Prediction of pile settlement using artificial neural networks based on standard penetration test data

In recent years artificial neural networks (ANNs) have been applied to many geotechnical engineering problems with some degree of success. With respect to the design of pile foundations, accurate prediction of pile settlement is necessary to ensure appropriate structural and serviceability performance. In this paper, an ANN model is developed for predicting pile settlement based on standard penetration test (SPT) data. Approximately 1000 data sets, obtained from the published literature, are used to develop the ANN model. In addition, the paper discusses the choice of input and internal network parameters which were examined to obtain the optimum model. Finally, the paper compares the predictions obtained by the ANN with those given by a number of traditional methods. It is demonstrated that the ANN model outperforms the traditional methods and provides accurate pile settlement predictions.     

Neural network based prediction of ground surface settlements due to tunnelling

Ground surface settlement due to tunnel excavation varies in magnitude and trend depending on several factors such as tunnel geometry, ground conditions, etc. Although there are several empirical and semi-empirical formulae available for predicting ground surface settlement, most of these do not simultaneously take into consideration all the relevant factors, resulting in inaccurate predictions. In this study, an artificial neural network (ANN) is incorporated with '113' of monitored field results to predict surface settlement for a tunnel site with prescribed conditions. To achieve this, a standard format (a protocol) for a database of monitored field data is first proposed and then used for sorting out a variety of monitored data sets available in KICT. Using the capabilities of pattern recognition and memorization of the ANN, an attempt is made to capture the rich physical characteristics smeared in the database and at the same time filter inherent noise in the monitored data. Here, an optimal neural network model is suggested through preliminary parametric studies. It is shown that preliminary studies for generating an optimal ANN under given training data sets are necessary because no analytical method for this purpose is available to date. In addition, this study introduces a concept of relative strength of effects (RSE) [Yang Y, Zhang Q. A heirarchical analysis for rock engineering using artificial neural networks. Rock Mechanics and Rock Engineering 1997; 30(4): 207–22] in sensitivity analysis for various major factors affecting the surface settlement in tunnelling. It is seen in some examples that the RSE rationally enables us to recognize the most significant factors of all the contributing factors. Two verification examples are undertaken with the trained ANN using the database created in this study. It is shown from the examples that the ANN has adequately recognized the characteristics of the monitored data sets retaining a generality for further prediction. It is believed that an ANN based hierarchical prediction procedure shown in this paper can be further employed in many kinds of geotechnical engineering problems with inherent uncertainties and imperfections.     

Evaluation of water quality parameters for the Mamasin dam in Aksaray City in the central Anatolian part of Turkey by means of artificial neural networks

Sustaining the human ecological benefits of surface water requires carefully planned strategies for reducing the cumulative risks posed by diverse human activities. The municipality of Aksaray city plays a key role in developing solutions to surface water management and protection in the central Anatolian part of Turkey. The responsibility to provide drinking water and sewage works, regulate the use of private land and protect public health provides the mandate and authority to take action. The present approach discusses the main sources of contamination and the result of direct wastewater discharges into the Melendiz and Karasu rivers, which recharge the Mamasın dam sites by the use of artificial neural network (ANN) modeling techniques. The present study illustrates the ability to predict and/or approve the output values of previously measured water quality parameters of the recharge and discharge areas at the Mamasin dam site by means of ANN techniques. Using the ANN model is appreciated in such environmental research. Here, the ANN is used for estimating if the field parameters are agreeable to the results of this model or not. The present study simulates a situation in the past by means of ANN. But in case any field measurements of some relative parameters at the outlet point “discharge area” have been missed, it could be possible to predict the approximate output values from the detailed periodical water quality parameters. Because of the high variance and the inherent non-linear relationship of the water quality parameters in time series, it is difficult to produce a reliable model with conventional modeling approaches. In this paper, the ANN modeling technique is used to establish a model for evaluating the change in electrical conductivity (EC) and dissolved oxygen (DO) values in recharge (input) and discharge (output) areas of the dam water under pollution risks. A general ANN modeling scheme is also recommended for the water parameters. The modeling process includes four main stages: (1) source data analysis, (2) system priming, (3) system fine-tuning and (4) model evaluation. Results of the ANN modeling scheme indicate that the output values are agreeable to the water quality parameters, which were measured at the field in the static water mass of the Mamasın dam lake. Water contamination at the dam site is caused by the continuous increase of nutrient contents and decrease of the O₂ level in water causing an anaerobic condition. It may stimulate algae growth flow in such water bodies, consequently reducing water quality.     

