زمین آمار

Calculation of variograms and spatial continuity is one of the first and most important processes in geostatistical modeling, which is a long and experience-oriented process. Due to the complexities of calculating experimental variograms, interpretation and fitting the appropriate model are always the main challenges in this field. This article presents an intelligent variogram modeling method using deep learning that can increase the speed of variogram modeling and also prevent common errors in manual variogram model fitting. In this method, two convolutional neural networks are used. The first CNN network converts the initial data into a 2D simulated map based on various variogram models. For this purpose, it is necessary to train the first network with initial data and their corresponding simulations. The output of this model is entered into the second convolutional neural network as input, and the variogram parameters (including range, azimuth, ratio, and nugget effect) are predicted. In this article, the proposed algorithm is implemented on synthetic 2D data and the parameters of the CNN models are optimized. The accuracy of the proposed model was 97 %, and then the proposed algorithm was used for variogram modeling of Nouchon area geochemical data, which included the elements Cu, Zn, and Pb. the accuracy of the obtained model compared to manual fitting was 90%.

The geostatistical method is one of the most advanced techniques to assess and survey groundwater quality.  Using unsuitable water results in some problems such as soil salinity, reduction of soil permeability, decrease in water absorption by plant roots, diminution of crop productivity, or even menace agricultural production. Accordingly, in this study, the groundwater quality in the study's area for agricultural uses as well as industrial applications by using common diagrams and components related to the classification of water quality was investigated. ISATIS and Surfur software were used to illustrate changes in quality characteristics in the study's area graphically and Consecutive Gaussian simulations to analyze spatial connection among variables and estimate some of the quality indexes such as PH, RSC, Na%, SAR, and Ec. After data normalization, the variogram was calculated, and a sufficient model was selected to fit the experimental variogram based on the lowest SSR Error. Afterward, Consecutive Gaussian simulation in 100 samples in the study's area was examined using the fitted model based on the experimental model, and the maps resulting through simulation in different thresholds were provided. The results showed that the majority of samples have fairly suitable quality in terms of SAR, and in general, they are appropriate for agricultural uses. Percentage risk of Sodium in agricultural water in the study's area was also classified into the appropriate class. Except for four samples whose spatial connection of the forgoing cases have been interpolated and the interpolated probability maps have been provided, the water quality for industrial uses is susceptible to sedimentation in all statutes.

High gas volume is the main reason of explosion, rock burst and fatal coal mining disasters. In a coal mining project, it is necessary to model the gas volume at all parts of the target seam. Amoung coal resources of Iran, Parvadeh I has high gas bearing seams with an average of 14 m3/t of gas. C1 coal seam is the main mining target in Parvadeh I and it was explored by 134 core drilling boreholes. Gas study has been carried out for 35 boreholes, in only central and eastern parts of the deposit. Gas volume is not measured at the deep western parts. In this study, fractured zones are identified and their gas data removed from the gas modelling procedure. It is found that gas volume and seam depth are highly correlated and also, two separated populations are evident in the gas vol. – depth cross plot. The first population is related to oxidized shallow, and the second one is interpreted as deep methane bearing zones of the seam. Using the regression kriging method with the seam depth as the auxiliary variable, the gas volume is estimated for the whole mining blocks of the C1 seam. The validity of the estimations is evaluated as 94% by the Jackknife method. In order to avoid the smoothing effect of kriging, the probability of critical gas zones (> 20 m3/t) is modelled by the conditional indicator cosimulation. Results demonstrated that 1.43 million m2 of the C1 seam with the probability of 75%, and 0.38 million m2 with the probability of 100% fell into the critical category. These critical gas zones are located in central, southern and western Parvadeh I. These zones are considered to be the main targets for the future gas drainage studies.

Geological modeling of heterogeneous facies plays an important role in the detection of stratigraphic uncertainty. In this research, three methods, Indicator kriging (IK), Sequential Indicator Simulation (SIS), and Generalized Coupled Markov Chain (GCMC) were applied to predict geological categories at unknown locations. Then the results of all three methods were compared.

There are various methods for estimating and determining the spatial variation of categorical variables using geological data and exploratory wells. One of the best of these methods is geostatistical methods. As new Geostatistical methods, the GCMC algorithm, one of the Markov chain models, has been used in the earth sciences to simulate categorical variables of sedimentary deposits. This method is based on the calculation of transition probability matrixes with respect to the direction and spatial variations between classes. Due to the realistic results and easy implementation, the GCMC method is a suitable tool for the initial predicting and modeling of categorical variables in sedimentary environments.

