Artificial Neural Network Methodology Research Articles

Comparison of artificial neural network and response surface methodology for evaluation of the predictive capability of bio-oil yield from pyrolysis of Mangifera indica wood sawdust

In this research article, application of artificial neural network (ANN) and response surface methodology (RSM) is employed to predict the extraction yield (EY) of bio-oil from the pyrolysis of Mangifera indica wood (MIW) sawdust, developing the two optimized models (i.e., ANN-FFBP and RSM-CCD). The effect of three input parameters with their ranges, viz. particle size (0.2–0.8 mm), temperature (430–630 °C), and N2-gas flow rate (60–160 mL/min) on EY of MIW bio-oil are investigated. The Levenberg–Marquardt (LM) algorithm is used to train the network and an ANN-FFBP based model is developed for predicting the EY. Sensitivity analysis confirmed the superiority of optimal ANN-FFBP [3–9–1] model over the optimized RSM-CCD model with the values (AARD%= 0.25945, MSE = 0.00001 and R2 = 0.99711). The maximum EY of bio-oil (45.50 mass%) was observed at optimum conditions of parameters (0.6 mm, 564 °C, and 120 mL/min). The analysis of bio-oil through GC–MS confirmed the presence of high-valued compounds, e.g., methylenecyclopropanecarboxylic acid (8.11 mass%) and p-Cresol (11.41 mass%). In addition to this, the commercial significance of the extracted pyrolysis products has been confirmed via FTIR, ignition point, elemental analyzer, 1H NMR, and FE-SEM-EDX. Moreover, the techno-economic assessment of bio-oil production at industrial scale of 10 TPD assured a profit margin of 109–128 % for per kg production of bio-oil.

Influence of cohesion on California bearing ratio of clay–gravel mixtures

An understanding of the behavior of cohesive sediment is required to solve various engineering problems such as scour around bridge elements, mitigation of soil erosion, pavement design, river bed degradation, stable channel design. Pavement foundation designers principally use the California bearing ratio (CBR) to describe the subgrade and subbase materials and their strength. Several laboratory experiments were done to study the variation in the CBR of cohesive mixtures comprised of clay–gravel mixtures. Nine different clay–gravel mixtures were used in which the clay content varies from 10% to 50% by weight. The variation of the CBR with clay percentage, moisture content, and undrained shear strength parameters was studied. The CBR value reduces with the increase in the moisture content and clay fraction in the mixtures and increases with an increase in the dry density of the mixture under unsoaked conditions. The CBR also increases with the increase of the angle of internal friction of clay–gravel mixtures. A functional relation has been identified to estimate the CBR of clay–gravel mixtures. Using multiple linear regression analysis (MLRA), a relation is proposed to estimate the CBR of clay–gravel mixtures under unsoaked conditions. A statistical analysis was done to judge the behavior of the pertinent variables on the CBR. The proposed relation predicts the CBR of clay–gravel mixtures very well. Artificial neural network (ANN) analysis using R programming was also done to determine the effects of the pertinent variables on the CBR. ANN methodology was applied to predict the contribution of each variable. Three different methods: Garson algorithm, Olden algorithm, and Lek's profile model are used to assess the influence of variable parameters. The Olden algorithm and Lek's profile both show positive association of cohesion with CBR in an unsoaked condition. The role of moisture content was found to be marginally negative in the Olden algorithm and Lek's profile results. It is found that both the ANN and MLRA models are accurate in predicting the CBR of clay–gravel mixtures. It was further found that the MLRA and ANN models are reliable and rapid tools for correct assessment of the CBR of cohesive soil mixtures using the basic soil properties.

Machinability investigation of Chromoly steel using TiN/Al2O3/TiCN/TiOCN multilayer coated tungsten carbide tool inserts with artificial neural network methodology

AbstractEnhancing the machining efficiency of hardened steel using four‐layer coated tungsten carbide tools has gained a lot of interest among manufacturing engineers. This paper reports the machining performance of AISI 4140 steel using TiN/Al2O3/TiCN/TiOCN multilayer coated tungsten carbide tool inserts. The impact of cutting speed, depth of cut, and the nose radius on machining force, tool chip interface temperature, surface roughness, and chip forms was ascertained using full factorial machining experiments and artificial neural network (ANN) methodology. The machining force evaluated using Logistic Sigmoid activation function resulted in the highest root mean square error (RMSE) of 4.076 and regression value of 0.998. The maximum error between the ANN predicted results and experimental results for machining force, cutting temperature and surface roughness was about 2.4%, 5.3%, and 2.07%, respectively. The best architecture for ANN was “3‐6‐4” which had a coefficient of regression value of 0.97. However, in case of chip form, some of the ANN predicted values could not be mapped to ISO standard chip form and hence prediction accuracy of 80% was achieved.

