The purpose of the paper is to present solution to design additional diagnostic system for, based on cutting-edge technology, purifying fumes installation. Neural networks, which determine the core of the system, were used as predictive models. Designed very efficient neural structures have served to build simulative diagnostic advisory system.
Keywords: artificial intelligence, neural networks, advisory systems, diagnostics.

Constraint-based scheduling is an approach for solving real-life scheduling problems by combining the generality of AI techniques with the efficiency of OR techniques. Basically, it describes a scheduling problem as a constraint satisfaction problem and then uses constraint satisfaction techniques to find a solution. In this paper we study three constraint models describing complex state transitions that are going beyond the existing models of resources (machines) used in scheduling. These models can naturally handle any setup/changeover/transition scheme as well as special counter constraints imposed on the sequence of activities. The proposed models have been implemented and tested in the commercial scheduling engine of Visopt ShopFloor system.
Keywords: constraint satisfaction, scheduling, machine setups.

Rotating machines are often described using linear methods with acceptable accuracy. Some malfunctions, however, are of non-linear nature. Accurate detection and identification of such malfunctions requires more accurate methods. One of such methods can be NARX- Non-linear AutoRegressive model with eXogenous input. The paper presents how NARX models can be applied for modeling rotating machinery malfunctions. Idea of the diagnostic algorithm based on such modeling is presented. Proposed algorithm was verified during research on a specialized test rig, which can generate vibration signals. The paper presents results of application of NARX models for detection of typical rotating machinery failures and the variations of NARX model parameters due to propagation of damage. In the paper authors present also a blade crack detection using the NARX models. The last chapter of the paper discusses the applicability of this method for damage detection in real machines.
Keywords: rotating machinery diagnostics, blade crack detection, neural networks, NARX models.

The aim of the paper is to prepare an efficient method of the optimization of the hybrid fibre-reinforced laminates. Since the several optimization criteria which cannot be satisfied simultaneously are proposed, the multi-objective optimization methods have been employed. Different optimization criteria connected with the laminates' cost, the modal properties and the stiffness are considered. The multi-objective evolutionary algorithm which uses the Pareto approach has been used as the optimization method. To solve the boundary-value problem the finite element method commercial software has been employed. Numerical examples presenting the effectiveness of the proposed method are attached.
Keywords: multi-objective optimization, evolutionary algorithm, multi-layered laminate, modal analysis.

This paper presents a new dual model combining binary and real-valued representations of samples for negative selection algorithms. Recent research show that the two types of encoding can produce quite good results for some types of datasets when they are applied separately in such algorithms. Besides a number of efficient algorithms, various affinity (or similarity) functions fitted to particular implementation was investigated. Basing on a series of experiments, we propose a dual representation enabling overcome some of the existing drawbacks of these algorithms, and allowing significant speed up the classification process. This new model was designed mainly for detecting anomalies in real-time applications, were the time of classification is crucial, e.g. intrusion detection systems.
Keywords: anomaly detection, negative selection, binary receptors, real-valued receptors, intrusion detection.

The authors of the paper took up Aoyama's concept of the integrated industrial processes intent analysis and linked it with the ideas of the designer's IPA (Intelligent Personal Assistant). In many cases it is not sufficient to analyze the engineer's intent depot when trying to explain the origin of a fault. Often computer models are built or real experiments are done with which the considered classes of problems can be analyzed better. The IPA concept was expanded for the process of building computer models and stands for the fault analysis.
Keywords: personal knowledge management, fault analysis.

Thermovision is more and more often used in machinery and apparatus diagnostics. With the aid of a thermographic camera non-contact simultaneous temperature measurements can be carried out at many points of an object and they can be recorded in a form of a thermographic image. The thermographic image can be a source of diagnostic information. Extraction of this information requires the necessity of application of different methods of the analysis of thermographic images. From thermographic image a huge amount of features can be extracted which causes problems with efficient assessment of technical state due to informational noise. There are methods which allow to search and find relevant features that are useful for diagnostic processes. In the paper application of evolutionary algorithm for selection of optimal diagnostic features has been shown. In case of assessment of selected features neural classifier has been used. A set of 259 features for each image has been considered. After searching process two features have been selected and the obtained classification results have been of very good quality. Efficiency of classifier has been in some cases 100% and not less than 97%. The results have shown that the evolutionary algorithm can be applied to selection of relevant diagnostic features.
Keywords: infrared thermography, evolutionary algorithms, neural networks, diagnostic.

