Coupled Simulated Annealing Research Articles

Toward a predictive model for estimating viscosity of ternary mixtures containing ionic liquids

Ionic liquids (ILs) are widely under research and being more and more used in many industrial fields due to their unique properties. The knowledge of physicochemical properties of ILs is vital for their applications. Viscosity is one of the most important physical properties of ILs. Since the experimental measurement of viscosity is expensive, time consuming and cumbersome, seeking for a reliable and accurate model for predicting the viscosity of IL mixtures is essential. To this end, a large data bank, covering a wide range of ternary mixtures containing ILs, was collected from open literature sources. Afterwards, a rigorous modeling approach, namely least square support vector machine (LSSVM) was employed to predict the viscosity of ternary mixtures. The parameters of the model were optimized using coupled simulated annealing (CSA) optimization tool. The inputs of the model are boiling point temperature, molecular weight and composition of mixture compounds. Statistical and graphical error analyses indicated that the developed model can satisfactorily predict ternary mixture viscosity. Moreover, it was demonstrated that the proposed model is able to predict the actual physical trend of viscosity with variation of molecular weight and boiling point of ILs, ethanol content and water content. Furthermore, the relevancy factor demonstrated that the average boiling point of the ternary mixture has the greatest impact on the viscosity. Finally, the Leverage approach was performed, in which the statistical Hat matrix, Williams Plot, and the residuals of the model results led to recognition of the probable outliers. All the experimental data seem to be reliable except five data points. Thus, the developed model could be reliable for the prediction of ternary mixture viscosity in its applicability domain.

Toward a predictive model for predicting viscosity of natural and hydrocarbon gases

Accurate knowledge of pure hydrocarbon and natural gas viscosity is essential for reliable reservoir characterization and simulation as well as economic design of natural gas processing and transport units. The most trustable sources of pure hydrocarbon and natural gas viscosity values are laboratory experiments. When there is no available experimental data for the required composition, pressure, and temperature conditions, the use of predictive methods becomes important. In this communication, a novel approach was proposed to develop for prediction of viscosity of pure hydrocarbons as well as gas mixtures containing heavy hydrocarbon components and impurities such as carbon dioxide, nitrogen, helium, and hydrogen sulfide, using over 3800 data sets. A robust soft computing approach namely least square support vector machine (LSSVM) modeling optimized with coupled simulated annealing (CSA) optimization technique was proposed (CSA-LSSVM model). Moreover, comparative studies between the CSA-LSSVM model and fourteen empirical correlations were done. To this end, statistical and graphical error analyses have been used simultaneously. Obtained results illustrated that the “two-parameter” CSA-LSSVM model is more robust, reliable, and consistent than the existing correlations for the prediction of pure and natural gas viscosity. Moreover, the relevancy factor illustrated that the molecular weight of gas has the greatest impact on gas viscosity.

Predicting the solubility of SrSO4 in Na–Ca–Mg–Sr–Cl–SO4–H2O system at elevated temperatures and pressures

Precipitation of strontium sulfate (or SrSO4) has already been distinguished as one of the most costly and critical problems which may occur in process industries and oilfield operations. Costs due to scaling and remedial actions that need to be taken afterward are generally high owing to low solubility of SrSO4 in aqueous solutions. Therefore, a thorough understanding of the SrSO4 thermodynamic behavior under various operating conditions is vital to predict or even avoid the overall damage caused by scaling. The primary aim of this work is to develop a model based on Least Squares Support Vector Machine (LSSVM) and Coupled Simulated Annealing (CSA) referred to as CSA-LSSVM to predict strontium sulfate solubility as a function of pressure, temperature and ionic compositions. In this context, we have employed almost 1641 experimental data regarding strontium sulfate solubility to build a comprehensive model and to evaluate its reliability. The results show that the proposed model has a better performance in comparison with pre-existing empirical correlations used for predicting SrSO4 solubility and it is also in well accordance with experimental measurements. A comparison was also drawn between predictions of Pitzer ion interaction model with CSA-LSSVM model forecasts in terms of sensitivity analysis and single point solubility predictions. The whole dataset was analyzed at the end to diagnose the possible outliers and to investigate the reliability of recorded measurements. Based on the results obtained from this study, developed model could successfully be used in predicting SrSO4 solubility in aqueous Na–Ca–Mg–Sr–Cl–SO4–H2O system over temperature ranges from 2 to 253.5°C, and pressures from 1 to 568.51atm.

