Details var a2a_config = a2a_config || {}; a2a_config.onclick = 1; a2a_config.locale = "pl"; a2a_config.color_main = "D7E5ED"; a2a_config.color_border = "9D9D9D"; a2a_config.color_link_text = "ffffff"; a2a_config.color_link_text_hover = "ffffff"; a2a_config.color_bg = "d9d9d9"; a2a_config.icon_color = "#4e4e4e"; a2a_config.exclude_services = ["dihitt", "delicious", "whatsapp", "dzone", "sms", "stumpedia", "viber", "yahoo_messenger", "google_plus"]; Details PDF BIBTEX RIS Title Application of Multidimensional Data Visualization by Means of Self-Organizing Kohonen Maps to Evaluate Classification Possibilities of Various Coal Types / Zastosowanie Wizualizacji Wielowymiarowych Danych Za Pomocą Sieci Kohonena Do Oceny Możliwości Klasyfikacji Różnych Typów Węgla Journal title Archives of Mining Sciences Yearbook 2015 Numer No 1 Authors Jamróz, Dariusz ; Niedoba, Tomasz Divisions of PAS Nauki o Ziemi Publisher Committee of Mining PAS Date 2015[2015.01.01 AD - 2015.12.31 AD] Identifier ISSN 0860-7001 DOI 10.1515/amsc-2015-0003

Visualization of multidimensional data is a new way of statistical analysis of so-called statistical graphical methods. These methods allow to classify some analyzed objects, including their various features. Facing grained materials problems, like coal or ores many characteristics have an influence on the quality of product. In case of coal, many features must be taken into consideration to determine quality of the material. Apart from most obvious characteristics like particle size, particle density or ash contents there are many others which cause significant differences between considered types of material. In the paper the application of Multidimensional Scaling Method is presented which is one of the multidimensional data visualization techniques. To this purpose, sampling of three types of coal was performed, which were 31, 34.2 and 35 (according to Polish classification of coal types). First, the material was screened on sieves and then divided into density fractions. Next step was to analyze chemically the obtained particle and size fractions of researched coal. Then, the Multidimensional Scaling Method was applied to visualize the investigated set of data. It was proved that the applied methodology allows to identify certain coal types efficiently and can be used as a qualitative criterion for grained materials. However, it was impossible to achieve such identification comparing all three types of coal together. The Multidimensional Scaling Method is new technique of data analysis concerning widely understood mineral processing.

Coal as energetic raw material features by many parameters determining its quality. In classification of coal types there are many of them with typical division of energetic, semi-coking and coking coal. The data concerning coal are usually treated as independent values while this kind of approach is not always right. Authors proposed new solutions in this aspect and performed the multidimensional analysis of three selected types of coal featuring by various properties which originated from three various hard coal mines located in Upper Silesia Region. The object of the research was so-called raw coal which was not processed before. For each type of coal the detailed statistical analysis of seven chosen properties of coal was performed. To perform adequate and complete statistical analysis it is necessary to analyze the chosen properties of coal together in multidimensional way. It was decided to apply new and modern visualizing methods of multidimensional data which were observational tunnels method and parallel coordinates method. The applied methods allowed to obtain visualization of seven-dimensional data describing coal. By means of these visualizations it was possible to observe the significant division of the features space between researched types of coal. These methods allowed to look at the investigated data from various perspectives and make possible to determine significant differences between researched materials. For the investigated coals such differences were determined clearly what proved that by means of these methods it is possible to successfully identify type of coal as well to analyze in details its individual properties and identify, for example, particle size fraction etc. The obtained results are innovative and are the basis for more detailed researches taking into consideration also other coal properties, including its structure and texture. This methodology can be also applied successfully for other types of raw materials, like ores.

The paper presents a way of combining neural networks with evolutionary algorithms in order to find optimal parameters of the copper flotation enrichment process. The neural network was used in order to build a model describing the flotation process. The network learning was carried out with the use of samples from previous empirical measurements of the actual process. The model created in this way made it possible to find optimal parameters not only from among the measurement spaces, but also those that go beyond the measurements. Then, evolutionary algorithms were used in order to find optimal flotation parameters. The learned neural network previously described was used to calculate the criterion in the evolutionary algorithm.