Wyniki wyszukiwania - BazTech

1

Classification and modelling of sound emission signals in selected tribosystems

Kekez Michał, Jurczak Wojciech, Ozimina Dariusz

Tribologia

|

2021

|

nr 3

19--26

EN

The paper presents an analysis of the sound level recorded during dry sliding friction conditions. Balls with a diameter of 6 mm placed on pins were made of 100Cr6 steel, silicon carbide (SiC), and corundum (Al2 O3 ), while rotating discs with a height of 6 mm and a diameter of 42 mm were made of 100Cr6 steel. Each pin and disc system was tested for two values of the relative humidity of the air (50 ± 5% and 90 ± 5%). Models of the A-sound level were developed using regression trees and random forest. The paper presents an analysis of the accuracy of the models obtained. Classifications of the six tests performed on the basis of sound level descriptors were also carried out.

PL

W pracy przedstawiono analizę poziomu dźwięku zarejestrowanego podczas tarcia technicznie suchego w ruchu ślizgowym. Podczas sześciu testów tribologicznych stosowano próbkę wykonaną ze stali 100Cr6 oraz trzy przeciwpróbki, wykonane ze stali 100Cr6, węglika krzemu (SiC) i korundu (Al2 O3 ), przy czym każdy układ próbka – przeciwpróbka był testowany dla dwóch wartości wilgotności względnej powietrza (50 ± 5% i 90 ± 5%). Opracowano modele poziomu dźwięku A z użyciem drzew regresji i lasu losowego. W pracy zamieszczono analizę dokładności otrzymanych modeli. Została również przeprowadzona klasyfikacja sześciu wykonanych testów w oparciu o deskryptory poziomu dźwięku.

2

Analyzing rear-end crash severity for a mountainous expressway in China via a classification and regression tree with random forest approach

Wang Yonggang, Luo Xianyu

Archives of Civil Engineering

|

2021

|

Vol. 67, nr 4

591--604

EN

To understand the contributory factors to rear-end accident severity on mountainous expressways, a total of 1039 rear-end accidents, occurring on G5 Jingkun Expressway from Hechizhai to Qipanguan in Shaanxi, China over the period of 2012 to 2017, were collected, and a non-parametric Classification and Regression Tree (CART) model was used to explore the relationship between severity outcomes and driver factors, vehicle characteristics, roadway geometry and environmental conditions. Then the random forest model was introduced to examine the accuracy of variable selection and rank their importance. The results show that driver’s risky driving behaviours, vehicle type, radius of curve, angle of deflection, type of vertical curve, time, season, and weather are significantly associated with rear-end accident severity. Speeding and driving while drunk and fatigued are more prone to result in severe consequences for such accidents and driving while fatigued is found to have the highest fatality probability, especially during the night period (18:00-24:00). The involvement of heavy trucks increases the injury probability significantly, but decreases the fatality probability. In addition, adverse weather and sharp curve with radius less than 1000 m are the most risk combination of factors. These findings can help agencies more effectively establish stricter regulations, adopt technical measures and strengthen safety education to ensure driver's driving safety on mountainous expressways for today and tomorrow.

3

On the use of acoustic emission to assess the wear in a tribosystem

Kekez Michał, Desaniuk Tomasz, Duszczyk Jurek, Ozimina Dariusz

Journal of Machine Construction and Maintenance - Problemy Eksploatacji

|

2019

|

no. 2

99--103

EN

The article focuses on the analysis of acoustic emission signals generated under dry sliding friction conditions. Two tests were conducted using a TRB3 tribometer with the disc made of 100Cr6 steel with a DLC coating, and pin made of corundum (Al2O3) and steel 100Cr6, respectively. Two tests with the disc without DLC coating were also carried out. The audio data written in the 16-bit linear pulse-code modulation (LPCM) format were analysed using the SpectraPLUS software. An A-weighting filter and 1/1 and 1/3-octave band filters were used for sound level measurements. The analysis of the equivalent sound level calculated for 10-second time intervals was carried out. The highest A-weighted sound level occurred during the first 2 hours of the test with the disc having a DLC coating and pin made of 100Cr6 steel. At the end of this test, the sound level dropped by about 40 dB compared to the maximum. The lowest A-weighted sound level was recorded during the last 2 hours of the test with disc having a DLC coating and pin made of corundum. The time-dependent variability of sound parameters was predicted using the regression tree and random forest models, which proved to be accurate and easy to follow.

