Publication: Kömürün Kısa Analiz Sonuçlarına Farklı Matematiksel Yöntemler Uygulayarak Elementel Bileşiminin Tahmini
Abstract
Bu tez çalışmasında, kömürün elementel analizinin deneysel olarak belirlenmesi uzun zaman gerektiren ve pahalı bir iş olduğundan dolayı, kömürün elementel bileşiminin tahmini üzerine bir çalışma gerçekleştirilmiştir. Geçmiş çalışmalarda elementel bileşim ile kısa analiz sonuçları arasındaki ilişkiler incelenmiş olup, karbon, hidrojen ve oksijen bileşimlerinin sırasıyla sabit karbon, uçucu madde ve külün nispi içerikleriyle orantılı olduğu görülmüştür. Kısa analiz sonuçlarını kullanılarak kömür ya da biyokütlenin elementel bileşimini tahmin etmek için çeşitli korelasyonlar sunulmuştur. Fakat sunulan çalışmalarda numune sayısının az ya da korelasyon değerlerinin doğru tahmin için yeterli olmadığı görülmüşür. Bu çalışmada, literatürde daha önce sunulmuş farklı çalışmalardan yararlanılarak büyük bir veri seti oluşturulmuştur. Farklı kalitede kömürlerden oluşturulan bu veri setine matematiksel modeller ve kısa analiz sonuçları uygulanarak kömürün elementel bileşiminin tahmini üzerine bir çalışma gerçekleştirilmiştir. Tahmin için çoklu lineer regresyon modeli ve yapay sinir ağlarından yararlanılmıştır. Elde edilen sonuçlar yaygın istatistiksel araçlar ile test edilmiştir. Çoklu lineer regresyon modeli ile regresyon katsayıları karbon, hidrojen ve oksijen için sırasıyla 0.99025, 0.64431 ve 0.98526 olarak elde edilmiştir. Yapay sinir ağları ile regresyon katsayıları karbon, hidrojen ve oksijen için sırasıyla 0.99177, 0.80452 ve 0.99113 olarak elde edilmiştir. Çalışma sonucu elde edilen modellerin, kısa analiz sonuçlarını kullanarak kömürün karbon, hidrojen ve oksijen yüzdesini yüksek doğruluk ile tahmin etme kabiliyetine sahip olduğu görülmüştür. Bu modeller zahmetli ve masraflı olan elementel analiz işleminin yerini alma potansiyeline sahiptir.
In this thesis study, because the experimental determination of the elementel analysis of coal is a long time and expensive process, a study on the estimation of the elementel composition of coal has been carried out. In the past studies, the relationships between the elementel composition and the proximate analysis results were examined and it was found that the carbon, hydrogen and oxygen compounds were proportional to the relative contents of the fixed carbon, volatile matter and ash, respectively. Various correlations are presented to estimate the elementel composition of coal or biomass using proximate analysis results. However, it was observed that the number of samples or the correlation values were not sufficient for accurate estimation. In this study, a large data set was created by using different studies previously presented in the literature. A mathematical model and proximate analysis results were applied to this data set composed of different quality coals, and a study was performed to estimate the elementel composition of coal. Multiple linear regression models and artificial neural networks were used for estimation. The results obtained were tested with common statistical tools. The regression coefficients were obtained as 0.99025, 0.64431 and 0.98526 for carbon, hydrogen and oxygen respectively with multiple linear regression model. The regression coefficients with artificial neural networks were obtained as 0.99177, 0.80452 and 0.99113 for carbon, hydrogen and oxygen, respectively. It was found that the models obtained as a result of the study were able to estimate the carbon, hydrogen and oxygen percentage of coal with high accuracy by using short analysis results. These models have the potential to replace the laborious and costly elementel analysis.
In this thesis study, because the experimental determination of the elementel analysis of coal is a long time and expensive process, a study on the estimation of the elementel composition of coal has been carried out. In the past studies, the relationships between the elementel composition and the proximate analysis results were examined and it was found that the carbon, hydrogen and oxygen compounds were proportional to the relative contents of the fixed carbon, volatile matter and ash, respectively. Various correlations are presented to estimate the elementel composition of coal or biomass using proximate analysis results. However, it was observed that the number of samples or the correlation values were not sufficient for accurate estimation. In this study, a large data set was created by using different studies previously presented in the literature. A mathematical model and proximate analysis results were applied to this data set composed of different quality coals, and a study was performed to estimate the elementel composition of coal. Multiple linear regression models and artificial neural networks were used for estimation. The results obtained were tested with common statistical tools. The regression coefficients were obtained as 0.99025, 0.64431 and 0.98526 for carbon, hydrogen and oxygen respectively with multiple linear regression model. The regression coefficients with artificial neural networks were obtained as 0.99177, 0.80452 and 0.99113 for carbon, hydrogen and oxygen, respectively. It was found that the models obtained as a result of the study were able to estimate the carbon, hydrogen and oxygen percentage of coal with high accuracy by using short analysis results. These models have the potential to replace the laborious and costly elementel analysis.
Description
Keywords
Citation
WoS Q
Scopus Q
Source
Volume
Issue
Start Page
End Page
67
