Committee Machine Research Articles

Drilling rate of penetration prediction through committee support vector regression based on imperialist competitive algorithm

Rate of penetration (ROP) is an important parameter affecting the drilling optimization during well planning. This is particularly important for offshore wells because, offshore rigs contain daily expensive cost and therefore ROP plays a critical role in minimizing time and cost of drilling. There are many factors that affect the ROP such as mud, formation, bit and drilling parameters. In the first step of this study, the best parameters to predict ROP, are selected by error analysis of multivariate regression and then ROP modeling is performed by means of various support vector regression (SVR) methods. Fundamental difference between the individual models is type of kernel function. Finally, a committee machine is constructed in power law framework and it is optimized with imperialist competitive algorithm (ICA). This novel technique is called committee support vector regression based on imperialist competitive algorithm (CSVR-ICA) in this study. Data set are gathered from three jack-up drilling rigs. Results show that CSVR-ICA model improved the results of individual SVR models and it has a good performance in the ROP estimation.

Fusing of optimized intelligence models by virtue of committee machine for estimation of the residual shear strength of clay

Owing to high dependency of landslide stability to residual shear strength (RSS) of clay, provide a sophisticated strategy for modeling of this parameter is advantageous. This paper present strategy based upon fusing of optimized intelligence models for estimation of RSS of clay as a function of readily available data. The developed model is achieved through implementing two following steps. In the first step, two optimized models including optimized neural network, and optimized fuzzy logic are developed for estimation of RSS of clay. Optimizing method which implanted in predictive models for improving those performance is bat-inspired algorithm. In second step, committee machine (CM) is employed for combining outputs of aforementioned optimized models. Bat-inspired is incorporated in CM for determining optimal contribution of optimized elements in final prediction. The superior performance of CM rather than its elements is ascertain through those evaluation based on statistical criteria. Results of this study infer that proposed methodology provide an alternative way for making quantitative formulation between RSS of clay and its index properties.

Combining the advantages of neural networks using the concept of committee machine in the groundwater salinity prediction

Salinity is one of the main factors in groundwater quality monitoring. The main objectives of this study are to investigate and compare the accuracy of three different neural computing techniques, multi-layer perceptron neural network (MLP), radial basis function neural network (RBFNN), and generalized regression neural network (GRNN), in prediction of groundwater salinity of the Tabriz plain confined aquifer, expressed by electrical conductivity [EC (µS/cm)], and to employ an integrated method to combine the advantages of neural network models utilizing the concept of committee machine. To develop the models, 93 data records of groundwater samples were collected from East Azarbaijan regional water company. The data set including Ca2+, Mg2+, Na+, SO4 2− and Cl− concentrations as the inputs and salinity [EC (µS/cm)] as an output were divided into two subsets; training and testing based on cross validation approach. After training and testing of the models, the performance of the models were evaluated using root mean square errors (RMSE), determination coefficient (R2) and mean absolute error (MAE). The performance criteria of the constructed neural network models showed that RBFNN model has the best performance in predicting salinity. The committee neural network (CNN) combined the results of salinity predicted from MLP, RBFNN and GRNN, each of them has a weight factor showing its contribution in overall prediction. The optimal weights were derived by a genetic algorithm (GA). The results of salinity prediction derived from CNN showed that the CNN performs better than any one of the individual ANNs acting alone for predicting groundwater salinity.

