W pierwszej części artykułu (Ch&K 3/2009) przedstawiono teorie działania, modele numeryczne i matematyczne Sztucznych Sieci Nauronowych (SSN) oraz scharakteryzowano rodzaje tych sieci. Dopełnieniem tematu będzie omówienie, w niniejszej publikacji, sposobu uczenia jednokierunkowej sieci neuronowej metodą wstecznej propagacji błędów.
Aktualna struktura sieci transportowych jest skutkiem długotrwałych i często przypadkowych procesów w przeszłości, jest więc niemal pewne, że nie jest ona optymalna dla obecnych potrzeb. Pilnym zadaniem jest zatem opracowanie metody umożliwiającej znalezienie optymalnej struktury sieci, odpowiadającej obecnym jak i prognozowanym potrzebom. Ze względu na brak analitycznego opisu i złożoność zagadnienia celowym wydaje się użycie metod sztucznej inteligencji. Jest bardzo prawdopodobne, że dla prawie każdego zestawu danych wejściowych istnieje kilka różnych rozwiązań o podobnej jakości, więc wiarygodna metoda powinna umożliwić jednoczesne znalezienie większości z nich. W pracy zaproponowano taką odmianę algorytmu genetycznego, która w sposób naturalny generuje alternatywne struktury sieci transportowej.
EN
The current network structure is usually the result of historical and often random long adaptation processes and it is almost sure not optimal for present demands. It is an urgent task to develop a method for searching the optimal transportation network structure, suitable for present and forecast demands. In the absence of analytical description and due to the complexity the usage of artificial inteligency methods seems to be the most appropriate. It is very likely, there are a few solutions of similar quality for almost each input data set. Therefore, the credible method should make possible a simultaneous obtainment most of them. In the paper the variant of genetic algorithm is presented, which in natural way generates alternative structures of a transportation network.
3
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
Purpose: The paper presents some results of the research connected with the development of new approach based on the neural network to predict the chemical composition and cooling rate to the mechanical properties of Al-Si-Cu cast alloys. The independent variables in the model are chemical composition of Al-Si-Cu cast alloys and cooling rate. The dependent parameters are hardness, microhardness, yield strength and apparent elastic limit. Design/methodology/approach: The experimental alloy used for training of neural network was prepared at the University of Windsor (Canada) in the Light Metals Casting Laboratory, in a 10 kg capacity ceramic crucible. Thermal analysis tests were conducted using the UMSA Technology Platform. Compression tests were conducted at room temperature using a Zwick universal testing machine. Prior to testing an extensometer was used to minimize frame bending strains. Compression specimens were tested corresponding to each of the three cooling rate. Rockwell F-scale hardness tests were conducted at room temperature using a Zwick HR hardness testing machine. Vickers microhardness tests were conducted using a DUH 202 microhardness testing machine. Findings: The results of this investigation show that there is a good correlation between experimental and predicted dates and the neural network has a great potential in mechanical behavior modeling of Al-Si-Cu castings. Practical implications: The worked out model can be applied in computer system of Al-Si-Cu casting alloys selection and designing for Al-Si-Cu casting parts. Originality/value: Original value of the work is applied the artificial intelligence as a tools for designing the required mechanical properties of Al-Si-Cu castings.
4
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
The objective of the paper is to present the concept of using artificial intelligence methods in simulation model s of selected logistics processes. The issues discussed in the paper refer to modelling selected logistics processes in the companies that produce electricity and thermal energy.
PL
Badano komfort jazdy pasażera pojazdu szynowego poprzez analizę poprzecznych i pionowych przyspieszeń. Badania przeprowadzono za pomocą numerycznych symulacji ruchu 4-osiowego pojazdu szynowego, który jest reprezentowany przez model matematyczny o 27 stopniach swobody. Uwzględniono losowe nierówności geometryczne toru zaburzające ruch pojazdu szynowego i badano ich wpływ na komfort jazdy. Szczegółowo analizowano, jak na komfort jazdy wpływają prędkość jazdy oraz modyfikacja parametrów określających zawieszenie pierwszego i drugiego stopnia. Ponadto wyznaczono przyspieszenie i analizowano komfort jazdy dla różnych położeń pasażera wewnątrz nadwozia pojazdu.
