Cascade Forward Neural Network Research Articles

Improved prediction accuracy for compressive strength of recycled aggregate concrete using optimization-based algorithms and cascade forward neural network

This study proposed a data driven approach to predict the compressive strength (CS) of recycled aggregate concrete (RAC) for sustainable construction using an elite single genetic optimization algorithm-based cascade forward neural network (ESGA-CFNN) model. It was applied to 272 RAC samples under different conditions and compositions focusing on key parameters for CS prediction: water-to-cement ratio (WCR), water absorption (WA), recycled coarse aggregate (RCA) density, fine aggregate (FA) density, naturally occurring coarse aggregate (NCA) density and water-to-total material ratio (WTMR). These parameters were used to develop the ESGA-CFNN model which was then evaluated for its performance. To compare the ESGA-CFNN model, two other models were developed and compared: particle swarm optimization-based CFNN (PSO-CFNN) and artificial bee colony-based CFNN (ABC-CFNN). K-fold cross-validation was used during model development to prevent overfitting. Results showed that ESGA-CFNN model performed better with an RMSE (root-mean-squared error) of 1.144, R2 (determination coefficient) of 0.991 and a10-index of 1.000. ABC-CFNN model had an RMSE of 1.434, R2 of 0.987 and a10-index of 0.982 while PSO-CFNN had an RMSE of 1.561, R2 of 0.984 and a10-index of 0.982. Practical validation with 6 RAC samples confirmed the real world applicability of these models. The findings of this study showed that the proposed ESGA-CFNN model is important for quality control in RAC production and optimizing mix designs to achieve required compressive strength to meet standards and reduce cost and increase sustainability in concrete construction. This study introduces a novel hybrid approach combining ESGA-CFNN, PSO-CFNN, and ABC-CFNN algorithms for accurately predicting the compressive strength of RAC. These models outperform traditional methodologies by offering enhanced predictive accuracy and generalization capability, especially in complex, real-world datasets.

Modeling drilling fluid density at high-pressure high-temperature conditions using advanced machine-learning techniques

The management of oil and gas wells under high-pressure high-temperature (HPHT) conditions is a significant challenge in the oil industry, leading to increased operational and maintenance costs. One method for calculating the density of drilling fluids involves the use of sensors placed at the bottom of the well, which measure the fluid density in real time. Despite their accuracy, these sensors are costly and perform suboptimal under HPHT conditions. In the present study, seven robust machine-learning models, namely generalized regression neural networks (GRNN), wavelet neural networks (WNN), and cascade forward neural networks (CFNN) combined with two different training algorithms containing Levenberg-Marquardt and Bayesian regression (CFNN-LM and CFNN-BR), respectively, and support vector regression (SVR) models combined with three optimization algorithms containing (farmland fertility algorithm (FFA) grasshopper optimization algorithm (GOA), and particle swarm optimization (PSO) were utilized to predict mud density at HPHT conditions. Moreover, mathematical models were developed using a group data management algorithm (GMDH) to predict this parameter. WNN, SVR model combined with (PSO, GOA, FFA), CFNN, and GRNN algorithms have better adaptability than other artificial intelligence models. These models can improve performance by changing the data and different situations. Also, these models can better resistance against noisy data, which means that they can show good performance compared to uneven and discontinuous data. To this aim, using 986 experimental data of drilling fluid at HPHT, including 117 water-base fluid data, 440 colloidal gas aphron fluid data, 219 oil-base fluid data, and 210 synthetic fluid data the drilling fluid density was modeled. Input parameters including temperature, pressure, initial mud density, and volume fraction of solid content were used for modeling. The results showed that the volume fraction of the solid content as an input variable improves the accuracy of intelligent models in this work. Based on average absolute percent relative error (AAPRE) the best outcomes for predicting the mud density without the input parameter of solid fraction were observed by the PSO-SVR model with an AAPRE of 0.5569%. Following, the GOA-SVR, WNN, and FFA-SVR delivered the next best results with AAPRE of 0.5860%, 0.5786%, and 0.6148%, respectively. Moreover, with the input parameter of solid fraction, the WNN model demonstrated the highest accuracy among all models, achieving an AAPRE of 0.2475%. Also, the correlation developed using the GMDH method based on input parameters of initial mud density, temperature, and pressure yielded satisfactory outcomes and can provide precise and swift estimates. Eventually, Sensitivity analysis revealed that initial mud density had the highest impact on mud density. Lastly, the leverage analysis indicates that most data points are trustworthy based on experimentation, with only a few falling outside the PSO-SVR and WNN models' scope.