Development of an imperialist competitive algorithm (ICA)-based committee machine to predict bit penetration rate in oil wells of Iran

Drilling operation of a well is one the most expensive and time consuming procedures of oil and gas exploitation. Oil companies are always seeking for safe and cost-effective techniques for drilling. The main goal and motivation of drilling optimization is achieving the highest efficiency of work. Optimization and minimization of operational costs is one of the most important prerequisites of any engineering project. Rate of penetration is a crucial factor n drilling controlling cost and time of drilling. In the current research, capabilities of single independent intelligent models are employed for developing a hybrid committee machine that can predict bit penetration bit with high accuracy. To get this goal, three single intelligent models, including neural network, fuzzy logic and neuro-fuzzy, are trained. In the second step, the outputs of these models are integrated by imperialist competitive algorithm (ICA). Finally, a linear equation is achieved which gets outputs of single models as inputs and integrate them somehow the final results is closer to the actual value. The developed ICA-based committee machine is tested by 145 real data points gathered from the drilled wells in an oil field. Correlation of actual and predicted value of ROP obtained from committee machine shows that the model predicts ROP with accuracy of 88 percent. Such model can be used for optimization of drilling parameters in future drilling operations.

Sparse Bayesian extreme learning committee machine for engine simultaneous fault diagnosis

The automotive engine is prone to various faults due to its complex structure and running conditions. Development of a fast response and accurate intelligent system for fault diagnosis of automotive engines is therefore greatly urged. The engine fault diagnosis faces challenges because of the existence of simultaneous-faults which are multiple single-faults appear concurrently. Another challenge is the high cost in acquiring the exponentially increased simultaneous-fault signals. Traditional signal-based engine fault diagnostic systems may not give reliable diagnostic results because they usually rely on single classifier and one kind of engine signal. To enhance the reliability of fault diagnosis and the number of detectable faults, this research proposes a new diagnostic framework namely probabilistic committee machine (PCM). The framework combines feature extraction (empirical mode decomposition and sample entropy), a parameter optimization algorithm, and multiple sparse Bayesian extreme learning machines (SBELM) to form an intelligent diagnostic framework. Multiple SBELM networks are built to form different domain committee members. The members are individually trained using air ratio, ignition pattern and engine sound signals. The diagnostic result from each fault detection member is then combined by using a new probabilistic ensemble method, which can improve the overall diagnostic accuracy and increase the number of detectable faults as compared to individual classifier. Experimental results show the proposed framework is superior to the existing single probabilistic classifier. Moreover, the proposed system can diagnose both single and simultaneous-faults for automotive engines while the system is trained by single-fault patterns only.

Medical image classification based on artificial intelligence approaches: A practical study on normal and abnormal confocal corneal images

Corneal images can be acquired using confocal microscopes which provide detailed views of the different layers inside a human cornea. Some corneal problems and diseases can occur in one or more of the main corneal layers: the epithelium, stroma and endothelium. Consequently, for automatically extracting clinical information associated with corneal diseases, identifying abnormality or evaluating the normal cornea, it is important to be able to automatically recognise these layers reliably. Artificial intelligence (AI) approaches can provide improved accuracy over the conventional processing techniques and save a useful amount of time over the manual analysis time required by clinical experts. Artificial neural networks (ANNs), adaptive neuro fuzzy inference systems (ANFIS) and a committee machine (CM) have been investigated and tested to improve the recognition accuracy of the main corneal layers and identify abnormality in these layers. The performance of the CM, formed from ANN and ANFIS, achieves an accuracy of 100% for some classes in the processed data sets. Three normal corneal data sets and seven abnormal corneal images associated with diseases in the main corneal layers have been investigated with the proposed system. Statistical analysis for these data sets is performed to track any change in the processed images. This system is able to pre-process (quality enhancement, noise removal), classify corneal images, identify abnormalities in the analysed data sets and visualise corneal stroma images as well as each individual keratocyte cell in a 3D volume for further clinical analysis.