5
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
Purpose: The paper presents optimization of technological processes according to MSTP (Modified Sustainable Technology Procedure) procedure by general analysis and estimation (with regard to BREF-BAT Reference Documents) of three realistic zinc plating technological processes. Design/methodology/approach: In this article one of the polyoptimization methods-genetic algorithms (GA) for technological processes optimization has been proposed. As well as the universal software Poli-Opt version 1.0 for technological processes optimization has been presented. Findings: Polyoptimization using simple genetic algorithms makes possible negative environmental influences minimization. It also leading to ideal process obtaining that in reality does not exist. Research limitations/implications: The optimum solution in polyoptimization process is not absolutely optimum. This solution is about the largest probability of optimum obtaining. This results from the fact of different criteria and different objective function values using. Practical implications: The simple and universal Poli-Opt version 1.0 software using elementary, but quite sufficient in practice, genetic algorithm could be in every company applied. Originality/value: The Modified Sustainable Technology Procedure with computer supported could be for every materials technological processes optimization used.
6
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
Purpose: The aim of this paper was developing a project of neural network for selection of steel grade with the specified CCT diagram - structure and of harness after heat treatment. Design/methodology/approach: The goal has been achieved in the following stages: at the first stage characteristic points of CCT diagram have been determined. At the second stage neural network has been developed and optimized. Findings: The neural network was developed in this paper, that allowed selection of steel grade with the assumed CCT diagram. Research limitations/implications: Created method for designing chemical compositions is limited by the established ranges of mass concentrations of elements. The methodology demonstrated in the paper makes it possible to add new steel grades to the system. Practical implications: The method worked out may be used in computer steel selection systems for the machine parts put to heat treatment. Originality/value: Presented computer aided method makes use of neural networks, and may be used for selecting the steel with the required structure after heat treatment.
7
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
W artykule przedstawiono metodę poszukiwania wytycznych do rozbudowy lub modernizacji sieci transportowej, tak aby zbliżyć ją do struktury optymalnej dla analizowanego obszaru, przy zadanych potrzebach transportowych. Zastosowanie wyspowego algorytmu genetycznego umożliwia jednoczesne znajdowanie kilku najlepszych rozwiązań, które mogą być następnie ponownie rozpatrywane pod kątem innych kryteriów, których nie można było ująć w proponowanym modelu. Łatwość modelowania struktury sieci transportowej i dość krótki czas obliczeń pozwalają na rozpatrywania szerokiego spektrum konkurencyjnych propozycji rozbudowy sieci. Aktualna postać sieci transportowej i zakres dopuszczalnych ingerencji w jej strukturę są danymi wejściowymi, zatem możliwe jest przeanalizowanie wielu opcji rozbudowy: od drobnych, koniecznych korekt, aż do gruntownej przebudowy w celu zbliżenia się do prawdziwie optymalnej struktury. W każdym rozpatrywanym wariancie uzyskiwane są informacje o korzyściach (zmniejszeniu kosztów użytkowania) i kosztach (nakładach na rozbudowę).
EN
The current transportation network structure is usually the result of historical and often long random adaptation processes and it is almost surely not optimal for the present needs. As always in the past, when modernization or extension was required, the current state was the base for changes. Cases, where the design of whole transportation network is possible, are very rare. Possible existence of several transportation network variants meeting new needs constitutes an additional difficulty. In this situation, development of a method allowing comparison of different variants of transportation network extension is an interesting and urgent task. In the absence of analytical description and due to the complexity of the problem. exploration of the whole range of solutions is impossible and it seems the most appropriate to use artificial intelligence methods, particularly evolutionary algorithms.
Rozwój metod sztucznej inteligencji i poszukiwania nowych oraz usprawnianie istniejących kanałów kontaktów z klientami zaowocowały zainteresowaniem programami symulującymi kompetencje komunikacyjne, w tym językowe człowieka. Obecnie programy tego typu rozwinęły się na tyle, że znalazły komercyjne zastosowania również w Polsce np. Adam, Hubert itd. W pracy przedstawiono chatterbota Chris opartego na metodzie wyszukiwania wzorców w wypowiedzi użytkownika. Prezentowane rozwiązanie jest oparte na mechanizmach wypracowanych w ramach projektu Alice fundacji A.L.I.C.E. AI. Wybór istniejącego szablonu rozwiązania pozwala skupić się na tym co w chatterbocie najważniejsze, czyli na jego bazie wiedzy. W ramach pracy przedstawiono wnioski i wskazówki, które pomogą zainteresowanym osobom tworzyć bazy wiedzy dla swoich wersji chatterbotów. Chatterbot Chris jest oparty na servletach Java, a jego wydajność jest na tyle duża, że nie stanowi elementu krytycznego.