Machine learning-powered performance monitoring of proton exchange membrane water electrolyzers for enhancing green hydrogen production as a sustainable fuel for aviation industry

Aviation is a major contributor to transportation carbon emissions but aims to reduce its carbon footprint. Sustainable and environmentally friendly green hydrogen fuel is essential for decarbonization of this industry. Using the extremely low temperature of liquid hydrogen in aviation sector unlocks the opportunity for cryo-electric aircraft concept, which exploits the advantageous properties of superconductors onboard. A significant barrier for green hydrogen adoption relates to its high cost and the immediate need for large-scale production, which Proton Exchange Membrane Water Electrolyzers (PEMWE) can address through optimal dynamic performance, high lifetimes, good efficiencies, and importantly, scalability. In PEMWE the cell is a crucial component that facilitates the electrolysis process and consists of a polymer membrane and electrodes. To control the required production rate of hydrogen, the output power of cell should be monitored which usually is done by measuring the cell’s potential and current density. In this paper, five different machine learning (ML) models based on different algorithms have been developed to predict this parameter. Findings of the work highlight that the model based on Cascade-Forward Neural Network (CFNN) is investigated to accurately predict the cell potential of PEMWE under different anodic material and working conditions with an accuracy of 99.998 % and 0.001884 in terms of R2 and root mean square error, respectively. It can predict the cell potential with a relative error of less than 0.65 % and an absolute error of below 0.01 V. The Standard deviation of 0.000061 for 50 iterations of stability analysis indicated that this model has less sensitivity to the random selection of training data. By accurately estimating different cell’s output with one model, and considering its ultra-fast response, CFNN model has the potential to be used for both monitoring and the designing purposes of green hydrogen production.

Investigation on the heat transfer estimation of subcooled liquid hydrogen for transportation applications using intelligent technique

Hydrogen, as a sustainable fuel for transportation sector as well as some other heavy industries, should be stored or transferred in liquid phase. To prevent the evaporation of hydrogen in the tanks and the consequent hazards, it is common to lower its temperature below the saturation point known as subcooling of the fluid. When the subcooled liquid hydrogen flows through pipes with higher wall temperatures, the hydrogen can evaporate and flow boiling occurs, resulting in a complex heat transfer behavior of the fluid. In this study, we propose an advanced intelligent model based on the Cascade Forward Neural Network (CFNN) to predict Heat Transfer Coefficient (HTC) of subcooled liquid hydrogen under various flow conditions. Our model leverages on a dataset that is nearly twice the size of the largest previously used in the literature, significantly enhancing its generalization, reliability, and accuracy. The proposed CFNN model incorporates 19 input features, including crucial dimensionless parameters, orientation and thermodynamic condition of the fluid, and characteristics of the pipe that hydrogen flow through to accurately estimate HTC in unseen fluid conditions. The results demonstrate that the proposed model achieves an R-squared value of 0.9978, substantially outperforming the best empirical methods, which have an R-squared value of only 0.8058 for the same testing data. This innovative approach provides a highly accurate and reliable tool for predicting HTC, offering significant improvements over conventional methods and contributing valuable insights for real-world applications in the energy and transportation sectors.

