Suitable Artificial Neural Network Model Research Articles

Decision support system for ship energy efficiency management based on an optimization model

This paper introduces an innovative decision support system for improving ship energy efficiency. It combines an engine optimization model and an artificial neural network (ANN). Real-time data regarding the ship and engine performance is obtained along a specific ship navigation, and an engine optimization model is built using Ricardo Wave software. This model is validated with high accuracy. Ship-specific parameters are derived, and fuel consumption is estimated using ANN models with different structures. The study identifies the most suitable ANN model with 1 hidden layer and 5 neurons based on error metric evaluation where the 0.99697 R2, 0.00035 RMSE, and 0.06470 MAPE scores are calculated. This approach offers a cost-effective solution for shipping companies to monitor critical engine parameters in real-time without investing in sensors or data collection systems. Thus, it contributes to dealing with the problem of the scarcity of analyzable data in the maritime literature, which is one of the significant issues in the papers related to energy efficiency, machine learning, and deep learning. It is an innovative approach since the parameters that are related to real-time operations rather than considering only instruction book information are aggregated, and the produced data is analyzed by an intelligent tool.

Performance Evaluation of Artificial Neural Network Modelling to a Ploughing Unit in Various Soil Conditions

Abstract The specific objective of this study is to find a suitable artificial neural network model for estimating the operation indicators (disturbed soil volume, effective field capacity, draft force, and energy requirement) of ploughing units (tractor disc) in various soil conditions. The experiment involved two different factors, i.e., (Ι) soil texture index and (ΙΙ) field work index, and included soil moisture content, tractor engine power, soil bulk density, tillage speed, tillage depth, and tillage width, which were linked to one dimensionless index. We assessed the effectiveness of artificial neural network and multiple linear regression models between the values predicted and the actual values using the mean absolute error criterion to test data points. When the artificial neural network model was applied, the mean absolute error values for disturbed soil volume, effective field capacity, draft force, and energy requirement were 69.41 m3·hr−1, 0.04 ha·hr−1, 1.24 kN, and 1.95 kw·hr·ha−1, respectively. In order to evaluate the behaviour of new models, the coefficient R 2 was used as a criterion, where R 2 values in artificial neural network were 0.9872, 0.9553, 0.9948, and 0.9718, respectively, for the aforementioned testing dataset. Simultaneously, R 2 values in multiple linear regression were 0.7623, 0.696, 0.492, and 0.5572, respectively, for the same testing dataset. Based on these comparisons, it was clear that predictions using the artificial neural network models proposed are very satisfactory.

RUNOFF ESTIMATION IN URBAN CATCHMENT USING ARTIFICIAL NEURAL NETWORK MODELS

Many types of physical models have been developed for runoff estimation with successful results. However, accurate estimation of runoff remains a challenging problem owing to the lack of field data and the complexity of its hydrological process. In this paper, a machine learning method for runoff estimation is presented as an alternative approach to the physical model. Various types of input variables and artificial neural network (ANN) architectures were examined in this study. Results showed that a two-layer network with the tansig activation function and the Levenberg–Marquardt learning algorithm performed the best. For this architecture, the most effective input vector consists of a catchment perimeter, canal length, slope, runoff coefficient, and rainfall intensity. However, results of multivariate analysis of variance indicated the significant interaction effect of input data and the ANN architecture. Thus, to create a suitable ANN model for runoff estimation, a systematic determination of the input vector is necessary

Estimativa de escoamento em captação urbana utilizando modelos de rede neural artificial

Many types of physical models have been developed for runoff estimation with successful results. However, accurate runoff estimation remains a challenging problem owing to the lack of field data and the complexity of its hydrological process. In this paper, a machine learning method for runoff estimation is presented as an alternative approach to the physical model. Various types of input variables and Artificial Neural Network (ANN) architectures were examined in this study. Results showed that a two-layer network with the tansig activation function and the Levenberg–Marquardt learning algorithm had the best performance. For this architecture, the most effective input vector consists of a catchment perimeter, canal length, slope, runoff coefficient, and rainfall intensity. However, results of multivariate analysis of variance indicated the significant interaction effect of input data and the ANN architecture. Thus, to create a suitable ANN model for runoff estimation, a systematic determination of the input vector is necessary.

