Wyniki wyszukiwania - BazTech

1

Influence of innovative three-element binder on permanent deformations in recycled mixtures with emulsion and foamed bitumen

Mazurek Grzegorz, Iwański Marek, Buczyński Przemysław, Horodecka Renata

Archives of Civil and Mechanical Engineering

|

2021

|

Vol. 21, no. 2

236--261

EN

This article is detected to the assessment of durable deformations of recycled mixtures made of foamed bitumen (MCAS) and emulsion (MCE). In the basic part of research and analyses, attention was focused on determining the scale of three-component composition modifying rheological phenomena of recycled mixtures and other selected features considering various methods of bituminous binder proportioning. Cement, hydrated lime, and dusts from cement dust extracting system were included in the composition hydraulic binder. In this paper, the effect of graining of recycled mixture was also taken into account. One of the main scientific aims of the paper was to evaluate the degree of changes in durable deformations described in the power model depending on proportions of elements making three-element hydraulic binder. In effect, it was pointed out that the influence of hydraulic binder differently affected the durable deformation of recycled mixture depending on ways of bitumen binder implementation. There-element binder exerted the highest influence on mechanical properties of mixtures with fine-grained mixtures made according to the MCAS technology. The presence of 4 groups of mixtures with different properties was demonstrated using the classification neuron net. Based on that information, a set of the most recommended solutions from the point of view of time deformation resistance, low sensitiveness to the load level at moderate stiffness was selected. The best representative among them was the arrangement with 20% of hydrated lime, less than 40% of CBPD dusts and 40 ÷ 60% of cement.

2

Projektowanie analogowych układów scalonych CMOS o strukturze sieci neuronowej do przetwarzania obrazów i sygnałów

Kowalski J.

Zeszyty Naukowe. Rozprawy Naukowe / Politechnika Łódzka

|

2012

|

Z. 420

1-135

PL

W monografii przedstawiono wyniki prac autora dotyczące projektowania i testowania prototypowych, analogowych układów scalonych CMOS, odpowiednich do neuronowego przetwarzania obrazów i sygnałów, na przykładzie trzech zaprojektowanych i przetestowanych układów scalonych. Układy zostały wykonane przez konsorcjum Europractice w różnych technologiach CMOS, tj. 2,4 μm, 0,8 μm oraz 0,35 μm W zaprojektowanych układach oprócz właściwej sieci neuronowej implementowano specjalne struktury testowe, które umożliwiły wykonanie pomiarów podstawowych bloków funkcjonalnych sieci. Pozwoliło to na porównanie wyników symulacji z pomiarami oraz na uzyskanie informacji wykorzystanych do budowy stanowiska do testowania poprawności działania wykonanych układów scalonych. Dla każdego układu zaprojektowano specjalne stanowisko pomiarowe, które umożliwiło weryfikację doświadczalną działania danej sieci neuronowej. Pierwszym prezentowanym układem scalonym jest sieć Kohonena, dedykowana do zadań identyfikacji parametrów układów dynamicznych, przetwarzająca dane w sposób analogowy. Przedstawiono architekturę układu realizującego sieć, jego implementację w technologii MIETEC 2,4 μm oraz wyniki pomiarów podstawowych bloków funkcjonalnych sieci. Drugim zaprezentowanym układem scalonym jest filtr ważonych statystyk porządkowych obrazu o architekturze sieci neuronowej komórkowej, zaprojektowany w technologii AMS 0,8 μm CYE. Omówiono model komórki tego filtru oraz jego architekturę. Podano też szczegółowy opis bloków funkcjonalnych wchodzących w skład filtru oraz wyniki badań eksperymentalnych. Ostatnią część monografii stanowi projekt sieci neuronowej zbudowanej z synchronizowanych oscylatorów, służącej do segmentacji obrazów binarnych. W pracy zaproponowano nowy model oscylatora oraz architekturę układu scalonego realizującego sieć. Przedstawiono również projekt układu scalonego wykonanego w technologii AMIS 0,35 μm C035M-D 5M/1P i wyniki pomiarów.