Optimization of soluble phosphate and IAA production using response surface methodology and ANN approach

Phosphorus (P) is often found inaccessible to plants, as it forms precipitates with cations and can be converted to accessible forms by using Phosphate solubilizing bacteria (PSB). In the present study, isolation and characterization of PSB from rhizospheric soil of coffee plants were performed. The influence of four independent variables (incubation temperature, incubation time, pH, and inoculum size) was investigated and optimized using an artificial neural network and response surface methodology on the solubility of phosphate and indole acetic acid production. The bacterium that can dissolve phosphate were isolated in Pikovskaya's agar containing insoluble tricalcium phosphate. Total, six Phosphate Solubilizing Bacteria were isolated and three of them (PSB1, PSB3, and PSB4) were found to be effectively solubilizing phosphate. Based on phosphate solubilizing index results Pseudomonas bacteria (PSB1) was selected for modeling. The results showed that both models performed reasonably well, but properly trained artificial neural networks have the more powerful modeling capability compared to the response surface method. The optimum conditions were found to be incubation temperature of 37.5 °C, incubation time of 9 days, pH of 7.2, and inoculum size of 1.89 OD. Under these conditions, the model predicted solubility of phosphate of 260.69 μg/ml and production of IAA of 80.00 μg/ml with a desirability value of 0.947. In general, the isolated Pseudomonas is expected to have phosphorus-degrading ability that promotes plant growth, and further field experimental work is required to use this bacterial strain as biofertilizer, as an alternative to synthetic fertilizer.

Hybrid optimisation studies on the microstructural properties and wear resistance of maraging steel 1.2709 parts produced by laser powder bed fusion

Improper selection of laser powder bed fusion (LPBF) process parameters tends to result in poor quality parts which imposes limitations with respect to the mechanical performance due to process induced defects. To address this LPBF processing challenge, this study employs a hybrid optimisation technique which combines artificial neural network (ANN) and response surface methodology (RSM) models. The models were employed for predicting the microstructural properties (porosity, microhardness and amount of martensite phase composition) and mechanical characteristic (wear resistance) of LPBF manufactured maraging steel 1.2709 parts as a function of a combination of process parameters (scan speed, laser power and hatch spacing). Both ANN and RSM models had a high tracking ability. However, ANN showed better prediction accuracy than RSM. The most desirable optimal LPBF processing parameters for minimum wear volume and porosity while maintaining maximum microhardness and martensite phase composition were found at volumetric energy density (VED) of 77 J/mm3 (laser power = 165 W, scan speed = 784 mm/s and hatch spacing = 91 μm). Optimal quality properties predicted by the RSM and ANN models were consistent with confirmatory experiment results.

Artificial neural network and response surface methodology for optimization of corrosion inhibition of mild steel in 1 M HCl by Musa paradisiaca peel extract

Banana (Musa paradisiaca) peel extract (BPE) was used as an environmentally benign corrosion inhibitor for mild steel in 1.0 M HCl. The efficiency of BPE was assessed by monitoring the pH of HCl solution and the quantity of hydrogen gas evolved during the reaction, using gasometric and thermometric methods. Moreover, the effect of concentration and temperature on the inhibition efficiency was modelled and optimized by response surface methodology (RSM) and artificial neural network (ANN). It was observed that the evolution of hydrogen gas decreases with increasing concentration of BPE, while it increases with time for the various concentrations up till 8 min before attaining constant values. By numerical optimization of RSM, the optimum corrosion inhibition efficiency of 60.08% was obtained at 308.08 K and concentration of 7.44 g/L for gasometric method, while an optimum of 61.25% was obtained at 308 K and 7.50 g/L for thermometric method. Optimization of inhibition parameters with ANN revealed the optimum number of neurons for both gasometric and thermometric methods to be 7; while the MSE are 2.2788 and 2.7306, and R2 are 96.21 and 98.86 respectively. Comparing the performance models of RSM and ANN: for gasometric method, R2 was 98.93 for RSM and 96.21 for ANN, while for thermometric method, R2 was obtained as 95.78 and 98.86 for RSM and ANN, respectively. Both RSM and ANN have demonstrated to be robust optimization techniques; particularly, ANN was found to give a good prediction of the available small dataset.

