Average Absolute Relative Error Values Research Articles

Determination of Gas–Oil minimum miscibility pressure for impure CO2 through optimized machine learning models

Minimum miscibility pressure (MMP) is one of the most important parameters for designing CO2 enhanced oil recovery (EOR) and associated storage in depleted oil reservoirs. The injection gas stream often contains a certain concentration of impurities such as N2, H2S, and CH4 depending on the source of CO2. These impurities have different effects on CO2 MMP, but there is a lack of widely accepted approaches to account for these effects on MMP calculation. In this study, a series of activities were conducted to develop a machine learning (ML)-based methodology for determining MMP for CO2 with various impurities. A database containing 234 CO2 MMP test sets with around 5000 data points was built based on the reported experimental measurements in the public domain. The database was then subgrouped by three specific criteria: CO2 concentration in the injection gas, type of impurities in the injection gas, and heavier hydrocarbon content in the oil. This subgrouping was essential to capture the impact of different factors on CO2 MMP. An ensemble ML approach with seven ML models, including random forest, adaptive boosting, light gradient boosting machine, extreme gradient boosting (XGBoost), stacking, artificial neural network, and voting regressor, was employed to calculate MMP based on the subgrouped database. The hyperparameters of these ML models were optimized by the grid search technique to minimize the relative errors between calculated and measured MMP values. The performance of each algorithm was assessed using three regression metrics: average absolute relative error (AARE), R-squared score (R2), and root mean square error (RMSE). All of these metrics exhibited satisfactory values for the optimized ML models. The average values of R2, RMSE, and AARE were 0.962, 1.571, and 4.55%, respectively, for the three subgroups, indicating a high accuracy of MMP calculations using the optimized ML models. The XGBoost model emerged as the top performer across the three metrics, with an R2 of 0.979, an AARE of 2.835%, and an RMSE of 1.183 for a dataset with 190 cases. The overall high level of accuracy confirmed the reliability of these ML models in calculating MMP for CO2 with different impurities as well as the importance of optimization in the modeling process.

High-temperature mechanical properties of additively manufactured 420 stainless steel

Martensitic stainless steels are indispensable alloys in various high stress and temperature applications such as plastic injection molds and components in steam generators. Subtractive manufacturing methods used to fabricate these parts, however, limits its functionality and performance due to design constraint of cooling channels. This limitation can be resolved by means of additive manufacturing while ensuring that acceptable high-temperature properties can be achieved. In this work, the mechanical behavior of additively manufactured 420 stainless steel (AM420SS) is explored through material constitutive modeling to determine the mathematical model that best describes its flow stress in extreme conditions. This is accomplished by subjecting the samples to hot compression under the strain rates of 0.1–1.0 s−1, and temperatures between 973–1423 K (700 °C–1150 °C) via Gleeble thermomechanical test. The experimental data were used to generate the predictive flow stress curves of constitutive models which includes Johnson-Cook, Zerilli-Armstrong, Zener-Hollomon, and Hensel-Spittel equations. Results showed that Zener-Hollomon and Hensel-Spittel models are the most accurate material constitutive equations with relatively high R values of 0.986 and 0.976, and low average absolute relative error values of 6.96% and 7.69%, respectively. The material constants derived from these models can be applied in finite element analysis simulations to assess the performance of using AM420SS parts at high temperature and strain conditions.

Hot Compression Behavior of 18Cr–11Mn–0.06Ni–0.19N High‐Mn Ultralow Nickel Duplex Stainless Steel under Large Deformation

To enhance the rolling production efficiency of high‐manganese low‐nickel duplex stainless steel (DSS) and optimize hot compression parameters, hot compression with a range of 0.01–10 s−1/850–1150 °C and deformations of 70% is used to investigate the hot deformation behavior of 18Cr–11Mn–0.06Ni–0.19 N DSS, which reveals the differences in the recrystallization mechanism of two phases and their thermoplastic properties. The results show that strain partitioning and softening occur first in the ferrite phase, with the austenite phase undergoing complex and incomplete softening. Most ferrite experiences continuous dynamic recrystallization and develops a <001>//ND texture when deformed at 0.1–10 s−1/950 °C, with higher temperatures facilitating the merging of high‐angle grain boundaries. A deformation‐induced phase transformation from δ to γ occurs in ferrite when deformed at 0.01 s−1/850–950 °C. However, austenite softening is dominated by discontinuous dynamic recrystallization, enhanced by twinning‐induced strain‐induced boundary migration when deformed at 0.1 s−1/950 °C; dynamic recovery controlled by dislocations is enhanced when deformed at 0.1 s−1/1150°C and 10 s−1/950 °C. A strain‐compensated constitutive model with an average absolute relative error value of 7.8% is established to better predict the flow stresses. Experimental DSS displays optimal hot workability in the parameter ranges of 0.1 s−1/950–1050 °C.