A Hybrid Intelligent Computational Scheme for Determination of Refractive Index of Crude Oil Using SARA Fraction Analysis

Asphaltene precipitation and consequent deposition may result in several operational problems ranging from the wellbore to transmission lines. Despite several studies, stability conditions of the asphaltene in crude oil are still a challenging issue and a potential area of investigation. Refractive Index (RI) is a parameter indicative of the region at which asphaltene becomes stable. In this study, a Committee Machine Intelligent System (CMIS) is incorporated to predict the RI of different crude oils through the existing SARA fractions experimental data. The CMIS itself utilizes different artificial neural networks: Multilayer Perceptron (MLP), Radial Basis Function (RBF), and Least Squares Support Vector Machine (LSSVM). By comparing the results of each artificial neural network with the final output, it was demonstrated that the CMIS increases the generalization capability of the utilized artificial network. The results were compared with two well‐known classical correlations. It was proven that the proposed intelligent system outperforms the classical correlations. At the end, outlier detection was performed to identify data which deviate from the bulk of the data points and obtain the applicability domain of the CMIS model.

Improving the estimation accuracy of titration-based asphaltene precipitation through power-law committee machine (PLCM) model with alternating conditional expectation (ACE) and support vector regression (SVR) elements

Asphaltene precipitation/deposition have been longstanding issues in petroleum industry which lead to decline in oil production and economical efficiency. Owing to severe undesirable issues associated with this phenomenon, it is crucial to develop a reliable, accurate, and robust approach for quantitative estimation of asphaltene precipitation. In the first section of this paper, amount of asphaltene precipitation from stock tank oil through titration process was estimated using two predictive methods of Support Vector Regression (SVR) as well as Alternating Conditional Expectation (ACE). A novel predictive method, the so-called Power-Law Committee Machine (PLCM) with constituents of SVR and ACE, was then employed for estimation of the amount of asphaltene precipitation. PLCM model assigns weight factors to each individual sub-model of SVR and ACE to specify the contribution of each particular model in the overall prediction of asphaltene precipitation. Optimal values of these weight factors were extracted by means of Genetic Algorithm (GA) since it was already inserted as the combiner in the structure of the PLCM model. To validate this predictive tool, experimental data collected from open source literature were compared against the model predictions. It was observed that PLCM model can estimate the amount of asphaltene precipitation with very high accuracy and it had more satisfactory prediction performance compared to the other models of SVR and ACE.

A supervised committee machine artificial intelligent for improving DRASTIC method to assess groundwater contamination risk: a case study from Tabriz plain aquifer, Iran

Vulnerability maps are designed to show areas of greatest potential for groundwater contamination on the basis of hydrogeological conditions and human impacts. The objective of this research is (1) to assess the groundwater vulnerability using DRASTIC method and (2) to improve the DRASTIC method for evaluation of groundwater contamination risk using AI methods, such as ANN, SFL, MFL, NF and SCMAI approaches. This optimization method is illustrated using a case study. For this purpose, DRASTIC model is developed using seven parameters. For validating the contamination risk assessment, a total of 243 groundwater samples were collected from different aquifer types of the study area to analyze \( {\text{NO}}_{ 3}^{ - } \) concentration. To develop AI and CMAI models, 243 data points are divided in two sets; training and validation based on cross validation approach. The calculated vulnerability indices from the DRASTIC method are corrected by the \( {\text{NO}}_{3}^{ - } \) data used in the training step. The input data of the AI models include seven parameters of DRASTIC method. However, the output is the corrected vulnerability index using \( {\text{NO}}_{3}^{ - } \) concentration data from the study area, which is called groundwater contamination risk. In other words, there is some target value (known output) which is estimated by some formula from DRASTIC vulnerability and \( {\text{NO}}_{3}^{ - } \) concentration values. After model training, the AI models are verified by the second \( {\text{NO}}_{3}^{ - } \) concentration dataset. The results revealed that NF and SFL produced acceptable performance while ANN and MFL had poor prediction. A supervised committee machine artificial intelligent (SCMAI), which combines the results of individual AI models using a supervised artificial neural network, was developed for better prediction of vulnerability. The performance of SCMAI was also compared to those of the simple averaging and weighted averaging committee machine intelligent (CMI) methods. As a result, the SCMAI model produced reliable estimates of groundwater contamination risk.