PL

W pracy przedstawiono analizę dźwięku zarejestrowanego podczas tarcia technicznie suchego w ruchu ślizgowym. Dwa testy przeprowadzono na tribometrze TRB3 dla próbek wykonanych ze stali 100Cr6 z powłoką DLC i przeciwpróbek wykonanych odpowiednio z korundu (Al2O3) i stali 100Cr6. Przeprowadzono również dwa testy dla próbek bez powłoki DLC. Dźwięk został zarejestrowany w standardzie 16-bitowego liniowego PCM, a następnie poddany analizie w programie SpectraPlus. Dla kolejnych chwil czasu wyznaczono wartości poziomu dźwięku A, a także poziomy dźwięku w wybranych pasmach oktawowych i 1/3-oktawowych. Przeprowadzono analizę równoważnego poziomu dźwięku obliczonego dla 10-sekundowych odcinków czasu. Najwyższy poziom dźwięku A występował podczas pierwszych 2 godzin testu próbki z powłoką DLC i przeciwpróbki wykonanej ze stali 100Cr6. Pod koniec tego testu poziom dźwięku spadł o około 40 dB względem dotychczasowego maksimum. Najniższy poziom dźwięku A zanotowano podczas ostatnich 2 godzin testu, w którym próbka miała powłokę DLC, a przeciwpróbka była wykonana z korundu. Utworzono modele opisujące zmienność w czasie wybranych parametrów dźwięku, oddzielnie dla każdej próbki. Do utworzenia modeli zastosowano drzewa regresji oraz Random Forest. W pracy zamieszczono analizę dokładności i przejrzystości otrzymanych modeli.

4

Decision tree for modeling survival data with competing risks

Dauda Kazeem Adesina, Pradhan Biswabrata, Uma Shankar B., Mitra Sushmita

Biocybernetics and Biomedical Engineering

|

2019

|

Vol. 39, no. 3

697--708

EN

This work considers decision tree for modeling survival data with competing risks. A Survival Classification and Regression Tree (SCART) technique is proposed for analysing survival data by modifying classification and regression tree (CART) algorithm to handle censored data for both regression and classification problems. Different performance measures for regression and classification tree are proposed. Model validation is done by two different cross-validation methods. Two real life data sets are analyzed for illustration. It is found that the proposed method improve upon the existing classical method for analysis of survival data with competing risks.

5

A comparison of advanced computational models and experimental techniques in predicting blast induced ground vibration in open pit coal mine

Nguyen Hoang, Bui Xuan Nam, Moayedi Hossein

Acta Geophysica

|

2019

|

Vol. 67, no. 4

1025--1037

EN

Green mining is an essential requirement for the development of the mining industry. Of the operations in mining technology, blasting is one of the operations that signifcantly affect the environment, especially ground vibration. In this paper, four artificial intelligence (AI) models including artificial neural network (ANN), k-nearest neighbor (KNN), support vector machine (SVM), and classification and regression tree (CART) were developed as the advanced computational models for estimating blast-induced ground vibration in a case study of Vietnam. Some empirical techniques were applied and developed to predict ground vibration and compared with the four AI models as well. For this research, 68 events of blasting were collected; 80% of the whole datasets were used to build the mentioned models, and the rest 20% were used for testing/checking the models’ performances. Mean absolute error (MAE), determination coefficient (R2 ), and root-mean-square error (RMSE) were used as the standards to evaluate the quality of the models in this study. The results indicated that the advanced computational models were much better than empirical techniques in estimating blast-induced ground vibration in the present study. The ANN model (2-6-8-6-1) was introduced as the most superior model for predicting ground vibration with an RMSE of 0.508, R2 of 0.981 and MAE of 0.405 on the testing dataset. The SVM, CART, and KNN models provided poorer performance with an RMSE of 1.192, 2.820, 1.878; R2 of 0.886, 0.618, 0.737; and MAE of 0.659, 1.631, 0.762, respectively.

6

Modeling the 2D space of emotions based on the poincare plot of heart rate variability signal

Moharreri S., Dabanloo N. J., Maghooli K.

Biocybernetics and Biomedical Engineering

|

2018

|

Vol. 38, no. 4

794--809

EN

Emotions mean accepting, understanding, and recognizing something with one's senses. The physiological signals generated from the internal organs of the body can objectively and realistically reflect changes in real-time human emotions and monitor the state of the body. In this study, the two-dimensional space-based emotion model was introduced on the basis of Poincare's two-dimensional plot of the signal of heart rate variability. Four main colors of psychology, blue, red, green, and yellow were used as a stimulant of emotion, and the ECG signals from 70 female students were recorded. Using extracted features of Poincare plot and heart rate asymmetry, two tree based models estimated the levels of arousal and valence with 0.05 mean square errors, determined an appropriate estimation of these two parameters of emotion. In the next stage of the study, four different emotions mean pleasure, anger, joy, and sadness, were classified using IF-THEN rules with the accuracy of 95.71%. The results show the color red is associated with more excitement and anger, while green has small anxiety. So, this system provides a measure for numerical comparison of mental states and makes it possible to model emotions for interacting with the computer and control mental states independently of the pharmaceutical methods.

