Hidden Layer Nodes Research Articles

Morphological characterization of Maize (Zea mays.) utilising the stage-wise structural and architectural perspective from temporal fully-polarimetric SAR

The morphological shape and structure of the crop vary with phenological stages. Model and eigen based decomposition model parameters extracted from the Radarsat-2 data and the trend with respect to ground truth crop phenology were analysed. Sensitive parameters were devised through stepwise approach under 7 combinations of polarimetric variables of increasing complexity were assessed. Compared under the three machine learning algorithms (ANN, RF and SVM) where ANN rendered the maximum correlation with 0.92 with a MAE of 4 days which was implemented on a large parcel of maize mask in the study area. SVM performed poorly with highly overlapping parameters such as backscatter but performed well (r = 0.85). For assessing the crop biophysical parameters, the three algorithms were evaluated and sensitivity analysis for statistically significant polarimetric variables for biophysical parameters was performed. The assessment was performed on Multi-Layer Perception (MLP) neural network. The networks were trained with algorithms and hidden layer nodes until the MAE achieved permissible limits. Plant height could be estimated more profoundly with an r = 0.8 with a considerably good MAE of 24.9 cm but other parameters (WB, DB and LAI) were estimated in moderate correlation of 0.6–0.65 where the MAE of WB, DB and LAI were found to be 1317gm−2, 553 gm−2 and 0.78 respectively. This is the first step towards understanding the complex scattering mechanisms in Indian maize scenario assessing the growth parameters from polarimetric data. Thus, the analytical findings brought out possess the potential to serve as the reference for the future research initiatives.

Prediction and simulation of wearable sensor devices for sports injury prevention based on BP neural network

In the research of sports injury prevention, the recognition of sports action plays an important role in the action recognition model and the prediction evaluation model. In view of the above problems, this paper constructs a new mathematical model through the idea of BP neural network. The model combines wearable technology and can improve the recognition accuracy of sports actions. The model uses the BP network classifier, which can be used in data. Processes such as feature extraction improve efficiency and reliability. In this paper, the algorithm simulation is carried out, and the experiments are carried out for three movements: running, running and static. The results show that for the recognition of running and running action, when the hidden layer node is 11, the BP neural network classifier shows the best recognition effect. For static motion, the recognition effect of each classifier is basically the same. This paper analyzes the wearable sports action recognition system, including perception layer, application layer and service layer, to realize the recognition and classification of sports actions, predict actions in advance and prevent sports injuries. Finally, this paper analyzes the causes of sports injury, puts forward specific measures to prevent sports injury, and further reduces sports injury events through wearable devices.

Research on renewable energy power demand forecasting method based on IWOA-SA-BILSTM modeling

This paper introduces a novel coupling method to enhance the precision of short- and medium-term renewable energy power load demand forecasting. Firstly, the Tent chaotic mapping incorporates the standard WOA and modifies its internal convergence factor to a nonlinear convergence mode, resulting in an improved IWOA. It is used for the weight optimization part of BILSTM. Then, the SA is introduced to optimize the learning rate, the number of nodes in hidden layers 1 and 2, and the number of iterations of BILSTM, constructing an IWOA-SA-BILSTM prediction model. Finally, through case analysis, the prediction model proposed in this paper has the highest improvement of 76.7%, 74.5%, and 45.9% in terms of Mean Absolute Error, Root Mean Square Error, and R2, respectively, compared to other optimal benchmark models, proving the effectiveness of the model.

Fast and noise-robust quantum state tomography based on ELM

This paper proposed a quantum state tomography approach based on the extreme learning machine (ELM), which is available in the reconstruction of quantum states via a lightweight neural network. The key step of the proposed tomography approach is to employ the ELM to approximate the complex mapping between the measurement values sequence and the real density matrix. After obtaining the output of the ELM-based estimator, a matrix transformation technique is used to make the network outputs satisfy quantum state constraints. Compared with deep learning-based tomography approaches, our proposed ELM-based approach enables both high-fidelity and high-efficiency quantum state tomography with only one training process under the condition of very few numbers of training samples, network layers and hidden layer nodes. In addition, the proposed tomography approach is robust to noisy measurement values, since the ELM-based estimator is quite lightweight. Simulations on the tomography of eigenstates, superposition states and mixed states are presented to verify our theoretical findings. Also, the superiority of the ELM-based tomography approach is demonstrated in comparison with that based on the radial basis function network, convolutional neural network and maximum likelihood estimation approach.