Methodology and Approaches:

In this study, one of the drilling profiles in block C of the Bostanabad Areshtenab limestone deposit was selected for modeling. At this point, three carbonate units can be distinguished from the 5 exploratory boreholes dataset. To build the prediction models, after transforming the coordinates into a stratigraphic coordinates system (unfolding the strata by vertical transformation), the vertical and horizontal variability and continuity structure of the three existing classes were modeled with indicator variograms and transition probabilities. Then the mentioned geostatistical prediction techniques were applied to generate the spatial variability models.

In general, this study suggested the application of three geostatistical prediction methods for constructing realistic subsurface models of the categorical variables. According to the results, the IK result represented the general occurrence trend better. However, the spatial variability structure could not be reflected sufficiently and clearly. Although, in the SIS results fine and subtle variations were reflected, the produced patterns were more scattered. As the result of this study, the GCMC method can reproduce the global statistics, spatial structural functions (transiograms), and more realistic subsurface models, especially with sparse data in sedimentary systems.

This study uses multiple indicator kriging estimator based on assayed data acquired from 48 exploration boreholes in Tabas coal mine in order to predict gas content of coal seams. Results of the estimated block models showed that approximately 12% of the total area has the gas content of less than 5 (m3/ton) (Low risk), 11% has 5-10 (m3/ton) (Medium Risk) and 15% has 10- 15 (m3/ton) (High Risk), while about 62% of total area has the gas content of more than 15(m3/ton) which is of a high risk. Therefore, according to the experiences from mined panels, in the zones with lower to median gasification risk, it is possible to keep the mining operation running with management and optimization of the ventilation system. However, in the zones with higher gasification rate, demethanization process is mandatory.

Sehezar area is located in southern Tonokabon in Mazandaran province, north of Iran, near the Tarom – Hashdtjin belt. The existence of granitoid masses in the region can be important in terms of mineralization potential. The geological complex exposed in the area mainly consists of Paleozoic to Cenozoic lithologic units. In this study, a number of samples were used for microscopic and petrographic studies, as well as some samples of granites in the area for the classification of rocks. According to the studies performed on polished sections, minerals of pyrite, chalcopyrite and magnetite were observed. The textures found in thin sections are granular, Hyaloporphyritic, and Vitrophyric. The presence of porphyry textures indicates the simultaneous intrusions with volcanics and clastics originating from them. The major minerals in the region include quartz, K-feldspar, and plagioclase, and in some cases, hornblende, biotite, and pyroxene appear as the major mineral. The minor minerals found in the samples include sphene, pyroxene, iron oxides, apatite, and opaque minerals. Petrographic studies showed that the intrusive rocks of the region are of granite, granodiorite, syenite to quartz-syenite and quartz-monzonite. The magma in this area is in the domain of alkaline to calc-alkaline series and is a magnesian series and in the domain of metaluminous to peraluminous. The granites of the studied area are of peraluminous type and have a tectonic setting of the active continental margin and are I-type. Variogram studies showed that the spherical model is the best fitted model, and the spatial correlation range for the three elements of Au, Cu and Fe are approximately 350 m. Evaluating the results of geostatistics by calculating the root mean square error (RMSE) and calculating the mean absolute error (MAE) indicates the acceptable accuracy of variogram model. According to the Meinert diagrams and spatial correlation of the elements, it is concluded that these masses can be related to the deposits of iron-gold-copper.

Determination of ore body domain is the main step in the beginning of a mining project. The mine designing, mining and processing plant planning would be implemented based on this domain. Traditional methods characterize this domain graphically and would not concentrate error and uncertainty, but Geostatistics as a powerful tool is able to determine this domain using spatial statistics. Ore domain is a district variable that can be modeled by different simulation methods, such as Truncated Gaussian simulation (TGS). TGS simulates the domain according to the simulation of one Gaussian function and rock type rule. Furthermore, simple relations can be applied for determination of ore domain, for example, Distance Function (DF). DF defines the ore boundary by the Euclidian distance between each point and the nearest dissimilar point and the distance calibration.
Statistical and geostatistical validations are applied to find the best realization of TGS. The DF and TGS results are compared by assessing confusion matrices, overall error and precision of the modeled domains. Moreover, different plans, and section are qualitatively compared. Validation and comparisons indicate that TGS is a more efficient way to model the ore domain. Therefore, Gol Gohar mine was studied and after the implementation of both methods showed that TGS compared to DF results in more precise domain simulation. TGS algorithm is able to reproduce the small scale anisotropies versus, better results at large scale anisotropies using DF algorithm.
Key words: Geostatistics, Ore Domain, Truncated Gaussian simulation, Distance Function,

