ANN Model Research Articles

An Integrated Machine Learning Workflow to Estimate In Situ Stresses Based on Downhole Sonic Logs and Laboratory Triaxial Ultrasonic Velocity Data

AbstractThe optimum performance of various subsurface operations such as stimulation treatments, wellbore drilling, horizontal well placement, underground mining, and tunneling rely on accurate estimation of the in situ stresses. This study presents an integrated machine learning (ML) workflow for the reliable determination of the in situ stress in subsurface rock formations. The study workflow was completed in three phases. In the first phase, six supervised ML regression techniques were employed to develop the stress prediction models using laboratory true triaxial ultrasonic velocity tests (labTUV) data of subsurface rock samples retrieved from well 16A(78)‐32 located at the Utah FORGE geothermal site. In the second phase, subsurface geological formations were classified into rock facies using an unsupervised K‐means clustering algorithm. Finally, in the third phase, the optimized ML models were employed for predicting the in situ stresses in the corresponding rock facies in the well 16A(78)‐32 using field sonic logs. A comparison of evaluation metrics revealed the superior performance of ANN models for horizontal and vertical stress predictions with root mean squared error of 1.5, 0.6, and 1.7 MPa, and determination coefficient (R2) of 0.98, 0.98, and 0.96, for the testing/validation phases, respectively. The generalization capability of ML models was explored by uncovering the underlying physics. The mathematical expressions of constitutive relations were extracted from three ANN models. Further, a total of five rock facies were identified in the subsurface geological formations. The novel workflow would be capable of delivering reliable in situ stress profiles in subsurface geological formations without performing expensive field tests.

An Improved Fick Model for Predicting Carbonation Depth of Concrete

Concrete carbonation can weaken its strength, cause the corrosion of steel reinforcement, and shorten its service life. Predicting the concrete carbonation depth is a critical aspect of assessing concrete durability. Currently, mathematical models for the concrete carbonation depth, exemplified by the Fick model, suffer from a low fitting accuracy and limited applicability due to the complexity and variability of concrete materials and service environments. In light of this, this work proposes an improved Fick model that incorporates a correction term to effectively enhance the curve fitting accuracy. The correction term in the improved model provides a reasonable adjustment for deviations in the development pattern of the concrete carbonation depth from the Fick model under different conditions, thereby broadening the applicability of the new model compared to the Fick model. Several sets of experimental data on the concrete carbonation depth are used to validate the universality and superiority of the new model. The results of the case studies indicate that the average prediction error and standard deviation of the new model are significantly smaller than those of the Fick model. For the first two examples, in most situations, the average prediction error and standard deviation of the new model are less than 50% of those of the Fick model, with the lowest average prediction error being only 4% and the lowest standard deviation being only 2% of the Fick model’s respective values. For the third example, the new model demonstrates superior predictive capability for the later-stage concrete carbonation depth compared to the Fick model and the ANN model. Specifically, for the carbonation depth of the concrete on the 56th day, the relative error between the predicted value of the new model and the measured value is only 2%, which is much smaller than the 27% of the Fick model and the 12% of the ANN model. These results demonstrate the unique advantage of the proposed model in predicting the carbonation depth, especially when only a limited amount of experimental data are available.

Use of Deep-Learning-Accelerated Gradient Approximation for Reservoir Geological Parameter Estimation

The estimation of space-varying geological parameters is often not computationally affordable for high-dimensional subsurface reservoir modeling systems. The adjoint method is generally regarded as an efficient approach for obtaining analytical gradient and, thus, proceeding with the gradient-based iteration algorithm; however, the infeasible memory requirement and computational demands strictly prohibit its generic implementation, especially for high-dimensional problems. The autoregressive neural network (aNN) model, as a nonlinear surrogate approximation, has gradually received increasing popularity due to significant reduction of computational cost, but one prominent limitation is that the generic application of aNN to large-scale reservoir models inevitably poses challenges in the training procedure, which remains unresolved. To address this issue, model-order reduction could be a promising strategy, which enables us to train the neural network in a very efficient manner. A very popular projection-based linear reduction method, i.e., propel orthogonal decomposition (POD), is adopted to achieve dimensionality reduction. This paper presents an architecture of a projection-based autoregressive neural network that efficiently derives an easy-to-use adjoint model by the use of an auto-differentiation module inside the popular deep learning frameworks. This hybrid neural network proxy, referred to as POD-aNN, is capable of speeding up derivation of reduced-order adjoint models. The performance of POD-aNN is validated through a synthetic 2D subsurface transport model. The use of POD-aNN significantly reduces the computation cost while the accuracy remains. In addition, our proposed POD-aNN can easily obtain multiple posterior realizations for uncertainty evaluation. The developed POD-aNN emulator is a data-driven approach for reduced-order modeling of nonlinear dynamic systems and, thus, should be a very efficient modeling tool to address many engineering applications related to intensive simulation-based optimization.

