Inputs Of Artificial Neural Network Research Articles

Effects of Model Structural Complexity and Data Pre-Processing on Artificial Neural Network (ANN) Forecast Performance for Hydrological Process Modelling

The choice of a particular Artificial Neural Network (ANN) structure is a seemingly difficult task; worthy of relevance is that there is no systematic way for establishing a suitable architecture. In view of this, the study looked at the effects of ANN structural complexity and data pre-processing regime on its forecast performance. To address this aim, two ANN structural configurations: 1) Single-hidden layer, and 2) Double-hidden layer feed-forward back propagation network were employed. Results obtained revealed generally that: a) ANN comprised of double hidden layers tends to be less robust and converges with less accuracy than its single-hidden layer counterpart under identical situations; b) for a univariate time series, phase-space reconstruction using embedding dimension which is based on dynamical systems theory is an effective way for determining the appropriate number of ANN input neurons, and c) data pre-processing via the scaling approach excessively limits the output range of the transfer function. In specific terms considering extreme flow prediction capability on the basis of effective correlation: Percent maximum and minimum correlation coefficient (Rmax% and Rmin%), on the average for one-day ahead forecast during the training and validation phases respectively for the adopted network structures: 8 7 5 (i.e., 8 input nodes, 7 nodes in the hidden layer, and 5 output nodes in the output layer), 8 5 2 5 (8 nodes in the input layer, 5 nodes in the first hidden layer, 2 nodes in the second hidden layer, and 5 nodes in the output layer), and 8 4 3 5 (8 nodes in the input layer, 4 nodes in the first hidden layer, 3 nodes in the second hidden layer, and 5 nodes in the output layer) gave: 101.2, 99.4; 100.2, 218.3; 93.7, 95.0 in all instances irrespective of the training algorithm (i.e., pooled). On the other hand, in terms of percent of correct event prediction, the respective performances of the models for both low and high flows during the training and validation phases, respectively were: 0.78, 0.96: 0.65, 0.87; 0.76, 0.93: 0.61, 0.83; and 0.79, 0.96: 0.65, 0.87. Thus, it suffices to note that on the basis of coherence or regularity of prediction consistency, the ANN model: 8 4 3 5 performed better. This implies that though the adoption of large hidden layers vis-à-vis corresponding large neuronal signatures could be counter-productive because of network over-fitting, however, it may provide additional representational power. Based on the findings, it is imperative to note that ANN model is by no means a substitute for conceptual watershed modelling, therefore, exogenous variables should be incorporated in streamflow modelling and forecasting exercise because of their hydrologic evolutions.

Prediction of biological activity of compounds containing a 1,3,5-triazinyl sulfonamide scaffold by artificial neural networks using simple molecular descriptors

Simple molecular descriptors of extensive series of 1,3,5-triazinyl sulfonamide derivatives, based on the structure of sulfonamides and their physicochemical properties, were designed and calculated. These descriptors were successfully applied as inputs for artificial neural network (ANN) modelling of the relationship between the structure and biological activity. The optimized ANN architecture was applied to the prediction of the inhibition activity of 1,3,5-triazinyl sulfonamides against human carbonic anhydrase (hCA) II, tumour-associated hCA IX, and their selectivity (hCA II/hCA IX).

Physics-inspired integrated space–time artificial neural networks for regional groundwater flow modeling

Abstract. An integrated space–time artificial neural network (ANN) model inspired by the governing groundwater flow equation was developed to test whether a single ANN is capable of modeling regional groundwater flow systems. Model-independent entropy measures and random forest (RF)-based feature selection procedures were used to identify suitable inputs for ANNs. L2 regularization, five-fold cross-validation, and an adaptive stochastic gradient descent (ADAM) algorithm led to a parsimonious ANN model for a 30 691 km2 agriculturally intensive area in the Ogallala Aquifer of Texas. The model testing at 38 independent wells during the 1956–2008 calibration period showed no overfitting issues and highlighted the model's ability to capture both the observed spatial dependence and temporal variability. The forecasting period (2009–2015) was marked by extreme climate variability in the region and served to evaluate the extrapolation capabilities of the model. While ANN models are universal interpolators, the model was able to capture the general trends and provide groundwater level estimates that were better than using historical means. Model sensitivity analysis indicated that pumping was the most sensitive process. Incorporation of spatial variability was more critical than capturing temporal persistence. The use of the standardized precipitation–evapotranspiration index (SPEI) as a surrogate for pumping was generally adequate but was unable to capture the heterogeneous groundwater extraction preferences of farmers under extreme climate conditions.