Short-term load forecasting: cascade intuitionistic fuzzy time series—univariate and bivariate models

AbstractShort-term load forecasting (STLF) is essential for developing reliable and sustainable economic and operational strategies for power systems. This study presents a forecasting model combining cascade forward neural network (CFNN) and intuitionistic fuzzy time series (IFTS) models for STLF. The proposed cascading intuitionistic fuzzy time series forecasting model (C-IFTS-FM) offers the advantage of CFNN using the links of both linear and nonlinear to model fuzzy relations between inputs and outputs. Moreover, it offers a more reliable and realistic approach to uncertainty, taking notice of also the degree of hesitation. C-IFTS-FM works in univariate structure when it uses only hourly load data, and in bivariate structure when it uses hourly load data and hourly temperature time series together. The conversion of time series into IFTS is realized with intuitionistic fuzzy c-means (IFCM). Thus, the membership and non-membership values for each data point are produced. In modelling process, membership and non-membership values, in addition to actual lagged observations, are used as input of the CFNNs. The effectiveness of C-IFTS-FM on test sets for both structures was discussed comparatively via different error criteria, in addition, the convergence time was examined, and also the fit of forecasts and observations was presented with different illustrations. Among different combinations of hyperparameters, in the best case, approximately 86% better accuracy is achieved than the best of the others, while even in the case of the worst of hyperparameters combination, the accuracy was improved by over 20% for the PSJM data sets. For HEXING, CHENGNAN, and EUNITE data sets, these progress rates reached approximately 90% in the best case.

Multi-objective Approach for Dynamic Economic Emission Dispatch Problem Considering Power System Reliability and Transmission Loss Prediction Using Cascaded Forward Neural Network

This study addresses the significant problem of Dynamic Economic Emission Dispatch (DEED), a critical consideration in power systems from both economic and environmental protection viewpoints. Reliability stands as another vital facet, impacting maintenance and operation perspectives. The integration of Artificial Neural Network (ANN)-based transmission loss prediction into the DEED model is also essential to address specific limitations and enhance the overall performance of the dispatch process. Traditionally, the DEED model relies on a single B-loss coefficient to estimate transmission losses. While this approach simplifies calculations, it fails to account for the significant variations in demand that occur throughout the dispatch period and it leads to inaccuracies in loss prediction, especially in dynamic environments. Using a single coefficient, the model cannot adequately capture the complex, non-linear relationships between power generation, load, and transmission losses under different operating conditions. To overcome this limitation, this study introduces an ANN-based loss prediction method integrated into the DEED model and uses trained ANN to replace the process of finding B-loss coefficients during each dispatch period. This paper also introduces a strategy leveraging the multi-objective northern goshawk optimizer algorithm, characterized by a non-dominated sorting and crowding distance mechanism, to enhance DEED considerations incorporating reliability (DEEDR). This novel algorithm improves the solution space effectively, maintains high population diversity and enables an even distribution of individuals sharing the same rank in the objective space. The fundamental objective of this study is to balance fuel cost, emission, and system reliability in power system operations. Compared with a few existing multi-objective optimization algorithms, this study demonstrates superior performance in generating a series of non-dominated solutions. The experimental results highlight its competitive and potential as an efficient tool in the DEED and DEEDR problems, promising a synergistic coordination of economy, environmental protection, and system reliability benefits in power system management.

Analyzing brain-activation responses to auditory stimuli improves the diagnosis of a disorder of consciousness by non-linear dynamic analysis of the EEG