Development of Artificial Neural Network for Predicting The Photodegradation of Reactive Black 5 Dye

We applied a multilayer artificial neural network (ANN) developed using a Lavenberg–Marquadt algorithm to predict the photodegradation activity of the Reactive Black 5 (RB5) dye. A copper-doped titanium dioxide was employed as a photocatalyst. A copper doped titanium dioxide was synthesized through a wet-impregnation method. To optimize the network the operational parameters including the RB5 initial concentration, photocatalyst dose, irradiation time, hydrogen peroxide concentration, and visible light intensity were used as the input parameter. Removal efficiency of RB5 was selected as output. The number of neurons in the second hidden layer was optimized to determine the suitable ANN model structure for the RB5 removal. ANN based through Levenberg-Marquadth algorithm with structure 1-10-21-1 gave the best performance in this study. The criteria for the applicability of the model were the root mean square error (0.1) and coefficient of correlation (0.98275).

Prediction of Volatile Organic Compounds (VOCs) From Decomposition of Local Household Food Waste Using the Artificial Neural Network

This study examines the potential of artificial neural network (ANN) to predict Total Volatile Organic Compounds (TVOCs) released via decomposition of local food wastes. To mimic the decomposition process, a bioreactor was designed to stimulate the food waste storage condition. The food waste was modeled based on the waste composition from a residential area. A feed forward multilayer back propagation (Levenberg – Marquardt training algorithm) was then developed to predict the TVOCs. The findings indicate that a two-layer artificial neuron network (ANN) with six input variables and these include (outside and inside temperature, pH, moisture content, oxygen level, relative humidity) with a total of eighty eight (88) data are used for the modeling purpose. The network with the highest regression coefficient (R) is 0.9967 and the lowest Mean Square Error (MSE) is 0.00012 (nearest to the value of zero) has been selected as the Optimum ANN model. The findings of this study suggest the most suitable ANN model that befits the research objective is ANN model with one (1) hidden layer with fifteen (15) hidden neurons. Additionally, it is critical to note that the results from the experiment and predicted model are in good agreement.

Future projection and the sales of industrial wood in Turkey: artificial neural networks

In this study, it was aimed to determine artificial neural network models with different architectures using artificial neural network (ANN) methods used in future prediction studies in recent times and forecast the sales quantities of industrial wood in Turkey with the help of models. The sales quantities of logs, mining poles, other industrial wood, pulpwood, fiber-chip wood, and the total of these five wood groups were analyzed separately. The data used in this study was obtained from the General Directorate of Forestry of Turkey and cumulative monthly data covering the period from January 2001 to December 2016 were used. The most suitable ANN models were determined using performance criteria such as mean absolute percentage error (MAPE), root mean square error (RMSE), and determination coefficient (R2). As a result, the R2 and MAPE values of the ANN models were found to be above 99% and below 6%, respectively. The ANNs can be used as a good tool in industrial wood sales forecasts.

Modelling of the Parameters of EDM in Gas Based on Back Propagation Neural Network

The aim of this study is to develop a predicted model of the machining parameters with relation to material removal rate (MRR) and surface roughness (SR) of electrical discharge machining (EDM) in gas. The experimental tasks were implemented by a specific design of experimental method named central composite design (CCD) method. The mathematical prediction models between operating parameters and machining characteristics based on artificial neural network (ANN) were established. The back propagation neural network (BPNN) was employed to construct the architecture of the input layer, the hidden layer and the output layer to build the ANN model. Moreover, the weight and the bias values were examined by the steepest descent method (SDM) with the training data. Thus, the suitable ANN models were established with the acquired weight and bias values. The essential parameters of the EDM in gas such as peak current (Ip), pulse duration (tp), gas pressure (GP), servo reference voltage (Sv) were chosen to investigate the effects on MRR and SR. The developed ANN model with 4 input variables on the input layer, one hidden layer with 5 neurons, and 2 response variables on the output layer was obtained by the training with 30 experimental data. Moreover, as the prediction values obtained from the ANN compared with the 5 testing data, the error falls in the rage of 5% indicating the developed ANN is appropriate and predictable. Moreover, the developed ANN model can be used to predict the machining characteristics such as MRR and SR for the EDM in gas with various parameter settings.