A Novel Assisted Artificial Neural Network Modeling Approach for Improved Accuracy Using Small Datasets: Application in Residual Strength Evaluation of Panels with Multiple Site Damage Cracks

An artificial neural network (ANN) extracts knowledge from a training dataset and uses this acquired knowledge to forecast outputs for any new set of inputs. When the input/output relations are complex and highly non-linear, the ANN needs a relatively large training dataset (hundreds of data points) to capture these relations adequately. This paper introduces a novel assisted-ANN modeling approach that enables the development of ANNs using small datasets, while maintaining high prediction accuracy. This approach uses parameters that are obtained using the known input/output relations (partial or full relations). These so called assistance parameters are included as ANN inputs in addition to the traditional direct independent inputs. The proposed assisted approach is applied for predicting the residual strength of panels with multiple site damage (MSD) cracks. Different assistance levels (four levels) and different training dataset sizes (from 75 down to 22 data points) are investigated, and the results are compared to the traditional approach. The results show that the assisted approach helps in achieving high predictions’ accuracy (<3% average error). The relative accuracy improvement is higher (up to 46%) for ANN learning algorithms that give lower prediction accuracy. Also, the relative accuracy improvement becomes more significant (up to 38%) for smaller dataset sizes.

Optimal Learning Behavior Prediction System Based on Cognitive Style Using Adaptive Optimization-Based Neural Network

Widespread development of system software, the process of learning, and the excellence in profession of teaching are the formidable challenges faced by the learning behavior prediction system. The learning styles of teachers have different kinds of content designs to enhance their learning. In this learning environment, teachers can work together with the students, but the learning materials are designed by the teachers. The cognitive style deals with mental activities such as learning, remembering, thinking, and the usage of language. Therefore, being motivated by the problems mentioned above, this paper proposes the concept of adaptive optimization-based neural network (AONN). The learning behavior and browsing behavior features are extracted and incorporated into the input of artificial neural network (ANN). Hence, in this paper, the neural network weights are optimized with the use of grey wolf optimizer (GWO) algorithm. The output operation of e-learning with teaching equipment is chosen based on the cognitive style predicted by AONN. In experimental section, the measures of accuracy, sensitivity, specificity, time (sec), and memory (bytes) are carried out. Each of the measure is compared with the proposed AONN and existing fuzzy logic methodologies. Ultimately, the proposed AONN method produces higher accuracy, specificity, and sensitivity results. The results demonstrate that the algorithm proposed in this study can automatically learn network structures competitively, unlike those achieved for neural networks through standard approaches.

Accuracy and Transferability of Artificial Neural Networks in Predicting in Situ Root-Zone Soil Moisture for Various Regions across the Globe

This paper explores the accuracy in using an artificial neural network (ANN) to estimate root-zone soil moisture (RZSM) at multiple worldwide locations using only in situ surface soil moisture (SSM) as a training dataset. The paper also addresses the transferability of the trained ANN across climatic and soil texture conditions. Data from the International Soil Moisture Network (ISMN) were collected for several networks with variable soil texture and climate classes. Several scaling, feature extraction, and training approaches were tested. An artificial neural network employing rolling averages (ANNRAV) of SSM over 10, 30, and 90 days was developed. The results show that applying a standard scaling (SSCA) to the ANN input features improves the correlation, Nash–Sutcliffe efficiency (NSE), and root mean square error (RMSE) for 52%, 91%, and 87%, respectively, of the tested stations, compared to MinMax scaling (MMSCA). Different training sets are suggested, namely, training on data from all networks, data from one network, or data of all networks excluding one. Based on these trainings, new transferability (TranI) and contribution (ContI) indices are defined. The results show that one network cannot provide the best prediction accuracy if used alone to train the ANN. They also show that the removal of the less contributing networks enhances performance. For example, elimination of the densest network (SCAN) from the training enhances the mean correlation by 20.5% and the mean NSE by 42.5%. This motivates the implementation of a data filtering technique based on the ANN’s performance. A median, max, and min correlation of 0.77, 0.96, and 0.65, respectively, are obtained by the model after data filtering. The performances are also analyzed with respect to the covered climatic regions and soil texture, providing insights into the robustness and limitations of the approach, namely, the need for complementary information in highly evaporative regions. In fact, the ANN using only SSM to predict RZSM has low performance when decoupling between the surface and root zones is observed. The application of ANN to obtain spatialized RZSM will require integrating remote sensing-based surface soil moisture in the future.

