Separate Artificial Neural Network Models Research Articles

Data science integrated with computational fluid dynamics for particle collision modeling in fluidized bed

Fluidized beds are critical in numerous industrial applications, including chemical processing and energy production. Accurate modeling of solid particles and fluids within these beds is essential for process optimization and improved efficiency. In this study, we developed a machine learning model to predict the coefficient of restitution (COR) of solid particles in fluidized beds. Four variables—collision velocity, effective temperature, effective mass, and effective elastic modulus—were used in the model. Our dataset comprised 2446 data points from previous literature. We employed self-organizing map (SOM) and artificial neural network (ANN) approaches for data analysis. The initial ANN model, which did not incorporate data clustering, exhibited an impressive coefficient of determination (R-squared) of 0.989, indicating its high accuracy. To enhance the model further, we clustered the data into four groups using SOM and developed separate ANN models for each group. All four models achieved R-squared greater than 0.99, illustrating the effectiveness of data clustering. The resulting model can be integrated into simulation software, such as MFIX, to provide a more precise representation of fluidized bed behavior across various industrial settings. The findings emphasize the potential of machine learning models to enhance fluidized bed simulations, leading to increased efficiency and cost-effectiveness in industrial processes. Future studies should explore the inclusion of additional variables and extend the model's application to different industrial processes. Additionally, incorporating recommendations for optimizing fluidized bed behavior based on the model's predictions would provide valuable insights for process engineers.

Implementing Artificial Neural Network-Based Gap Acceptance Models in the Simulation Model of a Traffic Circle in SUMO

The impact of various operational and design alternatives at roundabouts and traffic circles can be evaluated using microscopic simulation tools. Most microscopic simulation software utilizes default underlying models for this purpose, which may not be generalized to specific facilities. Since the effectiveness of traffic operations at traffic circles and roundabouts is highly affected by the gap rejection–acceptance behavior of drivers, it is essential to accurately model drivers’ gap acceptance behavior using location-specific data. The objective of this paper was to evaluate the feasibility of implementing an artificial neural network (ANN)-based gap acceptance model in SUMO, using its application programming interface. A traffic circle in New Jersey was chosen as a case study. Separate ANN models for one stop-controlled and two yield-controlled intersections were trained based on the collected ground truth data. The output of the ANN-based model was then compared with that of the SUMO model, which was calibrated by modifying the default gap acceptance parameters to match the field data. Based on the results of the analyses it was concluded that the advantage of the ANN-based model lies not only in the accuracy of the selected output variables in comparison to the observed field values, but also in the realistic vehicle crossings at the uncontrolled intersections in the simulation model.

Estimation of in-situ rock strength from borehole geophysical logs in Australian coal mine sites

This study aims to improve the conventional empirical rock strength estimation method that is widely adopted in the Australian coal mining industry and provide more accurate rock uniaxial compression strength (UCS) predictions based on artificial neural network (ANN) models. 274 laboratory-measured UCS data (sample dimension: 60 mm (diameter) × 150 mm (height)) were collected from two longwall coal mines in Australia, along with the four geophysical logs (sonic, gamma, neutron, and porosity logs) and rock density of the tested core specimens. A two-layer ANN model was developed from these data based on the Levenberg–Marquardt algorithm as a base model. As compared to conventional sonic-UCS fitting equations, the mean percentage error, root mean squared error and maximum absolute error of the proposed ANN model were reduced from 34.27%, 15.23 MPa and 70.66 MPa to 20.67%, 11.02 MPa and 47.66 MPa, respectively. Further investigations were conducted by splitting the dataset based on rock lithology and mine locations, and separate ANN models were developed based on the divided subsets. Both lithology-specific and site-specific models exhibited improved estimation accuracy in comparison to the base model, indicating lithology and local geological conditions could considerably affect the estimation accuracy. Among the three types of models, the lithology-specific models yielded the most accurate predictions over the 274 data (MPE: 17.55%; RMSE: 8.84 MPa), followed by the site-specific models and the base model in descending order. Overall, all three models significantly outperformed the current empirical methods adopted in the studied mine sites. The outcomes of this study could provide more accurate UCS predictions for further geotechnical analysis in underground constructions, such as tunnel stability analysis and estimating in-situ stresses based on borehole breakout.

