Engineering Applications of Artificial Intelligence

Published by Elsevier
Online ISSN: 0952-1976
Publications
Article
Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) has become an important tool in breast cancer diagnosis, but evaluation of multitemporal 3D image data holds new challenges for human observers. To aid the image analysis process, we apply supervised and unsupervised pattern recognition techniques for computing enhanced visualizations of suspicious lesions in breast MRI data. These techniques represent an important component of future sophisticated computer-aided diagnosis (CAD) systems and support the visual exploration of spatial and temporal features of DCE-MRI data stemming from patients with confirmed lesion diagnosis. By taking into account the heterogeneity of cancerous tissue, these techniques reveal signals with malignant, benign and normal kinetics. They also provide a regional subclassification of pathological breast tissue, which is the basis for pseudo-color presentations of the image data. Intelligent medical systems are expected to have substantial implications in healthcare politics by contributing to the diagnosis of indeterminate breast lesions by non-invasive imaging.
 
Article
This paper presents an application of genetic programming (GP) to optimally select and fuse conventional features (C-features) for the detection of epileptic waveforms within intracranial electroencephalogram (IEEG) recordings that precede seizures, known as seizure-precursors. Evidence suggests that seizure-precursors may localize regions important to seizure generation on the IEEG and epilepsy treatment. However, current methods to detect epileptic precursors lack a sound approach to automatically select and combine C-features that best distinguish epileptic events from background, relying on visual review predominantly. This work suggests GP as an optimal alternative to create a single feature after evaluating the performance of a binary detector that uses: 1) genetically programmed features; 2) features selected via GP; 3) forward sequentially selected features; and 4) visually selected features. Results demonstrate that a detector with a genetically programmed feature outperforms the other three approaches, achieving over 78.5% positive predictive value, 83.5% sensitivity, and 93% specificity at the 95% level of confidence.
 
Conference Paper
Modelling and control of 300 MW steam-boiler combustion system using neuro-fuzzy methodology is discussed in this paper. An associate memory network (AMN) is chosen to represent the nonlinear model of the steam pressure system based on local mechanism model and dynamic experiments. With the established neuro-fuzzy model, a relative fuzzy PI controller is constituted. The performance of the control system has been verified by the simulation process and then tested on real-time process in a distributed control system.
 
Conference Paper
3D object reconstruction from 2D orthographic views has been a major research issue during the past two decades. Existing algorithms assume coordinate-based, error-free input and expert acknowledgment. The approach presented in this work proposes an automatic procedure for 3D object reconstruction from 2D engineering drawings which mimics skilled human intelligence. Combining elements of variational geometry, matrix algebra and graph theoretic methods, the approach incorporates high level understanding of 2D engineering drawings, topological relations and dimensional scheme analysis for each 2D view. The dimensioning schemes of each view are merged into a common dimensioning scheme of the entire object. We present the principles of the methodology and demonstrate it on a simple example
 
Conference Paper
An approach to admittance control using fuzzy logic based reinforcement learning is proposed for the robotic automation of typical manufacturing operations. Use of fuzzy logic enables the knowledge of the manufacturing process operator to be incorporated into the controller design, which is then further refined using reinforcement learning techniques. Automated robotic deburring offers an attractive alternative to manual deburring in terms of reduced costs and improved quality of the finished parts, and hence it is used as an example of a typical manufacturing task. Simulation results are presented which demonstrate the effectiveness of the proposed controller in controlling the automated robotic deburring task
 
Conference Paper
A software environment that supports the implementation of image understanding applications is described. The environment, called I-see, is implemented on top of the object-oriented KEE system and focuses on the iconic representation and exploration of visual data. The system offers an interface between the high-level symbolic reasoning mechanisms of KEE and the raw image data and iconically represented image models or segmentation results. An example problem of interpreting satellite images shows the use of this iconically represented knowledge in the form of a terrain elevation image. In this specific example, the iconic representation could be used directly without the need for a complementary symbolic description. The iconic and symbolic representation schemes can be used to supplement each other and can be related by the use of label images
 
Conference Paper
A twin register architecture has been developed to improve the backtracking speed of Prolog programs. The twin register architecture is intended to realize a virtual infinite register set. The features of the architecture are: (1) only a small amount of hardware is needed, including a pair of register files, and (2) data transfer between the register and the memory is automatically executed. A register saving/restoring operation and the Prolog instruction are executed in parallel to reduce the overhead of memory accesses. The twin register architecture has been implemented in the IP704 AI processor to determine its effectiveness. Experimental results have shown that the execution time of the 8-Queen program is reduced by 15% in the case of the twin register architecture as compared with that for the ordinary architecture, in which saving/restoring are done by software. Also, the architecture is useful for register saving/restoring of the CALL/RETURN procedure in general procedural programs
 