广义回归神经网络预测加筋土支挡结构高度

土工合成材料加筋支挡结构（Geosythetics-Reinforced Retaining Wall, 简称GRW）设计方法主要是建立在似粘聚力理论基础之上的半经验设计法。由于土性及加筋机理的复杂性，常常要对它们进行人为假定，导致计算结果差强人意。神经网络方法与传统方法的不同之处在于不需要主观假定，而是模拟人脑思维，通过数据样本的学习来获得预测结果。引入神经网络技术来预测加筋土支挡结构的设计高度是一种新尝试。由于本问题具有样本容量非常有限、影响因素复杂多样的特点。因此，采用适用于稀土样本数据的广义回归网络（General Regression Neural Network)来预测加筋土支挡结构设计高度。基于MATLAB神经网络工具箱及文献[1]的挡墙离心模型试验结果，建立了一个可用于加筋支挡结构设计高度预测的GRNN网络。通过对足尺试验，实际工程及模型试验结果的检验，表明网络的学习是成功的，具有一定指导意义。     

土工合成材料耐久性的神经网络预测

以BP人工神经网络为工具,利用其强大的非线性映射能力,在综合分析土工合成材料耐久性影响因素的基础上将土工合成材料所处环境的温度、湿度、紫外线照射情况以及土工合成材料的老化时间作为网络的输入参数,以描述土工合成材料耐久性状态的强度和延伸率作为网络的输出,建立了神经网络模型。采用大量试验数据对网络进行了训练和检验,并对土工合成材料的耐久性进行了预测。结果表明,预测值和试验结果比较接近,该网络能较好地反映土工合成材料耐久性与其影响因素之间的非线性映射关系。     

Numerical simulation of the seismic response of the Zipingpu concrete face rockfill dam during the Wenchuan earthquake based on a generalized plasticity model

The strong ground motion of the 2008 Wenchuan earthquake in China caused considerable damage to the Zipingpu concrete face rockfill dam (CFRD). The maximum crest settlement was approximately 1.0 m, and compressive failure and joint dislocations were observed in the face slabs. This damage has made it necessary to understand the damage pattern and safety of high CFRDs subjected to strong earthquake shaking, and the response of the Zipingpu CFRD during the Wenchuan earthquake can be used as a benchmark for this purpose. In this study, a 3D dynamic procedure was employed to simulate the dynamic responses of the Zipingpu CFRD. The rockfill materials were described using a generalized plasticity model, while the interfaces between the face slabs and cushions were modeled using zero-thickness interface elements that follow a perfect elasto-plastic stress–strain model in the tangential direction using Coulomb’s friction law. Dam deformation, face-slab stress, and face joint dislocations were simulated, and the results were compared with the field measurements. Using the generalized plastic model, the residual deformation of the dam during the earthquake could be directly obtained without being complemented by separate, semi-empirical procedures. The rockfill materials shrank to the center of the valley due to the strong shaking, causing crushing damage in the zone of the slabs. The dislocation of construction joints was also duplicated by the numerical procedure. The results of this study indicate that a 3D finite element procedure based on a generalized plasticity model can be used to evaluate the dynamic responses of CFRDs during strong earthquakes.     

A neural network approach for attenuation relationships: An application using strong ground motion data from Turkey

This paper presents an application of neural network approach for the prediction of peak ground acceleration (PGA) using the strong motion data from Turkey, as a soft computing technique to remove uncertainties in attenuation equations. A training algorithm based on the Fletcher–Reeves conjugate gradient back-propagation was developed and employed for three sample sets of strong ground motion. The input variables in the constructed artificial neural network (ANN) model were the magnitude, the source-to-site distance and the site conditions, and the output was the PGA. The generalization capability of ANN algorithms was tested with the same training data. To demonstrate the authenticity of this approach, the network predictions were compared with the ones from regressions for the corresponding attenuation equations. The results indicated that the fitting between the predicted PGA values by the networks and the observed ones yielded high correlation coefficients (R²). In addition, comparisons of the correlations by the ANN and the regression method showed that the ANN approach performed better than the regression. Even though the developed ANN models suffered from optimal configuration about the generalization capability, they can be conservatively used to well understand the influence of input parameters for the PGA predictions.     