Analysis and estimation of cross-flow heat exchanger fouling in phosphoric acid concentration plant using response surface methodology (RSM) and artificial neural network (ANN)

The production of phosphoric acid by dehydrated process leads to the precipitation of unwanted insoluble salts promoting thus the crystallization fouling build-up on heat transfer surfaces of the exchangers. During the acid concentration operation, the presence of fouling in heat exchangers results in reducing the performance of this equipment, in terms of heat transfer, while increasing energy losses and damaging the apparatus. To mitigate these adverse effects of fouling, it is necessary to forecast the thermal resistance of fouling to schedule and perform exchanger cleaning. In this context, artificial neural network and response surface methodology were used to estimate thermal resistance of fouling in a cross-flow heat exchanger by using the operating data of the concentration loop. The absolute average relative deviations, mean squared errors, root mean squared errors and correlation coefficients were used as indicators error between the experimental and estimated values for both methods. The best fitted model derived from response surface methodology method was second order polynomial while the best architecture topology, for the artificial neural network method, consists of three layers: input layer with six input variables, hidden layer with six hidden neurons and an output layer with single output variable. The interactive influences of operating parameters which have significant effects on the fouling resistance were illustrated in detail. The value of correlation coefficient for the output parameter from the response surface methodology is 0.9976, indicating that the response surface methodology as an assessment methodology in estimating fouling resistance is more feasible compared with the artificial neural network approach.

Prediction of Friction Coefficient for Ductile Cast Iron Using Artificial Neural Network Methodology Based on Experimental Investigation

The key objective of the present study is to analyze the friction coefficient and wear rate for ductile cast iron. Three different microstructures were chosen upon which to perform the experimental tests under different sliding time, load, and sliding speed conditions. These specimens were perlite + ferrite, ferrite, and bainitic. Moreover, an artificial neural network (ANN) model was developed in order to predict the friction coefficient using a set of data collected during the experiments. The ANN model structure was made up of four input parameters (namely time, load, number, and nodule diameter) and one output parameter (friction coefficient). The Levenberg–Marquardt back-propagation algorithm was applied in the ANN model to train the data using feed-forward back propagation (FFBP). The results of the experiments revealed that the coefficient of friction reduced as the sliding speed increased under a constant load. Additionally, it exhibits the same pattern of action when the test is run with a heavy load and constant sliding speed. Additionally, when the sliding speed increased, the wear rate dropped. The results also show that the bainite structure is harder and wears less quickly than the ferrite structure. Additionally, the results pertaining to the ANN structure showed that a single hidden layer model is more accurate than a double hidden layer model. The highest performance in the validation stage, however, was observed at epochs 8 and 20, respectively, for a double hidden layer and at 0.012346 for a single layer at epoch 20.

Pyrolysis of de-fatted microalgae residue: A study on thermal-kinetics, products’ optimization, and neural network modelling

The present work focused on the pyrolysis-based valorization of microalgae biomass residual for the generation of sustainable fuel and value-added chemicals. Key pyrolysis factors, including temperature, residence time, particle size, and heating rate, were modeled via an artificial neural network (ANN) and response surface methodology (RSM) models. The use of such an integrated technique was able to conquer the individual constraints of both modeling approaches. RSM model for H2-rich syngas demonstrated that the value; R2 = 0.99, minimum p = 0.00, and maximum F = 3877.16 has close relation among the statistical parameters. The ANN model for H2-rich syngas revealed that higher R2 = 0.9985 and lower RSME = 0.1038 were obtained for the training phase, while; the higher R2 and lower RSME values of 0.9862 and 0.2661 were estimated for the training phase hence showed a better agreement among the parameters. Optimum H2 production of 44.46 vol% was produced at temperature = 516.76 °C, residence time = 17.7 min, particle size = 0.23 mm, and heating rate = 17.37 °C/min. All the pyrolytic products have been characterized in detail and are recommended for high end-use. The GC/MS technique revealed that bio-oil was constituted of various organic complexes, which could be used as a substitute for hydrocarbon fuels after undergoing certain upgradation procedures (e.g., hydrotreating, hydrodeoxygenation, hydrodenitrogenation, and hydrocracking) and extracted into different chemicals. In addition, the biochar was characterized using the SEM technique, which also demonstrated its potential as fuel and in a variety of other applications. The study showed that de-fatted Chlorella sorokiniana residue (De-CR) could be efficiently employed to produce bio-energy precursors.