Modified Fields-Backofen and Zerilli-Armstrong constitutive models to predict the hot deformation behavior in titanium-based alloys

This work presents modifications for two constitutive models for the prediction of the flow behavior of titanium-based alloys during hot deformation. The modified models are the phenomenological-based Fields-Backofen and the physical-based Zerilli-Armstrong. The modifications are derived and suggested by studying the hot deformation of titanium-based alloy Ti55531. The predictability of the modified models along with the original Fields-Backofen and another modified Zerilli-Armstong models is assessed and evaluated using the well-known statistical parameters correlation coefficient (R), Average Absolute Relative Error (AARE), and Root Mean Square Error (RMSE), for the Ti55531 alloy, and validated with other two different titanium-based alloys SP700 and TC4. The results show that the modified Fields-Backofen gives the best performance with R value of 0.996, AARE value of 3.34%, and RMSE value of 5.64 MPa, and the improved version of the modified Zerilli-Armstrong model comes in the second-best place with R value of 0.992, AARE value of 3.52%, and RMSE value of 9.15 MPa for the Ti55531 alloy.

Response surface optimization and support vector regression modeling of microwave-assisted essential oil extraction from cumin seeds

The current research involved creating models using Response Surface Methodology (RSM) and Support Vector Regression (SVR) to forecast the amount of extractable essential oil that can be obtained from powdered cumin seeds. Influence of microwave power (140–280–420–560–700 W), amount of water (500–600–700–800–900 ml), duration of distillation (30–45–60–75–90 min) and soak time (15–30–45–60–75 min) on essential oil yield were investigated. Microwave Assisted Extraction (MAE) allowed higher recoveries compared to conventional Soxhlet extraction, without altering the chemical components of the extract. A five-level four FCC experimental design was developed using Minitab (15.1.20.0). A total of 31 runs were performed in microwave-assisted extraction apparatus. Experimental data obtained was then used for developing RSM and SVR models for the prediction of the yield of essential oil. The optimum conditions for maximum yield of cumin oil were given by RSM. Maximum yield of 3.4 ml (0.017 ml/g) was found at 140 W of microwave power, 500 ml of water, 90 min duration of distillation, and 15 min of soak time. In this work, epsilon SVR with RBF kernel was used. The grid search (depth-first search) methodology was applied for tuning the values of epsilon, gamma, and cost using the LIBSVM module on the MATLAB interface. The statistical parameters namely, average absolute relative error (AARE), coefficient of determination (R2), standard deviation (SD), and root mean square error (RMSE) were selected as the performance parameters. The developed SVR model was compared with the RSM model. The AARE values of 2.27% and 1.29%, R2 values of 0.86 and 0.99, SD values of 1.73 and 0.29, and RMSE values of 0.0284 and 0.0132 were obtained for RSM and SVR models respectively. It is found that SVR is more accurate and better tool for modeling of MAE process.

Modeling interfacial tension of surfactant–hydrocarbon systems using robust tree-based machine learning algorithms