Although auditory stimuli benefit patients with disorders of consciousness (DOC), the optimal stimulus remains unclear. We explored the most effective electroencephalography (EEG)-tracking method for eliciting brain responses to auditory stimuli and assessed its potential as a neural marker to improve DOC diagnosis. We collected 58 EEG recordings from patients with DOC to evaluate the classification model’s performance and optimal auditory stimulus. Using non-linear dynamic analysis (approximate entropy [ApEn]), we assessed EEG responses to various auditory stimuli (resting state, preferred music, subject’s own name [SON], and familiar music) in 40 patients. The diagnostic performance of the optimal stimulus-induced EEG classification for vegetative state (VS)/unresponsive wakefulness syndrome (UWS) and minimally conscious state (MCS) was compared with the Coma Recovery Scale-Revision in 18 patients using the machine learning cascade forward backpropagation neural network model. Regardless of patient status, preferred music significantly activated the cerebral cortex. Patients in MCS showed increased activity in the prefrontal pole and central, occipital, and temporal cortices, whereas those in VS/UWS showed activity in the prefrontal and anterior temporal lobes. Patients in VS/UWS exhibited the lowest preferred music-induced ApEn differences in the central, middle, and posterior temporal lobes compared with those in MCS. The resting state ApEn value of the prefrontal pole (0.77) distinguished VS/UWS from MCS with 61.11% accuracy. The cascade forward backpropagation neural network tested for ApEn values in the resting state and preferred music-induced ApEn differences achieved an average of 83.33% accuracy in distinguishing VS/UWS from MCS (based on K-fold cross-validation). EEG non-linear analysis quantifies cortical responses in patients with DOC, with preferred music inducing more intense EEG responses than SON and familiar music. Machine learning algorithms combined with auditory stimuli showed strong potential for improving DOC diagnosis. Future studies should explore the optimal multimodal sensory stimuli tailored for individual patients.Trial registration: The study is registered in the Chinese Registry of Clinical Trials (Approval no: KYLL-2023-414, Registration code: ChiCTR2300079310).

Predicting bentonite swelling pressure: optimized XGBoost versus neural networks

The swelling pressure of bentonite and bentonite mixtures is critical in designing barrier systems for deep geological radioactive waste repositories. Accurately predicting the maximum swelling pressure is essential for ensuring these systems' long-term stability and sealing characteristics. In this study, we developed a constrained machine learning model based on the extreme gradient boosting (XGBoost) algorithm tuned with grey wolf optimization (GWO) to determine the maximum swelling pressure of bentonite and bentonite mixtures. A dataset containing 305 experimental data points was compiled, including relevant soil properties such as montmorillonite content, liquid limit, plastic limit, plasticity index, initial water content, and soil dry density. The GWO-XGBoost model, incorporating a penalty term in the loss function, achieved an R2 value of 0.9832 and an RMSE of 0.5248 MPa in the testing phase, outperforming feed-forward and cascade-forward neural network models. The feature importance analysis revealed that dry density and montmorillonite content were the most influential factors in predicting maximum swelling pressure. While the developed model demonstrates high accuracy and reliability, it may have limitations in capturing extreme values due to the complex nature of bentonite swelling behavior. The proposed approach provides a valuable tool for predicting the maximum swelling pressure of bentonite-based materials under various conditions, supporting the design and analysis of effective barrier systems in geotechnical engineering applications.

Intelligent surrogate model of a high-temperature superconducting cable for reliable energy delivery: short-circuit fault performance

High-temperature superconducting (HTS) cables are promising solutions for electric power transmission of renewable energy resources, where their fault performance study is vital to avoid power interruptions in the grid. In this study, a fast intelligent surrogate model was presented to estimate the fault performance of a 22.9 kV/50 MW HTS cable to make fast fault performance analysis of the HTS cables possible during the design stage. Different fault scenarios were considered under different fault durations, fault resistances, and types of faults. Then, the fault energy, fault current, fault type, fault duration, and fault resistance were fed into the surrogate model as inputs. The outputs were the temperature of the rare-earth barium copper oxide (ReBCO) tapes, the former temperature, the ReBCO layer current, and the total resistance of each phase. For surrogate modelling, cascade forward neural networks (CFNNs) were used. The results show that the CFNN-based model estimated the fault performance of the cable with an average accuracy of 99.1%. Finally, the impact of considering fault energy, fault current, and both, as the inputs of the models, on the final accuracy were explored. The results show that by considering the fault energy, the accuracy of the surrogate model can be increased.