Settlement Prediction of Model Piles Embedded in Sandy Soil Using the Levenberg–Marquardt (LM) Training Algorithm

This investigation aimed to examine the load carrying capacity of model piles embedded in sandy soil and to develop a predictive model to simulate pile settlement using a new artificial neural network (ANN) approach. A series of experimental pile load tests were carried out on model concrete piles, comprised of three piles with slenderness ratios of 12, 17 and 25. This was to provide an initial dataset to establish the ANN model, in attempt at making current, in situ pile-load test methods unnecessary. Evolutionary Levenberg–Marquardt (LM) MATLAB algorithms, enhanced by T-tests and F-tests, were developed and applied in this process. The model piles were embedded in a calibration chamber in three densities of sand; loose, medium and dense. According to the statistical analysis and the relative importance study, pile lengths, applied load, pile flexural rigidity, pile aspects ratio, and sand-pile friction angle were found to play a key role in pile settlement at different contribution levels, following the order: P?>?δ?>?lc/d?>?lc?>?EA. The results revealed that the optimum model of the LM training algorithm can be used to characterize pile settlement with good degree of accuracy. There was also close agreement between the experimental and predicted data with a root mean square error, (RMSE) and correlation coefficient (R) of 0.0025192 and 0.988, respectively.     

Tenfold cross validation artificial neural network modeling of the settlement behavior of a stone column under a highway embankment

Construction of embankments in engineering structures on soft clay soils normally encounters problems related to excessive settlement issues. The conventional methods are inadequate to analyze and predict the surface settlement when the necessary parameters are difficult to determine in the field and in the laboratory. In this study, artificial neural network systems (ANNs) were used to predict settlement under embankment load using soft soil properties together with various geometric parameters as input for each stone column (SC) arrangement and embankment condition. Data from a highway project called Lebuhraya Pantai Timur2 in Terengganu, Malaysia, were investigated. The FEM package of Plaxis v8 program analysis was utilized. The actual angle of internal friction, spacing between SC, diameter of SC, length of SC, and height of embankment were used as the input parameters, and the settlement was used as the main output. Non cross validation (NCV) and tenfold cross validation (TFCV) were used to build the ANN model. The results of the TFCV model were more accurate than those of the NCV model. Comparisons made with the predictions of the Priebe model showed that the proposed TFCV model could provide better predictions than conventional methods.     

Study on collapse settlement and cracks of core wall rockfill dams under wetting deformation

The safety and stability of core wall rockfill dams during impoundment are threatened by the wetting deformation of up-stream shell materials. The serious wetting deformation not only aggravates the collapse settlement of upstream rockfill but also intensifies differential settlement of the dam crest during impoundment, and then causes cracks on the dam crest. On the basis of the proposed wetting deformation model and its simulation method of wetting deformation, this paper simulated the impoundment process of Guanyinyan core wall rockfill dam and studied the deformation characteristic of the dam during impoundment. In addition, the smeared cracking model was used to simulate the crack propagation on the dam crest, and the crack develop and spread mechanism was analysed. The results show that the simulated deformation can fit the in-situ data well, and the simulated crack propagation is in good agreement with the actual situation. Once watered, the upstream rockfill and core wall have significant settlement, and the whole dam crest has obvious horizontal displacement towards the upstream. It is on the same order of magnitude that the increment of horizontal displacement and settlement at the top of the dam during impoundment. In the process of impoundment, the upper part of the dam tends to deform towards the reservoir area, which will lead to tensile cracks appearing in the rockfill areas on both upstream and downstream sides of the core wall of the dam crest, and the propagation direction of the cracks is basically parallel to the adjacent core wall surface. With the water level rising, the cracks on the downstream side of the dam crest mainly extend vertically, and the cracks on the upstream side of the dam crest not only extend vertically, but also extend horizontally.     