Prediction of the aqueous solubility of BaSO4 using pitzer ion interaction model and LSSVM algorithm

Deposition of barium sulfate (or BaSO4) has already been recognized as a devastating problem facing process industries and oilfield operations, mainly owing to its low solubility in aqueous solutions. Predicting and also preventing the overall damage caused by BaSO4 precipitation requires a profound knowledge of its solubility under different thermodynamic conditions. The main aim of this study is to develop a solubility prediction model based on a hybrid of least squares support vector nachines (LSSVM) and coupled simulated annealing (CSA) aiming to predict the solubility of barium sulfate over wide ranges of temperature, pressure and ionic compositions. Results indicate that predictions of the presented model are in well accordance with experimental measurements yielding overall correlation coefficient (R2) of 0.996 and total RMSE of 0.00077. Constructed model was also found to outperform a previously well-known correlation employed for predicting BaSO4 solubility in aqueous solutions. Solubility predictions based on Pitzer ion interaction model were also compared to CSA-LSSVM predictions in terms of single point estimations and sensitivity analysis. Results of this study suggest that CSA-LSSVM model could be implemented in assessing the solubility of BaSO4 in Na–Mg–Ca–K–Ba–Cl–SO4–H2O system over wide ranges of temperature from 0 to 279°C and pressures ranging from 1 to 1517 bars with an acceptable degree of accuracy.

Rigorous modeling for prediction of barium sulfate (barite) deposition in oilfield brines

Barium sulfate (barite) has been recognized to be a major operational problem in surface and subsurface oil and gas production operations. Therefore, accurate estimation of this deposition type can result in increasing the efficiency of oil and gas production. In this work, a novel approach is implemented to develop a predictive model for the estimation of solubility product data of barite in oilfield brines.The model is proposed using a robust soft computing approach, namely, least-squares support vector machine (LSSVM) modeling optimized with the coupled simulated annealing (CSA) optimization approach. Our results indicate that there is good agreement between predictions based on the CSA-LSSVM model and literature data on the solubility product of barite in oilfield brines. Furthermore, performance of the developed model is compared with the performance of an artificial neural network, available correlation in the literature and standard software (OLI Scalechem) for predicting barite deposition.The model perfectly fits the literature data with a squared correlation coefficient of 0.999.

Automatic Identification of Formation Iithology from Well Log Data: A Machine Learning Approach

Determination of the hydrocarbon content and also the successful drilling of petroleum wells are highly contingent upon the lithology of the underground formation. Conventional lithology identification methods are either uneconomical or of high uncertainties.The main aim of this study is to develop an intelligent model based on Least Squares Support Vector Machine (LSSVM) and Coupled Simulated Annealing (CSA) algorithm simply called CSA-LSSVM for predicting the lithology in one of the Iranian oilfields. To this end, photoelectric index (PEF) values were simulated by CSA-LSSVM algorithm based on valid well logging data generally known as lithology indicators. Model predictions were compared to the real data obtained from well logging operation and the overall Correlation Coefficient (R2) of 0.993 and Average Absolute Relative Deviation (AARD) of 1.6% were obtained for the total dataset (3243 data points) which shows the robustness of the CSA-LSSVM algorithm in predicting accurate PEF values. In order to check the validity of the employed well log data,value statistical method was implemented in this study for detecting the possible outliers. However, diagnosing only one single data point as the suspected data or probable outlier reveals the validity of recorded data points and shows high applicability domain of the proposed model.