EN
Development of artificial intelligence methods and looking for Improvement of existing contacts channels with clients are main reasons to increase of interests of programs for simulation human communication competences including spelling. Now this kind of programs are developed so far that have commerce application also in Poland (Adam, Hubert etc). In this paper chatterbot Chris based on looking for the pattern in user statement is presented. Presented solution is based on technologies created at the Alice project (A.L.I.C.E. foundation). The creation of the adequate knowledge base is the main part of programmers work. Some conclusions and advices which can be helpful to crate knowledge base adequate to programmers needs are presented. Presented bot is based on Java servlets, so its efficiency is satisfactory.
9
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
Purpose: The paper is concerned about intelligent motion of the mobile robot in the production area. The robot is self-learning and gathers the data from the environment by means of sensors. It processes the acquired information and utilizes it for making decisions. Design/methodology/approach: The concept imitates the natural selection of living organisms, where in the struggle for natural resources the fit individuals become more and more dominant and adaptable to the environment in which they live, whereas the less fit ones are present in the generations rarely. Some of the improved genetic operations were used for the robot motion. Findings: The use of those improved genetic operations has proved to be appropriate. By means of them the robot became more and more intelligent in the course of evolution and performed the set task successfully. Research limitations/implications: The tests were limited only to the space with static barriers. In future, it would be appropriate to test the proposed system also in the space with moving objects and to enable the robot to have full autonomy. Practical implications: The proposed system enables the robot to move completly independently in the space. The robot complies with simple instructions: come to the goal fastest possible (shortest path) without causing damage to youself and to the environment. Originality/value: Originality value is the implementation of the non-deterministic principles in the decision making strategy of the mobile robot. In learning and independent decision making the robot used some of the improved genetic operations.
10
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
The Applied Behavioral Analysis (ABA) is one of the very successful therapies on autistic children. There are no software systems supporting such a complex therapy either in Polish or international ABA rehabilitation centers. This article presents an idea of an integrated system to help managing the vast amount of data collected during behavioral therapy. There are four main modules in the system: a database containing therapy data, system for data analysis based on machine learning techniques, an expert system to support therapist and a semantic based search.
PL
Stosowana analiza behawioralna jest jedną ze skuteczniejszych terapii dla dzieci z autyzmem. Zarówno w Polskich jak i światowych centrach rehabilitacyjnych nie stosuje się systemów oprogramowania, które wspierałby tą skomplikowaną terapię. Niniejszy artykuł prezentuje ideę zintegrowanego systemu wspomagającego zarządzenie ogromną ilością danych gromadzonych podczas terapii. W systemie zaplanowano cztery moduły: bazę danych zawierającą dane z terapii, system analizy danych wykorzystujący techniki z maszynowego uczenia się, system ekspertowy dla terapeutów oraz oparty na semantyce.
11
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
Praca dotyczy oceny jakości elementów np. korpusów silników samochodowych ze stopów aluminium wytworzonych metodą Cosworth, przy wykorzystaniu analizy obrazów zdjęć cyfrowych uzyskanych w wyniku rentgenograficznych badań defektoskopowych tychże elementów. Uzyskane wyniki decydują o tym, czy wytworzony produkt jest odpadem poprodukcyjnym czy też nie, co ma na celu zmniejszenie liczby wytwarzanych produktów niespełniających kryteriów kontroli technicznej. W pracy przedstawiono opracowaną, komputerowo wspomaganą metodę oceny jakości odlewów, na podstawie morfologii wad, przy zastosowaniu narzędzi sztucznej inteligencji.
EN
This paper presents the analysis of automotive alloy castings of engine heads and blocks. The technological process applied for these Al-Si alloy castings, EN AC-AISi7Cu3Mg type, was Cosworth. AIso, it includes the development of the methodology of image analysis of casting defects. They were achieved by the defectoscopic X-ray analysis of AI-Si-Cu alloy elements, EN AC-AISi7Cu3Mg type. Besides, it contains both the manner of classification of casting defects by means of artificial intelligence, including neural networks and the application of the developed methodology in the form of a system of computer programmes for quality controI.
12
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
Purpose: The purpose of this paper is presentation of the computer program for calculating the Continuous Cooling Transformation diagrams for constructional and engineering steels. Design/methodology/approach: The computer program uses the artificial neural networks for prediction steel properties after heat treatment. Input data are chemical composition and austenitizing temperature. Results of calculation consist of temperature of the beginning and the end of transformation in the cooling rate function, the volume fraction of structural components and hardness of steel cooled from austenitizing temperature with a fixed rate. Findings: The algorithm can be use in designing new chemical compositions of steels with assumed hardness after heat treatment. Research limitations/implications: The created method for designing chemical compositions is limited by ranges of mass concentrations of elements. The methodology demonstrated in the paper makes possibility to add new steels to the system. Practical implications: The method may be used in computer steel selection systems for machines parts manufactured from constructional or engineering steels subjected to heat treatment. Originality/value: The presented computer program can be used for selecting steel with required structure after heat treatment.