Interfacial tension (IFT) between surfactants and hydrocarbon is one of the important parameters in petroleum engineering to have a successful enhanced oil recovery (EOR) operation. Measuring IFT in the laboratory is time-consuming and costly. Since, the accurate estimation of IFT is of paramount significance, modeling with advanced intelligent techniques has been used as a proper alternative in recent years. In this study, the IFT values between surfactants and hydrocarbon were predicted using tree-based machine learning algorithms. Decision tree (DT), extra trees (ET), and gradient boosted regression trees (GBRT) were used to predict this parameter. For this purpose, 390 experimental data collected from previous studies were used to implement intelligent models. Temperature, normal alkane molecular weight, surfactant concentration, hydrophilic–lipophilic balance (HLB), and phase inversion temperature (PIT) were selected as inputs of models and independent variables. Also, the IFT between the surfactant solution and normal alkanes was selected as the output of the models and the dependent variable. Moreover, the implemented models were evaluated using statistical analyses and applied graphical methods. The results showed that DT, ET, and GBRT could predict the data with average absolute relative error values of 4.12%, 3.52%, and 2.71%, respectively. The R-squared of all implementation models is higher than 0.98, and for the best model, GBRT, it is 0.9939. Furthermore, sensitivity analysis using the Pearson approach was utilized to detect correlation coefficients of the input parameters. Based on this technique, the results of sensitivity analysis demonstrated that PIT, surfactant concentration, and HLB had the greatest effect on IFT, respectively. Finally, GBRT was statistically credited by the Leverage approach.

Effect of Solution Heat Treatment on the Porosity Growth of Nickel-Based P/M Superalloys

Thermal-induced porosity (TIP) is one of the major defects in powder metallurgy (P/M) superalloys, and it seriously affects the performance of P/M superalloys. The effects of solution heat treatment on the growth of the TIP of the nickel-based P/M superalloy FGH97 were investigated. A series of solution heat treatment tests were carried out at holding temperatures ranging from 1150 to 1200 °C, with holding times ranging from 0.5 to 8 h. The results showed that the holding time, temperature, and the initial volume of porosity are the primary factors influencing porosity growth, and the volume fraction of TIPs increases by increasing the temperature or extending the holding time. The porosity growth models were constructed based on the porosity statistics combined with a nonlinear fitting method. To evaluate the accuracy of the proposed models, the correlation coefficient (R) and average absolute relative error (AARE) were calculated between the predicted and experimental values. The unbiased AARE values were 2.06% and 3.99% for the average value of TIP and the worst value of TIP, respectively, which imply that the proposed porosity growth models have greater accuracy and can be used to illustrate TIP behavior in solution heat treatment.

Modeling the solubility of light hydrocarbon gases and their mixture in brine with machine learning and equations of state

Knowledge of the solubilities of hydrocarbon components of natural gas in pure water and aqueous electrolyte solutions is important in terms of engineering designs and environmental aspects. In the current work, six machine-learning algorithms, namely Random Forest, Extra Tree, adaptive boosting support vector regression (AdaBoost-SVR), Decision Tree, group method of data handling (GMDH), and genetic programming (GP) were proposed for estimating the solubility of pure and mixture of methane, ethane, propane, and n-butane gases in pure water and aqueous electrolyte systems. To this end, a huge database of hydrocarbon gases solubility (1836 experimental data points) was prepared over extensive ranges of operating temperature (273–637 K) and pressure (0.051–113.27 MPa). Two different approaches including eight and five inputs were adopted for modeling. Moreover, three famous equations of state (EOSs), namely Peng-Robinson (PR), Valderrama modification of the Patel–Teja (VPT), and Soave–Redlich–Kwong (SRK) were used in comparison with machine-learning models. The AdaBoost-SVR models developed with eight and five inputs outperform the other models proposed in this study, EOSs, and available intelligence models in predicting the solubility of mixtures or/and pure hydrocarbon gases in pure water and aqueous electrolyte systems up to high-pressure and high-temperature conditions having average absolute relative error values of 10.65% and 12.02%, respectively, along with determination coefficient of 0.9999. Among the EOSs, VPT, SRK, and PR were ranked in terms of good predictions, respectively. Also, the two mathematical correlations developed with GP and GMDH had satisfactory results and can provide accurate and quick estimates. According to sensitivity analysis, the temperature and pressure had the greatest effect on hydrocarbon gases’ solubility. Additionally, increasing the ionic strength of the solution and the pseudo-critical temperature of the gas mixture decreases the solubilities of hydrocarbon gases in aqueous electrolyte systems. Eventually, the Leverage approach has revealed the validity of the hydrocarbon solubility databank and the high credit of the AdaBoost-SVR models in estimating the solubilities of hydrocarbon gases in aqueous solutions.