Conference Paper
Several approximate algorithms have been reported to solve large constraint satisfaction problems (CSPs) in a practical time. While these papers discuss techniques to escape from local optima, the present paper describes a method that actively performs global search. The present method is to improve the rate of search of genetic algorithms using viral infection instead of mutation. The partial solutions of a CSP are considered to be viruses and a population of viruses is created as well as a population of candidate solutions. Search for a solution is conducted by crossover infection substitutes the gene of a virus for the locus decided by the virus. Experimental results using randomly generated CSPs prove that the proposed method is faster than a usual genetic algorithm in finding a solution when the constraint density of a CSP is low
 
Conference Paper
A parallel shape analysis by simultaneous merging of elementary features detected through calculation of the number of objects within neighboring windows is presented. The parallel algorithm can be embedded on a SIMD (single-instruction, multiple-data) mesh architecture. Elementary image features are detected inside partly overlapping windows fixed in an image plane. Each window content is processed by a separate processing element (PE). Two neighboring elementary feature elements are merged by adjacent PEs, and the joined feature chunks are merged in the next step by every 2<sup>2</sup>th PE possessing the same feature. Feature ends are propagated through 2<sup>n</sup> PEs in each n th parallel step toward opposite edges of the mesh array of PEs. The shape coding is completed if all the feature limits (e.g. edges) meet together on one PE. The approach has the property of mapping an image fragment directly into words and phrases
 
Conference Paper
A RISC (reduced instruction set computer)-based chip set architecture for LISP is presented which contains an instruction fetch unit (IFU) and three processing units: integer processing unit (IPU), floating-point processing unit (FPU), and list processing unit (LPU). The IFU feeds instructions to the processing units and provides the branch handle mechanism to reduce branch penalty; the IPU is optimized for integer operations, string manipulation, operand address calculations, and some cooperation affairs for constructing a multiprocessor architecture; the FPU handles the floating point data type, which conforms to IEEE standard 754; and the LPU handles LISP runtime environment, dynamic type checking, and fast list access. In this architecture, the critical path of complex register file access and ALU (arithmetic and logic unit) operation is distributed into LPU and IPU, and the tracing of a list can be done quickly by the nondelayed car or cdr instructions of LPU. In addition, by using a new branch control mechanism (called branch peephole), this architecture can achieve almost-zero-delay branch and super-zero-delay jump. Performance simulation shows that this architecture would be about 4.1 times faster than SPUR and about 2.2 times faster than MIPS-X
 
Conference Paper
Controlling multifingered robot hands makes high demands on the control algorithms and the speed of the control computer. The nonlinear friction, the impact problem and other plant uncertainties require a special kind of control and tuning of the controller. Some simple linear and nonlinear controllers for the Karlsruhe Dexterous Hand are presented and the results and advantages of the controllers are shown. Also, a new adaptive fuzzy controller is presented to overcome the time consuming process of fine-tuning the membership functions. Finally, a short glance at the hardware platform is taken in order to show the control system architecture
 
Conference Paper
A family of basis functions, generated from the evolving states of cellular automata (CA), is used to compress and encrypt data. The operations required in encoding and decoding the data are described under the umbrella cellular automata transforms (CAT). There is a huge number of these transform bases. CAT which can be used in the way other mathematical transforms (e.g. Fourier, Discrete Cosine, Laplace, Wavelet, etc.) are utilized. In data compression applications, the rules and pertinent keys used to generate the CA are selected in favour of those which yield basis functions with the best information packing characteristics. On the other hand, for encryption the selection is biased towards those with the tendency to yield an avalanche effect. In the latter case the transform process must be error free
 
Conference Paper
Fast and accurate diagnosis of faults in computer integrated manufacturing systems is essential in order to avoid excessive equipment downtime, and to take full advantage of these systems. Traditional approaches to diagnosis have yielded to artificial intelligence approaches over recent years, as system complexity has increased; but results have been mixed. Symptom-based approaches have been too limited, while structural-based approaches have required excessive computational resources. This paper presents a hybrid model for diagnostics that is computationally efficient, and at the same time incorporates the potential to improve its performance with use through a learning scheme.
 