Artificial Neural Network based Harmonics Estimator for a Power Electronics Converter

Objectives: This paper presents harmonics estimation using Artificial Neural Network (ANN) for a 2 pulse Un-controlled power electronics converter. Methods/Analysis: Feed-forward architecture is chosen to model ANN-based Harmonics Estimator. The Feedforward architecture trained with various Learning algorithms is investigated. The suitable ANN model is identified. The performance of ANN based harmonics estimator is compared with conventional Fourier series method. Findings: The feed-forward architecture trained with LM algorithm is identified to be suitable for harmonics estimation in 2-pulse uncontrolled rectifier. Novelty/Improvement: The suitability of feed-forward architecture with different learning algorithms is investigated which is novel in this paper.

Applications of Artificial Neural Networks: A Review

Objectives: The objective of this research paper is study the different methods related to artificial neural network used for prediction and detection of the lung cancer in its early stages so that survival rate of lung cancer patients can be increased. Methods: Lung cancer is the leading cause of death in India so early detection of lung cancer is very important. The detection and prediction of lung cancer was determined with image prepossessing method where segmentation, smoothing and enhancement steps were processed and features were extracted from images and stages of lung cancer were identified with suitable artificial neural network model and also survival rate of lung cancer patients was determined. Findings: Artificial neural network has a significant role in medical area. In these days most of the disease cure methods are process with the help of artificial intelligence to increase the performance of output. In lung cancer disease the artificial neural network model is very useful because detection of lung cancer in its early stages can be determine and it is very important to cure this disease initially because with the increasing stages of lung cancer it is very difficult to cure this disease and also the survival rate of lung cancer patients in higher stages is very low Improvements: The ultimate goal of this paper is to study different methods of Artificial Neural Networks model that can help for detection, prediction and find the survival rate of lung cancer patients.

Computational design and development of novel Al-Mg-Sc-Cr alloy

Purpose – The purpose of this paper is to design and develop precipitation hardened Al-Mg alloy imparting enhanced strength with acceptable ductility through minor addition of Sc and Cr by using multi-objective genetic algorithm-based searching. In earlier attempts of strengthening aluminum alloys, owing to the formation of Al3Sc and Al7Cr phase, addition of Sc and Cr have yielded attractive precipitation hardening, respectively. Both the Al-Sc and Al-Cr system are quench sensitive due to presence of a sloping solvus in their phase diagrams. It is also known that both the Al3Sc and Al7Cr phases nucleate directly from the supersaturated solid solution without formation of GP-zones or transient phases prior to the formation of the Al3Sc and Al7Cr. Sc also found to have beneficial effect on the corrosion property of such alloys. In view of the above, it is of interest to explore the possibility of enhancing the age hardening effect in Al-Mg alloy by addition of Sc and Cr. Design/methodology/approach – The paper uses an approach where experimental information of two different alloy systems (namely, Al-Mg-Sc and Al-Cr) has been combined to generate a single database involving the potential features of both the systems with the aim to formulate the suitable artificial neural network (ANN) models for strength and ductility. The models are used as the objective functions for the optimization process. The patterns of the optimized Pareto front are analyzed to recognize the optimal property of the alloy system. The hitherto unexplored Al-Mg-Sc-Cr alloy, designed from the Pareto solutions and suitably modified on the basis of prior knowledge of the system, is then synthesized and characterized. Findings – The paper has demonstrated the ANN- and genetic algorithm (GA)-based design of a hitherto unexplored alloy by utilizing the existing information concerning the component alloy systems. The paper also established that analyses of the Pareto solutions generated through multi-objective optimization using GA provide an insight of the variation of the parameters at different combination of strength and ductility. It also revealed that the Al-Mg-Sc-Cr alloy has exhibited a two-stage age hardening effect. The first and second stages are due to the precipitation of Al3Sc and Al7Cr phases, respectively. Research limitations/implications – In the present study the two alloy systems are used in tandem to develop models to describe the properties involving the distinct mechanistic features of phase evolution inherent in both the systems. Though the ANN models having the capability to capture huge non-linearity of a system have been employed to predict the convoluted effects of those characteristics when an alloy containing Mg, Sc and Cr are added simultaneously, but the ANN models predictions can be checked experimentally by the future researchers. Practical implications – The paper demonstrates the role of scandium and chromium addition on the ageing characteristics of the alloy by analyzing the age hardening behavior of the designed alloy in cast and cold rolled condition clearly. Originality/value – The approach stated in this paper is a novel one, in the sense that experimental data of two different alloy systems have been clubbed to generate a single database with the aim to formulate the suitable ANN models for strength and ductility.