Comparison of Modified Johnson–Cook Model and Strain–Compensated Arrhenius Constitutive Model for 5CrNiMoV Steel during Compression around Austenitic Temperature

Isothermal compression behaviors of 5CrNiMoV steel were studied at temperatures of 870, 800, 750, and 700 °C, with strain rates of 0.001, 0.005, 0.01, 0.05, and 0.1 s−1, the compression temperatures 870 and 800 °C are above Ac3, as well as 750 and 700 °C below Ac3 temperature. The Modified Johnson–Cook (MJC) model and the Strain–Compensated Arrhenius (SCA) model were employed to demonstrate the relationship between the flow stress and the compression parameters. The correlation coefficient (R) and average absolute relative error (AARE) between the calculational and experimental flow stress were used to evaluate the accuracy of the two models. The results show that the effect of dynamic softening on flow stress is much more significant at higher temperatures and lower strain rates, while this effect is not obvious when the strain rate exceeds 0.005 s−1 with the temperature below Ac3. The MJC model has a good accuracy close to the reference conditions (0.001 s−1 and 700 °C), and it is suitable to predict the plastic behavior when the flow stress is lower than 200 Mpa. The unbiased AARE values were 6.82 and 5.71 for MJC model and SCA model, respectively, which implied the SCA model has a higher accuracy than the MJC model. The SCA model was believed to be capable of being used to illustrate the thermomechanical behavior of 5CrNiMoV tool steel in a wide range of plastic deformation conditions.

A Comparative Study on Hot Deformation Behaviors of Niobium Microalloyed Low‐Carbon and Medium‐Carbon Steels by Physical Constitutive Analysis

Hot compression tests are performed on low‐carbon (LC) and medium‐carbon (MC) niobium microalloyed steels at temperatures of 900–1100 °C and strain rates of 0.01–10 s−1. The constitutive equations are studied by a physical method based on creep theory considering the relationship between the self‐diffusion coefficient, Young's modulus, and temperature. It is found that carbon addition in niobium microalloyed steels shows an obvious softening effect. The physical constitutive analysis indicates that the deformation mechanism of MC steel is the slide and climb of dislocation; however, other deformation mechanisms may occur in LC steel. The accuracy of the physical constitutive equations is quantified by employing correlation coefficient (R) and average absolute relative error (AARE). For LC steel, the R value of the equation containing exponent 5 and exponent n is 0.98 and 0.99; the AARE value is 9.06% and 4.15%, respectively, and the accuracy of the latter equation is significantly higher. For MC steel, the R value of the two equations is the same and equal to 0.99, the AARE value is 5.96% and 4.91%, respectively, the accuracy of the equations is quite close to each other. The accuracy analysis is also in reasonable agreement with the speculation of the deformation mechanism.

Application of copper as a pulse shaper in SHPB tests on brittle materials- experimental study, constitutive parameters identification, and numerical simulations

Generally, optimum dimensions of a pulse shaper (PS) in SHPB tests on brittle materials like rock are chosen by performing numerous experiments with different sizes of PS. The dimensions that give the best force equilibrium and constant strain rate loading in the specimen are selected. Experimental efforts can be saved if the constitutive behavior of the PS can be simulated numerically. In the current work, the behavior of pure copper is investigated at various strain rates (up to ∼4000 /s) and temperatures (till 250 °C) in as-received and annealed conditions. The mechanical behavior of copper is assessed in terms of strain hardening rate, strain rate sensitivity, and temperature sensitivity. It is found that the strain hardening rate in annealed samples is almost twice that of as-received ones. A strain rate increase strengthens copper while a temperature rise weakens it. Fractographs of quasi-static samples tested at various temperatures indicate that all samples undergo ductile failure. The experimental data is used to derive constitutive parameters of the Johnson-Cook (JC) model using a multi-objective optimization method. It is concluded that the modified JC model fits better than the original form, as substantiated by higher R2 and lower AARE (average absolute relative error) values. The derived parameters are successfully applied to model the dynamic behavior of the copper PS in SHPB tests on Kota sandstone samples. It is observed that the experimental and numerical incident pulses are qualitatively similar, and the Russell comprehensive error is well within 0.15 for all the cases. It is concluded that investigating copper behavior till ∼4000 /s yields sufficiently accurate results to predict PS behavior during SHPB tests, although the actual strain rates experienced by the pulse shapers are higher. A sensitivity and probabilistic analysis are carried out to understand the effect of parameters variations on the incident rise time.