Development of an artificial neural network (ANN) for the prediction of a pilot scale mobile wastewater treatment plant performance

Productive activities such as pig farming are a fundamental part of the economy in Mexico. Unfortunately, because of this activity, large quantities of wastewater are generated that have a negative impact in the environment. This work shows an alternative for treating piggery wastewater based on advanced oxidation processes (Fenton and solar photo Fenton, SPF) that have been probed successfully in previous works. In the first stage, Fenton and SPF were carried out on a laboratory scale using a Taguchi L9-type experimental design. From the statistical analysis of this design, the operating parameters: pH, time, hydrogen peroxide concentration [H2O2], and iron ferrous concentration [Fe2+] that maximize the response variables: Chemical Oxygen Demand (COD), Total Organic Carbon (TOC), and color were chosen. From these, a cascade forward neural network was implemented to establish a correlation between data from the variables to the physicochemical parameters to be measure being that a great fit of the data was obtained having a correlation coefficient of 0.99 which permits to optimize the pollutant degradation and predict the removal efficiencies at pilot scale but with a projection to a future industrial scale. A relevant result, it was found that the optimal values for maximizing the removal of physicochemical parameters were pH = 3, time = 60 min, H2O2/COD = 1.5 mg L−1, and H2O2/Fe2+ = 2.5 mg L−1. With these conditions degradation percentages of 91.44%, 47.14%, and 97.89% for COD, TOC, and color were obtained from the Fenton process, while for SPF the degradation percentage increased moderately. From the ANN analysis, the possibility to establish an intelligent system that permits to predict multiple results from operational conditions has been achieved.

Security monitoring via sound analysis and voice identification with artificial intelligence

The article demonstrates the possibility of monitoring user access through authentication based on voice profiles using the means of Artificial Intelligence. A two-stage approach is proposed for sound analysis and voice recognition using Feed-Forward Neural Networks (FFNNs) and Cascade-Forward Neural Networks (CFNNs). Seven test voice profiles were pre-processed to extract quantitative sound features. The procedure involves registration of a set of sound parameters concerning three categories, respectively, for all audio and acoustic measurements in the entire sound spectrum, measurements up to and above 100 dBA. The neural architectures were trained with Scaled Gradient Descent (SCG) and Levenberg Marquardt (LM) algorithms, using different transfer functions in the output structural layers. In the initial phase of neural training, the entire sound spectrum of registered indicators was used, and high levels of Accuracy around 90.0% were reached. Subsequently, steps were taken to reduce the informative features when searching for similar levels of accuracy in order to limit the necessary computational procedures in neural training, but maintain the threshold of successful user authentication. In the analysis of neural performance, in addition to accuracy, additional criteria were used, namely Mean-Squared Error (MSE) and Root Mean Squared Error (RMSE). About the achieved and analyzed results, a synthesis was conducted of a set of four informative features with the highest significance, respectively LAE (A-weighted, sound exposure level), Laeq (A-weighted, equivalent sound level), LAF (A-weighted, fast time-constant, sound level) and LAS (A-weighted, slow time constant response, sound level). In the course of subsequent neural training processes, unsuitability was found when using the Log-sigmoid activation type with greatly underestimated accuracy readings and errors below 58.0% and above thresholds of 0.2300 and 0.4800. Positive performance indicators of voice recognition were achieved with Softmax and Hyperbolic tangent sigmoid activations in SCG and LM training procedures in levels of accuracy of 98.7 % and 96.1 % at FFNN models. Successful correct recognition of the test voice profiles on access and security personalization with a quantitative equivalent of 100.0 % accuracy was achieved in the Linear transfer function for Cascade-Forward Neural Networks. The proposed method and the synthesized neural models in the research can be used as units and modules in access control systems with biometric diagnostics and intelligent recognition of employees in company departments to electronically store classified information and physical access control.

