The neural network-based approach in this study, because of associating the star distribution patterns, turns out to decrease computational time in actual space missions. It also minimizes potential noise effects due to clustering algorithm error and/or CCD pixels. Accuracy and performance of the proposed star identification approach have been examined by simulations using the Bright Star Catalog (BSC) J2000 master catalog.

Lithology determination from well logs with fuzzy associative memory neural network

An artificial intelligence technique of fuzzy associative memory is used to determine rock types from well-log signatures. Fuzzy associative memory (FAM) is a hybrid of neural network and fuzzy expert system. This new approach combines the learning ability of neural network and the strengths of fuzzy linguistic modeling to adaptively infer lithologies from well-log signatures based on (1) the relationships between the lithology and log signature that the neural network have learned during the training and/or (2) geologist's knowledge about the rocks. The method is applied to a sequence of the Ordovician rock units in northern Kansas. This paper also compares the performances of two different methods, using the same data set for meaningful comparison. The advantages of FAM are: (1) expert knowledge acquired by geologists is fully utilized; (2) this knowledge is augmented by the neural network learning from the data, when available; and (3) FAM is transparent in that the knowledge is explicitly stated in the fuzzy rules.

A hybrid fuzzy neural network-expert system for a short term unit commitment problem

The major aim of this paper is to develop a hybrid model to determine the unit commitment schedule of generating units. It is proposed to produce the unit commitment schedules in two steps. In the first step, the input layer of the neural network is designed to adopt nonlinear fuzzy membership values of the load demand profile. In the second step, an adaptive expert system is used to manipulate the schedules by using rule-base and inference mechanism. To demonstrate the effectiveness of this model, extensive studies have been performed for different load profiles belonging to the power systems of Tamilnadu consisting of thermal, nuclear, hydro, wind and diesel power plants.

Real-time diagnosis of semiconductor manufacturing equipment using a hybrid neural network expert system

This paper presents a tool for the real-time diagnosis of integrated circuit fabrication equipment. The approach focuses on integrating neural networks into an expert system. The system employs evidential reasoning to identify malfunctions by combining evidence originating from equipment maintenance history, on-line sensor data, and in-line post-process measurements. Neural networks are used in the maintenance phase of diagnosis to approximate the functional form of the failure history distribution of each component. Predicted failure rates are then converted to belief levels. For on-line diagnosis in the ease of previously unencountered faults, a CUSUM control chart is implemented on real sensor data to detect very small process shifts and their trends. For the known fault case, continuous hypothesis testing on the statistical mean and variance of the sensor data is performed to search for similar data patterns and assign belief levels. Finally, neural process models of process figures of merit (such as etch uniformity) derived from prior experimentation are used to analyze the in-line measurements, and identify the most suitable candidate among faulty input parameters (such as gas flow) to explain process shifts. A working prototype for this hybrid diagnostic system has been implemented on the Plasma Therm 700 series reactive ion etcher located in the Georgia Tech Microelectronics Research Center.

Fuzzy neural network and fuzzy expert system for load forecasting

A hybrid neural network fuzzy expert system is developed to forecast short-term electric load accurately. The fuzzy membership values of the load and other weather variables are the inputs to the neural network, and the output comprises the membership values of the predicted load. An adaptive fuzzy correction scheme is used to forecast the final load by using a fuzzy rule base and fuzzy inference mechanism. Extensive studies have been performed for all seasons, and a few examples are presented in the paper, including average, peak and hourly load forecasts.

Neural network-expert system models of relative-depth perception

Degree of binocular, horizontal disparity was used by two hybrid neural network/expert system computer models to make relative-depth judgments for pairs of stimulus points. These judgments were then correlated with the actual depth relationships of the points. Results from Simulation 1 showed that horizontal disparity could be computed by the shift in activated cortical hypercolumns evoked by a particular stimulus, and that, in general, multiple disparities could be compared to make accurate judgments about relative depth. However, these results also indicated that stimuli toward the periphery of the visual field were inaccurately perceived as being more distant. Simulation 2 corrected for this inaccuracy by appropriately weighting a stimulus point’s disparity value as a function of its horizontal position in the visual field.

Shades of Braitenberg’s Vehicle 5

A model with a hybrid neural network-expert system brain was designed. This model proved to have better memory than that of Braitenberg’s (1984) original simulated vehicles, on which it was based. The parallel nature of the neural network component provided more storage capacity with the same number of neurons, whereas the serial expert system component was able to use the increased memory to make fast, accurate decisions.

Neural network learning and expert systems

Neural network learning and expert systems

On-line fault diagnosis of power substation using connectionist expert system

This paper proposes a new connectionist (or neural network) expert system for the online fault diagnosis of a power substation. The connectionist expert diagnosis system has similar profile to an expert system, but can be constructed much more easily from elemental samples. These samples associate the faults with their protective relays and breakers as well as the bus voltages and feeder currents. Through an elaborately designed structure, alarm signals are processed by different connectionist models. The output of the connectionist models is then integrated to provide fine final conclusion with a confidence level. The proposed approach has been practically verified by testing on a typical Taiwan Power (Taipower) secondary substation. The test results show that rapid and exactly correct diagnosis is obtained even for fault conditions involving multiple faults or failure operation of protective relays and circuit breakers. Moreover, the system can be transplanted into various substations with little additional implementation effort. >

A shell environment for developing connectionist decision support systems*

Abstract: This paper presents the architecture of a neural network expert system shell. The system captures every rule as a rudimentary neural network, which is called a network element (netel). The aim is to preserve the semantic structure of the expert system rules, while incorporating the learning capability of neural networks into the inferencing mechanism. These netel rules are dynamically linked up to form the rule‐tree during the inferencing process, just as a conventional expert system does. The system is also able to adjust its inference strategy according to different users and situations. A rule editor is provided to enable easy maintenance of the netel rules. These components are housed under a user‐friendly interface. An application

Artificial neural network techniques for estimating heavy convective rainfall and recognizing cloud mergers from satellite data

Abstract This research presents an artificial neural network (ANN) technique for heavy convective rainfall estimation and cloud merger recognition from satellite data. An Artificial Neural network expert system for Satellite-derived Estimation of Rainfall (ANSER) has been developed in the NOAA/NESDIS Satellite Applications Laboratory. Using artificial neural network group techniques, the following can be achieved: automatic recognition of cloud mergers, computation of rainfall amounts that will be ten times faster, and average errors of the rainfall estimates for the total precipitation event that will be reduced to less that 10 per cent.