NMR Parameters Determination through ACE Committee Machine with Genetic Implanted Fuzzy Logic and Genetic Implanted Neural Network

Free fluid porosity and rock permeability, undoubtedly the most critical parameters of hydrocarbon reservoir, could be obtained by processing of nuclear magnetic resonance (NMR) log. Despite conventional well logs (CWLs), NMR logging is very expensive and time-consuming. Therefore, idea of synthesizing NMR log from CWLs would be of a great appeal among reservoir engineers. For this purpose, three optimization strategies are followed. Firstly, artificial neural network (ANN) is optimized by virtue of hybrid genetic algorithm-pattern search (GA-PS) technique, then fuzzy logic (FL) is optimized by means of GA-PS, and eventually an alternative condition expectation (ACE) model is constructed using the concept of committee machine to combine outputs of optimized and non-optimized FL and ANN models. Results indicated that optimization of traditional ANN and FL model using GA-PS technique significantly enhances their performances. Furthermore, the ACE committee of aforementioned models produces more accurate and reliable results compared with a singular model performing alone.

On accurate determination of PVT properties in crude oil systems: Committee machine intelligent system modeling approach

This study is focused on accurate determination of PVT properties of reservoir oil using an ensemble approach referred to as committee machine intelligent system. PVT properties of interest are oil formation volume factor (Bo) and bubble point pressure (Pb). Committee machine intelligent system model developed in this work combine multi-layer perceptron network, radial basis function network, and least squares support vector machine via a modified weighted averaging method, while optimizing these weights using genetic algorithm. Developed committee machine intelligent system model was found to accord excellently with experimental data yielding correlation coefficients (R2) of 0.980 and 0.976 for bubble point pressure (Pb), and oil formation volume factor (Bo), respectively. Comprehensive comparisons were also carried out between a variety of PVT prediction models and committee machine intelligent system model developed in this study. At the end, leverage value statistics method was applied to detect and identify some probable outlying points from the gathered databases.

The Prediction of Photoelectric Factor, Formation True Resistivity, and Formation Water Saturation from Petrophysical Well Log Data: A Committee Neural Network Approach

Photoelectric factor, formation true resistivity, and formation water saturation are three functional parameters of a hydrocarbon reservoir that could provide invaluable data for reservoir characterization and formation evaluation. The present study proposes an improved strategy for making a quantitative formulation between conventional well log data and the mentionewd parameters. At the first stage of this study, three architectures of artificial neural networks, including generalized regression neural network, radial basis neural network, and Bayesian regulation backpropagation neural network, were employed to predict the aforementioned parameters from conventional well log data. Consequently, a committee neural network was constructed by virtue of hybrid genetic algorithm-pattern search technique. The propounded committee neural network combines the results of generalized regression neural network, radial basis neural network, and Bayesian regulation backpropagation neural network to improve the accuracy of final prediction. It assigns a weight factor to each of the individual artificial neural networks indicating its contribution in overall prediction. A set of data points was used for model construction and another set was employed to assess the model performance. The results showed that integration of artificial neural networks using the concept of committee machine could improve the precision of target prediction, although each of the artificial neural networks has performed adequately for prediction of photoelectric factor and formation true resistivity. The values obtained for formation water saturation are not as accurate as results obtained for photoelectric factor and formation true resistivity, although the correlation coefficient between measured and predicted values for formation water saturation is higher.

Fracture density determination using a novel hybrid computational scheme: a case study on an Iranian Marun oil field reservoir