Hybrid Neuro-Genetic Machine Learning Models for the Engineering of Ring-spun Cotton Yarns

ABSTRACT Yarn engineering is a long-standing problem for the cotton spinning industry as the functional relationship between fiber and yarn properties is quite complex. The objective of this research is to develop a hybrid machine learning-based prescriptive yarn engineering system that can foretell the properties of cotton fiber for achieving desired yarn properties. Artificial neural network (ANN) and genetic algorithm (GA) were used to develop the predictive model for cotton yarn properties and optimization of cotton fiber properties, respectively. Two separate ANN models were developed for predicting yarn tenacity and yarn unevenness. The functional relationships approximated by the ANN models were used to formulate the fitness function for GA. The validation of the ANN-GA system demonstrated good accuracy as cotton fiber strength, length and length uniformity were predicted with very good accuracy (mean error < 5%). The developed machine learning system can supplant the intuition-based decision making in textile spinning industry and pave the way for yarn engineering.

Prediction of wind pressure coefficients on building surfaces using artificial neural networks

Knowing the pressure coefficient on building surfaces is important for the evaluation of wind loads and natural ventilation. The main objective of this paper is to present and to validate a computational modeling approach to accurately predict the mean wind pressure coefficient on the surfaces of flat-, gable- and hip-roofed rectangular buildings. This approach makes use of artificial neural network (ANN) to estimate the surface-average pressure coefficient for each wall and roof according to the building geometry and the wind angle. Three separate ANN models were developed, one for each roof type, and trained using an experimental database. Applied to a wide variety of buildings, the current ANN models were proved to be considerably more accurate than the commonly used parametric equations for the estimation of pressure coefficients. The proposed ANN-based methodology is as general and versatile as to be easily expanded to buildings with different shapes as well as to be coupled to building performance simulation and airflow network programs.

Effective Prediction and Management of a CO2 Flooding Process for Enhancing Oil Recovery Using Artificial Neural Networks

The injection of CO2 has been in global use for enhanced oil recovery (EOR) as it can improve oil production in mature fields. It also has environmental benefits for reducing greenhouse carbon by permanently sequestrating CO2 (carbon capture and storage (CCS)) in reservoirs. As a part of numerical studies, this work proposed a novel application of an artificial neural network (ANN) to forecast the performance of a water-alternating-CO2 process and effectively manage the injected CO2 in a combined CCS–EOR project. Three targets including oil recovery, net CO2 storage, and cumulative gaseous CO2 production were quantitatively simulated by three separate ANN models for a series of injection frames of 5, 15, 25, and 35 cycles. The concurrent estimations of a sequence of outputs have shown a relevant application in scheduling the injection process based on the progressive profile of the targets. For a specific surface design, an increment of 5.8% oil recovery and 4% net CO2 storage was achieved from 25 cycles to 35 cycles, suggesting ending the injection at 25 cycles. Using the models, distinct optimizations were also computed for oil recovery and net CO2 sequestration in various reservoir conditions. The results expressed a maximum oil recovery from 22% to 30% oil in place (OIP) and around 21,000–29,000 tons of CO2 trapped underground after 35 cycles if the injection began at 60% water saturation. The new approach presented in this study of applying an ANN is obviously effective in forecasting and managing the entire CO2 injection process instead of a single output as presented in previous studies.

Development of a PSO-ANN Model for Rainfall-Runoff Response in Basins, Case Study: Karaj Basin

Successful daily river flow forecasting is necessary in water resources planning and management. A reliable rainfall-runoff model can provide useful information for water resources planning and management. In this study, particle swarm optimization algorithm (PSO) as a metaheuristic approach is employed to train artificial neural network (ANN). The proposed PSO-ANN model is applied to simulate the rainfall runoff process in Karaj River for one and two days ahead. In this regard, different combinations of the input variables including flow and rainfall time series in previous days have been taken under consideration in order to obtain the best model's performances. To evaluate efficiency of the PSO algorithm in training ANNs, separate ANN models are developed using Levenberg-Marquardt (LM) training algorithm and the results are compared with those of the PSO-ANN models. The comparison reveals superiority of the PSO algorithm than the LM algorithm in training the ANN models. The best model for 1 and 2 days ahead runoff forecasting has R2 of 0.88 and 0.78. Results of this study shows that a reliable prediction of runoff in 1 and 2 days ahead can be achieved using PSO-ANN model. Overall, results of this study revealed that an acceptable prediction of the runoff up to two days ahead can be achieved by applying the PSO-ANN model.