In this work diagnostics of railway point drive supported by Data Mining methods was considered. Results of FEM calculations of switching forces acting on the considered point are qualitatively correct, so Data Mining methods efficiency was examined on data obtained from FEM multi-body model. Hidden structures in data and patterns describing particular faults were identified. Proposed algorithms of Kohonen's neural networks and k-means clustering are easy to apply to classifying. Their implementation on the Digital Signal Processor is not difficult and memory consumption is low so diagnostic module supported by implemented Data Mining methods was proposed in order to preliminary asses technical state of railway points and to assure current state monitoring and supporting maintenance activities.
Keywords: artificial neural networks, data mining, diagnostics.

This paper describes results of the research concerning an augmented reality (AR) system for CAD design, developed within the framework of MSc Thesis in the Department of Fundamentals of Machinery Design. The authors present advantages resulted from utilization of AR systems allowing to combine the interactive computer-generated world with an interactive real world in such a way that they appear as one environment, especially in CAD design. The authors decided to apply the system to aid the designer of the machinery systems by choosing standard parts. The system enables the user to easily view the 3D models of standard parts from any perspective, in more natural, intuitive way than traditional one (on the computer screen). Next, the user can export the model to the modeling software and use it to model some machinery system. This paper presents possibilities of using AR technology in CAD design with the hope that maybe someday it would become an integral part of a standard design process.
Keywords: augmented reality, image processing, CAD design, visualization, marker, VRML.

Numerical modeling of the human pelvic bone is a complex process in which many important factors are taken into account. One of them concerns material properties. Numerical calculations require the characteristics of the material properties and the material parameters from the beginning. The material properties of the living body depend on age, health, gender, environment and many others factors. To determine correct material parameters, health details of a group of patients need to be taken into consideration. In this paper authors assumed interval values of the selected material parameters and proposed interval and fuzzy analysis of the pelvic bone.
Keywords: human pelvic bone, interval and fuzzy analysis, finite element method, material properties.

The article contains presentation of application of three-dimensional vision methods in realization of vibration measurements and their analysis. For this purpose algorithms were developed of discrete epipolar geometry and structure from motion, with the usage of one camera. Vibration amplitude is determined for selected measurement points on the analyzed object. Each point is represented by flat or three-dimensional marker attached on a construction. The article includes algorithms of the discussed methods and verification of those methods based upon simulation data, as we as preliminary experimental tests carried out on a test bed.
Keywords: 3D vision techniques, epipolar geometry, structure from motion, vibration measurements.

An adaptive information system is constructed in order to approximate a set of multidimensional data. To get better approximation properties a pre-processing stage of data is proposed in which the set of points, forming the multidimensional data base and called a training set TRE, undergoes a clustering analysis. In the analysis two independent clustering algorithms are used; on each cluster a feed-forward neural network is trained and a membership function of a fuzzy set is constructed. The constructed system contains a module of two-conditional fuzzy rules consequent parts of which are of the functional type. Each rule is designed on a pair of clusters.

The integration of electronic units, sensors and actuators into complex function-oriented systems is one of the key points in the development of intelligent fibre composites. A wide range of materials is available for that purpose, for example piezoelectric textile sensors, fibre-optic fibres as well as shape memory alloys and prefabricated information elements. These elements can be used to create active fibre composites ("smart composites") with selective properties, which are suitable especially for application in stressed lightweight components. While the functionality of these solutions could be proved on a laboratory scale, appropriate manufacturing strategies for a competitive series production of components in automotive and mechanical technologies have not been realised so far. One crucial obstacle which impedes a breakthrough for such active lightweight components is the lack of technologies suitable for large-scale production. Generally, these complex systems are manually integrated into the fibre composite with either prefabricated layer materials or as individual elements, or applied to the surface of the fibre composite compound, thus preventing process automation. Therefore, the goal of the Institute of Mechanical Engineering and Plastics Technology (Chemnitz University of Technology) is to develop application-oriented technological solutions for series production.