Development of an improved method for finding a solution for neuro-fuzzy expert systems

Nowadays, artificial intelligence has entered into all spheres of human activity. However, there are some problems in the analysis of objects, for example, there is a priori uncertainty about the state of objects and the analysis takes place in a difficult situation against the background of intentional (natural) interference and uncertainty. The best solution in this situation is to integrate with the data analysis of information systems and artificial neural networks. This paper develops an improved method for finding solutions for neuro-fuzzy expert systems. The proposed method allows increasing the efficiency and reliability of making decisions about the state of the object. Increased efficiency is achieved through the use of evolving neuro-fuzzy artificial neural networks, as well as an improved procedure for their training. Training of evolving neuro-fuzzy artificial neural networks is due to learning their architecture, synaptic weights, type and parameters of the membership function, as well as the application of the procedure of reducing the dimensionality of the feature space. The analysis of objects also takes into account the degree of uncertainty about their condition. In the proposed method, when searching for a solution, the same conditions are calculated once, which speeds up the rule revision cycle and instead of the same conditions of the rules, references to them are used. This reduces the computational complexity of decision-making and does not accumulate errors in the training of artificial neural networks as a result of processing the information coming to the input of artificial neural networks. The use of the proposed method was tested on the example of assessing the state of the radio-electronic environment. This example showed an increase in the efficiency of assessment at the level of 20–25 % by the efficiency of information processing

Development and Application of Ion Current/Cylinder Pressure Cooperative Combustion Diagnosis and Control System

The application of advanced technologies for engine efficiency improvement and emissions reduction also increase the occurrence possibility of abnormal combustions such as incomplete combustion, misfire, knock or pre-ignition. Novel promising combustion modes, which are basically dominated by chemical reaction kinetics show a major difficulty in combustion control. The challenge in precise combustion control is hard to overcome by the traditional engine map-based control method because it cannot monitor the combustion state of each cycle, hence, real-time cycle-resolved in-cylinder combustion diagnosis and control are required. In the past, cylinder pressure and ion current sensors, as the two most commonly used sensors for in-cylinder combustion diagnosis and control, have enjoyed a seemingly competitive relationship, so all related researches only use one of the sensors. However, these two sensors have their own unique features. In this study, the idea is to combine the information obtained from both sensors. At first, two kinds of ion current detection system are comprehensively introduced and compared at the hardware level and signal level. The most promising variant (the DC-Power ion current detection system) is selected for the subsequent experiments. Then, the concept of ion current/cylinder pressure cooperative combustion diagnosis and control system is illustrated and implemented on the engine prototyping control unit. One application case of employing this system for homogenous charge compression ignition abnormal combustion control and its stability improvement is introduced. The results show that a combination of ion current and cylinder pressure signals can provide richer and also necessary information for combustion control. Finally, ion current and cylinder pressure signals are employed as inputs of artificial neural network (ANN) models for combustion prediction. The results show that the combustion prediction performance is better when the inputs are a combination of both signals, instead of using only one of them. This offline analysis proves the feasibility of using an ANN-based model whose inputs are a combination of ion current and pressure signals for better prediction accuracy.

An assessment of shallow neural networks for stress updates in computational solid mechanics

How do computational costs of shallow Artificial Neural Networks (ANNs) compare with conventional schemes for three-dimensional stress updates in elastic, nonlinear elastic, and elastic-plastic solids? We implement ANNs in an explicit finite element solver using strain increments and total strains as inputs for the first two materials, respectively. For elastic-plastic solids, ANN inputs are derivable from three-dimensional displacement fields. Though we minimized size of network architectures and input data sets, we conclude that obtaining reliable multi-axial training data remains a substantial challenge; and for all three materials, computational costs for ANNs exceed costs of conventional methods.