7

Wspomaganie decyzji w zakresie optymalizacji walcowania blach na gorąco z wykorzystaniem metod eksploracji danych

Regulski K., Szeliga D., Rojek G., Kusiak J.

Hutnik, Wiadomości Hutnicze

|

2015

|

Vol. 82, nr 4

248--255

PL

Optymalizacja parametrów procesu produkcji innowacyjnych materiałów, np. stali dwufazowej, wymaga modelu procesu termomechanicznego i dużych zbiorów danych pokrywających możliwą przestrzeń wyników. Trudności w optymalizacji parametrów procesu wynikają z dużej liczby zmiennych sterujących, które należy uwzględnić w projektowaniu technologii. W pracy przedstawiono możliwości zastosowania drzew regresyjnych we wstępnych analizach pozwalających na odnalezienie zależności pomiędzy parametrami procesu walcowania na gorąco blachy ze stali dwufazowej, a tym samym na wskazanie najważniejszych zmiennych w procesie, oraz konstrukcję reguł optymalizacji tego procesu. Wykazano, że indukcja drzew regresyjnych nie wymaga dużych zbiorów danych, ani wiedzy a priori o procesach termomechanicznych.

EN

The optimization of parameters of innovative steels manufacturing, e.g. dual phase steel, requires model of thermomechanical processes and large datasets that covers whole surface of results. Difficulties in the optimization of process parameters correspond with large number of control variables, which should be considered in the technology design. Presented work concerns possibility of application of regression trees in preliminary analysis related to discovering the relationships among parameters of the hot rolling process of dual phase steel strips − indicating the most important variables and construction of rules that refer to the optimization of this process. The regression trees induction neither required large datasets, nor any a priori knowledge about thermomechanical processes, what will be proved.

8

The spontaneous succession in a sand-pit: the role of life history traits and species habitat preferences

Kompała-Bąba A., Bąba W.

Polish Journal of Ecology

|

2013

|

Vol. 61, nr 1

13--22

EN

The prediction of species response to human activity is of great interest in contemporary restoration ecology. The purpose of the article was to analyse which species life history traits and species habitat preferences are important during succession after the abandonment of mining activity in a sand-pit. During a 15-year period (1996–2010), 176 phytosociological relevés were placed within vegetation patches of different ages and divided according to soil moisture, thus forming two series of chronosequences that ranged from 0 to ca. 50 years. The datasets were analysed using both DCA/CCA ordinations and regression trees. The successional gradient, apart from the theoretically predicted replacement R by C strategists, revealed the occurrence of stresstolerant ruderals and competitive ruderals in the early successional stages. However, differences were observed between the wet and dry series. On dry soils the ruderals, anemochorous and windpollinated species dominated in the early phases. In the middle successional phases, a preponderance of anemochorous species (nanophanerophytes, light or semi-shade demanding species), nitrogen-poor and competitive ruderals or species typical for nitrogen-rich soils was recorded. In the late successional stages, species with both the ability of vegetative and generative reproduction appeared. On the other hand, on wet sites in the early successional phases, species with vegetative growth, hydrophytes, chamaephytes and stresscompetitors prevailed. Later in the succession, they were replaced by insect-pollinated species in nitrogen rich habitats and stress-tolerant ruderals on less fertile habitats. Finally, competitors started to prevail. Knowledge of the environmental conditions of a given site, the ecological processes and species biology can assist in achieving the desired goals or in initiating or enhancing succession on some disturbed sites.

9

Sub-pixel classification of middle-resolution satellite images – evaluation of regression trees applicability to monitor impervious surfaces coverage

Drzewiecki W.