Conference Paper
This paper suggests a new genetic algorithm (GA) for VLSI circuit partitioning problem. In a genetic algorithm, the encoding of a solution plays an important role. The key feature of the new genetic algorithm is a technique to provide dynamically many encodings in which encodings themselves undergo evolution. Before generating every new solution, we first generate a new encoding by combining two encodings chosen from a pool containing diverse encodings. The new solution is generated by a crossover which combines two parent solutions which are temporarily encoded by the generated encoding scheme. That is, a new solution is generated by a two-layered crossover. Depending on the new solution's quality and its improvement over the parents solutions, a fitness value is assigned to the underlying encoding. The encoding is discarded or enter the pool based on the fitness. Two populations are maintained for this purpose: one for solutions and the other for diverse encodings. On experiments with the public ACM/SIGDA benchmark circuits, the new genetic algorithm significantly outperformed recently published state-of-the-art approaches
 
Conference Paper
The explosion of on-line information has given rise to many manually constructed topic hierarchies (such as Yahoo!!). But with the current growth rate in the amount of information, manual classification in topic hierarchies results in an immense information bottleneck. Therefore, developing an automatic classifier is an urgent need. However, the classifiers suffer from the enormous dimensionality, since the dimensionality is determined by the number of distinct keywords in a document corpus. More seriously, most classifiers are either working slowly or they are constructed subjectively without learning ability. In this paper, we address these problems with a fair feature Subset selection algorithm and an adaptive fuzzy learning network (AFLN) for classification. The fair feature subset selection algorithm is used to reduce the enormous dimensionality. It not only gives fair treatment to each category but also has ability to identify useful features, including both positive and negative features. On the other hand, the AFLN provides extremely fast training and testing and, more importantly, it has the ability to learn the human knowledge. Experimental results show that our proposed fair feature subset selection algorithm is effective in recognizing useful keywords for classification. It indeed can be used to reduce a surprising number of dimensions in classification models. Besides, experimental results also show the adaptive fuzzy learning network for classification with high-speed classification and high accuracy rate
 
Conference Paper
The paper describes the analysis, design, and fast prototyping of MINT, an information and decision support system for railway traffic control. MINT is a complex system that tightly integrates information management and problem solving functionalities, by means of an object oriented approach. The work is characterised by several issues: (i) object oriented analysis and design; (ii) knowledge based application modeling, by means of a powerful conceptual language (TQL++); (iii) advanced search techniques in the problem solving component; (iv) fast prototyping by automatic generation of executable code; (v) use of an advanced knowledge based modeling and prototyping environment (Mosaico). The paper starts with a description of the railway traffic control problem, then it focuses no the architecture of MINT, with a particular attention to the database component and its train conflict solving capabilities
 
Conference Paper
Jet engines are nonlinear dynamical systems for which an exact mathematical model cannot be used for estimator design, because it is either not available or so complex that it does not fit the necessary assumptions. Thus, classical analytical tools for studying standard system properties like observability, which is very important in estimator design, cannot be directly applied. Generally, for practical jet engine applications, the designer faces two closely related problems: first, given an unmeasurable parameter, find the minimal set of estimator inputs that facilitates achieving a satisfactory estimation performance (input selection); second, given a predetermined set of inputs, derive an “observability” measure that characterizes the estimation feasibility of a specific unmeasurable parameter. In the paper, techniques for solving these two problems are developed and applied to estimator design for jet engine thrust, stall margins, and an unmeasurable state
 
Conference Paper
Most of the process control and monitoring problems cannot be solved by either operations research (OR) or expert systems (ES) techniques alone. We present an integrated approach in which expert systems and operations research techniques complement each other for optimizing natural gas pipeline operations. Typically, expert systems are used in problems where mathematical calculations cannot but the knowledge of an experienced human expert can provide a satisfactory solution. The ES approach has shown exceptional performance in process control and monitoring when the working knowledge of the system is non-linear and incomplete. The OR approach, on the other hand, can be used to handle problems in which well constructed mathematical models are available or can be developed
 
Conference Paper
This paper investigates the generation of neural networks through the induction of binary trees of threshold logic units (TLUs). Initially, we describe the framework for our tree construction algorithm and show how it helps to bridge the gap between pure connectionist (neural network) and symbolic (decision tree) paradigms. We also show how the trees of threshold units that we induce can be transformed into an isomorphic neural network topology. Several methods for learning the linear discriminant functions at each node of the tree structure are examined and shown to produce accuracy results that are comparable to classical information theoretic methods for constructing decision trees (which use single feature tests at each node), but produce trees that are smaller and thus easier to understand. Moreover, our results also show that it is possible to simultaneously learn both the topology and weight settings of a neural network simply using the training data set that we are initially given
 
Conference Paper
The emergence of the theory of dynamic neural computing has made it possible to develop neural learning and adaptive schemes that can be used to obtain feasible solutions to complex control problems, such as inverse kinematic control for robotic systems. In this paper, such a neural learning scheme using a multilayered dynamic neural network (MDNN) is proposed. The basic dynamic computing element of MDNN is a dynamic neural unit (DNU) developed in this paper. The learning and adaptive capabilities of DNU can be used for developing complex dynamic structures. In this paper, we have used DNU for developing a MDNN for the inverse kinematic control of a two-link robot. The validity of the proposed scheme is demonstrated through computer simulation studies
 