Automatic Extraction of the Non-Urban Roads from Satellite Images using Artificial Neural Network

Road detection from the satellite images can be considered as a classification process in which pixels are divided into the road and non-road classes. In this research, an automatic road extraction using an artificial neural network (ANN) based on automatic information extraction from satellite images and self-adjusting of the hidden layer proposed. Parameters of non-urban road networks from satellite images using a histogram-based binary image segmentation technique are also presented. The segmentation method is implemented by determining a global threshold, which is obtained from a statistical analysis of a number of sample satellite images and their ground truths. The thresholding method is based on two major facts: first, the points corresponding to non-asphalt roads are brighter than other areas in non-urban images. Second, it is observed that in an aerial image, the area covered by roads is only a small fraction of total pixels. It is also observed that pixels corresponding to roads are generally populated at the very bright end of the image greyscale histogram. In this method, at first, the possible road pixels are selected by the proposed segmentation method. Then different parameters, including color, gradient, and entropy, are computed for each pixel from the source image. Finally, these features are used for the artificial neural network input. The results show that the accuracy of the proposed road extraction method is around 80%.

Efficient Structural Design of a Prefab Concrete Connection by Using Artificial Neural Networks

In the built environment, one of the main concerns during the design stage is the selection of adequate structural materials and elements. A rational and sensible design of both materials and elements results not only in economic benefits and computing time reduction, but also in minimizing the environmental impact. Nowadays, Artificial Neural Networks (ANNs) are showing their potential as design tools. In this research, ANNs are used in order to foster the implementation of efficient tools to be used during the early stages of structural design. The proposed networks are applied to a dry precast concrete connection, which has been modelled by means of the Finite Element Method (FEM). The parameters are: strength of concrete and screws, diameter of screws, plate thickness, and the posttensioning load. The ANN input data are the parameters and nodal stresses obtained from the FEM models. A multilayer perceptron combined with a backpropagation algorithm is used in the ANN architecture, and a hyperbolic tangent function is applied as an activation function. Comparing the obtained predicted stresses to those of the FEM analyses, the difference is less than 9.16%. Those results validate their use as an efficient structural design tool. The main advantage of the proposed ANNs is that they can be easily and effectively adapted to different connection parameters. In addition, their use could be applied both in precast or cast in situ concrete connection design.

Recognizing of the best different artificial intelligence method for determination of local scour depth around group piers in equilibrium time

Unlike calculation of scour depth around of a single pier, estimation of scour depth around group piers is a complex problem. In this research, four types of artificial neural networks (ANNs) were applied multi layer perceptron (MLP), radial basis function (RBF), neuro fuzzy inference system (ANFIS), and support vector machine (SVM). The inputs of ANNs were coordinates of considered points in the channel (X, Y), size of pier, flow depth, flow discharge, flow velocity, and critical velocity for inception of sediment transport. The trained methods were momentum and Levenberg–Marquardt (LM) training methods. Also, genetic algorithm (GA) was applied for optimization of MLP and RBF. Two group piers were used (three circular piers and three square piers) in a channel with stable flow conditions, scouring of clear water, and uniform bed particles (with D50 = 1.31 mm). Outputs of different ANNs (scour depth) were compared with observed data using of performance criteria correlation coefficient (R), mean square error (MSE), and mean absolute error (MAE). For circular piers, between square piers and behind of gear square pier, the best ANN is RBF-GA, while in front of first square pier, the best ANN is MLP-GA. The best ANNs have two hidden layers, and their training method is the Levenberg–Marquardt (LM). The values of R, MSE, and MAE of RBF-GA are 0.997, 1.66 mm2, and 0.9 mm, and the values of R, MSE, and MAE of MLP-GA are 0.997, 2.55 mm2, and 1.16 mm. Also, the sensitive analysis shows that threshold velocity (Uc) is the most effective factor on scour depth

Analysis and detection of electric field distribution characteristics of degraded insulators

Abstract After long-term operation, the electrical performance of UHV transmission line insulator string decreases and zero-value insulators are prone to occur. The existence of zero-value insulators leads to the flashover of insulator strings, which causes large-scale power outages in serious cases. In order to improve the power quality and detect the zero-value insulators of transmission lines more timely and efficiently, the influence of zero-value insulators on the electric field distribution of insulator strings is simulated and calculated by finite element method. At the same time, taking the grounding current of insulator string as the cut-in point, the principal component analysis method is used to screen the effective characteristic quantities, and then as the input of artificial neural network, and the fault area of insulator string is determined by the neural network. The simulation results show that the recognition accuracy of artificial neural network based on insulator string grounding current is high, and it has certain engineering application value.