기계학습 방법론을 활용한 아파트 매매가격지수 연구*

This study proposes to calculate new housing price indices through machine learning techniques. Our research is conducted focusing on the random forest and artificial neural network methodologies that proved excellence in existing real estate studies. We use micro-level real estate transaction data and housing characteristics information to train our models. As a result, the forecasting powers of the machine learning based models are found to be much superior in terms of explanatory powers and estimation performances compared to the hedonic methodology- based model. And the random forest model shows the best explanatory power and performance. Our results show that the housing price indices based on the machine learning models have greater volatility than currently used indices at the time of the housing price increase. Considering the limitation that the existing indices have a smoothing problem, our results can be interpreted that the new machine learning based indices reflect the market trend successfully.

Analysis and prediction of thermal runaway propagation interval in confined space based on response surface methodology and artificial neural network

Thermal runaway (TR) of lithium-ion batteries (LIBs) and its propagation in battery packs may bring significant losses and restrict the wide application of LIB. It is important to study the propagation characteristics of TR. Based on a series of experiments, this work analyses the influence of state of charge, environment temperature, and heating power on the thermal runaway propagation interval (TRPI) between two adjacent cells in a confined space. The results show that they all have a significant impact on TRPI. Furthermore, response surface methodology (RSM) is employed to study the interactions among these three factors. The minimum TRPI is predicted to be 61.08 s. Based on artificial neural network (ANN), a prediction model trained by back-propagation algorithm is constructed for temperature variations of two cells. The results show that the model is effective in prediction, with the maximum prediction error of 6.88 % and the average prediction error of 3.42 %. It is found that TR can be propagated within 37 s, which brings great challenges to the management of battery packs. This research provides effective methods for identifying the safety problems of LIB packs based on RSM and predicting the temperature variations of cells based on ANN methodology.

Prediction and optimization of diesel engine characteristics for various fuel injection timing: Operated by third generation green fuel with alumina nano additive

Azolla pinnata is a macroalgae commonly known as mosquito weed that is widely utilized for biomass production and biodiesel because of its capacity to thrive in low nitrogen environments. In this study, the various concentration (25 ppm, 50 ppm and 75 ppm) of alumina nanoparticles dosed Azolla pinnata biodiesel blend performance and emission parameters were analyzed on a single-cylinder compression ignition engine for different load conditions and fuel injection timings. Using the experimental results, the artificial neural network and response surface methodology regression models were developed and compared its prediction capabilities of output parameters as brake thermal efficiency, brake specific fuel consumption, hydrocarbon, carbon monoxide, nitrogen oxide and smoke. The identified best model was applied to optimize the input parameters of nano additive, engine load and fuel injection timing. The attained R2 (0.9992) and root mean square error 2.437 values exposed that the developed response surface methodology model was more accurate than artificial neural network. The best possible responses provided through the desirability function approach were brake thermal efficiency (19.05 %), brake specific fuel consumption (785.66 g/kWh), hydrocarbon (25.18 ppm), carbon monoxide (0.0576 %), nitrogen oxide (443.79 ppm), smoke (7.48 %) respectively for optimized working factors of nano additive (44 ppm), engine load (1.368 kW) and fuel injection timing (26° before top dead centre). The overall error percentage calculated through the validation study was observed to be under 5 %. The developed response surface methodology model produced good results, which are helpful in predicting and optimizing engine emissions and performance parameters.