Parameter Design of LS-SVM Based on QPSO and its Application to Node Localization

Regularization parameter and kernel parameters play a crucial role on the accuracy and rapidity of the least squares support vector machine (LS-SVM) algorithm. Existing methods of tuning the parameters for LS-SVM have some disadvantages, such as difficulty of getting global optimum and poor convergent speed. In this work, we develop the parameter design of LS-SVM using quantum particle swarm optimization (QPSO) and apply the designed LS-SVM into node localization. According to the initial values, the optimal parameter values are obtained using the leave-10-out cross validation method. Simulation results show that the proposed method is compared to the coupling simulated annealing (CSA) LS-SVM and the improved particle swarm optimization (IPSO) LS-SVM.

Robust Model for the Determination of Wax Deposition in Oil Systems

Wax deposition is a serious problem during oil production in the petroleum industry. Therefore, accurate prediction of this solid deposition problem can result in increasing the efficiency of oil/gas production. In this article, a novel approach is proposed to develop a predictive model for the estimation of wax deposition. An intelligent reliable model is proposed using a robust soft computing approach, namely, least-squares support vector machine (LSSVM) modeling optimized with the coupled simulated annealing (CSA) optimization approach. Our results demonstrate that there is good agreement between predictions based on the CSA-LSSVM model and experimental data on wax deposition. Furthermore, the performance of the newly developed model is compared with the performance of neural network and multisolid models for predicting wax deposition. The results of this comparison indicate that the proposed method is superior, in terms of both accuracy and generality, to the neural network and multisolid models. Finally, to check whether the newly developed CSA-LSSVM model is statistically correct and valid, the leverage approach, in which the statistical Hat matrix, the Williams plot, and the residuals of the model results lead to the identification of probable outliers, is applied. It is found that all of the wax deposition experimental data used in the present study seem to be reliable and that only one point is outside the applicability domain of the developed models for wax deposition.

Utilization of support vector machine to calculate gas compressibility factor

The compressibility factor (Z-factor) is considered as a very important parameter in the petroleum industry because of its broad applications in PVT characteristics. In this study, a meta-learning algorithm called Least Square Support Vector Machine (LSSVM) was developed to predict the compressibility factor. In addition, the proposed technique was examined with previous models, exhibiting an R2 and an MSE of 0.999 and 0.000014, respectively. A significant drawback in the conventional LSSVM is the determination of optimal parameters to attain desired output with a reasonable accuracy. To eliminate this problem, the current study introduced coupled simulated annealing (CSA) algorithm to develop a new model, known as CSA-LSSVM. The proposed algorithm included 4756 datasets to validate the effectiveness of the CSA-LSSVM model using statistical criteria. The new technique can be utilized in chemical and petroleum engineering software packages where the most accurate value of Z-factor is required to predict the behavior of real gas, significantly affecting design aspects of equipment involved in gas processing plants.

Toward a predictive model for estimating dew point pressure in gas condensate systems

Dew-point pressure is one of the most important quantities for characterizing and successful prediction of the future performance of gas condensate reservoirs. The objective of this study is to present a reliable, computer-based predictive model for prediction of dew-point pressure in gas condensate reservoirs. An intelligent approach based on least square support vector machine (LSSVM) modeling was developed for this purpose. To this end, the model was developed and tested using a total set of 562 experimental data points from different retrograde gas condensate fluids covering a wide range of variables. Coupled simulated annealing (CSA) was employed for optimization of hyper-parameters of the model. The results showed that the developed model significantly outperforms all the existing methods and provide predictions in acceptable agreement with experimental data. In addition, it is shown that the proposed model is capable of simulating the actual physical trend of the dew-point pressure versus temperature for a constant composition fluid on the phase envelope.