Hot Deformation Behavior of the 25CrMo4 Steel Using a Modified Arrhenius Model.

25CrMo4 steel is widely used in the manufacturing of high-speed train axles due to its excellent mechanical properties. The purpose of this study is to develop an accurate modified constitutive model to describe the hot deformation behavior of the steel. Isothermal compression experiments were performed at different strain rates (0.01, 0.1, 0.5, and 1 s−1) and different temperatures (950, 1000, 1050, and 1100 °C) using a Gleeble-3800 thermal simulator. The microstructure after hot deformation was observed by the electron backscatter diffraction (EBSD), and the effects of temperature and strain rate were analyzed. The results showed that the coupling effect of temperature and strain rate on the dislocation density led to the change in the shape of the true stress–strain curve and that dynamic recovery (DRV) and dynamic recrystallization (DRX) caused the macroscopic softening phenomenon, with DRX being the main mechanism. Based on the true stress–strain curves, the strain-compensated Arrhenius constitutive model was calibrated. To improve prediction ability, a modified Arrhenius constitutive model was proposed, in which the temperature and strain rate coupling correction functions were incorporated. The original, modified Arrhenius models were evaluated according to the absolute relative error (ARE), the average absolute relative error (AARE), and the correlation coefficient (R2). Compared with the original model, the modified Arrhenius model has a higher prediction accuracy, with the ARE value mostly below 4%, the AARE value of 1.91%, and the R2 value of 0.9958.

Application of Artificial Intelligence Techniques for the Determination of Groundwater Level Using Spatio-Temporal Parameters.

Increasing the depth of mining leads to the location of the mine pit below the groundwater level. The entry of groundwater into the mining pit increases costs as well as reduces efficiency and the level of work safety. Prediction of the groundwater level is a useful tool for managing groundwater resources in the mining area. In this study, to predict the groundwater level, multilayer perceptron, cascade forward, radial basis function, and generalized regression neural network models were developed. Moreover, four optimization algorithms, including Bayesian regularization, Levenberg–Marquardt, resilient backpropagation, and scaled conjugate gradient, are used to improve the performance and prediction ability of the multilayer perception and cascade forward neural networks. More than 1377 data points including 12 spatial parameters divided into two categories of sediments and bedrock (longitude, latitude, hydraulic conductivity of sediments and bedrock, effective porosity of sediments and bedrock, the electrical resistivity of sediments and bedrock, depth of sediments, surface level, bedrock level, and fault), and besides, 6 temporal parameters are used (day, month, year, drainage, evaporation, and rainfall). Also, to determine the best models and combine them, 165 extra validation data points are used. After identifying the best models from the three candidate models with a lower average absolute relative error (AARE) value, the committee machine intelligence system (CMIS) model has been developed. The proposed CMIS model predicts groundwater level data with high accuracy with an AARE value of less than 0.11%. Sensitivity analysis indicates that the electrical resistivity of sediments had the highest effect on the groundwater level. Outliers’ estimation applying the Leverage approach suggested that only 2% of the data points could be doubtful. Eventually, the results of modeling and estimating groundwater level fluctuations with low error indicate the high accuracy of machine learning methods that can be a good alternative to numerical modeling methods such as MODFLOW.

Constitutive Model Parameter Identification Based on Optimization Method and Formability Analysis for Ti6Al4V Alloy.