EN

This monograph summarizes Author's research in the field of designing and testing CMOS prototype analog-integrated-circuit neural networks for image and signal processing. Three chips are presented which implement three various types of neural networks. The circuits have been designed using different CMOS technologies offered by Europractice, i.e. 2,4 μm, 0,8 μm and 0,35 μm ones. Apart from a main neural network, special test structures have been implemented in the circuits. The test structures enable the neural-network basic building blocks to be measured. This allows us to compare simulation with measurement results and provides some information needed for proper designing the integrated-circuit functional-test set-up. A special test set-up has been realized for each integrated circuit to perform functional verification of a given neural network. The first ASIC circuit considered in this monograph is a Kohonen network, operating with analog signals, dedicated for estimation of dynamic-system parameters. Architecture of the circuit, its implementation in the MIETEC 2,4 μm technology, as well as measurement results has been presented. The second integrated circuit presented in the monograph is a filter, based on a cellular neural network architecture, suitable for weighted-order-statistic image processing. It has been designed in the AMS 0,8 μm CYE technology. The filter cell model and structure have been described. Detailed description of its basic building blocks and the chip test results have been shown. The final part of this monograph is a description of a synchronized-oscillators-based neural network implemented in an ASIC form, which is well suited for binary-image-segmentation tasks. A new oscillator model and architecture of the designed circuit have been proposed. The AMIS 0,35 μm C035M-D 5M/1P technology has been used. Design, simulation and measurement results have been presented as well.

3

Automatic prolongation recognition in disordered speech using CWT and Kohonen network

Codello I., Kuniszyk-Jóźkowiak W., Smołka E., Kobus A.

Journal of Medical Informatics & Technologies

|

2012

|

Vol. 20

137--144

EN

Automatic disorder recognition in speech can be very helpful for the therapist while monitoring therapy progress of the patients with disordered speech. In this article we focus on prolongations. We analyze the signal using Continuous Wavelet Transform with 18 bark scales, we divide the result into vectors (using windowing) and then we pass such vectors into Kohonen network. Quite large search analysis was performed (5 variables were checked) during which, recognition above 90% was achieved. All the analysis was performed and the results were obtained using the authors' program - "WaveBlaster". It is very important that the recognition ratio above 90% was obtained by a fully automatic algorithm (without a teacher) from the continuous speech. The presented problem is part of our research aimed at creating an automatic prolongation recognition system.

4

Disordered sound repetition recognition in continuous speech using CWT and Kohonen network

Codello I., Kuniszyk-Jóźkowiak W., Smołka E., Kobus A.

Journal of Medical Informatics & Technologies

|

2011

|

Vol. 17

123--130

EN

Automatic disorders recognition in speech can be very helpful for therapist while monitoring therapy progress of patients with disordered speech. This article is focused on sound repetitions. The signal is analyzed using Continuous Wavelet Transform with 16 bark scales, the result is divided into vectors and passed into Kohonen network. Finally, the Kohonen winning neuron result is put on the 3-layer perceptron. The recognition ratio was increased by about 20% by adding a modification into the Kohonen network training process as well as into CWT computation algorithm. All the analysis was performed and the results were obtained using the authors' program ”WaveBlaster“, The problem presented in this article is a part of our research work aimed at creating an automatic disordered speech recognition system.

5

Badanie kategoryzacji zagrożenia tąpaniami z wykorzystaniem sieci neuronowych

Kabiesz J.

Prace Naukowe GIG. Górnictwo i Środowisko / Główny Instytut Górnictwa

|

2008

|

Wyd. spec. nr 7

93-104

PL

Celem przedstawionych w niniejszym artykule badań było sprawdzenie przydatności sieci neuronowych jako narzędzia umożliwiającego kategoryzację zagrożenia tąpaniami w kopalniach węgla kamiennego. Sprawdzano na rzeczywistym przykładzie eksploatacji pokładu węgla kamiennego w jednej z kopalń możliwość klasyfikacji tego zagrożenia przez sieć neuronową Kohonena. Sprawdzano wyniki uczenia tej sieci na zestawach danych (zmiennych wejściowych) niewątpliwie wpływających na stan zagrożenia oraz danych rozszerzanych o zestawy zmiennych o wartościach losowych lub/i wartościach stałych. Badano w ten sposób czułość i odporność wyników uczenia sieci na występowanie informacji niezwiązanych z klasyfikowanym zagrożeniem reprezentowanym przez stały zestaw danych.