The aim of the paper is to combine the evolutionary algorithms method, optimal dynamic filtration method and measurement data for the simultaneous identification of the thermal properties or their temperature characteristics of anisotropic solids. The idea of the proposed method depends on measuring the time-dependent temperature distribution at selected points of the sample and identification of the thermal parameters (heat conductivity and specific heat) by solving a transient inverse heat conduction problem. In the paper the discrete mathematical model has been formulated basing on the control volume method. The inverse problem was solved by using a hybrid method. Information about measurement data which are necessary to solve the inverse heat conduction problem was obtained by solving the direct heat conduction problem. The chosen results of analysis have been presented.
Keywords: control volume method, discrete mathematical model, parameter inverse problem, evolutionary algorithms method, optimal dynamic filtration method, thermal properties of anisotropic materials.

The mathematical model of solidification process can be formulated using the macro or micro-macro approach. In this paper the second generation model (micro-macro one) is considered. The driving force of crystallization is the local and temporary undercooling below solidification point T_cr. The nucleation and nuclei growth are proportional to the second power of undercooling. Formulas determining the phenomena previously mentioned contain coefficients called the nucleation coefficient and nuclei growth one. These coefficients are assumed to be interval values. For above assumptions the problem has been solved by means of interval boundary element method. In the final part of the paper the results of computations are shown.
Keywords: interval boundary element method, interval arithmetic, crystallization

The exploration of protein conformation can be supported by methods of similarity searching that allow seeking the 3D patterns in a database containing many molecular structures. We developed a novel search method called EAST (Energy Alignment Search Tool), which serves as a tool for finding strong structural similarities of proteins. It differs from other algorithms that concentrate on fold similarities. We use the EAST to find protein molecules representing the same structural protein family and inspect conformational modifications in their molecular structures as an effect of biochemical reactions or environmental influences. The similarity searching is performed through the comparison and alignment of protein energy profiles. Energy profiles are received in the computational process based on the molecular mechanics theory. These profiles are stored in the special database (Energy Distribution Data Bank, EDB) and can be used later by the search engine to find similar fragments of protein structures on the energy level. In order to optimize the alignment path we use modified, energy-adapted Smith-Waterman method, which is one of the main phases of the EAST. The use of fuzzy techniques improves the fault tolerance of presented method and allows to measure the quality of the alignment. In the paper, we present the main idea of the EAST algorithm and brief discussion on its basic parameters. Finally, we give an example of the system usage regarding proteins from the RAB family that play an important role in intracellular reactions in living organisms.
Keywords: bioinformatics, proteins, soft computing, data mining, protein structure comparison.

The paper is devoted to applications of evolutionary algorithms in identification of structures being under the uncertain conditions. Uncertainties can occur in boundary conditions, in material parameters or in geometrical parameters of structures and are modelled by three kinds of granularity: interval mathematics, fuzzy sets and theory of probability. In order to formulate the optimization problem for such a class of problems by means of evolutionary algorithms the chromosomes are considered as interval, fuzzy and random vectors whose genes are represented by: (i) interval numbers, (ii) fuzzy numbers and (iii) random variables, respectively. Description of evolutionary algorithms with granular representation of data is presented in this paper. Various concepts of evolutionary operator such as a crossover and a mutation and methods of selections are described. In order to evaluate the fitness functions the interval, fuzzy and stochastic finite element methods are applied. Several numerical tests and examples of identification of uncertain parameters are presented.
Keywords: evolutionary algorithms, granular computing, intervals, fuzzy sets, theory of probability, identification.

Inside the EU countries significant investments are expected in both the electricity production and energy transfers within the next 15 years. Such investments will need the precise decision-making processes, supported with very versatile engineering tools. The major objective of this paper is to propose an application of a new methodology to design power systems in a fully automatic way. The proposed methodology utilizes such artificial intelligence tools like genetic algorithms and expert systems.
Keywords: genetic algorithms, expert systems, combined heat and power (CHP), power plant.