Most oil production all over the world is from carbonated reservoirs. Carbonate reservoirs are abundant in the Middle East, the Gulf of Mexico and in other major petroleum fields that are regarded as the main oil producers. Due to the nature of such reservoirs that are associated with low matrix permeability, the fracture is the key parameter that governs the fluid flow in porous media and consequently oil production. Conventional methods to determine the fracture density include utilizing core data and the image log family, which are both time consuming and costly processes. In addition, the cores are limited to certain intervals and there is no image log for the well drilled before the introduction of this tool. These limitations motivate petroleum engineers to try to find appropriate alternatives. Recently, intelligent systems on the basis of machine learning have been applied to various branches of science and engineering. The objective of this study is to develop a mathematical model to predict the fracture density using full set log data as inputs based on a combination of three intelligent systems namely, the radial basis function neural network, the multilayer perceptron neural network and the least square supported vector machine. The developed committee machine intelligent system (CMIS) is the weighted average of the individual results of each expert. Proper corresponding weights are determined using a genetic algorithm (GA). The other important feature of the proposed model is its generalization capability. The ability of this model to predict data that have not been introduced during the training stage is very good.

Downhole Pressure Estimation Using Committee Machines and Neural Networks

In gas-lifted oil wells the monitoring of downhole pressure plays an important role. However, the permanent downhole gauge (PDG) sensor often fails. Because maintenance or replacement of PDGs is usually unfeasible, soft-sensors are promising alternatives to monitor the downhole pressure in the case of sensor failure. In this paper, a data-driven soft-sensor is implemented to estimate the downhole pressure using committee machines composed by finite impulse response (FIR) neural networks. Experimental results in three real datasets of the same oil well indicate that the identified soft-sensor is able to predict the downhole pressure with satisfactory accuracy. The model input variables were selected by statistical tests which increased insight concerning such variables. Committee machines outperformed single-model soft-sensors on experimental data.

Supervised Intelligence Committee Machine to Evaluate Field Performance of Photocatalytic Asphalt Pavement for Ambient Air Purification

The ability of a titanium dioxide (TiO2) photocatalytic nanoparticle to trap and to decompose organic and inorganic air pollutants makes it a promising technology as a pavement coating to mitigate the harmful effects of vehicle emissions. Statistical models and artificial intelligence (AI) models are two applicable methods to quantify photocatalytic efficiency. The objective of this study was to develop a model based on field-collected data to predict the nitrogen oxide (NOx) reduction. To achieve this objective, the supervised intelligent committee machine (SICM) method as a combinational black box model was used to predict NOx concentration at the pavement level before and after TiO2 application on the pavement surface. SICM predicts NOx concentration by a nonlinear combination of individual AI models through an artificial intelligent system. Three AI models—Mamdani fuzzy logic, artificial neural network, and neuro-fuzzy—were used to predict NOx concentration in the air as a function of traffic count and climatic conditions, including humidity, temperature, solar radiation, and wind speed before and after the application of TiO2. In addition, an intelligent committee machine model was developed by combining individual AI model output linearly through a set of weights. Results indicated that the SICM model could provide a better prediction of NOx concentration as an air pollutant in the complex and multidimensional air quality data analysis with less residual mean square error than that given by multivariate regression models.

How committee machine with SVR and ACE estimates bubble point pressure of crudes

Bubble point pressure (Pb), one of the most important parameters of reservoir fluids, plays an important role in petroleum engineering calculations. Accurate determination of Pb from laboratory experiments is time, cost and labor intensive. Therefore, the quest for an accurate, fast and cheap method of determining Pb is inevitable. In this communication, a sophisticated approach was followed for formulating Pb to temperature, hydrocarbon and non-hydrocarbon compositions of crudes, and heptane-plus specifications. Firstly, support vector regression (SVR), a supervised learning algorithm plant based on statistical learning (SLT) theory, was employed to construct a model estimating Pb. Subsequently, an alternating conditional expectation (ACE) was used to transform input/output data space to a highly correlated data space and consequently to develop a strong formulation among them. Eventually, SVR and ACE models are combined in a power-law committee machine structure by virtue of genetic algorithm to enhance accuracy of final prediction. A comparison among constructed models and previous models using the concepts of correlation coefficient, mean square error, average relative error and absolute average relative error reveals power-law committee machine outperforms all SVR, ACE, and previous models.