Neural Network Learning and Expert Systems

Neural Network Learning and Expert Systems

Embedding neural nets and expert systems in diagnostic microbiology laboratories

Describes a project for developing a reliable computer vision system that analyzes bacterial growths in real time on inexpensive personal computers, and an expert advising system that adds treatment recommendations to laboratory reports. The system embeds a neural network and multiple expert systems in the normal workflow of a clinical microbiology laboratory. Such laboratories receive specimens from patients with suspected bacterial infections, identify the organisms, determine which antibiotics will kill the bacteria or prevent growth, and convey the results to the treating physician, who can use the findings as a basis for treatment. Clearly this information is critical, since in the more extreme cases, lives are at stake. >

Book review: Neural Network Learning and Expert Systems by Stephen L. Gallant (MIT Press, Cambridge, MA, 1993)

There is an abundance of literature on Neural Networks (NNs). However, it is hard to find a text on NNs that provides a happy balance among good pedagogy, theory, and practice. For example, most introductory NN courses are taught with a collection of loosely related papers, excerpts from books, or software packages. A novice attempting to learn the subject without a good instructor inevitably becomes lost in an array of inconsistent terminology, complicated mathematics, and diverse methodologies and usually does not obtain a firm grasp of the basic foundations of the subject. Steve Gallant's book "Neural Network Learning and Expert Systems" is a good effort towards alleviating these problems. In addition, the book introduces a method to integrate the best features of NNs and expert systems (ESs). The goals of the book are (1) to provide a systematic development of NN learning algorithms suitable for researchers and students and (2) to present NN expert systems. The former goal is largely met and thus renders the book well-suited for self-directed study or for primary reading in a graduate or upper-level undergraduate NN course. The latter goal is met also, although the book lacks sufficient comparison with alternative approaches to the integration of NNs and ESs.

Neural Network Learning and Expert Systems

Neural Network Learning and Expert Systems

Neural network learning and expert systems

Neural Network Learning and Expert Systems is the first book to present a unified and in-depth development of neural network learning algorithms and neural network expert systems. Especially suitable for students and researchers in computer science, engineering, and psychology, this text and reference provides a systematic development of neural network learning algorithms from a computational perspective, coupled with an extensive exploration of neural network expert systems which shows how the power of neural network learning can be harnessed to generate expert systems automatically. Features include a comprehensive treatment of the standard learning algorithms (with many proofs), along with much original research on algorithms and expert systems. Additional chapters explore constructive algorithms, introduce computational learning theory, and focus on expert system applications to noisy and redundant problems. For students there is a large collection of exercises, as well as a series of programming projects that lead to an extensive neural network software package. All of the neural network models examined can be implemented using standard programming languages on a microcomputer.

Neural network-Expert system hybrid approach for tunnel design : Sterling, R L; Lee, C A Proc 33rd US Symposium on Rock Mechanics, Santa Fe, 3–5 June 1992P501–510. Publ. Rotterdam: A A Balkema, 1992

Neural network-Expert system hybrid approach for tunnel design : Sterling, R L; Lee, C A Proc 33rd US Symposium on Rock Mechanics, Santa Fe, 3–5 June 1992P501–510. Publ. Rotterdam: A A Balkema, 1992

Expert system development using ellipsoid-based clustering

Most classification systems classify given data into one of a number of predefined groups, which have an associated response or action. These groups form a knowledge base, whose acquisition is crucial for successful data classification. This paper describes the design and implementation of an expert system, whose knowledge base is made up of elliptical clusters. The clusters are acquired using a clustering technique specifically designed and developed for mineral processing applications. The use of clustering techniques provides significant advantages over rule-based and neural network expert systems for many applications. The expert system is demonstrated using the classification of colour density tracer samples used in spiral separator experiments.

A multi-stage intelligent system for unit commitment

A heuristic approach to power system generation scheduling is discussed. The proposed short-term unit commitment employs a multistage neural network expert system approach to achieve real-time processing results. The operating constraints are presented as, heuristic rules in the system where a feasible solution is obtained through inference. The neural networks are used at the preprocessor and postprocessor stages. At the preprocessor stage, a load pattern matching scheme is used to retrieve an optimal schedule that represents the closest solution to the given load profile from the database. At the postprocessor stage, a trained neural network performs considerable adjustments to achieve the optimal solution.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Expert Systems: Guidelines for Managers

Enumerates the factors which decide the feasibility of an expert system (ES) solution based on certain domain characteristics. Explains the points to be given importance while going for an ES solution. An attempt has been made to explain the major differences between DSSs (decision support systems) and ESs. The integration of symbolic and connectivist theories have resulted in the neural network ES, which can learn and adapt to situations, becoming an emerging trend. Certain facts about ESs which should be remembered by managers, in general, which can be treated as a set of guidelines for ES application, are also enumerated in the course of the article.