Modeling and forecasting of Turkey’s energy consumption using socio-economic and demographic variables

This study deals with the modeling of the energy consumption in Turkey in order to forecast future projections based on socio-economic and demographic variables (gross domestic product-GDP, population, import and export amounts, and employment) using artificial neural network (ANN) and regression analyses. For this purpose, four diverse models including different indicators were used in the analyses. As the result of the analyses, this research proposes Model 2 as a suitable ANN model (having four independent variables being GDP, population, the amount of import and export) to efficiently estimate the energy consumption for Turkey. The proposed model predicted the energy consumption better than the regression models and the other three ANN models. Thus, the future energy consumption of Turkey is calculated by means of this model under different scenarios. The predicted forecast results by ANN were compared with the official forecasts. Finally, it was concluded that all the scenarios that were analyzed gave lower estimates of the energy consumption than the MENR projections and these scenarios also showed that the future energy consumption of Turkey would vary between 117.0 and 175.4 Mtoe in 2014.

Free vibration analysis of adhesively bonded single lap joints with wide and narrow functionally graded plates

This study investigates the three-dimensional free vibration behaviour of an adhesively bonded functionally graded single lap joint. The functionally graded plates of the adhesive joint are composed of ceramic (Al 2 O 3 ) and metal (Ni) phases varying through the plate thickness. The effects of geometrical parameters, such as plate width, plate thickness and overlap length, especially the effect of the similar and dissimilar material composition variations through-the-thicknesses of both upper and lower plates on the natural frequencies and corresponding mode shapes of the adhesive joint were also investigated using both the finite element method and the back-propagation artificial neural network (ANN) method. A series of the free vibration analyses were carried out for various random values of the geometrical parameters and the through-the-thickness material composition so that a suitable ANN model could be trained successfully. The proposed ANN models indicated that increasing plate thickness and compositional gradient exponent resulted in increases in the first 10 frequencies whereas the overlap length has negligible effect. In contrast, the natural frequencies decrease suddenly with increasing the plate width. For the plate width >50 mm, the natural frequencies become very low and the effect of the other design parameters on the natural frequencies becomes minor. In case the upper and lower plates have similar or dissimilar material composition variations the mode shapes were affected considerably, but the natural frequencies.

The free vibration analysis and optimal design of an adhesively bonded functionally graded single lap joint

In this study, three-dimensional free vibration and stress analyses of an adhesively bonded functionally graded single lap joint were carried. The effects of the adhesive material properties, such as modulus of elasticity, Poisson's ratio and density were found to be negligible on the first ten natural frequencies and mode shapes of the adhesive joint. Both the finite element method and the back-propagation artificial neural network (ANN) method were used to investigate the effects of the geometrical parameters, such as overlap length, plate thickness and adhesive thickness; and the material composition variation through the plate thickness on the natural frequencies, mode shapes and modal strain energy of the adhesive joint. The suitable ANN models were trained successfully using a series of free vibration and stress analyses for various random geometrical parameters and compositional gradient exponents. The ANN models showed that the support length, the plate thickness and the compositional gradient exponent played important role on the natural frequencies, mode shapes and modal strain energies of the adhesive joint whereas the adhesive thickness had a minor effect. In addition, the optimal joint dimensions and compositional gradient exponent were determined using genetic algorithm and ANN models so that the maximum natural frequency and the minimum modal strain energy conditions are satisfied for each natural frequency of the adhesively bonded functionally graded single lap joint.