Hybrid machine learning system based on multivariate data decomposition and feature selection for improved multitemporal evapotranspiration forecasting

Evapotranspiration is an essential component of the hydrological cycle. Forecasting the reference crop evapotranspiration (ETo) using a reliable and generalized framework is crucial for agricultural operations, especially irrigation. This study was aimed at evaluating the performance of a hybrid system including the K-Best selection (KBest), multivariate variational mode decomposition (MVMD), and Machine learning (ML) models for 1-, 3-, 7-, and 10-day-ahead forecasting of the daily ETo in twelve stations of California. The analysis covered a span of 20 years, from 2003 to 2022. Three stand-alone ML models, namely Cascade Forward Neural Network (CFNN), Extreme Learning Machine (ELM), and Bagging Regression Tree (BRT) are used and were integrated with various preprocessing techniques to construct three hybrid models, i.e., MVMD-KBest-CFNN, MVMD-KBest-ELM, and MVMD-KBest-BRT. According to the results obtained in the testing phase, averaged across all stations, all three stand-alone models (CFNN, ELM, and BRT) yielded similar outcomes. In contrast, the hybrid models exhibited significantly enhanced performances compared with the standalone models, and MVMD-KBest-CFNN and MVMD-KBest-ELM models outperformed MVMD-KBest-BRT model. The BRT-based models were vulnerable to overfitting. The performance of the best models is superior compared to similar existing studies. Examining the variations across stations, it was found that the stations located further from the coast and in arid regions could be susceptible to prediction errors and necessitate more attention.

Cascade computational model for prediction impact of transient depth change on combustion parameters of certain timber species under continuous heating rate

The purpose of this research is to examine the effects of depth variation and exposure duration on the combustion characterization of the Maple wood species (Acer platanoides L.) and Walnut (Juglans regia L.) according to ISO 5660–1 in case of exposing the specimens to a constant heating rate 50 kWm−2. The delivered experimental data by cone calorimeter apparatus is used to train, test, and validate the optimized cascade forward artificial neural network with topology 2–8–8–12–2. This structure is simulated to predict the target parameters: the heat release rate (HRR) and the mass burning rate (MBR). With real systems, it is possible to achieve a reasonable performance and the closest approximations for the cascade model. The experimental combustion parameters: time of flame out, average specific mass burning rate, total heat release, maximum average rate of heat emission, and fire growth index are considered to make a systematic analysis and assessment with change depth. Three regions are considered for investigation of output target curves, thermally thin (d ≤ 1.1 cm), thermally medium (1.1 cm < d< 1.5 cm), thermally thick (d≥ 1.5cm). The findings demonstrate that independent of exposure period, an increase in depth change causes a sizable reduction in the HRR and MBR. Beyond 1.5 cm of depth in thermally thick regions, there is no discernible effect of time instant on changes in target parameters. In contrast, HRR and MBR gradients across depth in thermally thin regions are striking.

Machine learning-based model for the intelligent estimation of critical heat flux in nanofluids

The rising demand for advanced energy systems requires enhanced thermal management strategies to maximize resource utilization and productivity. This is quite an important industrial and academic trend as the efficiency of energy systems depends on the cooling systems. This study intends to address the critical need for efficient heat transfer mechanisms in industrial energy systems, particularly those relying on pool boiling conditions, by mainly focusing on Critical Heat Flux (CHF). In fact, CHF keeps a limit in thermal system design, beyond which the efficiency of the system drops. Recent research materials have highlighted nanofluids’ superior heat transfer properties over conventional pure fluids, like water, which makes them a considerable substitution for improving CHF in cooling systems. However, the broad variability in experimental outcomes challenges the development of a unified predictive model. Besides, Machine Learning (ML) based prediction has shown great accuracy for modeling of the designing parameters, including CHF. Utilizing ML algorithms—Cascade Forward Neural Network (CFNN), Extreme Gradient Boosting (XGBoost), Extra Tree, and Light Gradient Boosting Method (LightGBM)— four predictive models have been developed and the benchmark shows CFNN’s superior accuracy with an average goodness of fit of 89.32%, significantly higher than any available model in the literature. Also, the iterative stability analysis demonstrated that this model with a 0.0348 standard deviation and 0.0268 mean absolute deviation is the most stable and robust method that its performance minorly changes with input data. The novelty of the work mainly lies in the prediction of CHF with these advanced algorithm models to enhance the reliability and accuracy of CHF prediction for designing purposes, which are capable of considering many effective parameters into account with much higher accuracy than mathematical fittings. This study not only explains the complex interplay of nanofluid parameters affecting CHF but also offers practical implications for the design of more efficient thermal management systems, thereby contributing to the broader field of energy system enhancement through innovative cooling solutions.