Conference Paper
Two types of learning networks for nonparametric regression problems are studied and compared: one is the parametric two-layer perceptron type neural network, which is well known in artificial neural network (ANN) literature; the other is the semiparametric projection pursuit network (PPN), which has emerged in recent years in the statistical estimation literature. From an algorithmic viewpoint, both the PPN and the ANN parametrically form projections of the data in directions determined from interconnection weights. However, unlike an ANN which uses a fixed set of nonlinear nodal functions to perform an explicit parametric estimate of a nonparametric model, the PPN nonparametrically estimates the nonlinear functions using a one-dimensional data smoother. From experimental simulations, ANNs and PPNs perform comparably in predicting independent test data but PPN training is much faster than that of an ANN
 
Article
The importance of coordinating product, process, and supply chain (PPSC) decisions has received much attention and popularity in academia and industry alike. This paper formulates PPSC coordination as a factory loading allocation problem (FLAP) from a constraint satisfaction perspective. A domain-based FLAP reference model is proposed for the conceptualization of a multi-site manufacturing supply chain, considering multiple domains, network structures, product characteristics, decision variables, along with various constraints. A decision propagation structure (DPS) incorporating with a connectionist approach is developed based on the concept of constraint heuristic search to facilitate the exploration of solution spaces. A case study in a multi-national company is presented to illustrate the FLAP framework, which implies practical insights into PPSC coordination.
 
Article
An important issue in application of fuzzy inference systems (FISs) to a class of system identification problems such as prediction of wave parameters is to extract the structure and type of fuzzy if–then rules from an available input–output data set. In this paper, a hybrid genetic algorithm–adaptive network-based FIS (GA–ANFIS) model has been developed in which both clustering and rule base parameters are simultaneously optimized using GAs and artificial neural nets (ANNs). The parameters of a subtractive clustering method, by which the number and structure of fuzzy rules are controlled, are optimized by GAs within which ANFIS is called for tuning the parameters of rule base generated by GAs. The model has been applied in the prediction of wave parameters, i.e. wave significant height and peak spectral period, in a duration-limited condition in Lake Michigan. The data set of year 2001 has been used as training set and that of year 2004 as testing data. The results obtained by the proposed model are presented and analyzed. Results indicate that GA–ANFIS model is superior to ANFIS and Shore Protection Manual (SPM) methods in terms of their prediction accuracy.
 
Article
Automated incident detection and alternative path planning form important activities within a modern expressway traffic management system which aims to ensure a smooth flow of traffic along expressways. This is done by adopting efficient technologies and processes that can be directly applied for the automated detection of non-recurrent congestion, the formulation of response strategies, and the use of management techniques to suggest alternative routes to the road-users, resulting in significant overall reductions in both congestion and inconvenience to motorists. A delicate balance has to be struck here between the incident detection rate and the false-alarm rate. This paper presents the development of a hybrid artificial intelligence technique for automatically detecting incidents on a traffic network. The overall framework, algorithm development, implementation and evaluation of this hybrid fuzzy-logic genetic-algorithm technique are discussed in the paper. A cascaded framework of 11 fuzzy controllers takes in traffic indices such as occupancy and volume, to detect incidents along an expressway in California. The flexible and robust nature of the developed fuzzy controller allows it to model functions of arbitrary complexity, while at the same time being inherently highly tolerant of imprecise data. The maximizing capabilities of genetic algorithms, on the other hand, enable the fuzzy design parameters to be optimized to achieve optimal performance. The results obtained for the traffic network give a high detection rate of 70.0%, while giving a low false-alarm rate of 0.83%. A comparison between this approach and four other incident-detection algorithms demonstrates the superiority of this approach.
 
Article
This paper presents a novel hardware framework of particle swarm optimization (PSO) for various kinds of discrete optimization problems based on the system-on-a-programmable-chip (SOPC) concept. PSO is a new optimization algorithm with a growing field of applications. Nevertheless, similar to the other evolutionary algorithms, PSO is generally a computationally intensive method which suffers from long execution time. Hence, it is difficult to use PSO in real-time applications in which reaching a proper solution in a limited time is essential. SOPC offers a platform to effectively design flexible systems with a high degree of complexity. A hardware pipelined PSO (PPSO) Core is applied with which the required computational operations of the algorithm are performed. Embedded processors have also been employed to evaluate the fitness values by running programmed software codes. Applying the subparticle method brings the benefit of full scalability to the framework and makes it independent of the particle length. Therefore, more complex and larger problems can be addressed without modifying the architecture of the framework. To speed up the computations, the optimization architecture is implemented on a single chip master–slave multiprocessor structure. Moreover, the asynchronous model of PSO gains parallel efficacy and provides an approach to update particles continuously. Five benchmarks are exploited to evaluate the effectiveness and robustness of the system. The results indicate a speed-up of up to 98 times over the software implementation in the elapsed computation time. Besides, the PPSO Core has been employed for neural network training in an SOPC-based embedded system which approves the system applicability for real-world applications.
 