Robust frequency–voltage stabilization scheme for multi-area power systems incorporated with EVs and renewable generations using AI based modified disturbance rejection controller

The advent of modern artificial intelligence methods for performance improvement of optimal control strategy has paved a way for providing a reliable operation of power systems. Based on the modern advancements in such techniques, the present paper provides a detailed comparison for finding the optimal control strategy using such techniques. This is exhibited by developing a novel control strategy in the form of a 2nd order active disturbance rejection controller for concurrent frequency-voltage control of a hybrid power system. The hybrid power system comprises of renewable generations in the form of solar-thermal, wind plants. Moreover, the modern day electric vehicles (EVs) are also incorporated as energy storage and operate in vehicle-to-grid mode. The developed control strategy is compared with established industrial controllers to prove its dominance based on concurrent frequency-voltage control of the hybrid power system. Firstly, the controller gains are optimized using magnetotactic bacteria optimization (MBO) technique. Then, the developed control strategy is tuned using artificial neural network (ANN) methodology. Based on the simulation outcomes, the results for frequency deviations, voltage deviations and tie-line power deviations are compared with MBO and ANN optimized 2nd order active disturbance rejection controller. The simulations are carried out on one-area, two-area and standard IEEE-39 bus power systems for in depth validation. Results show that the ANN optimized 2nd order active disturbance rejection controller has superior performance with respect to MBO optimized one. The effects of modern day EVs and renewable generations on the power system is studied broadly.

Deployment of the Microeconomic Consumer Theory in the Artificial Neural Networks Modelling: Case of Organic Food Consumption

Organic food consumption has become a significant trend in consumer behaviour, determined by various motives, among which the price does not play a major role, thus reflecting the Lancaster approach to the microeconomic consumer theory. Additionally, artificial neural networks (ANNs) have proven to have significant potential in providing accurate and efficient models for predicting consumer behaviour. Considering these two trends, this study aims to deploy the Lancaster approach in the emerging area of artificial intelligence. The paper aims to develop the ANN-based predictive model to investigate the relationship between organic food consumption, demographic characteristics, and health awareness attitudes. Survey research has been conducted on a sample of Croatian inhabitants, and ANN models have been used to assess the importance of various determinants for organic food consumption. A Three-layer Multilayer Perceptron Neural Networks (MLPNN) structure has been constructed and validated to select the optimal number of neurons and transfer functions. One layer is used as the first input, while the other two are hidden layers (the first covers the radially symmetrical input, sigmoid function; the second covers the output, softmax function). Three versions of the testing, training, and holdout data structures were used to develop ANNs. The highest accuracy was achieved with a 7-2-1 partition. The best ANN model was determined as the model that was showing the smallest percent of incorrect predictions in the holdout stage, the second-lowest cross-entropy error, the correct classification rate, and the area under the ROC curve. The research results show that the availability of healthy food shops and consumer awareness of these shops strongly impacts organic food consumption. Using the ANN methodology, this analysis confirmed the validity of the Lancaster approach, stating that the characteristics or attributes of goods are defined by the consumer and not by the product itself.

Dynamic managerial capability portfolios in early internationalising firms

It is acknowledged that dynamic capabilities can underlie internationalisation of firms, yet extant research has tended to examine the role of capabilities in international entrepreneurship in separate studies, neglecting that international enterprises tend to develop different bundles or portfolios of capabilities that may have differing impacts on their internationalisation. Applying the concept of dynamic managerial capabilities, this study illustrates how different capability portfolios of entrepreneurial decision-making and network capabilities explain early internationalisation of small- and medium-sized enterprises (SMEs). Employing a machine learning method through artificial neural network (ANN) analysis, the findings show how early internationalisation is predicted by a portfolio of different dynamic managerial capabilities. This study thus contributes to the international entrepreneurship literature by outlining the capability portfolios necessary for early internationalisation and proposing a taxonomy for capability portfolios for early and late internationalising firms, by clarifying the role of dynamic managerial capabilities in early and successful entrepreneurial internationalisation and by introducing the ANN methodology as a useful tool in the research corpus of the international entrepreneurship domain.