A comprehensive machine learning-based investigation for the index-value prediction of 2G HTS coated conductor tapes

Index-value, or so-called n-value prediction is of paramount importance for understanding the superconductors’ behaviour specially when modeling of superconductors is needed. This parameter is dependent on several physical quantities including temperature, the magnetic field’s density and orientation, and affects the behaviour of high-temperature superconducting devices made out of coated conductors in terms of losses and quench propagation. In this paper, a comprehensive analysis of many machine learning (ML) methods for estimating the n-value has been carried out. The results demonstrated that cascade forward neural network (CFNN) excels in this scope. Despite needing considerably higher training time when compared to the other attempted models, it performs at the highest accuracy, with 0.48 root mean squared error (RMSE) and 99.72% Pearson coefficient for goodness of fit (R-squared). In contrast, the rigid regression method had the worst predictions with 4.92 RMSE and 37.29% R-squared. Also, random forest, boosting methods, and simple feed forward neural network can be considered as a middle accuracy model with faster training time than CFNN. The findings of this study not only advance modeling of superconductors but also pave the way for applications and further research on ML plug-and-play codes for superconducting studies including modeling of superconducting devices.

A hybrid approach to soft sensor development for distillation-in-series plant under input data low variability

This paper presents a hybrid approach for integrating fundamental process knowledge with measurement data to soft sensor (SS) development with improved estimation capability. Measurement data from sensors are collected and used as inputs for a first-principles model to emulate the data close to restrictions of the operating regulations, thus addressing a low variability problem of the inputs. Next, variables from measurement data and results of the first-principles modeling are combined to extend the training dataset for SSs, which become of a hybrid type in nature. To improve an estimation capability, a cascade-forward neural network and algorithm for alternating conditional expectation for nonparametric SS development was used. It was shown that the estimation capabilities of the developed SS can be improved by extending the training dataset with first-principles model data approximating the upper and lower limits of the process regime, the size of which in total does not exceed 21% of industrial data alone. As a result, the designed hybrid SS demonstrates a better efficacy in predicting quality index of the targeted distillation product with significantly reduced mean absolute error.

Artificial intelligence-based surrogate model for computation of the electric field of high voltage transmission line ceramic insulator with corona ring

Purpose The ionization of the air surrounding the phase conductor in high-voltage transmission lines results in a phenomenon known as the Corona effect. To avoid this, Corona rings are used to dampen the electric field imposed on the insulator. The purpose of this study is to present a fast and intelligent surrogate model for determination of the electric field imposed on the surface of a 120 kV composite insulator, in presence of the Corona ring. Design/methodology/approach Usually, the structural design parameters of the Corona ring are selected through an optimization procedure combined with some numerical simulations such as finite element method (FEM). These methods are slow and computationally expensive and thus, extremely reducing the speed of optimization problems. In this paper, a novel surrogate model was proposed that could calculate the maximum electric field imposed on a ceramic insulator in a 120 kV line. The surrogate model was created based on the different scenarios of height, radius and inner radius of the Corona ring, as the inputs of the model, while the maximum electric field on the body of the insulator was considered as the output. Findings The proposed model was based on artificial intelligence techniques that have high accuracy and low computational time. Three methods were used here to develop the AI-based surrogate model, namely, Cascade forward neural network (CFNN), support vector regression and K-nearest neighbors regression. The results indicated that the CFNN has the highest accuracy among these methods with 99.81% R-squared and only 0.045468 root mean squared error while the testing time is less than 10 ms. Originality/value To the best of the authors’ knowledge, for the first time, a surrogate method is proposed for the prediction of the maximum electric field imposed on the high voltage insulators in the presence Corona ring which is faster than any conventional finite element method.