Parvadeh coalfield which located in 85 km south-west of Tabas, South Khorasan province, is the largest coal reserve of Iran. This mine is producing 1.5 million tons of coking coal annually. Largest sub-region of Parvadeh coalfield is Parvadeh IV. 168 drill holes were drilled in Parvadeh IV to obtain core samples. Ash content of coal is the most important qualitative parameter in explorations and concentrate production. This study concentrates on estimation and modeling of ash content in B2 seam using geostatistical approach. An exploration database was made based on data gathered from 154 drill holes which crossed B2 seam. Basic statistical studies were carried out. Directional semi-variograms used to analyze and interpret the spatial heterogeneity of ash content. Due to trend existence in spatial raw data of ash content, universal kriging method was applied in order to removing the trend. As the result, ash content modeled and estimated for every 100×100m exploitation block using block kriging. Estimation results presents that ash content varies from 13.0% to 41.6% with an average of 28.0%. This estimation which has minimum errors, establishes a manner for exploitation and blending plans. Besides, the uncertainty map was generated and locations of proposed additional drillings were specified.

This work deals with the geostatistical simulation of mineral grades whose distribution exhibits spatial trends within the ore deposit. Such trends can be reproduced via the definition of conditioning data. These conditioning data can be available from hard data or some soft data، which are produced from the result of cokriging. The conditioning process forces its realizations to conform these data and therefore، their local properties، in particular the local mean (trend)، could be reproduced. In fact، soft data can compensate the lack of hard data in sampled areas with insufficient data and force the realizations to reproduce the trend in the study area. The proposed method was applied on obtained data، located at 1850 m، from exploration boreholes of Sungun mine. The study area was divided into two rock type domains: Sungun Porphyry (SP) and Dyke (DK) and it was shown that the incorporation of soft and hard conditioning data improves the modeling of the uncertainty in the actual copper grades and decreases the conditional variance of simulation significantly.

The beneficiation plants are designed and optimized for ore dressing with a permissible range of grade variation. Any fluctuation out of this range will considerably affect the concentrate quality and beneficiation recovery. This paper attempts to analyze the variability of feed grade of Esfordi phosphate plant using geostatistical method and introduces a procedure to overcome this problem. The first step was the collection of representative samples which was carried out during 960 hours on the feed stream. Based on the results obtained from this study, a periodic variation has been seen on the feed grade which might be due to ore exploitation from different mine faces and lack of any suitable blending. The high values of relative nugget effects of variograms (especially for phosphates (0.12 and 0.66 for iron and phosphate, respectively) confirm this undesirable fluctuation. In the next step, the grade homogenization of the feed was simulated using moving average technique at four time intervals (16, 24, 32 and 40 hours). By sensitivity analysis, the 32h time interval was chosen as optimized blending interval. Considering a feed rate of 140 t/h, it is necessary to blend about 4500 tons of the ore before feeding to the plant.

The investigation and estimation of surface subsidence and ground settlement is an important aspect in the design of shallow tunnels in urban areas. Proper exploration could lead to correct measures and avoiding any damage to structures over and underground.
In this study, four sections of the east-western part of Tehran’s Metro Line 7 with the most potential for surface subsidence were chosen and studied. This part of the metro project is positioned in the young alluvial deposits of Tehran and consists mainly of fine grained soil. Ground settlement requires three dimensional design. Hence, FLAC 3D was used to investigate the surface subsidence of the ground along the tunnel route in sections mentioned above. One of the most important advantages of three dimensional design is the possibility of stepwise excavation, and to explore the effect of each excavation step on the surrounding ground. Hence, the stepwise excavation was carried out and continued in 1.5 m advancing stages, until the advancement stopped affecting the ground settlement at the point over the tunnel portal considered in the model.
Even though, analytical and empirical methods are unable to evaluate the effect of ground water and the effect of overburden due to the presence of structures and traffic at the surface, because of not installing the monitoring system, analytical methods suggested by Sagaseta, Loganathan and Poulos, Verruijt and Booker, and Bobbet, and also Peck’s empirical method were used to investigate the ground settlement in this project.
After using the various methods to study the four chosen sections of the metro line, and analyzing the settlement curves in these sections, it was concluded that the ground settlements is permissible (except the second section).