Geomatics and Environmental Engineering

|

2010

|

Vol. 4, no. 4

61-75

EN

The aim of the presented research was to test the method of assessing the imperviousness index on the basis of middle-resolution satellite images with the use of regression trees. The task also included evaluation of the applicability of the method to monitor the changes of impervious surfaces coverage. The research has been done in the catchments of Prądnik and Dłubnia rivers (Malopolska region, Poland). The imperviousness index has been assessed for two time periods – current state (2007) and the mid-1990s. The training and verification data for both time periods have been obtained from aerial orthophotomaps for urban, suburban, rural, industrial and commercial areas. In both time states the best assessment of imperviousness index have been achieved for the variants where the regression trees were built on the basis of all satellite data accessible for the time period. However, it is worth notifying that the variant with the input data limited to three images from spring, summer and autumn provided comparable accuracy of the results. These models have the systematic error between 1.3–2.2%, the mean error between 15.8–16.4% and correlation coefficient between 0.85-0.86 for the mid-1990s. For the year 2009 these values are respectively: 1.4–1.7%, 15.7–16.0% and 0.86. The accuracy of the imperviousness index obtained in the present research is comparable with the accuracy obtained with the use of regression trees in research reported in the literature. The comparison has shown high accuracy of imperviousness index change assessment for the whole population of pixels in verification dataset. The systematic error is 0.1%. However, the obtained assessment accuracy for a single pixel (š14.5%) can be too low for some applications.

PL

Celem prezentowanych badań było sprawdzenie możliwości zastosowania drzew regresji do szacowania wskaźnika nieprzepuszczalności powierzchni terenu na podstawie średniorozdzielczych obrazów satelitarnych. W ramach badań przeprowadzona została również ocena stosowalności tej metody jako narzędzia monitoringu pokrycia terenu powierzchniami nieprzepuszczalnymi. Badania przeprowadzono na obszarze obejmującym zlewnie rzek Prądnik i Dłubnia. Wskaźnik nieprzepuszczalności oszacowano dla dwóch stanów czasowych – roku 2007 i połowy lat 90. XX wieku. W obu przypadkach dane treningowe i weryfikacyjne pozyskano z ortofotomap lotniczych dla obszarów o zróżnicowanym użytkowaniu (terenów zabudowy miejskiej, podmiejskiej, wiejskiej, przemysłowej oraz handlowej). W przypadku obu stanów czasowych najlepsze oszacowanie wskaźnika nieprzepuszczalności uzyskano w wariantach, w których do budowy drzew regresji użyto wszystkich dostępnych obrazów satelitarnych z poszczególnych okresów. Na uwagę zasługuje jednak fakt, iż porównywalną dokładność oszacowania uzyskano także w wariantach, w których dane wejściowe ograniczone były jedynie do trzech obrazów pozyskanych w okresie wiosennym, letnim i jesiennym. Zbudowane modele pozwalały na oszacowanie wskaźnika nieprzepuszczalności dla stanu z połowy lat 90. z błędem przeciętnym wynoszącym 1.3–2.2%, błędem średnim pomiędzy 15.8% a 16.4% oraz współczynnikiem korelacji w granicach 0.85–0.86. Dla roku 2007 wartości te wyniosły odpowiednio: 1.4–1.7%, 15.7–16.0% i 0.86. Uzyskany w prezentowanych badaniach poziom dokładności oszacowania wartości wskaźnika nieprzepuszczalności jest porównywalny z wynikami uzyskiwanymi z zastosowaniem drzew regresji przez innych autorów. Ocena dokładności oszacowania zmian wskaźnika nieprzepuszczalności wykazała bardzo wysoką dokładność ich określenia w odniesieniu do całości populacji pikseli w próbce weryfikacyjnej. Błąd systematyczny wyniósł w tym przypadku 0.1%. Należy jednak zauważyć, iż uzyskany dla pojedynczego piksela błąd średni na poziomie š14.5% może być zbyt duży dla niektórych zastosowań takiego podejścia jako narzędzia monitoringu zmian pokrycia powierzchni terenu.

10

Methods of development fuzzy logic driven decision-support models in copper alloys processing

Kluska-Nawarecka S., Górny Z., Mrzygłód B., Wilk-Kołodziejczyk D., Regulski K.

Archives of Foundry Engineering

|

2010

|

Vol. 10, iss. 1 spec.

23--28

EN

Development of a diagnostic decision support system using different then divalent logical formalism, in particular fuzzy logic, allows the inference from the facts presented not as explicit numbers, but described by linguistic variables such as the "high level", "low temperature", "too much content", etc. Thanks to this, process of inference resembles human manner in actual conditions of decision-making processes. Knowledge of experts allows him to discover the functions describing the relationship between the classification of a set of objects and their characteristics, on the basis of which it is possible to create a decision-making rules for classifying new objects of unknown classification so far. This process can be automated. Experimental studies conducted on copper alloys provide large amounts of data. Processing of these data can be greatly accelerated by the classification trees algorithms which provides classes that can be used in fuzzy inference model. Fuzzy logic also provides the flexibility of allocating to classes on the basis of membership functions (which is similar to events in real-world conditions). Decision-making in foundry operations often requires reliance on knowledge incomplete and ambiguous, hence that the conclusions from the data and facts may be "to some extent" true, and the technologist has to determine what level of confidence is acceptable, although the degree of accuracy for specific criteria is defined by membership function, which takes values from interval <0,1>. This paper describes the methodology and the process of developing fuzzy logic-based models of decision making based on preprocessed data with classification trees, where the needs of the diverse characteristics of copper alloys processing are the scope. Algorithms for automatic classification of the materials research work of copper alloys are clearly the nature of the innovative and promising hope for practical applications in this area.