EN

Investigations presented in the paper were targeted at the checking of neural networks' usefulness as a tool enabling the categorization of outbursts hazard in hard coal mines. A possibility of classification of this hazard by the Kohonen neuron network was checked on a real example of exploitation of hard coal seam in one of mines. The results of learning of this network were verified on sets of data (input variables) undoubtedly influencing the state of hazard, as well as data being extended with sets of variables of random values or/and the constant values. In this way, the sensitivity and resistance of network learning results was tested for the occurrence of information not associated with the classified hazard represented by a fixed data set.

6

Wykorzystanie sieci Kohonena do selekcji podobrazów na potrzeby dopasowania zdjęć lotniczych

Czechowicz A., Mikrut Z.

Automatyka / Akademia Górniczo-Hutnicza im. Stanisława Staszica w Krakowie

|

2007

|

T. 11, z. 3

329-343

PL

Automatyzacja procesu wyznaczania elementów orientacji wzajemnej zdjęć lotniczych jest jednym z kluczowych zadań w fotogrametrii. Artykuł przedstawia zastosowanie reprezentacji obrazu opartej na informacji o rozkładzie gradientu oraz sieci neuronowych Kohonena do selekcji podobrazów dla potrzeb dopasowania zdjęć lotniczych. Badania przeprowadzono, wykorzystując 904 podobrazy zdjęć lotniczych okolic Krakowa o różnym pokryciu terenu, grupując próbki w trzy kategorie: obszarów korzystnych, pośrednich i niekorzystnych pod względem wyszukiwania cech do orientacji wzajemnej. Dla każdego podobrazu pozyskano dwuwymiarowy histogram gradientu. Na jego podstawie wyznaczono reprezentację w postaci wektora wartości maksymalnych dla kierunku gradientu. Reprezentację wykorzystano do klasyfikacji obszarów siecią Kohonena. Poprawność uzyskanej klasyfikacji w stosunku do wykonanej manualnie otrzymano na poziomie 68,3%.

EN

Automatic relative orientation is one of the key problems in photogrammetric processing. This paper concerns the application of the representation based on the gradient distribution and Kohonen neural networks for the selection of sub-images for aerial photographs matching purposes. The examinations were conducted over 904 sub-images of the aerial photographs of the Krakow's surroundings with different land cover, grouped into three categories: advantageous, nondescript and disadvantageous in respect of searching features for relative orientation. The 2D histogram was acquired for every sub-image and on this basis the representation in form of the vector of maximum values for gradient direction has been determined. This representation was utilized for the classification of areas with Kohonen network. The correctness of the obtained classification, compared to manually done, achieved the Ievel of 68,3%.

7

The method of neuron weight vector initial values selection in Kohonen network

Chandzlik S.

Journal of Medical Informatics & Technologies

|

2006

|

Vol. 10

189--197

EN

Diagnosing of morbid conditions by means of automatic tools supported by computers is a significant and often used element in modern medicine. Some examples of these tools are automatic conclusion-making units of Parotec System for Windows (PSW). In the initial period of PSW system implementation, the units were used for recognition of orthopaedic diseases on the basis of the patient's walk and posture [15,17]. Subsequently, many additional options have been implemented, which have been used for purposes of diagnosing neurological diseases [1,2,3,9,12]. During automatic classification of diseases the additional units use elements of neural networks. The vectors based on normalised diagnostic measures [3] are inputs of the units. The measurements describe a patient's posture condition, his walk and overloads occurring on his feet. The Counter-Propagation (CP), two-layer network has been used in one of the automatic conclusion-making units. During CP network activity, we can see not only supervised but unsupervised learning processes as well. This is a characteristic feature of the CP network. The initial steps of the CP network learning process are very important, because the success of the network training process depends on them to a great extent. Therefore, a new method of weight vector initial values selection was proposed. The efficiency of the method was compared with classical methods. The results were very satisfactory. Owing to the proposed method, the time of the network training process as well as the mean-square error and the classification error was reduced. The research has been carried out using clinical cases of some neurological diseases: Parkinson's Disease, left-lateral hemiparesis and right-lateral hemiparesis after ischemic stroke. The measurements, which were made on a control group of patients without any neurological diseases, were the reference for these diagnostic classes.