Translating Sumerian Symbols into French Letters

Sumerian symbols, which look like cuneiform letters, were used in a very old writing style. In this paper, a new method to translate the Sumerian symbols into French letters is introduced. It is based on the Cascade-Forward Neural Network (CFNN). The CFNN is exploited and adapted for the translation issue. It accepts a cuneiform letter image as an input and produces an appropriate output that refers to the translated french letter. Reasonable image augmentations are employed. These augmentations are for the: left direction rotations, right direction rotations and multiple translation directions (to the left, right, bottom and top). Total of 780 images are utilized for rotations and 338 images are collected for translations. This work can successfully attain the performance of 100%.

Towards Reliable Barrier Systems: A Constrained XGBoost Model Coupled with Gray Wolf Optimization for Maximum Swelling Pressure of Bentonite

Bentonite and bentonite mixtures are used as buffer material for deep geological radioactive waste repositories. The proper determination of bentonite maximum swelling pressure is important as it influences the long-term stability and sealing characteristics of the barrier system. We developed a constrained machine learning model based on extreme gradient boosting (XGBoost) algorithm tuned with grey wolf optimization (GWO) to determine the maximum swelling pressure of bentonite and bentonite mixtures. We integrated a penalty term into the XGBoost algorithm’s loss function to constrain the model during the training phase ensuring meaningful predictions. To this end, we compiled a dataset containing 305 experimental data of bentonite maximum swelling pressure and other relevant soil properties including montmorillonite content, liquid limit, plastic limit, plasticity index, initial water content, and soil dry density to develop the model. The performance of the hybrid GWO-XGBoost model was benchmarked against feed-forward and cascade-forward neural network models. The GWO-XGBoost model showed a better performance in estimating the experimental values comparing to the other models, thereby demonstrating its effectiveness and reliability in determining the maximum swelling pressure of bentonite and bentonite mixtures.

Toward Estimating CO2 Solubility in Pure Water and Brine Using Cascade Forward Neural Network and Generalized Regression Neural Network: Application to CO2 Dissolution Trapping in Saline Aquifers.

Predicting carbon dioxide (CO2) solubility in water and brine is crucial for understanding carbon capture and storage (CCS) processes. Accurate solubility predictions inform the feasibility and effectiveness of CO2 dissolution trapping, a key mechanism in carbon sequestration in saline aquifers. In this work, a comprehensive data set comprising 1278 experimental solubility data points for CO2-brine systems was assembled, encompassing diverse operating conditions. These data encompassed brines containing six different salts: NaCl, KCl, NaHCO3, CaCl2, MgCl2, and Na2SO4. Also, this databank encompassed temperature spanning from 273.15 to 453.15 K and a pressure range spanning 0.06-100 MPa. To model this solubility databank, cascade forward neural network (CFNN) and generalized regression neural network (GRNN) were employed. Furthermore, three optimization algorithms, namely, Bayesian Regularization (BR), Broyden-Fletcher-Goldfarb-Shanno (BFGS) quasi-Newton, and Levenberg-Marquardt (LM), were applied to enhance the performance of the CFNN models. The CFNN-LM model showcased average absolute percent relative error (AAPRE) values of 5.37% for the overall data set, 5.26% for the training subset, and 5.85% for the testing subset. Overall, the CFNN-LM model stands out as the most accurate among the models crafted in this study, boasting the highest overall R2 value of 0.9949 among the other models. Based on sensitivity analysis, pressure exerts the most significant influence and stands as the sole parameter with a positive impact on CO2 solubility in brine. Conversely, temperature and the concentration of all six salts considered in the model exhibited a negative impact. All salts exert a negative impact on CO2 solubility due to their salting-out effect, with varying degrees of influence. The salting-out effects of the salts can be ranked as follows: from the most pronounced to the least: MgCl2 > CaCl2 > NaCl > KCl > NaHCO3 > Na2SO4. By employing the leverage approach, only a few instances of potential suspected and out-of-leverage data were found. The relatively low count of identified potential suspected and out-of-leverage data, given the expansive solubility database, underscores the reliability and accuracy of both the data set and the CFNN-LM model's performance in this survey.