Robust LS-SVM regression for ore grade estimation in a seafloor hydrothermal sulphide deposit

Due to the geological complexities of ore body formation and limited borehole sampling, this paper proposes a robust weighted least square support vector machine (LS-SVM) regression model to solve the ore grade estimation for a seafloor hydrothermal sulphide deposit in Solwara 1, which consists of a large proportion of incomplete samples without ore types and grade values. The standard LS-SVM classification model is applied to identify the ore type for each incomplete sample. Then, a weighted K-nearest neighbor (WKNN) algorithm is proposed to interpolate the missing values. Prior to modeling, the particle swarm optimization (PSO) algorithm is used to obtain an appropriate splitting for the training and test data sets so as to eliminate the large discrepancies caused by random division. Coupled simulated annealing (CSA) and grid search using 10-fold cross validation techniques are adopted to determine the optimal tuning parameters in the LS-SVM models. The effectiveness of the proposed model by comparing with other well-known techniques such as inverse distance weight (IDW), ordinary kriging (OK), and back propagation (BP) neural network is demonstrated. The experimental results show that the robust weighted LS-SVM outperforms the other methods, and has strong predictive and generalization ability.

Prediction of sour gas compressibility factor using an intelligent approach

Compressibility factor (z-factor) values of natural gasses are essential in most petroleum and chemical engineering calculations. The most common sources of z-factor values are laboratory experiments, empirical correlations and equations of state methods. Necessity arises when there is no available experimental data for the required composition, pressure and temperature conditions. Introduced here is a technique to predict z-factor values of natural gasses, sour reservoir gasses and pure substances. In this communication, a novel mathematical-based approach was proposed to develop reliable model for prediction of compressibility factor of sour and natural gas. A robust soft computing approach namely least square support vector machine (LSSVM) modeling optimized with coupled simulated annealing (CSA) optimization tool was proposed. To evaluate the performance and accuracy of this model, statistical and graphical error analyses have been used simultaneously. Moreover, comparative studies have been conducted between this model and nine empirical correlations and equations of state. The obtained results demonstrated that the proposed CSA-LSSVM model is more robust, reliable and efficient than the existing correlations and equations of state for prediction of z-factor of sour and natural gasses.

An empirical model for shear capacity of RC deep beams using genetic-simulated annealing

This paper presents an empirical model to predict the shear strength of RC deep beams. A hybrid search algorithm coupling genetic programming (GP) and simulated annealing (SA), called genetic simulated annealing (GSA), was utilized to develop mathematical relationship between the experimental data. Using this algorithm, a constitutive relationship was obtained to make pertinent the shear strength of deep beams to nine mechanical and geometrical parameters. The model was developed using an experimental database acquired from the literature. The results indicate that the proposed empirical model is properly capable of evaluating the shear strength of deep beams. The validity of the proposed model was examined by comparing its results with those obtained from American Concrete Institute (ACI) and Canadian Standard Association (CSA) codes. The derived equation is notably simple and includes several effective parameters.

Implementation of SVM framework to estimate PVT properties of reservoir oil

Through this work, a novel mathematical-based approach was proposed to develop reliable models for calculation of PVT properties of crude oils at various reservoir conditions. For this purpose, a new soft computing approach namely Least Square Support Vector Machine (LSSVM) modeling optimized with Coupled Simulated Annealing (CSA) optimization technique was implemented. The constructed models are evaluated by carrying out extensive experimental data reported in open literature. Results obtained by the proposed models were compared with the corresponding experimental values. Moreover, in-depth comparative studies have been carried out between these models and all other predictive models. The results indicate that the proposed models are more robust, reliable and efficient than existing techniques for prediction of PVT properties. Results from present research show that implementation of CSA-LSSVM in crude oil PVT calculations can lead to more accurate and reliable estimation of reservoir oil PVT properties.