Article
Unified Modeling Language (UML 2.0) is the upcoming standard of the Object Management Group for specifying object-oriented software systems. In this paper, we will show how UML 2.0 can be applied for the specification of agent-based systems. Moreover, we will give a short overview on existing agent methodologies to have a reference what has to be specified in such systems. The paper concludes with some outlook for further research and open issues for specifying agents with UML 2.0.
 
Article
This paper presents a novel real-time application of fuzzy logic for an integrated navigation avionics suite. A knowledge-based system, which uses a fuzzy rule-base for a real-time INS/DGPS integrated navigation system on-board a Bell 206 helicopter, has been designed and developed. This knowledge base is developed in such a way as to detect aircraft maneuvering and tune the integration algorithm (Kalman filter) of the INS/DGPS system accordingly. The signal processing method developed for the integration of INS and DGPS data provides accurate navigation even during dynamic maneuvering of the aircraft, while taking advantage of low-cost modular equipment rather than costly inertial navigation systems.
 
Article
The field of remote sensing and sensor technology has undergone tremendous development in the past decades. Sensors technologies of all kinds such as electro-optics, acoustic, active/passive UV to LWIR, ground-penetrating radar, passive mm wavelength, X-ray tomography, neutron activation imaging, multi-spectral, hyper-spectral, and ultra-spectral imaging, will provide valuable images that normal CCD camera cannot offer. By combining algorithms and images taken by sensors at different part of the electromagnetic spectrum, we will be able to extract valuable images automatically. By using multi-spectral images and processing them with neural network computing, our “Third Eye” team is able to extract human face features from those images. In this paper, we will present an application for detecting human facial parts, images taken by different imaging systems and sensors, and the current status of image processing applications.
 
Article
3D object reconstruction from 2D orthographic views has been a major research issue during the past two decades. Existing algorithms assume coordinate-based, noise- and error-free input without dimensioning annotation. The approach presented here assumes that the original input is a real-life engineering drawing, in which the 2D geometry of each orthographic view is annotated with dimensioning. Detected dimensions are translated into sets of constraints, one for each view, from which proper dimensioning is validated and 2D minimal graphs are obtained. The method combines elements from variational geometry, matrix algebra and graph theory to construct a composite network describing the structural and topological relations among the various entities that combine the 3D object. This graph provides the basis for a complete 3D object reconstruction. The paper describes the details of the method, and demonstrates it on a comprehensive example.
 
Article
In this paper we present a successful application of genetic algorithms to the registration of uncalibrated optical images to a 3D surface model. The problem is to find the projection matrices corresponding to the images in order to project the texture on the surface as precisely as possible. Recently, we have proposed a novel method that generalises the photo-consistency approach by Clarkson et al. to the case of uncalibrated cameras by using a genetic algorithm. In previous studies we focus on the computer vision aspects of the method, while here we analyse the genetic part. In particular, we use semi-synthetic data to study the performance of different GAs and various types of selector, mutation and crossover. New experimental results on real data are also presented to demonstrate the efficiency of the method.
 
Article
The integration of high-performance sensor systems with a robot is essential to enhance the “intelligent” capability of the robot, while force/torque sensors are especially important sources of feedback in robot applications, for proper monitoring, analysis and force/motion control. Being integrated with robots, force/torque sensors may suffer from non-linearity and various forms of uncertainty, resulting not only from the sensors themselves, but also from the workcells they integrate; therefore the conventional least-squares method for force/torque sensor calibration is unable to interpret the relationships between sensor readings and their represented outputs accurately enough for some applications. This paper therefore presents the development of a new approach, using neural networks for the efficient and accurate calibration of the F/T sensors integrated with robots. Apart from the neural-network-based force/torque sensor-calibration methodology, this paper also presents the calibration implementation using both the least-square method and the proposed method, and a comparison and discussion of the calibration results. These results show that the proposed neural-network-based calibration method is more efficient and accurate, which verifies the adaptability and applicability of this method to nonlinear and dynamic robot systems.
 
Article
The exploitation of wideband code division multiple access (W-CDMA) technology in third generation (3G) networks gives an inherent flexibility in managing the system capacity, although radio resource management (RRM), including congestion management, is more complicated. To guarantee the quality of service (QoS) provided to customers, the concept of a “service level agreement” (SLA) is introduced and these must be managed by the RRM. This work proposes the application of intelligent agents in SLA-based control in the RRM, essentially for congestion management and demonstrates the ability of intelligent agents to improve and maintain the QoS to meet the required SLA. A particularly novel aspect of this work is the use of learning (case-based reasoning—CBR) to predict the control strategies to be imposed. If there is no congestion, the network operates as provisioned, but, if congestion occurs, it is detected by the agent monitoring process and CBR will be used to provide a suitable policy either by recalling from experience or recalculating the solution from its knowledge. With this approach, the system performance will be monitored at all times and a suitable policy can be applied immediately as the system environment changes, resulting in the QoS being maintained.
 