Comparison of energy prediction models for residential buildings: A case study in Himachal Pradesh, India

Abstract The demand for consumption of energy has been increasing globally with the tremendous increase in population. Different studies have proved that inadequate energy management and planning may lead to energy crisis which is a result of inadequacies in energy prediction. Accurate prediction of energy demand is important as underestimation may lead to shortage in supply and overestimation may lead to overinvestment in energy generation. Various available literature has been reviewed for determining the various factors responsible for affecting the energy consumption of residential buildings. Based on the factors determined, survey questionnaire has been formulated and survey was conducted in 400 residential buildings in one of northern states of India, i.e., Himachal Pradesh. It was observed by reviewing various studies that different models developed for energy consumption by different researchers were based on either of the three approaches, namely, engineering-based, AI-based, and hybrid approaches. Three tools namely, case-based reasoning, artificial neural network, and multilinear regression, based on these approaches were individually used for developing the model in this study, and their prediction results were compared. It was observed that the accuracy in the overall predicted results was highest in the proposed ANN model, followed by CBR model, and MLR model, with an overall accuracy of 99.93%, 96.3%, and 91.7%, respectively. The error obtained in the predicted results using ANN, CBR and MLR ranges from -4.0 to +3.0%, -15.0 to +26.0%, -30.0% to 20.0%, respectively. The overall RMSE of ANN, CBR, and MLR model was 1.44%, 11.7%, and 19.5%, respectively. It is concluded that ANN model is best suitable for predicting the short and long-term energy consumption with very high accuracy, as compared to the CBR and MLR. The results discussed in this study can be advantageously used for enhancing the consumption of operational energy in existing as well as proposed buildings.

Assessing the impact of recycled concrete aggregates and metakaolin on the mechanical properties of pervious concrete

Population growth and urban expansion have increased impervious surfaces, leading to environmental issues such as stormwater runoff and groundwater depletion. Traditional concrete, while durable, contributes to these problems. There is a need for sustainable construction materials that can address environmental challenges and reduce the consumption of natural resources. Pervious concrete, enhanced with recycled aggregates (RCA) and supplementary cementitious materials (SCMs) like metakaolin (MK) present a potential solution. This study aims to improve the mechanical and durability properties of pervious concrete by incorporating RCA and MK, assessing their effects on workability, density, porosity, permeability, compressive strength and flexural strength. The incorporation of RCA resulted in increased aggregate porosity compared to natural aggregates. Workability was assessed using the slump test, density was measured, and porosity and permeability were evaluated. The results showed that incorporation of RCA generally decreased workability, density and strength, while the addition of MK improved these properties by refining the microstructure. Specifically, mixes with 10% MK demonstrated better performance in terms of increased compressive and flexural strengths, improved density and reduced porosity. The ANN model showed high accuracy in predicting compressive and flexural strengths. The study concludes that RCA and MK can enhance the mechanical and durability properties of pervious concrete.