13
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
The purpose of the paper is to present a computer program for calculating Continuous Cooling Transformation diagrams for constructional and engineering steels. The computer program uses artificial neural networks for prediction of steel parameters after heat treatment. Input data are chemical composition and austenitising temperature. The results of calculation consist of temperature of the beginning and end of transformation in the cooling rate function, the volume fraction of structural components and hardness of steel cooled from austenitising temperature at a fixed rate. The algorithm can be used in designing new chemical compositions of steels with required hardness after heat treatment.
PL
Celem niniejszego artykułu jest prezentacja komputerowego programu do obliczania diagramów CTPc dla stali konstrukcyjnych. Program ten korzysta ze sztucznych sieci neuronowych do predykcji parametrów stali po obróbce cieplnej. Danymi wejściowymi są skład chemiczny oraz temperatura austenityzowania. Wynikami obliczeń są temperatury początków i końców przemian w funkcji szybkości chłodzenia, udziały procentowe faz oraz twardość stali po chłodzeniu z temperatury austenityzowania. Algorytm może być użyty w projektowaniu nowych stali z wymaganą twardością po obróbce cieplnej.
14
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
Purpose: Most often used methods for prediction of austenite decomposition are described and analysed. Design/methodology/approach: The austenite decomposition prediction is usually based on continuous cooling transformation (CCT) diagrams. The next method is based on semi-empirical approach based on the Scheil's additivity rule. The third method is based on time, t8/5, relevant for microstructure transformation measured on Jominy-specimen. Very good results are obtained by artificial neural network (ANN) with learning rule based on the error backpropagation algorithm. Findings: By the comparison of application ability of investigated methods in mathematical modelling and computer simulation of austenite decomposition during the cooling of low-alloyed steel, it can be concluded that everyone method gives different results, and minimum variation in elemental composition and history of cooling may produce extremely different results in microstructure portion. Very good results were achieved by the method, which applies the Jominy-test results. In this method the additivity rule and specific performance of Jominy-test has been combined. Research limitations/implications: The investigation was performed on low-alloyed steels. Practical implications: The results of prediction of microstructure transformations could be used for prediction of mechanical properties after a heat treatment and of generation of stresses and strains during a heat treatment. Originality/value: The ability and applicability of potential methods of austenite decomposition prediction in general mathematical modelling of heat treatment of steel are carried out. The finding of this paper will be so useful in development new algorithms in mathematical modelling and computer simulation of heat treatment of low-alloyed steels.
15
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
Purpose: The purpose of this study is to develop a methodology for material design, enabling the selection of the chemical elements concentration, heat and plastic treatment conditions and geometrical dimensions to ensure the required mechanical properties of structural steels specified by the designer of machinery and equipment as the basis for the design of material components manufactured from these steels, by using a computational model developed with use of artificial intelligence methods and virtual environment. The model is designed to provide impact examinations of these factors on the mechanical properties of steel only in the computing environment. Design/methodology/approach: A virtual research environment built with use of computational model describing relationships between chemical composition, heat and plastic treatment conditions, product geometric dimensions and mechanical properties of the examined group of steels was developed and practical applied. This model enables the design of new structural steel by setting the values of mechanical properties based on material production descriptors and allows the selection of production descriptors on the basis of the mechanical properties without the need for additional tests or experimental studies in reality. Findings: Virtual computing environment allows full usage of the developed intelligent model of non-alloy and alloy structural steel properties and provides an easy, intuitive and user-friendly way to designate manufacturing descriptors and mechanical properties for products. Research limitations/implications:The proposed solutions allow the usage of developed virtual environment as a new medium in both, the scientific work performed remotely, as well as in education during classes. Practical implications: The new material design methodology has practical application in the development of materials and modelling of steel descriptors in aim to improve the mechanical properties and specific applications in the production of steel. Presented examples of computer aid in structural steel production showing a potential application possibility of this methodology to support the production of any group of engineering materials. Originality/value: The prediction possibility of the material mechanical properties is valuable for manufacturers and constructors. It ensures the customers quality requirements and brings also measurable financial advantages.