Applying a robust solution based on expert systems and GA evolutionary algorithm for prognosticating residual gas saturation in water drive gas reservoirs

In strong water drive gas reservoirs (WDGRs), the water encroachment in the gas zone has adverse effects on the gas mobility and causes considerable volume of gas to be trapped behind water front; therefore estimation of residual gas saturation after water influx is an important parameter in estimation of gas reservoirs with strong aquifer support. It is difficult to achieve a thorough and exact understanding of water drive gas reservoirs. It depends on several parameters of petrophysical and operational features. In majority of the previous studies about residual gas saturation, the correlations were depended on petrophysical properties such as porosity, permeability, and initial gas saturation. Most of these correlations are well applied on limited dataset that they are constructed based on, but they are not applicable to dataset from other references. In other words, they are not capable of generalization. One reason for this might be different experimental methods to determine the residual gas saturation. In the present study, the prediction of residual gas saturation is presented utilizing Committee Machine Intelligent Technique and some well-known correlations are used for comparison. The reviewed correlations in this study generally do not provide good results and some of them that exhibit reasonable results demand some experimental parameters that are usually unavailable. In this study, two different intelligent models are proposed for spontaneous imbibition and force flood. The suggested models provide good results for the two cases; however, the prediction for force flood is not as exact as the results for the spontaneous imbibition. At the end, an outlier approach detection based on Leverage method was applied to investigate the applicability domain of the proposed models as well as possible outlier data.

Topology and simulations of a Hierarchical Markovian Radial Basis Function Neural Network classifier

This paper presents a Hierarchical Markovian Radial Basis Function Neural Network (HiMarkovRBFNN) model that enables recursive operations. The hierarchical structure of this network is composed of recursively nested RBF Neural Networks with arbitrary levels of hierarchy. All hidden neurons in the hierarchy levels are composed of truly RBF Neural Networks with two weight matrices, for the RBF centers and the linear output weights, in contrast to the simple summation neurons with only linear weighted combinations which are usually encountered in ensemble models and cascading networks. Thus the neural network operation in every node is exactly the same at all levels of the hierarchical integration. The hidden RBF response units are recursive. The training methods also remain the same for all levels as in a typical single RBF Neural Network. The simplicity in the neural network construction process is demonstrated by means of three textbook algorithms, namely the well known k-means clustering, the classical tree-based recursion function and the standard regularized least squares solver. Determining centers can be performed top-down and calculation of linear output weights is performed bottom-up. The framework is rather general and optimization algorithms can also be applied. Experimental simulations on various benchmark datasets show that the recursive operation for the hidden RBF response units is promising. Comparisons with the two standard model meta-learning architectures, namely Committee Machines and Cascaded Machines, reveal that the proposed method produces similar results and compares well as a combiner that can merge many HiMarkovRBFNN child nodes into one higher level parent HiMarkovRBFNN node of the same functionality.

COMPARATIVE ANALYSIS OF NEURAL NETWORKS AND STATISTICAL APPROACHES TO REMOTE SENSING IMAGE CLASSIFICATION

This paper examines different approaches to remote sensing images classification. Included in the study are statistical approach, in particular Gaussian maximum likelihood classifier, and two different neural networks paradigms: multilayer perceptron trained with EDBD algorithm, and ARTMAP neural network. These classification methods are compared on data acquired from Landsat-7 satellite. Experimental results showed that to achieve better performance of classifiers modular neural networks and committee machines should be applied.

IMPLEMENTATION OF NEURAL NETWORKS AND BOOSTING ALGORITHMS FOR EFFECTIVE INTRUSION DETECTION

In this article the classification task in the domain of intrusion detection is considered. Often a chosen algorithm is not good enough for practical use. So the question arises how is it possible to improve the performance? In this case we can employ so-called Committee Machines that increase accuracy and reliability of the base classification model. These advantages are the result of dividing complex computational problems among several experts. The knowledge of each expert influences on the general conclusion of Committee Machine.