RSM, SVM and ANN modeling of the properties of self-compacting concrete with natural mordenite-rich tuff and recycled glass

RSM, SVM and ANN modeling of the properties of self-compacting concrete with natural mordenite-rich tuff and recycled glass

Gender Disparities in Book Circulation at a Chinese University Library (2012–2022)

The study explores the significance of gender differences in higher education, using book circulation data from Nanjing Normal University libraries between 2012 and 2022. It aims to analyze gender disparities and calculate the mathematical characteristics of gender prediction probability using the ANN model. The time-series ARIMA model also forecasts the average number of books that male and female teachers and teacher candidates are expected to borrow and renew in the next three years. The findings indicate that men tend to be more focused and persistent readers despite women having a more extensive network of book borrowing ties. Male graduate students and faculty members demonstrate significantly greater individual variability in book renewals than their female counterparts. Additionally, the maximum value of book renewals for male graduate students and male faculty members is notably higher than that of female graduate students and female faculty members. Book renewal frequency is a crucial factor in ANN gender prediction. On average, male readers incur higher overdue fines than female readers, and the overdue fines for both genders during the COVID-19 pandemic were considerably higher than usual. The actual monthly average renewal rates in 2022 exhibited significant fluctuations across genders and academic statuses, highlighting the limitations of the ARIMA model and other factors influencing renewal rates.

Forecasting Carbon Dioxide Emission Regional Difference in China by Damping Fractional Grey Model

The emission of carbon dioxide is the main reason for many global warming problems. Although China has made tremendous efforts to reduce carbon emission, the space–time dynamics of the carbon emission trend is still imbalanced. To forecast CDED in China, the Dagum Gini coefficient was applied to measure regional CDED. Then, a grey correlation model was used to select potential influence factors and a wrapping method for selecting the optimal subset. DGMC is proposed to forecast CDED. The research results showed that the DGMC generalization performance is significantly superior to other models. The MAPE of DGMC in six cases are 1.18%, 1.11%, 0.66%, 1.13%, 1.27% and 0.51%, respectively. The RMSPEPR of DGMC in six cases are 1.08%, 1.21%, 0.97%, 1.36%, 1.41% and 0.57%, respectively. The RMSPEPO of DGMC in six cases are 1.29%, 0.69%, 0.02%, 0.58%, 0.78% and 0.32%, respectively. In future trends, the eastern carbon dioxide emission intraregional differences will decrease. Additionally, the intraregional differences in western and middle-region carbon dioxide emissions will expand. Interregional carbon emission difference will display a narrowing trend. Compared with the traditional grey model and ANN model, integrating the influence factor information significantly improved forecasting accuracy. The proposed model will present better balanced historical information and accurately forecast future trends. Finally, policy recommendations are proposed based on the research results.

Elucidating synergistic effects and environmental value enhancement in infrared-Assisted Co-Pyrolysis of coal and polyvinyl chloride

Co-pyrolysis of high-alkali coal and polyvinyl chloride (PVC) through infrared heating is a promising approach for managing escalating PVC waste and converting low-grade coal resources efficiently. The synergistic effects during co-pyrolysis and thermal degradation of PVC were studied through thermogravimetric analysis (TG-FTIR) and spectral analysis. Furthermore, chlorine migration and chemical transformation integral to the process, along with catalytic interactions, were explored through chlorine mass balance assessment, employing X-ray photoelectron spectroscopy and ion chromatography to establish a groundbreaking understanding of these phenomena. The results of products exhibited a trend of initially increasing and then decreasing with increasing temperatures and mixing ratios. The maximum oil yield was 14.62 %, achieving at 600 °C and 20 %. Meanwhile, the content of aromatic hydrocarbons remained high level through the synergistic interaction of Lewis and Brønsted acid sites, nearly exceeding 60 %. The results showed that the Artificial Neural Network model had a high prediction accuracy with an R2 value of 0.99, while the decarboxylation effect dissociated the metal, resulting in an increase in the fixation of inorganic chlorine from 1 % to 66 % and a decrease in the chlorine content in the liquid phase. Innovatively, the study employs an ANN model for predicting the behavior of co-pyrolysis, exemplifying high accuracy in understanding complex thermal conversions of PVC and coal. Notably, the work challenges existing constraints by applying sophisticated analytical methods to elucidate the thermodynamics and kinetics of the co-pyrolysis processes, thereby enhancing the energetic and environmental value of the products.