Modeling LWR fuel Rod's gap thickness heat transfer coefficient by artificial neural network technique

During the lifetime of nuclear Light Water Reactors (LWRs), the fuel-cladding gap thickness heat transfer coefficient (hgap) is the most crucial parameter of fuel rod performance that determines the fuel rod thermo-mechanical behavior as well as the maximum level of fuel burn-up during the normal operation. The main objective of the present work is to develop a smart calculative method by using the artificial neural network (ANN) technique to predict the hgap for a fuel rod of LWRs. The parameters of ANN input include those of fuel design gap initial thickness and pressure, operation time, Linear Heat Generation (LHG), and the average fuel burn-up while hgap is considered as the only output parameter. The obtained results of FRAPCON-3.5 steady-state fuel rod performance code is used to form the training data set. A Multi-Layer Perceptron feed-forward ANN with one hidden layer is trained with the Levenberg–Marquardt training algorithm. It has been illustrated that the created artificial network can accurately predict the hgap. Through the weight matrix of the ANN, Garson's sensitive analysis procedure shows the significant role of fuel burn-up in determining the level of hgap.

LOAD SCHEDULING FOR SMART HOME USING DAY-AHEAD PREDICTION FROM ARTIFICIAL NEURAL NETWORK (ANN)

This paper presents load scheduling for smart home application using day-ahead prediction from an artificial neural network (ANN). In this study, load forecasting using ANN approach is embedded in the load scheduling scheme that is modeled using mixed integer linear programming (MILP). The main objective of the scheduling is to reduce the electricity bill by shifting peak load to off-peak period. A day-ahead energy consumption is predicted based on a previous yearly data set of hourly resolution. The dataset is normalized and injected as input in ANN and the result is then fed to the load scheduling optimization process. The results show that the integration process affects the allocation of load consumption in the load profile as well as the electricity cost. From the comparative study between before and after ANN integration, the total cost saving achieved is $1.53/day with the cost reduction of 38.44%.

Prediction of tire pattern noise in early design stage based on convolutional neural network

In early design stage of tire pattern, it is very useful to predict noise level associated with tire pattern. Artificial neural network (ANN) was used for development of the model for the prediction of tire pattern noise recently. The ANN used supervised training method which extracts the feature applying Gaussian curve fitting to the tread profile spectrum of tire pattern and used it as the input of ANN. This method requests laser scanning for tire pattern of a real tire. In early design, there is no real tire. In this study, the convolutional neural network (CNN) to predict tire pattern noise was developed based on non-supervised training method. Two Learning algorithms such as stochastic gradient descent (SGD) and RMSProp were studied in the CNN model for the comparison of their learning performance. RMSProp algorithm was suggested for the CNN model. In this case, a pattern image of a tire to be designed was used as the input of CNN. The CNN to predict tire pattern noise was developed and its utility in the early design stage of tire was discussed. In the study, pattern noise for 28 tires were measured in the semi- anechoic chamber and their pattern images were scanned. For the training of ANN and CNN, pattern noise for 24 tires and their pattern images were used. The trained ANN and CNN were validated respectively with 4 tires which were not used for the training of two neural networks. Finally, two networks were successfully developed and validated for the prediction of tire pattern noise. The trained CNN can be used for the prediction of pattern noise for a tire to be designed in early design stage using the only drawing image of tire whilst ANN can be used for the prediction of pattern noise for a real tire in development stage.