Article
In this paper, a fusion approach to determine inverse kinematics solutions of a six degree of freedom serial robot is proposed. The proposed approach makes use of radial basis function neural network for prediction of incremental joint angles which in turn are transformed into absolute joint angles with the assistance of forward kinematics relations. In this approach, forward kinematics relations of robot are used to obtain the data for training of neural network as well to estimate the deviation of predicted inverse kinematics solution from the desired solution. The effectiveness of the fusion process is shown by comparing the inverse kinematics solutions obtained for an end-effector of industrial robot moving along a specified path with the solutions obtained from conventional neural network approaches as well as iterative technique. The prominent features of the fusion process include the accurate prediction of inverse kinematics solutions with less computational time apart from the generation of training data for neural network with forward kinematics relations of the robot.
 
Article
Hourly temperature forecasts are important for electrical load forecasting and other applications in industry, agriculture, and the environment. Modern machine learning techniques including neural networks have been used for this purpose. We propose using the alternative abductive networks approach, which offers the advantages of simplified and more automated model synthesis and transparent analytical input–output models. Dedicated hourly models were developed for next-day and next-hour temperature forecasting, both with and without extreme temperature forecasts for the forecasting day, by training on hourly temperature data for 5 years and evaluation on data for the 6th year. Next-day and next-hour models using extreme temperature forecasts give an overall mean absolute error (MAE) of 1.68 °F and 1.05 °F, respectively. Next-hour models may also be used sequentially to provide next-day forecasts. Performance compares favourably with neural network models developed using the same data, and with more complex neural networks, reported in the literature, that require daily training. Performance is significantly superior to naive forecasts based on persistence and climatology.
 
Article
This paper proposes a method to solve the network fault diagnosis problem using the Realistic Abductive Reasoning Model. This model uses an abductive inference mechanism based on the parsimonious covering theory, and adds some new features to the general model of diagnostic problem-solving. The network fault-diagnosis knowledge is assumed to be represented in the form of causal chaining, namely, a hyper-bipartite graph. A layered graph is constructed from the given hyper-bipartite graph by the addition of a few dummy nodes. Then the diagnostic problem is solved, starting from the lowest layer of the layered graph, as a series of bipartite graphs, until the top-most layer is reached. The inference mechanism uses a Realistic Abductive Reasoning Model to diagnose the faults in a communication network, which is symptom-driven, based on some application programs. The hypothesis-test paradigm is used to refine the solution space. The fault-diagnostic capability of the proposed inference model is demonstrated by considering one node of a given network where the management information would be used to diagnose its local problems and the connectivity of the node in the network. The results obtained by the proposed model substantiate its effectiveness in solving network fault-diagnostic problems.
 
Article
This paper discusses the use of a back-propagation neural network for the purpose of modeling an automotive shock absorber. A brief description of the method is presented, as well as the structure of the network developed. The absorber model has been implemented in a numerical suspension simulation and comparisons are made between the simulation outputs and experimental test results.
 
Article
In previous approaches, the combination of multiple classifiers depends heavily on one of the three classification results; measurement scores (measurement level), ranking (rank level), and top choice (abstract level). For a more general combination of multiple classifiers, it is desirable that combination methods should be developed at the abstract level. In combining multiple classifiers at this level, most studies have assumed that classifiers behave independently. Such an assumption degrades and biases the classification performance, in cases where highly dependent classifiers are added. In order to overcome such weaknesses, it should be possible to combine multiple classifiers in a probabilistic framework, using a Bayesian formalism. A probabilistic combination of multiple decisions of K classifiers needs a (K + 1)st-order probability distribution. However, it is well known that such a distribution will become unmanageable to store and estimate, even for a small K. In this paper, a framework is proposed to optimally identify a product set of kth-order dependencies, where 1≤k⪯-K for the product approximation of the (K + 1)st-order probability distribution from training samples, and to probabilistically combine multiple decisions by the identified product set, using the Bayesian formalism. This framework was tested and evaluated using a standardized CENPARMI data base. The results showed superior performance over other combination methods.
 
Article
This paper describes a debugger which uses the design artifacts of the Prometheus agent-oriented software engineering methodology to alert the developer testing the system, that a specification has been violated. Detailed information is provided regarding the error which can help the developer in locating its source. Interaction protocols specified during design, are converted to executable Petri net representations. The system can then be monitored at run time to identify situations which do not conform to specified protocols. A process for monitoring aspects of plan selection is also described. The paper then describes the Prometheus Design Tool, developed to support the Prometheus methodology, and presents a vision of an integrated development environment providing full life cycle support for the development of agent systems. The initial part of the paper provides a detailed summary of the Prometheus methodology and the artifacts on which the debugger is based.
 