BP神经网络-灰色系统联合模型预测软基沉降量

目前软基沉降预测多采用指数曲线和双曲线延伸法,其结果不够理想,神经网络在此方面的运用也存在一定的局限,虽然GM(1,1)模型在软基沉降预测领域已得到运用,但在已有的案例中所使用的等时距模型都没有明确说明所采用的插值方法。以深圳湾西部通道填海软基沉降预测分析为例,建立BP神经网络-灰色系统联合模型来探讨解决这一问题的方法。采用BP神经网络逼近非线性插值方法构建等时距时间序列数据,在此基础上建立沉降变形时间序列的GM模型,并建立相应的时间响应函数,预测沉降量。计算实例表明,该模型短期沉降预测结果比较准确,其最终沉降预测结果具有一定的工程参考价值。     

Inversion of quasi-3D DC resistivity imaging data using artificial neural networks

The objective of this paper is to investigate the applicability of artificial neural networks in inverting quasi-3D DC resistivity imaging data. An electrical resistivity imaging survey was carried out along seven parallel lines using a dipole-dipole array to confirm the validation of the results of an inversion using an artificial neural network technique. The model used to produce synthetic data to train the artificial neural network was a homogeneous medium of 100Ωm resistivity with an embedded anomalous body of 1000Ωm resistivity. The network was trained using 21 datasets (comprising 12159 data points) and tested on another 11 synthetic datasets (comprising 6369 data points) and on real field data. Another 24 test datasets (comprising 13896 data points) consisting of different resistivities for the background and the anomalous bodies were used in order to test the interpolation and extrapolation of network properties. Different learning paradigms were tried in the training process of the neural network, with the resilient propagation paradigm being the most efficient. The number of nodes, hidden layers, and efficient values for learning rate and momentum coefficient have been studied. Although a significant correlation between results of the neural network and the conventional robust inversion technique was found, the ANN results show more details of the subsurface structure, and the RMS misfits for the results of the neural network are less than seen with conventional methods. The interpreted results show that the trained network was able to invert quasi-3D electrical resistivity imaging data obtained by dipole-dipole configuration both rapidly and accurately.     

Application of an Expert System for the Assessment of Blast Vibration

The purpose of this article is to evaluate and predict the blast induced ground vibration using different conventional vibration predictors and artificial neural network (ANN) at a surface coal mine of India. Ground Vibration is a seismic wave that spread out from the blast hole when detonated in a confined manner. 128 blast vibrations were recorded and monitored in and around the surface coal mine at different strategic and vulnerable locations. Among these, 103 blast vibrations data sets were used for the training of the ANN network as well as to determine site constants of various conventional vibration predictors, whereas rest 25 blast vibration data sets were used for the validation and comparison by ANN and empirical formulas. Two types of ANN model based on two parameters (maximum charge per delay and distance between blast face to monitoring point) and multiple parameters (burden, spacing, charge length, maximum charge per delay and distance between blast face to monitoring point) were used in the present study to predict the peak particle velocity. Finally, it is found that the ANN model based on multiple input parameters have better prediction capability over two input parameters ANN model and conventional vibration predictors.     

软基沉降的BP神经网络和灰色系统联合预测

使用BP神经网络插值方法对灰色数据进行了预处理,进而建立了预测软基沉降量的BP神经网络和灰色系统联合模型.实例分析表明,该模型短期沉降预测结果的最大相对误差小于2%,最终沉降预测结果的相对偏差小于5%,且灰色预测时取后期沉降瘦导颇算结果准确度高于取前期沉降数据的计算结果准确度.     

Microparameter Prediction for a Triaxial Compression PFC3D Model of Rock Using Full Factorial Designs and Artificial Neural Networks

This paper presents an integrated approach that predicts the microparameters of the particle flow code in three dimensions (PFC3D) model in triaxial compression simulations. The new approach combines a full factorial design (FFD) with an artificial neural network (ANN). The ANN model maps the input factors (triaxial compressive strength, Poisson’s ratio, and Young’s modulus) onto output variables, which are microparameters that affect the macroscopic responses in a PFC3D model. Emphasis is placed on data collection and optimization of the ANN model using FFD. The data for training and testing the ANN model were obtained from laboratory experiments and numerical simulations of a PFC3D model according to the principles of FFD. Using a backpropagation artificial neural network (BPNN) optimized with FFD principles, the object of the current study (to reliably predict the microparameters for a PFC3D model) has been achieved because the predicting data obtained by the BPNN model were in excellent agreement with the testing data.     

