Geri Dön

Patlatma kaynaklı yer sarsıntısının farklı regresyon modelleri ve yapay sinir ağı ile kestirimi

Blast induced ground vibration forecasting using different regression models and artificial neural network

PDF İndir

  1. Tez No: 553435
  2. Yazar: TAYLAN ÖZKAN
  3. Danışmanlar: DOÇ. DR. TÜRKER HÜDAVERDİ
  4. Tez Türü: Yüksek Lisans
  5. Konular: Maden Mühendisliği ve Madencilik, Mining Engineering and Mining
  6. Anahtar Kelimeler: Belirtilmemiş.
  7. Yıl: 2019
  8. Dil: Türkçe
  9. Üniversite: İstanbul Teknik Üniversitesi
  10. Enstitü: Fen Bilimleri Enstitüsü
  11. Ana Bilim Dalı: Maden Mühendisliği Ana Bilim Dalı
  12. Bilim Dalı: Belirtilmemiş.
  13. Sayfa Sayısı: 101

Özet

Patlatma, madencilik başta olmak üzere sert kayaçların parçalanmasında kullanılan ekonomik bir kazı yöntemidir. Patlatma ile yapılan bir kazı çalışmasında verimlilik büyük önem taşımaktadır. Patlatma verimliliği, istenilen tane boyut dağılımının elde edilmesiyle ve çevresel etkilerin olabildiğince düşük seviyede tutulmasıyla ölçülmektedir. Parçalanmanın istenilen düzeyde olmasının kayacın yükleme, nakliye ve kırma işlemlerinde olumlu etkisi olduğu gibi, bu aşamalarda kullanılan makinelerin bakım masraflarında da olumlu etkisi bulunmaktadır. Patlatma kaynaklı çevresel etkiler yer sarsıntısı, hava şoku, kaya fırlaması ve toz oluşumu olarak sıralanabilir. Çevresel etkiler, kontrolsüz patlatmalardaki aşırı patlatma enerjisinin bir sonucudur. Çevresel etkilerin minimum düzeyde tutulması kontrollü ve verimli bir patlatma ile mümkündür. Çevresel etkilerin en başında gelen yer sarsıntısının kontrol altına alınması özellikle yerleşim yerlerine yakın bölgelerdeki maden ocakları için büyük önem arz etmektedir. Aşırı yer sarsıntısı çevredeki yapılara zarar verdiği gibi bölge halkını da rahatsız etmektedir. Aynı zamanda ocak içerisinde istenmeyen geri çatlaklara sebep olması nedeniyle işletme içinde sorunlara sebep olmaktadır. İstenmeyen çatlaklar yer altı sularında yön değişimine de sebep olarak bölge ekolojisini olumsuz yönde etkileyebilmektedir. Yer sarsıntısının önceden tahmin edilmesi ile patlatma tasarımı güncellenebilir ve istenmeyen boyuttaki yer sarsıntılarının önüne geçilebilmektedir. Tez kapsamındaki saha çalışmaları Oyak Beton Çerkeşli Agrega İşletmesinde gerçekleştirilmiştir. İlk aşamada çalışma sahasının jeolojisi ve üretim faaliyetleri hakkında bilgi verilmiştir. Kullanılan patlayıcılar, ateşleme sistemleri ve özellikleri anlatılmıştır. İşletmede uygulanan genel patlatma tasarımı patlatma simülasyon programı ile modellenmiş ve zaman konturları, anlık şarj miktarı, parçalanacak kayaç hacmi ve patlayıcı enerji dağılımı incelenmiştir. Kullanılan titreşim ve hava şoku ölçüm cihazı hakkında bilgi verilmiştir. Titreşim tahmin modeli oluşturmak için sahadan toplamda 66 adet titreşim ölçüm verisi alınmıştır. Bu verilerden rastgele seçilmiş olan 44 veri ile titreşim tahmin modelleri kurulmuş, geriye kalan 22 veri bağımsız test verisi olarak kullanılmıştır. Oyak Beton Çerkeşli Agrega İşletmesinde patlatma kaynaklı yer sarsıntılarının önceden tahmini için, ölçekli mesafe konseptine dayanarak yer sarsıntısı tahmin denklemleri oluşturulmuştur. Ayrıca çok değişkenli regresyon analizi kullanılarak patlatma tasarım parametrelerini de içeren bir tahmin denklemi geliştirilmiştir. Son aşamada, çok değişkenli regresyon analizinin çıktıları esas alınarak, ileri beslemeli geri yayılımlı bir yapay sinir ağı modeli kurulmuştur. Geliştirilen modellerin tahmin kapasiteleri incelenmiş, detaylı hata analizleri yapılmıştır. Hata analizinde ortalama hata, ortalama yüzde hata, simetrik hata ve ölçekli hata kullanılmıştır. Modellerin yüksek titreşim tahminlerinde ve düşük titreşim tahminlerindeki kapasiteleri incelenmiştir. En iyi tahmin modelinin en düşük hata payı ile yapay sinir ağı ile oluşturulan model olduğuna karar verilmiştir.