16
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
Paper presents review of theoretical methods for added assembly schedule and systems design. There was example analytical modeling and schedule optimization operations, based on new clinching technology, for went trunk ventilating canal made of steel profile presented. Production is realized on special assembly, technological station. There were scheduling assembly operations problem in model of traveling salesman problem formulated. Implementation model was in excel spreadsheet elaborated and with program Risk Solver Platform V 9.0 solved.
17
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
Purpose: The paper presents the application of artificial neural network for evaluation of alloying elements effect on selected materials properties and austenite transformations during continuous cooling. Design/methodology/approach: Multi-layer feedforward neural networks with learning rule based on the error backpropagation algorithm were employed for modelling the steels properties. Then the neural networks worked out were employed for the computer simulation of the effect of particular alloying elements on the steels' properties. Findings: Obtained results show that neural network are useful in evaluation of synergic effect of alloying elements on selected materials properties when classical investigations' results do not provide evaluation of the effect of two or more alloying elements. Research limitations/implications: The results presented are valid in the ranges of mass concentrations of alloying elements presented in the paper. Practical implications: The worked out relationship may be used in computer systems of steels' designing for the heat-treated machine parts. Originality/value: The use of the neural networks as an tool for evaluation of the chemical composition effect on steels' properties.
18
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
Purpose: Modelling of the agent based systems is presented along with a brief description of the relevant software toolkit. Design/methodology/approach: Areas of activity of the agent systems are presented, split into their internal and external domains, recognition of their world's state of affairs as well as planning of actions to affect their environment. Findings: Overview of the approach to development of the agent based systems from the general goal notion to its constituent elements, like plans and tasks. The proposed hybrid control structure allows both the horizontal level negotiations among the local agents and the vertical intervention by supervisory agents; therefore, such attitude makes also possible the multi-task and many-to-many negotiations. The result is the global control and the possibility of the decentralized negotiation. Research limitations/implications: Modelling LPA Chimera toolkit is based on Prolog which supports mainly backward chaining. Originality/value: Selection of a powerful and flexible agent systems modeling tool has been made, exemplary analysis of the agent based system model design has been presented. There agents representing the physical system elements, like a humans, machine, production line, shop floor system, the entire plant, or simply a workpiece; it may also be the part-oriented scheduling, or even the complete scheduling process.
19
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
Purpose: The goal of this work was development of the computer aided system for selection of technological parameters for making the metallographic microsections for the microscopic examinations. Design/methodology/approach: Prolog language - the Artificial Intelligence tool was employed in the project, and the Logic Programming Associate's VisiRule system, which was used for saving the knowledge base - rules pertaining the objects and their relationships. These tools were used to develop the expert system for selection of the recommended technology for the metallographic microsections. Findings: Knowledge acquisition process was carried out to cover recommendations of manufacturers of the equipment for metallographic microsections. Formulation of the pertinent knowledge in the LPA-Prolog VisiRule tool is presented along with the results of technology instruction sheet presentation. Research limitations/implications: The expert system for selection of parameters for making the metallographic microsections is currently at the testing and development stage. Originality/value: Making the metallographic microsections is a time consuming task, sometimes material for examination is scarce, so any system that ensures good quality of microsections is very valuable. The system under development offers the best advice to anybody striving to obtain the best quality of the specimens.
W artykule omówiono podstawowe aspekty realizacji aktywnych strategii inwestycyjnych na rynkach finansowych z wykorzystaniem systemów wspomagania decyzji (systemów transakcyjnych), w kontekście klasycznych teorii zarządzania portfelem inwestycyjnym. Wskazano zasadnicze przesłanki zastosowania metod sztucznej inteligencji, takich jak sieci neuronowe i algorytmy genetyczne, do konstrukcji inwestycyjnych systemów decyzyjnych. Przedstawiono charakterystykę sieci neuronowych oraz algorytmów genetycznych jako efektywnych narzędzi w modelowaniu i prognozowaniu rynków finansowych.
EN
The paper discusses basic aspects of application of active investment strategies in financial markets - in the context of classic theories of portfolio management. Such active strategies are generated with the use of decision support systems (transaction systems). The main assumptions of utilisation of artificial intelligence methods, such as neural networks and genetic algorithms, in the construction of investment decision systems have been indicated. The characteristic of neural networks and genetic algorithms as effective tools in financial markets modelling and prediction has also been discussed here.
JavaScript jest wyłączony w Twojej przeglądarce internetowej. Włącz go, a następnie odśwież stronę, aby móc w pełni z niej korzystać.