Article
The design of a chemical plant is a difficult and time-consuming task that requires the co-operation of skilled personnel from many different disciplines. Once a plant has been designed and constructed, it is expected to last for many years. However, design modifications are often made to a plant in order to benefit from the advances made in technology and to meet the changing demands of the market. In order to avoid unsafe changes to a plant, it is very important that the design rationale behind a plant is captured and made easily accessible. This paper describes an Integrated Design Information System (IDIS) that supports the design of chemical plants. The system places particular emphasis on supporting the design process so that the recording of design rationale will be done easily. It provides an integrated framework for recording three different aspects of design rationale: exploration of design alternatives, reasons for design decisions and design constraints. A design example is used to illustrate the different aspects of the system, and to show how they are linked together. Novel facilities to aid the access of information are also described.
 
Article
The identification of the type of accident during the early stages of an accident in a nuclear power plant is crucial for the selection of the appropriate response actions. A plant accident can be identified by its time-dependent patterns, related to the principal variables. The Hidden Markov Model (HMM) can be applied to accident identification, which is a spatial and temporal pattern-recognition problem. The HMM is created for each accident from a set of training data by the maximum-likelihood estimation method, which uses an algorithm that employs both forward and backward chaining, and a Baum–Welch re-estimation algorithm. The accident identification is decided by calculating which model has the highest probability for the given test data. The optimal path for each model at the given observation is found by the Viterbi algorithm, and the probability of the optimal path is then calculated. The system uses a left-to-right HMM, including six states and 22 input variables, to classify eight types of accidents and a normal state. The simulation results show that the proposed system identifies the accident types correctly. It is also shown that the identification is performed well for incomplete input observations caused by sensor faults or by the malfunctioning of certain equipment.
 
Article
Neural networks have been employed in a multitude of transportation engineering applications because of their powerful capabilities to replicate patterns in field data. Predictions are always subject to uncertainty arising from two sources: model structure and training data. For each prediction point, the former can be quantified by a confidence interval, whereas total prediction uncertainty can be represented by constructing a prediction interval. While confidence intervals are well known in the transportation engineering context, very little attention has been paid to construction of prediction intervals for neural networks. The proposed methodology in this paper provides a foundation for constructing prediction intervals for neural networks and quantifying the extent that each source of uncertainty contributes to total prediction uncertainty. The application of the proposed methodology to predict bus travel time over four bus route sections in Melbourne, Australia, leads to quantitative decomposition of total prediction uncertainty into the component sources. Overall, the results demonstrate the capability of the proposed method to provide robust prediction intervals.
 
Article
Backpropagation networks are compared to radial basis function (RBF) networks when it comes to small signal modeling RF/microwave active devices. The modeled device is a 4×50 μm gate width, 0.25 μm gate length gallium arsenide (GaAs) Metal semiconductor field-effect transistor (MESFET). It is the authors’ intent to prove that RBF networks provide much better performance than backpropagation when it comes to this type of modeling. First, two separate backpropagation networks are created to determine the best training algorithm in terms of convergence speed. Then, the backpropagation network, using its best training algorithm, is compared to the RBF network in terms of speed and accuracy. Simulation results are presented in tables and figures for better understanding. All tests and simulations for the backpropagation and RBF networks are done using Matlab's Neural Network Toolbox.
 
Article
This paper presents a new diagnosis method of induction motor faults based on time–frequency classification of the current waveforms. This method is composed of two sequential processes: a feature extraction and a rule decision. In the process of feature extraction, the time–frequency representation (TFR) has been designed for maximizing the separability between classes representing different faults. The diagnosis is realised in two levels; the first one allows the detection of different faults—bearing fault, stator fault and rotor fault. The second one refines this detection by the determination of severity degree of faults, which are already identified on the previous level. The diagnosis is independent of the level of load. This method is validated on a 5.5 kW induction motor test bench.
 
Article
This paper describes the development of neural models and their industrial applications to the basic oxygen steel-making (BOS) plant of the Companhia Siderúrgica Nacional (CSN—Volta Redonda/Brazil). The BOS is a transient process, highly complex and is also subject to oscillations in raw material composition. A precise model is essential to adjust end-blow oxygen and coolant requirements to match with the targets of end-point temperature and carbon percentage in liquid steel. An inverse neural model was developed in order to calculate the end-blow process adjustments. At the end of 40 industrial runs, 82.5% of simultaneous agreement with the targets was obtained, against 66% obtained from the commercial model usually employed at CSN's plant. The inverse model was then on-line implemented to automatically control the BOS process. The neural model has been retrained from previous weights and biases as soon as the performance decreases. Average hitting rate decreased related to the previous industrial investigation, however, it is still higher than that obtained from the commercial model application. As a consequence, liquid steel reprocessing is avoided and a high level of steel productivity is obtained.
 