A signal-processing tool for non-destructive testing of inaccessible pipes

The design of Non-Destructive-Testing systems for fault detection in long and not accessible pipelines is an actual task in the industrial and civil environment. At this purpose, the diagnosis based on the propagation of guided ultrasonic waves along the pipes offers an attractive solution for the fault identification and classification. The authors studied this problem by means of suitable Artificial Neural Network models. Numerical techniques have been used to simulate the guided wave propagation in the pipes. In particular, the finite element method has been used to model different kinds of pipes and faults, and to obtain several returning echoes containing the faults information. Torsional wave modes have been used as excitation waves. The obtained signals have been processed in order to reduce the data dimensionality, and to extract suitable features. The features selected from the signals can be further processed in order to limit the size of the Neural Network models without loss of information. At this purpose, the principal component analysis has been investigated. Finally, the selected features have been used as input for the neural network models. In this paper, traditional feed-forward, multi-layer perceptron networks have been used to obtain the information on size and location of localized notches.

Development of bond strength model for FRP plates using back‐propagation algorithm

AbstractFor the flexural reinforcement of bridge and building structure, synthetic materials whose dynamic properties are superior and those containing the merit of corrosion‐proof are widely used as the substitute for a steel plate. Since FRP plate has improved bond strength owing to the fibers externally adhering to the plate, many researches regarding the bond strength improvement have been substantially performed. To search out such bond strength improvement, previous researchers had ever examined the bond strength of FRP plate through their experiment by setting up many variables. However, since the experiment for a research on the bond strength takes much of expenditure for setting up the equipment and is time‐consuming, also is difficult to be carried out, it is limitedly conducted. The purpose of this study was to develop the most suitable artificial neural network model by application of various neural network models and algorithm to the data of the bond strength experiment conducted by previous researchers. Many variables were used as input layers against bond strength: depth, width, modulus of elasticity, tensile strength of FRP plate and the compressive strength, tensile strength, and width of concrete. The developed artificial neural network model has been applied back‐propagation, and its error was learned to be converged within the range of 0.001. Besides, the process for the over‐fitting problem has been dissolved by Bayesian technique. The verification on the developed model was executed by comparison with the test results of bond strength made by other previous researchers, which was never been utilized to the learning as yet. © 2006 Wiley Periodicals, Inc. J Appl Polym Sci 100: 5119–5127, 2006

Optimal design of an adhesively-bonded corner joint with single support based on the free vibration analysis

This study investigates the three-dimensional free vibration behaviour of an adhesively-bonded corner joint with single support. The modulus of elasticity, Poisson's ratio and density of adhesive were found to have negligible effects on the first 10 natural frequencies and mode shapes of the corner joint. The effects of the geometrical parameters, such as support length, plate thickness, adhesive thickness and joint length, on the natural frequencies, mode shapes and modal strain energies of the adhesive joint were also investigated using both the finite element method and the back-propagation artificial neural network (ANN) method. The free vibration and stress analyses were carried out for the corner joints with various random geometrical parameters so that a suitable ANN model could be trained successfully. The support length, plate thickness and joint length all played important roles in the natural frequencies, mode shapes and modal strain energies of the corner joint, whereas the adhesive thickness for the range of adhesive thickness studied had only a minor effect. The Genetic Algorithm was also combined with the present ANN models in order to determine the optimum geometrical dimensions which satisfied the maximum natural frequency and minimum modal strain energy conditions for each natural frequency and mode shape of the adhesively-bonded corner joint.

Optimum design for artificial neural networks: an example in a bicycle derailleur system

The integration of neural networks and optimization provides a tool for designing network parameters and improving network performance. In this paper, the Taguchi method and the Design of Experiment (DOE) methodology are used to optimize network parameters. The users have to recognize the application problems and choose a suitable Artificial Neural Network model. Optimization problems can then be defined according to the model. The Taguchi method is first applied to a problem to find out the more important factors, then the DOE methodology is used for further analysis and forecasting. A Learning Vector Quantization example is shown for an application to bicycle derailleur systems.