Titanium alloy is widely applied in aerospace, medical, shipping and other fields due to its high specific strength and low density. The purpose of this study was to analyze the formability of Ti6Al4V alloys at elevated temperatures. An accurate constitutive model is the basic condition for accurately simulating the plastic forming of materials, and it is an important basis for optimizing the parameters of the hot forging forming process. In this study, the optimization algorithm was used to accurately identify the high-temperature constitutive model parameters of Ti6Al4V titanium alloy, and the hot working diagram was established to optimize the hot forming process parameters. The optimal forming conditions of Ti6Al4V titanium alloy are given. Ti6Al4V alloy was subjected to high-temperature compression tests at 800–1000 °C and at strain rates of 0.01–5 s−1 on a Gleeble-1500D thermal/mechanical simulation machine. Each parameter of the Hansel–Spittel constitutive model was taken as an independent variable, and the accumulated error between the stress calculated by the constitutive model and the stress obtained by experimentation was used as an objective function. Based on response surface methodology, an inverse optimization method for identifying the parameters of the high-temperature constitutive model of Ti6Al4V alloy is proposed in this paper. An orthogonal test design was adopted to obtain sample point data, and a third-order response surface approximate model was established. The genetic algorithm (GA) was applied to reversely optimize the parameters of the constitutive model. To verify the accuracy of the optimized constitutive model, the average absolute relative error (AARE) and correlation coefficient (R) were used to evaluate the reliability of optimized constitutive model. The R value of the model was 0.999, and the AARE value was 0.048, respectively, indicating that the established high-temperature constitutive model for Ti6Al4V alloy has good calculation accuracy. The flow stress behavior of the material could be accurately delineated. Meanwhile, in order to study the formability of Ti6Al4V alloy, the hot processing map of the alloy, based on a dynamic material model, was established in this paper. The optimum hot working domains of the Ti6Al4V alloy were determined within 840–920 °C/0.01–0.049 s−1 and 940–980 °C/0.11–1.65 s−1; the hot processing map was verified in combination with the microstructure, and the fine and equiaxed grains and a large amount of β phase could be found at 850 °C/0.01 s−1.

Robust and general predictive models for condensation heat transfer inside conventional and mini/micro channel heat exchangers

To design, scale-up, and optimize the two-phase flow heat exchangers, accurate information about the heat transfer coefficient (HTC) is required. Here, an extensive database including 11,128 experimental data points from 80 sources, covering 37 condensing fluids over a wide range of operating conditions, were gathered for developing general and accurate models to forecast the HTC in mini/micro and conventional channel heat exchangers. The conventional intelligent approaches, namely, gaussian process regression (GPR), radial basis function (RBF), and hybrid radial basis function (HRBF) and also the least square fitting method (LSFM) were utilized for development of the new models. Firstly, the gathered data were used in evaluation of earlier two-phase HTC models, and these models showed average absolute relative error (AARE) values higher than 29%. Thus, more accurate models are necessary for these systems. Based on Pearson’s correlation analysis, eight dimensionless groups were selected as the most important input parameters. The GPR model showed the best results in estimation of Nu number for the test process with AARE of 4.50%. However, RBF and HRBF models estimated the HTC with AARE values 19.41% and 24.36% for testing data points, respectively. In addition, a general explicit correlation was developed by the least square fitting method (LSFM) for estimating the HTC, with acceptable results at an AARE of 23.40% for all analyzed data. The performance of the new general correlation, GPR and the previous models were assessed for different channel sizes, flow regimes, fluid types, and operating conditions. The new models and well-known earlier ones were examined with about 372 additional data samples from four new independent sources. The proposed models showed excellent results for prediction of HTC. Finally, a sensitivity analysis was studied based on the experimental data and the outcomes of the LSFM.

Applying intelligent approaches to estimate the removal efficiency of heat stable salts from lean amine via electrodialysis

Intelligent approaches based on radial basis function (RBF) neural networks and genetic programming (GP) were used to establish accurate models for estimating the removal efficiency of heat stable salts from lean amine via electrodialysis. The operating time, current intensity, membrane types, HSS concentration, and kind of concentrated solution were lumped into dimensionless groups. The groups with the most influence were selected based on the Pearson's correlation matrix for the models’ inputs. The RBF model showed an excellent agreement with real data with average absolute relative error (AARE) of 1.90% and R2 of 99.21%. Then, an explicit empirical correlation was developed for the removal efficiency using the GP technique, which yielded AARE and R2 values of 5.74% and 96.35%, respectively. The performance of the GP and RBF models for estimating the removal efficiency of different ions for different types of membranes and operating conditions were assessed and reasonable results were achieved. Finally, to identify the most effective dimensionless groups to describe the removal efficiency, a sensitivity analysis based on the developed GP and RBF models was accomplished.