Artificial neural network applied forecast on a parametric study of Calophyllum inophyllum methyl ester-diesel engine out responses

This experimental work presents a parametric investigation of Calophyllum inophyllum methyl ester (CIME)-diesel engine operations and artificial neural network (ANN) applied forecast of the engine out responses. The engine tests were performed for five test fuels from idle to full load conditions with the stipulated increment of 25% of the load for every run at three selected injection timings (21°, 23° and 25° CA bTDC) for 220bar, 260bar and 300bar injection pressures. The experimental outcomes indicated that twenty percentage blend of the biodiesel in neat diesel (CIME20) showed the highest brake thermal efficiency (BTE) among the CIME-diesel operations for 300bar injection pressure at 23° CA bTDC injection timing whereas BTE for the test fuels reduced at advanced and retarded injection timings at full load. CO, UBHC, dry soot and engine out O2 emissions were reduced at the advanced injection timing whereas NO and CO2 emissions increased. Using steady state experimental data, separate ANN models were proposed to forecast performance (BTE, BSEC, EGT) and emission (CO, CO2, UBHC, NO, dry soot and engine out O2) parameters with percentage load, blend percentage, injection pressure and injection timing as selected input control variables. The proposed ANN models indicated an impressive agreement as correlation coefficient (R) and mean absolute percentage error (MAPE) values perceived in the range of 0.99879–0.99993 and 0.87–4.62% respectively with remarkably lower root mean squared errors. Besides, lower values of mean squared relative error (MSRE) and noteworthy Nash-Sutcliffe coefficient of Efficiency (NSE) indices reasonably demonstrated robustness of the proposed models. Moreover, observed values of forecasting uncertainty Theil U2 indicated more effective outcomes for a credible model forecasting ability.

Performance Enhancement of Rainfall Pattern – Water Level Prediction Model Utilizing Self-Organizing-Map Clustering Method

During recent two decades, Artificial Neural Network (ANN) has become one of the most widely used methods in hydrology. One solution for better capturing the existing non-linear and complex nature of data is to develop new hybrid approaches. These hybrid models can be developed in a way that two or more techniques are combined in order to benefit from the advantages of these available approaches and eliminate their limitations. The main scope of this paper is to improve the performance of rainfall-water level modeling by combining ANN with Self Organizing Map (SOM) as an unsupervised clustering method. The proposed method in this study consists of two phases. In the first phase, with the aim of reducing the complexity and dimensionality of input data, a two-step clustering using SOM technique is carried out. Then, in the second phase, separate ANN models are used to model each cluster of data, and final results are obtained by combining the outputs of all models. The proposed new hybrid approach is evaluated using real hydrological data of Johor River. The results of the study indicate that the new proposed SOM-ANN hybrid model has a better performance in daily rainfall-water level forecasting compared to ANN model alone.

Prediction of Land Surface Temperature Under Cloudy Conditions Using Microwave Remote Sensing and ANN

Land surface temperature (LST) is an important variable in climate, hydrologic, ecological, biophysical and biochemical studies (Mildrexler et al., 2011). The most effective way to obtain LST measurements is through satellites. Presently, LST from moderate resolution imaging spectroradiometer (MODIS) sensor is applied in various fields due to its high spatial and temporal availability over the globe, but quite difficult to provide observations in cloudy conditions. This study evolves of prediction of LST under clear and cloudy conditions using microwave vegetation indices (MVIs), elevation, latitude, longitude and Julian day as inputs employing an artificial neural network (ANN) model. MVIs can be obtained even under cloudy condition, since microwave radiation has an ability to penetrate through clouds. In this study LST and MVIs data of the year 2010 for the Cauvery basin on a daily basis were obtained from MODIS and advanced microwave scanning radiometer (AMSR-E) sensors of aqua satellite respectively. Separate ANN models were trained and tested for the grid cells for which both LST and MVI were available. The performance of the models was evaluated based on standard evaluation measures. The best performing model was used to predict LST where MVIs were available. Results revealed that predictions of LST using ANN are in good agreement with the observed values. The ANN approach presented in this study promises to be useful for predicting LST using satellite observations even in cloudy conditions.

Developing an artificial neural network model to evaluate chloride diffusivity in high performance concrete

Chloride attack is one of the major causes of deterioration of reinforced concrete structures. In order to evaluate chloride behavior in concrete, a reasonable prediction for the diffusion coefficient of chloride ion (Dc), which governs the mechanism of chloride diffusion inside concrete, is basically required. However, it is difficult to obtain chloride diffusion coefficients from experiments due to time and cost limitations. This study focuses on the artificial neural network (ANN) as an alternative approach to evaluate the chloride diffusivity of high performance concrete (HPC). A total of 300 different data of fly ash (FA) and slag (GGBFS) concrete were collected from the literature. Two separate ANN models were developed for two types of HPC. The data used in the ANN model consisted of four input parameters which include W/B ratio, cement content, FA or GGBFS content and curing age. Output parameter is determined as diffusion coefficient of chloride ion. Back propagation (BP) algorithm was employed for the ANN training in which a Tansig function was used as the nonlinear transfer function. Through the comparison of the estimated results from ANN models and experimental data, it was clear that ANN models give high prediction accuracy. In addition, the research results demonstrate that using ANN models to predict chloride diffusion coefficient is practical and beneficial.