11

Application of regression trees in the analysis of electricity load

Gładysz B., Kuchta D.

Badania Operacyjne i Decyzje

|

2008

|

nr 4

19-28

EN

In the paper electricity load analysis was performed for a power region in Poland. Identifying the factors that influence the electricity demand and determining the nature of the influence is a crucial element of an effective energy management. In order to analyse the electricity load level the CART (Classification and Regression Tree) method has been used. The data for the analysis are hourly observations of the electricity load and weather throughout one year period. Two categories of factors were taken as predictor variables, on which the demand for the electricity load depends: variables describing weather and variables representing structure days in a year. An analysis of the errors of the presented models was carried out.

PL

W artykule zbadano wpływ warunków atmosferycznych na poziom obciążenia systemu elektroenergetycznego. Identyfikacja czynników warunkujących wielkość popytu na energię elektryczną jest podstawowym elementem systemu zarządzania energią elektryczną. W badaniach zastosowano metodę k-średnich oraz technikę drzew klasyfikacyjnych i regresyjnych. W pierwszym etapie badań metodą k-średnich wyróżniono jednorodne - z uwagi na obciążenie systemu elektroenergetycznego - grupy godzin w skali doby. Dla każdej z grup dla wybranej godziny (reprezentanta) zbudowano drzewo regresyjne, przyjmując jako czynniki dane meteorologiczne oraz typ dnia w skali roku. Przeprowadzona analiza pokazała, że czynnikami warunkującymi poziom obciążenia systemu energetycznego są: temperatura, punkt rosy, wilgotność oraz rodzaj opadów. Informacja o rodzaju oraz wartościach progowych tych czynników meteorologicznych może zostać wykorzystana w procesie prognozowania poziomu obciążenia systemu elektroenergetycznego i tym samym przyczynić się do poprawy efektywności procesów zarządzania energią. Przeprowadzono analizę błędów skonstruowanych drzew regresyjnych.

12

Modelowanie emisji NOx w zawiesinowym procesie wytwarzania miedzi przy wykorzystaniu metody drzew regresyjnych

Talar J., Jarosz P., Stanisławczyk A., Kusiak J.

Rudy i Metale Nieżelazne

|

2007

|

R. 52, nr 3

129-135

PL

Przedstawiono problem przewidywania emisji NOx w zawiesinowym procesie wytwarzania miedzi. Algorytm drzew regresyjnych CART został wykorzystany do przewidywania poziomu NOx w gazach. W modelowaniu tego zjawiska wykorzystano przemysłowe dane pomiarowe. Opracowany model na bazie drzew decyzyjnych pozwolił na identyfikację zmiennych niezależnych, które mają decydujące znaczenie dla przewidywania poziomu stężenia NOx w gazach. Wyniki modelowania uzyskane przez algorytm CART porównano z wynikami sztucznych sieci neuronowych oraz metod regresji liniowej i nieliniowej. Opracowane modele oparte o sztuczne sieci neuronowe oraz drzewo regresyjne mogą być zastosowane w optymalizacji i sterowaniu procesu wytwarzania miedzi pod kątem redukcji szkodliwej emisji NOx.

EN

The problem of prediction of NOx emission in the copper flash smelting process is presented. The CART algorithm was applied to prediction of the NOx content in exhausts. The industrial data were used to modelling of this phenomenon. The model developed on the base of the decision trees allows to identify the independent variables, which are significant for prediction of NOx content in gases. The results of CART algorithm were compared with the artificial neural networks and the linear and non-linear regression models. The elaborated models based on the artificial neural networks and regression tree method can be applied in optimisation and control of the copper production process for reduction of harmful emission of NOx.

13

Dipolar regression trees in survival analysis

Krętowska M.

Biocybernetics and Biomedical Engineering

|

2004

|

Vol. 24, no. 3

25-33

EN

In this paper a new method for induction of multivariate regression trees is presented. The technique is designed for the survival time prediction and based on given data. The proposed method aims at identification of subgroups of patients with homogenous survival experience i.e. homogenous response for a given treatment. The method allows using information from censored cases for which the exact failure time is unknown. An appropriate degree of generalization is obtained by using a pruning algorithm, which is based on rank correlation coefficient D.