An improved generic Johnson-Cook model for the flow prediction of different categories of alloys at elevated temperatures and dynamic loading conditions

This paper presents a generic model for material flow prediction based on the well-known Johnson-Cook model. The model is developed to precisely predict the flow behavior of various categories of alloys. The coupled effects between strain, strain rate, and temperature were taken into consideration. The proposed model is developed and assessed using the hot deformation data of four different categories of alloys; with four different base elements. Besides, the data of two different alloys under dynamic loading are used for assessment. The proposed modification is compared to the original Johnson-Cook model and three different modifications for the Johnson-Cook model; a model for the hot deformation and two models for the dynamic loading, giving that all alloys have different base elements. The well-known statistical correlation coefficient parameters (R), Average Absolute Relative Error (AARE), and Root Mean Squared Error (RMSE) are used to assess the prediction accuracy of the proposed model. The results show that the proposed model outperforms the other addressed models as it can precisely predict the flow behavior of the six alloys, with the highest R-value of 0.9940±0.0126 and lowest AARE and RMSE values of 1.95±1.08 % and 4.15±3.47 MPa, respectively.

Comparative Study of Accurate Descriptions of Hot Flow Behaviors of BT22 Alloy by Intelligence Algorithm and Physical Modeling

The nonlinear flow behaviors of BT22 alloy were investigated by thermal simulation experiments at different temperature and strain rates. Taking the experimental stress-strain data as samples, the support vector regression (SVR) model and back propagation artificial neural network (BPANN) model were established by cross-validation (CV) method to describe the nonlinear flow behaviors of BT22 alloy. Genetic algorithm (GA) was used to optimize the parameters of the SVR model and establish the GA-SVR model. At the same time, the physical model optimized by GA algorithm is compared with the machine learning model. Average absolute relative error (AARE), absolute relative error (ARE), and correlation coefficient (R) were used to evaluate the predictive ability of the four models. The results show that the order of model accuracy and generalization ability is GA-SVR > BPANN > SVR > physical model. The AARE value of the GA-SVR model is 1.5752%, and the R value is as high as 0.9984, which can accurately predict the flow behaviors of BT22 alloy. According to the GA-SVR model, the flow behaviors under other conditions could be predicted to expand the experimental stress-strain data and avoid a large number of artificial tests.

Development and analyses of data-driven models for predicting the bed depth profile of solids flowing in a rotary kiln

Soft computing data-driven modeling (DDM) techniques have attracted the attention of many researchers across the globe as they do not require deep knowledge of the complex physical process. In the present research, data-driven based models have been developed using support vector regression (SVR), multilayer perceptron neural network (MLP), radial basis function neural network (RBFNN) and general regression neural networks (GRNN) techniques for predicting the bed depth profile of solids flowing in a rotary kiln. The performances of the developed models were compared and evaluated against the experimental results in terms of statistical measures such as coefficient of determination (R2), and average absolute relative error (AARE). The obtained results and findings from this research have revealed that data-driven models can predict the bed depth profile of solids flowing in a rotary kiln quite accurately. The SVR-based model exhibited the lowest AARE value of 1.72% and highest R2 value of 0.9981 while GRNN, RBFNN, and MLP models gave corresponding values of AARE as 3.69%, 55.13%, 98.15% and those of R2 as 0.9898, 0.0052 and 0.0081, respectively. Moreover, the developed DDM-based models i.e. GRNN, RBFNN, and MLP models overcame the limitations of the existing solutions which involved iterative numerical procedure entailing high degree of computational complexity.

Evolving artificial intelligence techniques to model the hydrate-based desalination process of produced water

Abstract In this study, two artificial intelligence models based on an adaptive neuro-fuzzy inference system (ANFIS) and a support vector machine (SVM) technique have been successfully developed to predict the desalination efficiency of produced water through a hydrate-based desalination treatment process. A genetic algorithm as an evolutionary optimization method has been used to determine the optimal values of SVM model coefficients. To this end, compressed natural gas and CO2 hydrate formation experiments were carried out, and the desalination efficiency of produced water was measured and utilized for model training and validation. After model development, graphical and statistical analysis approaches have been applied to evaluate the performance of suggested models by a comparison of model predictions with measured experimental data. For the ANFIS model, the coefficient of determination (R2) and average absolute relative error (AARE) are 0.9927 and 0.58%, respectively. The values of AARE and R2 for the SVM model are obtained 0.35% and 0.9985, respectively. These statistical criteria confirm excellent accuracy and robustness of intelligent models in predicting the desalination efficiency of produced water through the hydrate-based desalination treatment process. Furthermore, the Leverage statistical technique has been carried out to define the outliers. The obtained results demonstrate that all experimental data are reliable and both ANFIS and SVM models are statistically valid.