Search published articles


Showing 7 results for Regression


Volume 3, Issue 1 (11-2009)
Abstract

(Paper pages 513-522) Estimation of engineering properties of rocks and flow rate is an important issue in rock engineering. Properties of discontinuities have considerable effect on rock mass inflow, because they are the main pass of water flow in fracture rock masses. Despite the bulky research about water flow in rock mass, there is no clear evidence as to relationships between all of these parameters and water inflow in rock masses. Neural network systems have a great advantage in dealing with complicated problems such as forecasting, classification and pattern recognition. In this paper, artificial neural network techniques were used in order to forecast Lugeon amount and Hydraulic conductivity behavior of Granodioritic rock mass of Shoor-Jiroft dam site from some characterization of discontinuities such as Rock quality designation, Fracture frequency, Aperture, Weighted joint density, Fracture zone and depth. Relationships between these factors were analyzed with Simple Linear Regression, Multivariate Regression and Stepwise Regress-ion. A Multilayer Perceptron Neural Network (MLPNN) with back propaga-tion procedure was developed for training the network. A Dataset containing 304 values of water pressure test in Granodioritic rock mass of Shoor-Jiroft Dam project was used to train and test the network with the Levenberg-Marquardt training algorithm. The results indicated that neural network forecast hydraulic conductivity considerably better than regression methods do.
A Ghorbani, F Kalantari, M Zohori,
Volume 7, Issue 2 (3-2014)
Abstract

Determining the precise shear strength parameters of the fine grained soils is always a difficult task. In order to conduct the shear strength tests and determine the mechanical parameters of the soil, achieving an untreated high quality sample is a problem with a high degree of importance. Therefore, during the recent decades many researchers have attempted to provide relations between strength parameters and soil physical characteristics in a specific structure and so to provide the possibility of estimating the strength parameters based on these characteristics. The aim of this research is to estimate the shear strength parameters of a wide region of fine grained alluvial soil located in southern Tehran, Iran. In this regard, the geotechnical data including physical and shear strength parameters of 294 boreholes were firstly collected from the site. Then, the obtained data were statistically and independently analyzed. Based on the results of analyzes, the soil geotechnical parameters were presented for various depths with an acceptable level of reliability. Moreover, they were considered as a basis for providing a nonlinear regression model to estimate the soil shear strength parameters and based on the index physical characteristics of the fine grained soil (water content and plasticity index). The developed model is capable to predict the soil drained shear strength parameters and also other similar soil properties with a very good accuracy
Mohammad Moghadas, Ali Raeesi Estabragh, Amin Soltani,
Volume 13, Issue 1 (8-2019)
Abstract