Özet (Çeviri)

Blasting is an economical excavation method used for breaking of hard rocks, mainly in mining activities. When the main rock and side rock characteristics are suitable for underground mining activities, drilling and blasting applications is used in the progression of production and preparation galleries, sublevel cavings and creating many underground structures. In surface mining activities, blasting applications are widely used for excavation of the overburden and ore. The first step in mining activities is to decide the method of excavation. The first step in blasting method is to decide the design of blasting operation. Several factors should be taken into consideration in preparing the blasting design, such as desired fragment size, geology of the region, the structures in the nearby area, ground vibration caused by blasting and air shock waves. In surface mining; in front of the bench, there should be a free face for rock movement during detonation. Blast design parameters are calculated by empirical formulas and must be updated based on the expereince of blasting engineer. Basic blast design parameters can be sorted as burden (B), distance between holes (S), hole length (L), bench height (H), sub drilling (U), stemming height (T), hole diameter and specific charge (PF). Efficiency is very important for an excavation operation by blasting. Blasting efficiency is measured by achieving the desired fragment size distribution and keeping environmental effects as low as possible. The optimum rock fragmentation has a positive effect on the loading process, transportation operations and crushing processes of the rock as well as the maintenance costs of the machines used in these stages. Blast induced environmental effects can be sorted as ground vibration, air blast, fly rock and dust. Environmental effects are the result of excessive blast energy occuring during detonation. Minimization of environmental effects is possible by controlled blasting operation. The most important environmental effect of blasting is ground vibrations. Vibrations that move in the rock block with surface and body waves reach very long distances. Incorrect blast design, insufficient stemming height and unsuitable stemming material cause air shock waves. Ground vibration, which is the most important environmental effects is of great importance especially for mines located near to residential areas. Excessive ground vibration damages to the surrounding buildings as well as disturbing the people of the region. At the same time, it causes problems in the quarry due to the undesired back break in the bench. By predicting ground vibration, the blast design can be updated and undesirable ground vibration can be prevented. Field studies within the scope of the thesis were carried out at Oyak Beton Çerkeşli Aggregate Quarry. As a first stage, the geological formation and the production operations in the quarry are expressed. Explosives types, initiation systems and their specifications are described. The blast pattern applied in the quarry was modeled and examined by a blast simulation software. The specifications of the blasting seismograph used to measure ground vibration is explained. A total of 66 vibration measurement data were obtained from the field to create a vibration estimation model. The vibration prediction models were established by using 44 randomly selected data, and the remaining 22 data were kept as test data. All the developed models created by same blast database. The validation of models were performed under equal terms. The prediction models based on scale distance, multiple regression analysis and artificial neural networks were developed in order to predict blast induced ground vibration at Oyak Beton Çerkeşli Aggregate Quarry. It is very difficult to determine the behavior of ground vibrations due to the variation in the rock mass properties and geological structure of study area. Several researchers have developed different empirical models to estimate ground vibration by using statistical techniques. Most of the models rely on scaled distance concept. All the scaled distance equations were considered for the Çerkeşli Quarry. The scale distance based models were analyzed according to their correlation coefficients, coefficient of determinations. Additonaly, a vigorous validation was performed. In the multivariate regression analysis, design parameters burden, spacing, bench height, stemming, subdrilling, powder factor, charge weight per hole and vibration measurement distance were selected as independent variables. Peak particle velocity (PPV), which is the value to be estimated, was chosen as the dependent variable. Stepwise regression was used to determine the dominant parameters on the occurance of ground vibration. In stepwise regression analysis, regression was performed step by step. Firstly, the independent variable, which has the highest correlation coefficient with particle velocity, is included in the regression. Then the parameter with the highest partial correlation with the particle velocity from the remaining independent variables was included in the regression by checking the first independent variable. This process continues in steps and the coefficient of determination of the regression (R2) is increased with each added independent variable. The stepwise regression was completed when there was no increase in the coefficient of determination by addition of a new independent variable to the regression. It was determined that the effective blast design parameters on the occurrence of vibrations are measurement distance (D), charge weight per delay (W), Bench height (H) and Burden (B). In this study, 66 blast-induced vibration data were collected. Fourty four vibration data were used to create artifical neural nerwork (ANN) model. The 44 data is divided into 3 groups as training data (70%), test data (15%) and validation data (15%). Data were splitted into groups randomly. For each input set, an output set has been created. The parameters are D, W, H and B selected by stepwise regression as input data. Target data is the peak particle velocity (PPV). A network was created using the Matlab nntool function. Different numbers of neurons and layers were tried. Four neurons in the input layer, 10 neurons in the hidden layer and 1 neuron in the output layer gave the best estimation results. The feed-forward back-propagation network type has been used in the model training because of its success in the estimation of nonlinear equations. Trainlm has been chosen because of the fastest back-propagation algorithm. The Trainlm function updates the weight and bias values according to the Levenberg-Marquardt optimization. Learngdm has been selected as a learning function because it allows training of a large number of input data. Model performance was measured with the mean square error (MSE). Tansig function is selected as transfer function. Then the artificial neural network model is trained. Estimation results obtained from prediction models were compared using various error metrics. As a result, the best vibration prediction model was decided for the quarry. The error metrics used to evaluate ground vibration estimations are grouped as absolute error, percentage error, symmetric error and scale error. Absolute errors and Percent errors are widely used in the mining literature. All the error criteria have different advantages and disadvantages. Therefore, the error criteria must be considered together. In this study, Variance Accounted For, determination coefficient and standard error of estimation were also considered to compare the models. In addition, the number of tests blasts, which was predicted with an error less than 2 mm/s, was examined. It is easy to use the absolute error because it gives directly error size. The percentage error is easy to interpret because it refers to the error as a percentage. It is found that the symmetric error and scale error criteria can be used efficiently in estimating ground vibrations. The scaled error was decreased to 0.26 for ANN, indicating the success of the model. Correlation and R2 values indicate the relationship between measured and estimated values. These values should not be used alone without considering error criteria. The smallest error values were obtained for artificial neural network model. ANN provided best error values for the 14 out of 15 error metrics. The coefficient of determination (R2) between measured vibration values and ANN predictions is higher than 0.90. Artificial neural network was found to be the best predictor model for estimation of ground vibration in Çerkeşli Quarry.