STATIC AIR-GAP ECCENTRICITY FAULT DETECTION OF INDUCTION MOTOR USING ARTIFICIAL NEURAL NETWORK (ANN)

Induction motor plays an important role in the industrial, commercial and residential industries, owing to its immense advantages over the opposite types of motors. Such motors have to operate under different operating conditions that cause engine degradation leading to fault occurrences. There are numerous fault detection techniques available. There are numerous fault detection techniques available. The technique used in this paper to prove the effect of static air gap eccentricity on behaving or performing of the three-phase induction motor is the artificial neural network (ANN) as ANN depends on detecting the fault on the amplitude of positive and negative harmonics of frequencies. In this paper, we used two motors to achieve real malfunctions and to get the required data and for three different load tests. In this paper, we adopted MCSA to detect the fault based on the stator current. The ANN training algorithm used in this paper is back propagation and feed forward. The inputs of ANN are the speed and the amplitudes of the positive and the negative harmonics, and the type of fault is the output. To distinguish between healthy and faulty motor, the input data of ANN are well-trained via experiments test. The methodology applied in this paper was MATLAB and present how we can distinguish between healthy and faulty motor.

The flood prediction model using Artificial Neural Network (ANN) and weather Application Programming Interface (API) as an alternative effort to flood mitigation in the Jenelata Sub-watershed

Jenelata is a sub of the Jeneberang watershed in South Sulawesi which has rainfall intensity from 2,800 mm to 4,000 mm per year on the upstream area, some of the rainfall occurs in a short period with high intensity, resulting significant rise to river water level. It potentially to cause floods downstream. Topography of the upstream area is mountainous with an average slope of ± 0.024 along the river, with flow length of 38,314 km it has a large velocity flow. Purpose of this research is to providing alternative for flood mitigation using rainfall predictions and runoff calculations, hypoteticaly it can reduce the impact of possible flood. This research was conducted using Artificial Neural Network (ANN) as rainfall predictor and input variable for runoff calculation using SCS method. ANN input variable will use weather prediction data from the global weather API, while SCS method will calculate maximum runoff in the catchment for the next 24 hours. Results of the rainfall prediction model get deviation of 28.81 mm and accuracy of 58% compared to the observations. Meanwhile, runoff model discharge acquires deviation of 1181.7 m3 and water level ± 0.19 m at the designated location of water level gauge.

Forecasting Iraqi oil production using artificial neural networks

Artificial neural networks (ANNs) are bendy computational structures and yet ordinary approximations as the execute stand applied to day selection predicting especially correct problems. The paper aims to address the effect on the outputs of Artificial Neural Network of the number of input layer nodes, and to minimize error. Time series views included the amount of Iraqi oil production for the 2011–2019 period. The most important finding of this paper is the greater the number of artificial neural network inputs, the smaller the error, and the improved performance results, which reflect positively on the prediction results.

DEVELOPMENT OF AN ADVANCED METHOD OF FINDING SOLUTIONS FOR NEURO-FUZZY EXPERT SYSTEMS OF ANALYSIS OF THE RADIOELECTRONIC SITUATION

Nowadays, artificial intelligence has entered into all spheres of our life. The system of analysis of the electronic environment is not an exception. However, there are a number of problems in the analysis of the electronic environment, namely the signals. They are analyzed in a complex electronic environment against the background of intentional and natural interference. Also, the input signals do not match the standards due to the influence of different types of interference. Interpretation of signals depends on the experience of the operator, the completeness of additional information on a specific condition of uncertainty. The best solution in this situation is to integrate with the data of the information system analysis of the electronic environment and artificial neural networks. Their advantage is also the ability to work in real time and quick adaptation to specific situations. These circumstances cause uncertainty in the conditions of the task of signal recognition and fuzzy statements in their interpretation, when the additional involved information may be incomplete and the operator makes decisions based on their experience. That is why, in this article, an improved method for finding solutions for neuro-fuzzy expert systems of analysis of the electronic environment is developed. Improving the efficiency of information processing (reducing the error) of evaluation is achieved through the use of neuro-fuzzy artificial neural networks that are evolving and learning not only the synaptic weights of the artificial neural network, but also the type and parameters of the membership function. High efficiency of information processing is also achieved through training in the architecture of artificial neural networks by taking into account the type of uncertainty of the information that has to be assessed and work with clear and fuzzy products. This reduces the computational complexity of decision-making and absence of accumulation of an error of training of artificial neural networks as a result of processing of the arriving information on an input of artificial neural networks. The use of the proposed method was tested on the example of assessing the state of the electronic environment. This example showed an increase in the efficiency of assessment at the level of 20–25 % on the efficiency of the processing information