Introduction
Improving the mechanical behavior of clay soil by stabilization agents is a mean of fulfilling geotechnical design criteria. The method of stabilization can be divided into chemical, mechanical, or a combination of both methods. Chemical stabilization is performed by adding chemical agents such as cement, lime or fly ash to the soil (Bahar et al., 2004). Soil reinforcement is one of the mechanical methods that is used for improving the behavior of soils (Tang et al., 2007). Reinforcement of soil achieved by either inclusion of strips, bars, grids and etc. within a soil mass in a preferred direction or mixing discrete fibers randomly with a soil mass.
Mixing of cement with soil is made a production that is called soil-cement and results in chemical reaction between soil, cement, and water. The compressive strength of soil-cement is increased by increasing the cement content and this leads to brittle behavior or sudden failure. On the other hand, by increasing the cement to soil ratio for cohesive soils, shrinkage micro-cracks may develop in the soil as a result of the loss of water content during drying or hydration of cement. Therefore, if the tensile strength of these materials is not sufficient cracks will develop under loading and damage will be resulted (Khattak and Alrashidi, 2006). Consoli et al. (2003) and Tang et al. (2007) indicated that adding the fiber to soil can prevent from occurrence of these cracks and increases the tensile strength of the soil.
The focus of this paper is on the statistical analysis of the results and development of regression models. Regression relationships are developed based on the experimental results that were presented by Estabragh et al. (2017). These relationships relate the compressive and tensile strengths of the soil to percent of used fiber, cement and curing time.
Material and methods of testing
Unconfined compression and tensile strength tests were carried on unreinforced and reinforced soil, soil cement according to ASTM standards. Samples of soil-cement were made by mixing a clay soil and two different weight percent of cement (8 and 10%). Reinforced soil samples were also prepared by mixing 0.5 and 1 weight percent of Polypropylene fibers with 10, 15, 20 and 25 mm lengths. The dry unit weight and water content of prepared samples were the same as optimum water content and maximum dry unit weight that were resulted from standard compaction test. The compressive and tensile strength tests were conducted on the samples by considering the curing time according to ASTM standards until the failure of the sample is achieved.
Results and discussion
The experimental tests showed that reinforcement of the soil and soil cement increase the peak compressive and tensile strength. The peak compressive strength of reinforced soil is increased by increasing the fiber content at a constant length of the fiber. It can be said that by increasing the percent of fiber, the number of fibers in the sample is increased and contact between soil particle and fibers is increased which result in increase in the strength (Maher 1994). However, by increasing the length of the constant fiber inclusion there will be no significant increase in strength because the number of shorter fiber is more than longer fiber in a specific sample (Ahmad et al., 2010). Inclusion of fibers can greatly increase the tensile strength of clay soil. In addition to reinforcement of soil cement showed the same trend. When fiber is added to soil cement, the surface of fiber adheres to the hydration products of cement and some clay particle. Therefore, this combination increases the efficiency of load transfer from the composition to the fibers which increase the peak strength (Tang et al., 2007). In addition, the tensile strength shows the same trend.
Based on the experimental data on the behavior of a randomly reinforced clay soil and soil cement multiple regression models (linear and non-linear) were developed for calculating the peak compressive and tensile strength (dependent variables) based on the value of the coefficient of determination (R2). The proposed regression models were functions of independent variables including weight percent of fiber, length of fiber (length/diameter of fiber), weight percent of cement, and curing time. Finally, the comparison is made between the predicted results from proposed models and experimental results. In order to investigate the model accuracy, the Root Mean Square Error (RMSE) and Normalized Root Mean Square Error (NRMSE) are used.
 The Multiple Linear Regression models (MLR) was very suitable for the study of the effect of independent variables on the quantitative analytic dependent variable. The NRSME for peak compressive and tensile strength is was 3.59% and 5.11% respectively for these models. Also, the Multiple Nonlinear Regression models (MNLR) had a much lower error than the linear model because of the quadratic equation, the equation will be able to predict the increase and decrease of the output variable in terms of the increase of the independent input variable. Therefore, The NRMSE for peak compressive and tensile strength was 1.02% and 4.04% for MNLR models respectively.
Conclusion
The following conclusions can be drawn from this study:
- The strength of reinforced soil and soil cement is increased by increasing the fiber content.
- Increasing the length of the fibers in the soil and soil cement has no significant effect on increasing the peak compressive strength, but it will be effective in increasing the tensile strength.
- The Multiple Nonlinear Regression models (MNLR) have more accuracy for prediction of output variable (peak strength) because of lower normalized root mean square error../files/site1/files/131/7Extended_Abstract.pdf


 
Ehsan Amjadi Sardehaei, Gholamhosein Tavakoli Mehrjardi,
Volume 13, Issue 5 (12-2019)
Abstract

This paper presents a feed-forward back-propagation neural network model to predict the retained tensile strength and design chart to estimate the strength reduction factors of nonwoven geotextiles due to the installation process. A database of 34 full-scale field tests was utilized to train, validate and test the developed neural network and regression model. The results show that the predicted retained tensile strength using the trained neural network is in good agreement with the results of the test. The predictions obtained from the neural network are much better than the regression model as the maximum percentage of error for training data is less than 0.87% and 18.92%, for neural network and regression model, respectively. Based on the developed neural network, a design chart has been established. As a whole, installation damage reduction factors of the geotextile increases in the aftermath of the compaction process under lower as-received grab tensile strength, higher imposed stress over the geotextiles, larger particle size of the backfill, higher relative density of the backfill and weaker subgrades.