The focus of this paper is on the problems of system identification, process modeling and time series forecasting which can be met during the use of locally recurrent neural networks in heuristic modeling technique. However, the main interest of this paper is to survey the properties of the dynamic neural processor which is developed by the author. Moreover, a comparative study of selected recurrent neural architectures in modeling tasks is given. The results of experiments showed that some processes tend to be chaotic and in some cases it is reasonable to use soft computing models for fault diagnosis and control.
Keywords: chaotic dynamic systems, recurrent neural networks, gradient-based and soft computing learning algorithms, nonlinear system identification, time-series forecasting.

This paper considers decision systems with decision attributes which are hierarchical. Atomic queries are built only from values of decision attributes. Queries are constructed from atomic queries the same way as we construct terms in logic using functors {+,*, ¬}. Negation symbol "¬" is only used on the atomic level. Queries are approximated by terms built from values of classification attributes. We only consider rule-based classifiers as the approximation tool for queries. When a user query fails, then the cooperative module of the query answering system (QAS) constructs its smallest generalization which does not fail and which is approximated by rules of the highest confidence discovered by the classifier. Two interpretations of queries are proposed: user-based and system-based. They are used to introduce the precision and recall of QAS. The implementation of QAS follows system-based interpretation. Automatic indexing of music by instruments and their types is an example of the application area for the proposed approach.
Keywords: information systems, knowledge discovery, cooperative query answering, music information retrieval.

Humans find it extremely easy to say if two words are related or if one word is more related to a given word than another one. For example, if we come across two words - "car" and "bicycle", we know they are related since both are means of transport. Also, we easily observe that "bicycle" is more related to "car" than "fork" is. In the paper we describe our approach on quantifying the semantic relatedness of concepts based on the theory of associative learning of concepts.
Keywords: semantic surrounding, associative learning, concept similarity, grammar, relatedness.

This paper presents an Ant Colony Optimization (ACO) approach to the resource-constrained project scheduling problem (RCPSP). RCPSP as a generalization of the classical job shop scheduling problem belongs to the class of NP-hard optimization problems. Therefore, the use of heuristic solution procedures when solving large problem is well-founded. Most of the heuristic methods used for solving resource-constrained project scheduling problems either belong to the class of priority rule based methods or to the class of metaheuristic based approaches. ACO is a metaheuristic method in which artificial ants build solutions by probabilistic selecting from problem-specific solutions components influenced by a parametrized model of solution, called pheromone model. In ACO several generations of artificial ants search for good solution. Every ant builds a solution step by step going through several probabilistic decisions. If ant find a good solution mark their paths by putting some amount of pheromone (which is guided by some problem specific heuristic) on the edges of the path.
Keywords: resources constrained scheduling problem, project scheduling, multi-project scheduling, ant colony optimization, swarm intelligence.

The process of engineering analysis, especially its preprocessing stage, comprises some knowledge-based tasks which influence the quality of the results greatly, require considerable level of expertise from an engineer; the support for these tasks by the contemporary CAE systems is limited. Analysis of the knowledge and reasoning involved in solving these tasks shows that the appropriate support for them by an automated system can be implemented using case-based reasoning (CBR) technology and ontological knowledge representation model. In this paper the knowledge-based system for intelligent support of the preprocessing stage of engineering analysis in the contact mechanics domain is presented which employs the CBR mechanism. The knowledge representation model is formally represented by the OWL DL ontology. Case representation model, case retrieval and adaptation algorithms for this model and the automated system are described.
Keywords: artificial intelligence, case-based reasoning, ontologies, finite element analysis.

In the paper an intelligent monitoring system of local water supply system is described. The main task of this system concerns water leakages detecting. For inputs, this system uses information from few pressure or flow sensors, mounted on the pipeline network, the output is a piece of information about leakage detection and localization. A heuristic model of water supply network makes the main part of intelligent diagnostic system. The model was built with the use of artificial neural networks. This paper presents the structure and optimization of a heuristic model. The authors took advantage of methods of artificial intelligence and methods known from model-based process diagnostics to increase the accuracy with which system detects of water leakages.
Keywords: water supply systems, diagnostics, genetics algorithm, artificial neural network.