Article
This paper presents a comparative study of genetic algorithms (GA) and ant colony optimization (ACO) applied the online re-optimization of a logistic scheduling problem. This study starts with a literature review of the GA and ACO performance for different benchmark problems. Then, the algorithms are compared on two simulation scenarios: a static and a dynamic environment, where orders are canceled during the scheduling process. In a static optimization environment, both methods perform equally well, but the GA are faster. However, in a dynamic optimization environment, the GA cannot cope with the disturbances unless they re-optimize the whole problem again. On the contrary, the ant colonies are able to find new optimization solutions without re-optimizing the problem, through the inspection of the pheromone matrix. Thus, it can be concluded that the extra time required by the ACO during the optimization process provides information that can be useful to deal with disturbances.
 
Article
Ant Colony Optimization (ACO) is used to obtain rules that can classify the data into pre-defined classes. It can be used to classify acoustic emission (AE) signals to their respective sources. ACO based technique has an advantage over conventional statistical techniques like maximum likelihood estimate, nearest neighbor classifier, etc., because they are distribution free, i.e., no knowledge is required about the distribution of data. AE test is carried out using pulse, pencil and spark signal source on the surface of solid steel block. The signal parameters are measured using AET 5000 system. Classification of AE signal is done using Ant Colony Optimization, and the simplicity of the rules generated is emphasized.
 
Article
Feature extraction and feature selection are two important issues in sensor-based condition monitoring of any engineering systems. In this study, acoustic emission signals were first collected during grinding operations, next processed by autoregressive modeling or discrete wavelet decomposition for feature extraction, and then the best feature subsets are found by three different feature selection methods, including two proposed ant colony optimization (ACO)-based method and the famous sequential forward floating selection method. Posing monitoring as a classification problem, the evaluation is carried out by the wrapper approach with four different algorithms serving as the classifier. Empirical test results were shown to illustrate the effectiveness of feature extraction and feature selection methods.
 
Article
Knowledge acquisition is important and machine learning techniques can be used to achieve automated knowledge acquisition. This article examines how knowledge acquisition can be assisted by programming using CLIPS (an acronym for C Language Integrated Production System). A machine learning preprocessor has been developed for the CLIPS environment, so that the CLIPS rule-base can be expanded by adding rules generated through machine learning techniques. The paper also shows how knowledge updating can be supported in the CLIPS environment itself. Operational engineering knowledge is captured in a data structure called a decision tree, and its structure can be updated when new knowledge is acquired. In addition, some advanced features are also briefly discussed, including using COOL (the CLIPS object-oriented language) for knowledge acquisition in a software product recommendation system, as well as the design of a self-evolving knowledge-acquisition tool.
 
Article
Searching of state transitions is an important subject of problem solving in artificial intelligence, computer science, engineering and operations research. In artificial intelligence, a breadth-first search is optimal, with uniform cost, but it takes considerable time to obtain a solution. Neural networks process state transitions in parallel with learning ability. The authors have developed a search procedure for state transitions, that resembles a breadth-first search, using neural networks. First, the input pattern states are self-organized in the neural network, which consists of a Kohonen layer followed by a state-planning layer. The state-planning layer makes lateral connections between the cells of transitions. Then, the initial and the target states are given as a problem. The network shows an optimal transition pathway of states in the neuron firings. Next, the state-transition procedure is developed for the formation of a concept for action planning. Here, as the action planning, an integration between the symbols and the action pattern is carried out in the extended neural network.
 
Article
This paper proposes a support vector machine-based fuzzy rules acquisition system (SVM-FRAS) for modeling of the gas tungsten arc welding (GTAW) process. The character of SVM in extracting support vector provides a mechanism to extract fuzzy IF–THEN rules from the training data set. We construct the fuzzy inference system using fuzzy basis function. The gradient technique is used to tune the fuzzy rules and the inference system. Theoretical analysis and comparative tests are performed comparing with other fuzzy systems. Modeling is one of the key techniques in the automatic control of the arc welding process, and is still a very difficult problem. Comprehensibility is one of the required characteristics in modeling for the complex GTAW process. We use the proposed SVM-FRAS to obtain the rule-based model of the aluminum alloy pulse GTAW process. Experimental results show the SVM-FRAS model possesses good generalization capability as well as high comprehensibility.
 
Top-cited authors
Ajith Abraham
  • Machine Intelligence Research Labs (MIR Labs), Auburn, WA, United States
Dan Simon
  • Cleveland State University
Haiping Ma
  • Shaoxing University
Tak-Chung Fu
  • Vocational Training Council
Kwok Wing Chau
  • The Hong Kong Polytechnic University