Artificial neural network methodology: application to predict plasticity of clay soil treated with sand

Artificial neural network methodology: application to predict plasticity of clay soil treated with sand

Thermal Performance Study of Solar Air Dryers for Cashew Kernel: A Comparative Analysis and Modelling Using Response Surface Methodology (RSM) and Artificial Neural Network (ANN)

In this study, a new procedure for drying of cashew kernels is proposed. The new methodology involves the use of solar air dryers to remove the dark sticky coating called testa on the surface of the cashew kernel to get the final product. There are several challenges in the new procedures, especially with respect to the moisture level to be retained in the cashew kernel post drying for facilitating the optimal peeling activity. In this regard, thermal performance is carried out through a set of experimental and computational trials. The experiments have revealed that the natural and forced convection solar dryers with drying speeds of 1.0 kg/h and 1.66 kg/h, respectively, showed drying efficiencies of 51.7% and 50.9%, which is within the permissible acceptable limit, thus ascertaining a moisture reduction of 40 to 42% in each of case and keeping the moisture content within the band of approximately 5%, required for effective peeling of the testa from the cashew kernel to obtain the final product. This new method has resulted in batch drying of cashew kernels of up to 30 kg capacity in a time span of 360 minutes of solar irradiation with an average consumption of 255 kJ. The results of the experiments are also validated from the artificial neural network (ANN) and response surface methodology (RSM) modelling, resulting in better prediction and optimization of the performance parameters. The error between the actual and experimental values is well within the permissible limit of +/-5%, thereby correlating the experimental and statistical values for deriving an ANN model for predicting the results for different time duration and solar irradiation. In this way, the thermal analysis of the solar dryers for drying cashew kernels has resulted in findings, which shall be utilized to improve the overall performance efficiency and the methodology adopted for maximum yield.

Environmental sustainability in the city of Shanghai municipal solid waste treatment system: an integrated framework of artificial neural network (ANN) and LCA-emergy methodology

Environmental sustainability in the city of Shanghai municipal solid waste treatment system: an integrated framework of artificial neural network (ANN) and LCA-emergy methodology

DMLAR: Distributed Machine Learning-Based Anti-Collision Algorithm for RFID Readers in the Internet of Things

Radio Frequency Identification (RFID) is considered as one of the most widely used wireless identification technologies in the Internet of Things. Many application areas require a dense RFID network for efficient deployment and coverage, which causes interference between RFID tags and readers, and reduces the performance of the RFID system. Therefore, communication resource management is required to avoid such problems. In this paper, we propose an anti-collision protocol based on feed-forward Artificial Neural Network methodology for distributed learning between RFID readers to predict collisions and ensure efficient resource allocation (DMLAR) by considering the mobility of tags and readers. The evaluation of our anti-collision protocol is performed for different mobility scenarios in healthcare where the collected data are critical and must respect the terms of throughput, delay, overload, integrity and energy. The dataset created and distributed by the readers allows an efficient learning process and, therefore, a high collision detection to increase throughput and minimize data loss. In the application phase, the readers do not need to exchange control packets with each other to control the resource allocation, which avoids network overload and communication delay. Simulation results show the robustness and effectiveness of the anti-collision protocol by the number of readers and resources used. The model used allows a large number of readers to use the most suitable frequency and time resources for simultaneous and successful tag interrogation.

Process optimization and kinetics study for photocatalytic ciprofloxacin degradation using TiO2 nanoparticle: A comparative study of Artificial Neural Network and Surface Response Methodology

The photocatalytic degradation of ciprofloxacin was investigated by developing a predictive mathematical model using response surface methodology and an artificial neural network. The four independent variables involve solution pH, reaction time, catalyst dose, and initial antibiotic concentration considered as factors in central composite design to observe the response in the form of antibiotic degradation. Accordingly, at an optimum antibiotic concentration of 5.02 mg/L, catalyst dose of 44.51 mg/L, solution pH of 5.04, and reaction time of 75.80 min, the photocatalysis method achieved a ciprofloxacin degradation of 88.30%. The experimental outputs were very much consistent along with the predicted output of experiments through response surface methodology (R2 = 0.9969) and artificial neural network (R2 = 0.975). The adsorption isotherm and kinetic study reveal that Langmuir isotherm and pseudo-second-order kinetic models respectively were best fitted for degradation of ciprofloxacin through photocatalysis. The finding provides a novel method for evaluating the photocatalysis process for the optimization of ciprofloxacin antibiotic removal from pharmaceutical waste using experiments and computer simulation tools.