 


Maryam Mokhtari,
Volume 16, Issue 1 (5-2022)
Abstract

In geotechnical engineering, rock mechanics and engineering geology, depending on the project design, uniaxial strength and static Youngchr('39')s modulus of rocks are of vital importance. The direct determination of the aforementioned parameters in the laboratory, however, requires intact and high-quality cores and preparation of their specimens have some limitations. Moreover, performing these tests is time-consuming and costly. Therefore, in this study, it was tried to precisely predict the desirable parameters using physical characteristics and ultrasonic tests. To do so, two methods, i.e. principal components regression and support vector regression, were employed. The parameters used in modelling included density, P- wave velocity, dynamic Poisson’s ratio and porosity. Accordingly, the experimental results conducted on 115 limestone rock samples, including uniaxial compressive and ultrasonic tests, were used and the desired parameters in the modelling were extracted using the laboratory results. By means of correlation coefficient (R2), normalized mean square error (NMSE) and Mean absolute error (MAE), the developed models were validated and their accuracy were evaluated. The obtained results showed that both methods could estimate the target parameters with high accuracy. In support vector regression, Particle Swarm Optimization method was used for determining optimal values of box constraint mode and epsilon mode, and the modelling was conducted using four kernel functions, including linear, quadratic, cubic and Gaussian. Here, the quadratic kernel function yielded the best result for UCS and cubic kernel function yielded the best result for Es. In addition, comparing the results of the principal components regression and the support vector regression indicated that the latter outperformed the former.
Mr. Mehdi Abbasi, Prof. Gholamreza Lashkaripour, Prof. Naser Hafezi Moghaddas, Dr. Hossein Sadeghi,
Volume 19, Issue 1 (6-2025)
Abstract

The elastic modulus is considered one of the most essential parameters in the analysing and designing deep foundations and underground structures. Accurate determination of this parameter usually requires expensive and time-consuming in-situ testing, and validating its accuracy poses significant challenges. Therefore, researchers have consistently focused on developing  empirical models based on geotechnical parameters. In the present study, multiple linear regression models, including general, coarse-grained soil, and fine-grained soil models, were developed to predict the elastic modulus using data obtained from 180 boreholes totaling 5,783 meters in the Mashhad Metro Line 3 project.. Out of 489 pressuremeter tests, 160 datasets were selected based on the availability of complete geotechnical parameters at the same depth. The analysis incorporated the influence of various parameters, including the percentage of sand, silt, and fine particles; grain size characteristics (D10, D30, D60, uniformity coefficient, and coefficient of curvature); Atterberg limits; moisture content; natural and dry density; specific gravity; and cementation indicators (gypsum, carbonate, and organic matter), as well as depth and in-situ stress. The final regression models were developed using a backward stepwise method, implemented through Python programming. The resulting regression equations were derived, and comparative plots between predicted and actual elastic modulus values were presented. The findings demonstrate that the proposed model offers reliable accuracy in estimating the elastic modulus. To evaluate the accuracy of the proposed models in predicting soil elastic modulus, an independent dataset of 39 pressuremeter test results, including both fine- and coarse-grained soils, was used. Statistical indicators demonstrated that the overall model performed best (R²=0.79, MAPE=9.86%). Additionally, the low values of normalized RMSE confirmed the stability and acceptable accuracy of all models.

Mojtaba Rahimi Shahid, Gholam Reza Lashkaripour, Naser Hafezi Moghaddas,
Volume 19, Issue 6 (12-2025)
Abstract

The Sanandaj–Sirjan Structural-Sedimentary Zone is one of the most important geological regions in Iran, where the limestone formations play a key role in civil engineering and mining projects. Precise knowledge of the mechanical properties of these rocks, especially the Uniaxial Compressive Strength (UCS Dry) and the Dry Point Load Index (Is ₅₀-Dry), is essential for the safe and economical design of structures. Due to the high cost and time-consuming nature of direct testing methods, this study employs indirect modeling techniques, including regression and neural networks, to predict these properties.Initially, a comprehensive database was compiled by collecting physical, mechanical, dynamic, and chemical data of limestone samples from the region. Then, univariate, bivariate, and multivariate regression analyses were conducted to extract statistical relationships among the variables. Subsequently, multilayer perceptron neural network models with various architectures based on the Levenberg–Marquardt learning algorithm were developed. The comparison results of model performance indicated that neural networks, due to their ability to identify complex and nonlinear relationships between parameters, provide more accurate predictions of the limestone mechanical properties compared to statistical models. A comparison of the correlation coefficients of multivariate regression equations and neural network models showed that, overall, using neural network models improves the accuracy of UCS Dry predictions by 14.89% and the Is ₅₀-Dry predictions by 4.70%. The application of these models can contribute to cost reduction, increased speed of rock engineering studies, and improved safety in civil projects.
 


Page 1 from 1     

© 2025 CC BY-NC 4.0 | Journal of Engineering Geology

Designed & Developed by : Yektaweb