One of the estimate methods in mining is geostatistical methods. The advantage of the statistical method is good mathematical background and also in results acquired through their practical application. But through the geostatistal methods, choosing the most useful method is very important. One of most important thing, that operators usually don’t use it, is the ability of using liner kriging. Therefore the concentrating on non-liner kriging has been seen. One of the powerful methods in non-liner kriging is Co-kriging. According to complicated mathematical theory and untouchable usefully software, this method hasn’t been used. This method based on hermitian polynomial and isofactorial model that use for estimating the black points with the primary data histogram then the value of hermitian polynomial are estimated by this histogram. Finally with liner composited of this value in hermitian polynomial and estimated value in black points the value of parameter is determined. In this paper, using by computer programming in Co-kriging the grade of copper distribution was determined in Sungon mine and produced data was compared with producing data by kriging. According to this study, the soft western-north and eastern-south districts have lower grades and the middle district of mine has higher grade than other parts. Also the normalizing and data inversing by hermitian polynomial has high capability. Using this method improves some problems in kriging such as smoothing and compability in non-liner estimated function

Grade deviation between the feed and the consumption of concentration plant was a great debate from the beginning of the Sarcheshmeh copper mine operations until now. However، the ore dressing plant is designed in such a way to work optimally base on specific input feed. Therefore، the error identification of different sampling procedure and grade deviation investigation between the feed of the mine and the consumption of the concentrating plant is essential. With determination of limit allowance of adequate parameters in deviation variance and make use of them in daily product schedual، it is possible to reduce the grade deviation. In this research with utilizing the statistical and geostatistical implement an evaluation method of accurate grade deviation of concentrating plant feed is presented so that with the help of its result، possibility these deviations with the predicted deviations of feed of mine gets provided in terms of shift or daily scheduals. To apply this research in a particular day، 50000 ton feed compared with grade 0. 89± 0. 05 that this grade lies within the limitation of predicted grade of 0. 93±0. 06 and limitation of operation grade 0. 89± 0. 05 (at 95% confidence limit). Finally suggested that in order to reduce the grade deviation and to optimize the recovery of the concentrating plant of Sarcheshmeh copper ore، no sampling should be done from the concentrating plant feed but instead the operation grade of mine should be counted for the concentrating metallurgical calculations، provided from the 100 continuous days، 95 grade cases of concentrating plant feed should lie in the operation grade of confidence limit of the mine.

The proper characterization of reservoir heterogeneities plays an important role in reservoir evaluation and forcasting of its performance. However, the heterogeneity of a reservoir and its property distributions in the large scale of reservoir volume could not be properly determined and estimated with known data from a few wells. In such a case, one of the newest and roboust methods in reservoir characterization is the use of fractal methods whenever the geostatistics fails. The objective of this paper is to introduce and discuss a new method that is the combination of fractal geometry (small scale) and geostatistics (large scale) to characterize the reservoir property distributions in a gian reservoir with a few known well data. The developed method was applied to a real reservoir in the south west of Iran with three exploration well data. In this method, first of all, the reservoir zonation and formation units including carbonate, sandstone and shale must be identified from the petrophysical data by using multivariate statistics analysis. So that, the porosity, permeability, and water saturation for the known wells could be estimated from the core and log data. Then, spatial correlation using variogram analysis and quantify heterogeneities using R/S analysis, can be determined for each zone in the wells. Eventually, merging kriging method and fGn noise of fractal will lead to develope our new method to estimates and predict the reservoir property distributions between wells and throughout the reservoir more accurate and very precise than the previouse methods. The produced error is in acceptable and accurate range that was less than 4.93 percent for porosity for instance. The output of the model may be used as an input for geomodeling and simulation softwares. Also, heterogeneity distributions of reservoir characteristics are well visualized by fractal distributions.