Assessment of asphaltene deposition due to titration technique

Due to problems followed by asphaltene deposition, which cause many remedial processes and costs, it seemed necessary to develop equations for determining asphaltene precipitation quantitatively or qualitatively. In this study a new scaling equation as a function of temperature, molecular weight, and dilution ratio (solvent) has been developed. This equation can be used to determine the weight percent of precipitated asphaltene in the presence of different precipitants (solvents). The proposed methodology utilizes least square support vector machines/regression (LSSVM/LSSVR) to perform nonlinear modeling. This paper proposes a new feature selection mechanism based on coupled simulated annealing (CSA) optimization in an attempt to tune the optimal parameters. CSA-LSSVM has the good capability of characterizing the nonlinear behavior. The performance of the proposed LSSVM algorithm is highly satisfactory and demonstrated by residuals and statistical indicator and was compared with previous works. The results showed its superiority to previous and highly dependent performance.

Prediction of principal ground-motion parameters using a hybrid method coupling artificial neural networks and simulated annealing

In this study, new models are derived to predict the peak time-domain characteristics of strong ground-motions utilizing a novel hybrid method coupling artificial neural network (ANN) and simulated annealing (SA), called ANN/SA. The principal ground-motion parameters formulated are peak ground acceleration (PGA), peak ground velocity (PGV) and peak ground displacement (PGD). The proposed models relate PGA, PGV and PGD to earthquake magnitude, earthquake source to site distance, average shear-wave velocity, and faulting mechanisms. A database of strong ground-motion recordings released by Pacific Earthquake Engineering Research Center (PEER) is used to establish the models. For more validity verification, the ANN/SA models are employed to predict the ground-motion parameters of a part of the database beyond the training data domain. ANN and multiple linear regression analyses are performed to benchmark the proposed models. Contributions of the input parameters to the prediction of PGA, PGV and PGD are evaluated through a sensitivity analysis. The ANN/SA attenuation models give precise estimations of the site ground-motion parameters. The proposed models perform superior than the single ANN, regression and existing attenuation models. The optimal ANN/SA models are subsequently converted into tractable design equations. The derived equations can readily be used by designers as quick checks on solutions developed via more in-depth deterministic analyses.

An improved hybrid topology optimization approach coupling simulated annealing and SIMP (SA-SIMP)

The Solid Isotropic Material with Penalization (SIMP) methodology has been used extensively due to its versatility and ease of implementation. However, one of its main drawbacks is that resulting topologies exhibit areas of intermediate densities which lack any physical meaning. This paper presents a hybrid methodology which couples simulated annealing and SIMP (SA-SIMP) in order to achieve solutions which are stiffer and predominantly black and white. Under a look-ahead strategy, the algorithm gradually fixes or removes those elements whose density resulting from SIMP is intermediate. Different strategies for selecting and fixing the fractional elements are examined using benchmark examples, which show that topologies resulting from SA-SIMP are more rigid than SIMP and predominantly black and white.

Coupled Simulated Annealing

We present a new class of methods for the global optimization of continuous variables based on simulated annealing (SA). The coupled SA (CSA) class is characterized by a set of parallel SA processes coupled by their acceptance probabilities. The coupling is performed by a term in the acceptance probability function, which is a function of the energies of the current states of all SA processes. A particular CSA instance method is distinguished by the form of its coupling term and acceptance probability. In this paper, we present three CSA instance methods and compare them with the uncoupled case, i.e., multistart SA. The primary objective of the coupling in CSA is to create cooperative behavior via information exchange. This aim helps in the decision of whether uphill moves will be accepted. In addition, coupling can provide information that can be used online to steer the overall optimization process toward the global optimum. We present an example where we use the acceptance temperature to control the variance of the acceptance probabilities with a simple control scheme. This approach leads to much better optimization efficiency, because it reduces the sensitivity of the algorithm to initialization parameters while guiding the optimization process to quasioptimal runs. We present the results of extensive experiments and show that the addition of the coupling and the variance control leads to considerable improvements with respect to the uncoupled case and a more recently proposed distributed version of SA.