Fusing Transformer and diffusion for high-resolution prediction of daylight illuminance and glare based on sparse ceiling-mounted input

Accurate prediction of workplane daylight illuminance and eye-height glare is crucial for lighting control. Existing studies used machine learning to predict illuminance at predetermined locations based on indoor sensors, but they may encounter challenges in scenarios 1) with flexible seating arrangements, 2) with dynamic shading devices, and 3) requiring the prediction of glare. To address these challenges, we proposed a novel method fusing Transformer and Diffusion models, with the input being data collected from sparse ceiling-mounted illuminance sensors, and the outputs being high-resolution workplane illuminance and glare. The model works well for rooms without and with dynamic roller shades. For the former, the mean absolute errors for illuminance below 3000 lx and Daylight Glare Index (DGI) are only 20.77 lx and 0.20, and the error rates in detecting illuminance <500 lx and DGI>22 are only 0.85 % and 5.55 %. For the more complicated latter case, the aforementioned four numbers are 34.78 lx, 0.59, 2.47 % and 23.13 %. The model significantly outperforms the linear and the ANN models, particularly in glare prediction. The influence of sensor number and placement strategy on model performance was also revealed. The model can potentially enhance lighting control, especially in cases with dynamic shading, with flexible seating arrangements, and where glare is of interest.

Designing of a proficient macro porous metallo-polymer of iron with anion functionality: A sustainable approach for arsenic remediation from water

In this study, a novel porous metallo-polymeric network named poly(Ferric trimethacrylate-co-4-Vinyl benzyl chloride), incorporating an iron moiety, was developed for arsenic remediation from water. The synthesis of crosslinked q(pFeVBC) was carried out via suspension polymerization and functionalizing it with hexamethylenetetramine, resulting in a porous three-dimensional structure. Characterization using FTIR, FESEM, XPS, XRD, and BET-SA analysis confirmed its properties. Various monomer ratios were tested to optimize q(pFeVBC) for arsenic removal. The q(pFeVBC) exhibited significant adsorption capacities (qe max exp) of 8.33 mg g−1 for As(V) and 7.9 mg g−1 for As (III), owe to its unique architecture (integrated iron moiety and anion functionality) and porous texture (SA: 112 m2 g−1, pore size: 1.27–2.07 µm), which is close to the capacity calculated by Langmuir model (qe max theo) 8.719 mg g−1 for As (V) and 8.463 mg g−1 for As (III). The PSO kinetic model showed excellent fitting for both arsenic forms. Thermodynamic studies indicated a spontaneous and endothermic adsorption process. Additionally, a 4:10:1 ANN model accurately predicted adsorption behavior.

A comparative study of RSM and ANN models for predicting spray drying conditions for encapsulation of Lactobacillus casei

A comparative study of RSM and ANN models for predicting spray drying conditions for encapsulation of <i>Lactobacillus casei</i>

Harnessing artificial neural networks and linear regression models for modeling thermal modification processes: Characterization by FTIR and prediction of the mechanical properties of eucalyptus wood

In the present study, an artificial neural network model and multiple linear regression method were constructed to establish a relationship between the parameters of the thermal modification process and the mechanical properties of Eucalyptus camaldulensis wood samples. Three influencing parameters on the mechanical properties during heat treatment were considered input variables, namely water absorption, volumetric mass, and mass loss; the other parameters were kept constant (treatment temperature: 200, 220, 240, and 260 °C, and the duration: 5, 60, and 90 minutes). There were five neurons in the hidden layer that were used, as well as an output layer for the mechanical property. According to the results obtained, the mean square error (MSE) for the training, validation, and testing datasets were determined to be 0.21, 0.25, and 0.22 in the prediction modulus of rupture (MOR) and 0.019, 0.017, and 0.023 in the prediction modulus of elasticity (MOE). Higher coefficients of determination (R2) ranging from 0.93 to 0.98 were obtained for all datasets with the proposed ANN models, while the multiple linear regression models found the MSE to be 1.03 and 1.40 for MOR and MOE, respectively, as well as R2 to be 0.97 and 0.80, respectively. These results show that ANN models can be successfully used to predict the change in mechanical properties of wood during heat treatment.