RBFNN模型在渗透系数反演中的应用

针对经典的BP网络存在的一些缺陷,采用了径向基函数神经网络（RBFNN）。在相同的收敛条件下,用RBFNN和经典算法的BP网络进行了比较,表明前者具有优越性。在工程实例中,基于人工神经网络的非线性特点,在三维渗流有限元的基础上,利用RBFNN反演了大坝的渗透系数。并利用反演结果进行渗流场分析,水头预报值也有很高的精度,说明反演结果是正确的,从而,验证了RBFNN应用于反演分析中的可靠性。     

人工神经网络氡气灾害模型在铀矿山的应用

对多层前馈神经网络模型(BP神经网络)的结构特点、数据分析、学习方法和过程等内容做了分析。以中核抚州金安铀业有限公司铀矿山氡气状态为学习训练样本及预测样本,建立铀矿山氡气灾害模型。讨论了基于BP神经网络技术的氡气灾害模型分析方法及其有效性。通过实例样本的训练检验表明,采用人工神经网络方法对铀矿山氡灾害预测取得了比较满意的效果,为神经网络在铀矿山氡气灾害预报的应用提供了可行性。     

Artificial neural network and liquefaction susceptibility assessment: a case study using the 2001 Bhuj earthquake data,Gujarat, India

This study pertains to prediction of liquefaction susceptibility of unconsolidated sediments using artificial neural network (ANN) as a prediction model. The backpropagation neural network was trained, tested, and validated with 23 datasets comprising parameters such as cyclic resistance ratio (CRR), cyclic stress ratio (CSR), liquefaction severity index (LSI), and liquefaction sensitivity index (LSeI). The network was also trained to predict the CRR values from LSI, LSeI, and CSR values. The predicted results were comparable with the field data on CRR and liquefaction severity. Thus, this study indicates the potentiality of the ANN technique in mapping the liquefaction susceptibility of the area.     

An evaluation of existent methods for estimation of embankment dam breach parameters

The study of dam-break analysis is considered important to predict the peak discharge during dam failure. This is essential to assess economic, social and environmental impacts downstream and to prepare the emergency response plan. Dam breach parameters such as breach width, breach height and breach formation time are the key variables to estimate the peak discharge during dam break. This study presents the evaluation of existing methods for estimation of dam breach parameters. Since all of these methods adopt regression analysis, uncertainty analysis of these methods becomes necessary to assess their performance. Uncertainty was performed using the data of more than 140 case studies of past recorded failures of dams, collected from different sources in the literature. The accuracy of the existing methods was tested, and the values of mean absolute relative error were found to be ranging from 0.39 to 1.05 for dam breach width estimation and from 0.6 to 0.8 for dam failure time estimation. In this study, artificial neural network (ANN) was recommended as an alternate method for estimation of dam breach parameters. The ANN method is proposed due to its accurate prediction when it was applied to similar other cases in water resources.

     

Prediction of the unconfined compressive strength of compacted granular soils by using inference systems

Adaptive neuro-fuzzy inference system (ANFIS) and artificial neural network (ANN) models have been extensively used to predict different soil properties in geotechnical applications. In this study, it was aimed to develop ANFIS and ANN models to predict the unconfined compressive strength (UCS) of compacted soils. For this purpose, 84 soil samples with different grain-size distribution compacted at optimum water content were subjected to the unconfined compressive tests to determine their UCS values. Many of the test results (for 64 samples) were used to train the ANFIS and the ANN models, and the rest of the experimental results (for 20 samples) were used to predict the UCS of compacted samples. To train these models, the clay content, fine silt content, coarse silt content, fine sand content, middle sand content, coarse sand content, and gravel content of the total soil mass were used as input data for these models. The UCS values of compacted soils were output data in these models. The ANFIS model results were compared with those of the ANN model and it was seen that the ANFIS model results were very encouraging. Consequently, the results of this study have important findings indicating reliable and simple prediction tools for the UCS of compacted soils.