Benzer Tezler

  1. Tez No

    338192

    Patlatma sahalarında farklı jeofizik yöntemler kullanılarak çevre kaya kütle özelliklerinin tespiti ve yer sarsıntısı analizlerinde kullanılabilirliği

    Determination of rock masses properties using different geophysical methods and their usability for ground vibration analysis at blasting applications

    ARZU KOÇASLAN

    Doktora

    Türkçe

    Türkçe

    2013

    Jeofizik MühendisliğiCumhuriyet Üniversitesi

    Maden Mühendisliği Ana Bilim Dalı

    DOÇ. DR. KAZIM GÖRGÜLÜ

    DOÇ. DR. AYDIN BÜYÜKSARAÇ

  2. Tez No

    791299

    Patlatma kaynaklı yer sarsıntısı tahmininde uyarlamalı bulanık çıkarım sistemi (ANFIS), destek vektör makineleri (SVM) ve gauss süreç regresyonu (GPR) tekniklerinin kullanımı

    Application of adaptive-network based fuzzy inference system (ANFIS), support vector machines (SVM) and gaussian process regression (GPR) techniques for prediction of blast-induced ground vibrations

    YAŞAR AĞAN

    Yüksek Lisans

    Türkçe

    Türkçe

    2023

    Maden Mühendisliği ve Madencilikİstanbul Teknik Üniversitesi

    Maden Mühendisliği Ana Bilim Dalı

    PROF. DR. TÜRKER HÜDAVERDİ

  3. Tez No

    434842

    Patlatma kaynaklı yer sarsıntısının sonlu eleman yöntemiyle modellenmesi

    The modeling of blast induced ground vibration with the finite element method

    MEHMET ÖZEL

    Yüksek Lisans

    Türkçe

    Türkçe

    2016

    Maden Mühendisliği ve MadencilikAtatürk Üniversitesi

    İnşaat Mühendisliği Ana Bilim Dalı

    DOÇ. DR. İLKER KAZAZ

  4. Tez No

    184040

    Patlatma kaynaklı yer sarsıntılarının yönsel değişiminin araştırılması

    The investigation of directional changes of the blast-induced ground vibrations

    HAKAN AK

    Doktora

    Türkçe

    Türkçe

    2006

    Maden Mühendisliği ve MadencilikEskişehir Osmangazi Üniversitesi

    Maden Mühendisliği Ana Bilim Dalı

    PROF. DR. ADNAN KONUK

  5. Tez No

    178797

    Konya Çimento Fabrikası kireç taşı ocağındaki patlatma kaynaklı yer sarsıntılarının değerlendirilmesi

    Assessment of the blast induced vibrations in limestone quarry of Konya Cement Factory

    HAMDİ LEVENT YÜCEL

    Yüksek Lisans

    Türkçe

    Türkçe

    2008

    Maden Mühendisliği ve MadencilikSelçuk Üniversitesi

    Maden Mühendisliği Ana Bilim Dalı

    YRD. DOÇ. DR. MURAT ÜNAL

Geri Dön