Density Functional Theory (DFT) Study on the Nontoxic Alternative of Bisphenol A (BPA) Derivatives: A Comprehensive Review

AbstractBisphenol A is an oil‐derived, large market volume chemical with a wide spectrum of applications in plastics, adhesives, and thermal papers. However, bisphenol A and its derivative are not considered safe due to its endocrine disrupting properties and reproductive toxicity. A nontoxic alternative of bisphenol analogus has been proposed in this study using plant biomass. A study of different DFT‐based QSAR approaches has been done to show the significance of the conceptual DFT‐based selected descriptors with different QSAR models in the prediction of toxicity and to establish meaningful correlations between the molecular structure of the proposed compounds and their toxicological properties. Multiple regression analysis and ANN model are also suggested to use to relate the biological activity with the global and local reactivity descriptors.

Experimental Analysis of EDM Parameters on D2 Die Steel Using Nano-aluminum Composite Electrodes

This paper presents an experimental study on the electrical discharge machining (EDM) of AISI D2 die steel using an Al-Ni composite electrode. The investigation focuses on the influence of input parameters like current, pulse duration, and pulse interval, on key output parameters such as material removal rate (MRR), tool wear rate (TWR), and surface roughness (SR). EDM oil was employed as the dielectric fluid. Grey relational analysis (GRA) was utilized for designing and conducting the experiments using the Taguchi L9 method. The ANN model showed excellent predictive accuracy with a perfect correlation coefficient (R) of 1.00, indicating strong capability in forecasting MRR based on machining parameters. GRA further confirmed that higher current settings and longer pulse-off times effectively reduce tool wear, suggesting that the ANN model accurately reflects the conditions that minimize TWR. The ANN model achieved strong predictive accuracy for SR, with high correlation coefficients, although with slightly higher mean squared error (MSE) in testing.

DIAGNOSTIC EVALUATION OF URBAN METRO TRANSIT SYSTEM POST-COVID-19

Public transportation usage in Delhi has declined, with the Delhi Metro having a significant share. However, due to fare hikes and COVID-19 restrictions, the DM's share has been decreasing further. To improve ridership, a study is being conducted to evaluate the DM's performance and identify areas for improvement in passenger convenience and comfort. The Magenta line is investigated through an on-board survey to collect primary data. The survey covers commuter perceptions of safety &amp; security, financial &amp; economic factors, infrastructure &amp; comfort and functional &amp; operational features. The Relative Importance Index approach is used to analyse the data and evaluate DM performance. An ANN model is also presented to determine the factors influencing the choice to travel on the DM, with the “metro fare per trip” factor being a key consideration. Based on the analysis results, recommendations are made to improve the DM's performance. The study found that safety and security had the highest RII, followed by efficiency and viability, functional and operational features, infrastructure and comfort, and financial and economic factors. The subway fare had the lowest RII. The ANN model is adapted to understand the reasons behind low metro ridership.

Machine Learning and Linear Regression Approach to Model Unconfined Compressive Strength of Ceramic Waste Modified Soil as Subgrade Pavement Material

An effective application of artificial intelligence involves artificial neural networks. Artificial neural networks and linear regression models were developed to simulate the effects of using discarded ceramic waste as a subgrade for pavement. The ceramic waste was used at 2.5%, 5%, 7.5%, 10%, 12.5%, and 15%. A sample with 0% ceramic waste was tested to serve as a reference sample. The dataset was produced from laboratory experimentation findings used to train, test, and evaluate the model. A training set, a target set, and a prediction set were created from the dataset. The artificial neural network MSE was 0.42-1.40, while the linear regression model range was 1.74 to 3.63 for ceramic modified samples. The R2 range for the ANN model was 0.85-0.92, and the linear regression model exhibited a range of 0.71-0.78. The ANN model was more accurate than the linear regression model. Future studies are required to compare different machine-learning approaches for predicting soil mechanical properties.