Estimating compaction parameters of fine- and coarse-grained soils by means of artificial neural networks

This paper presents artificial neural network (ANN) prediction models for estimating the compaction parameters of both coarse- and fine-grained soils. A total number of 200 soil mixtures were prepared and compacted at standard Proctor energy. The compaction parameters were predicted by means of ANN models using different input data sets. The ANN prediction models were developed to find out which of the index properties correlate well with compaction parameters. In this respect, the transition fine content ratio (TFR) was defined as a new input parameter in addition to traditional soil index parameters (i.e. liquid limit, plastic limit, passing No. 4 sieve and passing No. 200 sieve). Highly nonlinear nature of the compaction data dictated development of two separate ANN models for maximum dry unit weight (γdmax) and optimum water content (ωopt). It was found that generalization capability and prediction accuracy of ANN models could be further enhanced by sub-clustered data division techniques.

Forecasting highway casualties under the effect of railway development policy in Turkey using artificial neural networks

This study presents forecast of highway casualties in Turkey using nonlinear multiple regression (NLMR) and artificial neural network (ANN) approaches. Also, the effect of railway development on highway safety using ANN models was evaluated. Two separate NLMR and ANN models for forecasting the number of accidents (A) and injuries (I) were developed using 27 years of historical data (1980–2006). The first 23 years data were used for training, while the remaining data were utilized for testing. The model parameters include gross national product per capita (GNP-C), numbers of vehicles per thousand people (V-TP), and percentage of highways, railways, and airways usages (TSUP-H, TSUP-R, and TSUP-A, respectively). In the ANN models development, the sigmoid and linear activation functions were employed with feed-forward back propagation algorithm. The performances of the developed NLMR and ANN models were evaluated by means of error measurements including mean absolute percentage error (MAPE), mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R 2). ANN models were used for future estimates because NLMR models produced unreasonably decreasing projections. The number of road accidents and as well as injuries was forecasted until 2020 via different possible scenarios based on (1) taking TSUPs at their current trends with no change in the national transport policy at present, and (2) shifting passenger traffic from highway to railway at given percentages but leaving airway traffic with its current trend. The model results indicate that shifting passenger traffic from the highway system to railway system resulted in a significant decrease on highway casualties in Turkey.

Combination of artificial neural network technique and linear free energy relationship parameters in the prediction of gradient retention times in liquid chromatography

In this work multiple linear regression (MLR) and artificial neural network (ANN) were used to predict the gradient retention times of diverse sets of organic compounds in four separate data sets. Descriptors which were used as inputs of these models are five linear free energy relationship (LFER) solute parameters including E, S, A, B and V. In the first step eight separate multiple linear regression and artificial neural network models were used to predict the gradient retention time for each gradient condition separately. Results obtained in this step reveal that there are significant relations between LFER parameters and gradient retention times of solutes in liquid chromatography. Then MLR and ANN were applied to develop more general models in which several different gradient elution conditions were used. The performances of these models are compared in terms of their standard errors and also correlation analysis. The results obtained reveal that although there are no significant differences between ANN and MLR in separate modeling of the gradient retention times, ANN has a significant superiority over MLR models in developing the general models for various gradient elution conditions. The results of sensitivity analysis on ANN models indicate that the order of importance for input terms in separate ANN models is V x > B > S > E > A and in the case of combined ANN model is V x > B > t g > S > E > A, which are in agreement with the order of percentage of significance terms that obtained from the MLR models.

M5 Model Trees and Neural Networks: Application to Flood Forecasting in the Upper Reach of the Huai River in China

The applicability and performance of the so-called M5 model tree machine learning technique is investigated in a flood forecasting problem for the upper reach of the Huai River in China. In one of configurations this technique is compared to multilayer perceptron artificial neural network (ANN). It is shown that model trees, being analogous to piecewise linear functions, have certain advantages compared to ANNs—they are more transparent and hence acceptable by decision makers, are very fast in training and always converge. The accuracy of M5 trees is similar to that of ANNs. The improved accuracy in predicting high floods was achieved by building a modular model (mixture of models); in it the flood samples with special hydrological characteristics are split into groups for which separate M5 and ANN models are built. The hybrid model combining model tree and ANN gives the best prediction result.