Daily reference evapotranspiration prediction using empirical and data-driven approaches: A case study of Adana plain

Precise determination of the reference evapotranspiration (ET0) is vital to studying the hydrological cycle. In addition, It plays a significant role in properly managing and allocating water resources in agriculture. The objective of this research was to examine the effectiveness of five different data-driven techniques, including artificial neural networks "multilayer perceptron" (ANN), gene expression programming (GEP), random forest (RF), support vector machine "radial basis function" (SVM), and multiple linear regression (MLR) to model the daily ET0. These methods were also compared with Hargreaves-Samani (HS), Oudin, Ritchie, Makkink (MAK), and Jensen Haise (JH) empirical models and their calibrated versions. The empirical models JH and MAK performed better than the models HS and Oudin after being calibrated by linear regression. All data-driven methods with four inputs were superior to the original and calibrated empirical models. The RF and ANN methods generally demonstrated better estimation accuracy than other data-driven methods. The performance of the RF and ANN models that utilized Tmax, Tmin, and Rs inputs, as well as those that incorporated Tmax, Tmin, Rs, and U2 inputs, proved to be superior to their corresponding MLR-based and GEP-based models for predicting ET0 in the Adana plain, which is characterized by a Mediterranean climate. Nevertheless, the GEP and MLR methods have the advantage of utilizing explicit algebraic equations, making them more convenient to apply, especially in the context of agricultural irrigation practices.

Damage of thin target plates impacted by conical projectiles with varying apex angles in ballistic application: experimental, FEA and ANN integrated approach

In the current study, an experimental setup consisting of smoothbore 30-caliber powder gun was employed to launch spherical projectiles (3 g/ϕ9) in the ballistic range of velocities from 500 to 1700 m/s and obliquity (0, 33 and 65°), impacting 2 mm thin Steel 1006 target plates. Crater dimensions (major and minor dia.) obtained from series of FE simulations run for the above configuaration was validated by corresponding experimental data. Subsequently, a fullscale 3D FE model considering 8-noded hexahedral elements was used to discretize SS304 conical projectiles (replacing spherical projectile) with various apex angles viz. 40°, 60°, 80° and 100° impacting on single (3 mm) and double layer (1.5 mm each) steel 1006 targets (replacing 2 mm target) in LS dyna. The material behavior was characterized using J–C strength and damage models, along with Gruneisen Equation of State. An erosion algorithm was used in the explicit FE code LS-DYNA to remove undesirable elements. In the next stage, Adam and Nadam optimizers have been employed in ANN models developed using Python code within the Tensor-Flow framework. The Keras Tuner library in the Tensor-Flow framework was used for hyper parameter tuning. The ANN models trained using simulation data successfully predicted the residual KE of conical projectiles with intermediate apex angles of 90°, 70°, and 50° across twelve different impact scenarios. The models with Adam and Nadam optimizers achieved mean squared error (MSE)/coefficient of determination (R2)/mean absolute percentage error (MAPE) values of 109.44/0.998/6.72 and 91.19/0.998/7.76, respectively.

Mineral policy and sustainable development goals: Volatility forecasting in the Global South's minerals market

This paper aims to evaluate the volatility of precious metals, specifically Palladium, Gold, and Platinum, within the context of the global minerals market. The research focuses on understanding the price dynamics of these metals and their implications for sustainable development, particularly in the Global South. The study employs a comprehensive approach, utilizing advanced machine learning and deep learning models such as GRU, Huber, Lasso, LSTM, Random Forest, Ridge Regression, SVM, ANN, and XGBoost. These models are assessed based on their forecasting accuracy for different time horizons, using metrics such as RMSE and MAPE. The findings reveal that the ANN, XGBoost, and LSTM models exhibit robust performance in forecasting the volatility of precious metals across various time horizons. The research highlights the unique volatility patterns of each metal and underscores the effectiveness of machine learning techniques in capturing these dynamics. The study acknowledges limitations such as the exclusion of macroeconomic and geopolitical factors in the forecasting models. Future research is suggested to integrate these factors to enhance forecasting accuracy. The study's findings are pivotal for investors, policymakers, and market regulators, especially in the context of the Global South and sustainable development. The research offers valuable insights for risk management strategies, investment planning, and policy formulation aimed at promoting market stability and sustainable economic growth. The study emphasizes the importance of selecting appropriate forecasting models based on specific time horizons and market requirements.