Developing a hybrid model for accurate short-term water demand prediction under extreme weather conditions: a case study in Melbourne, Australia

Accurate prediction of short-term water demand, especially, in the case of extreme weather conditions such as flood, droughts and storms, is crucial information for the policy makers to manage the availability of freshwater. This study develops a hybrid model for the prediction of monthly water demand using the database of monthly urban water consumption in Melbourne, Australia. The dataset consisted of minimum, maximum, and mean temperature (°C), evaporation (mm), rainfall (mm), solar radiation (MJ/m2), maximum relative humidity (%), vapor pressure (hpa), and potential evapotranspiration (mm). The dataset was normalized using natural logarithm and denoized then by employing the discrete wavelet transform. Principle component analysis was used to determine which predictors were most reliable. Hybrid model development included the optimization of ANN coefficients (its weights and biases) using adaptive guided differential evolution algorithm. Post-optimization ANN model was trained using eleven different leaning algorithms. Models were trained several times with different configuration (nodes in hidden layers) to achieve better accuracy. The final optimum learning algorithm was selected based on the performance values (regression; mean absolute, relative and maximum error) and Taylor diagram.

Data-Mining-Based Hardware-Efficient Neural Network Controller for DC–DC Switching Converters

For the exiting neural network (NN)-based controllers, since the training data are not optimized using different control algorithms, their controlling performance improvements are limited. This paper proposes a data-mining-based hardware-efficient NN controller for DC–DC switching converters. Firstly, the data mining strategy for NN controllers is proposed to optimize the training data. In data mining process, the optimal control data are selected from different control algorithms to create an optimized dataset for training NN controller. In addition, to reduce the hardware resources, the structure of a hardware-efficient NN controller is proposed, which comprises coarse-tuning NN (NN-ct) and fine-tuning NN (NN-ft) controllers, and the switching between two NN controllers is conducted based on the output voltage variation. The NN-ct and NN-ft controllers are trained using the optimized fine-tuning and coarse-tuning sub-datasets, respectively. By switching the weights and biases and disabling some hidden-layer nodes, two separate NN controllers can share some hardware resources, thus reducing the hardware cost without sacrificing the performance. The simulation and experimental results demonstrate that data-mining-based NN controller has better performance than other NN controllers. Furthermore, the hardware resource requirement of a hardware-efficient NN controller can be reduced by at least 17% when compared with an ordinary NN controller.

Super Typhoon Rai’s Impacts on Siargao Tourism: Deciphering Tourists’ Revisit Intentions through Machine-Learning Algorithms

Super Typhoon Rai damaged Siargao’s tourism industry. Despite the reconstruction projects, there was still evidence of limited resources, destructed infrastructures, and destroyed natural resources. Therefore, this study aimed to examine the significant factors influencing tourists’ intentions to revisit Siargao after Super Typhoon Rai using feature selection, logistic regression (LR), and an artificial neural network (ANN). It employed three feature-selection techniques, namely, the filter method’s permutation importance (PI), the wrapper method’s Recursive Feature Elimination (RFE), and the embedded method’s Least Absolute Shrinkage and Selection Operator (LASSO). Each feature-selection technique was integrated into LR and the ANN. LASSO-ANN, with a 97.8146% model accuracy, was found to be the best machine-learning algorithm. The LASSO model performed at its best with a 0.0007 LASSO alpha value, resulting in 35 subfeatures and 8 primary features. LASSO subsets underwent the ANN model procedure, and the optimal parameter combination was 70% training size, 30% testing size, 30 hidden-layer nodes, tanh hidden-layer activation, sigmoid output-layer activation, and Adam optimization. All eight features were found to be significant. Among them, hedonic motivation and awareness of Typhoon Rai’s impact were considered the top-tier post-typhoon tourism factors, as they maintained at least 97% prediction accuracy. The findings could be elaborated by combining feature-selection techniques, utilizing demographic characteristics, assessing Siargao’s tourism before the typhoon, and expanding the context and participant selection. Nevertheless, none of the existing studies explored the combination of feature selection, LR, and ANNs in a post-typhoon tourism context. These unique methods and significant findings represent the study’s novelty. Furthermore, practical contributions were provided through economic resolutions focusing on tourism activities and communication revamping by the government, media outlets, and transportation companies.

Low-voltage series arc fault detection based on ECMC and VB-SCN

The currents of low-voltage series arc faults are non-stationary and nonlinear, so the selection of fault features and the adaptability of the detection algorithm will affect the detection results. In response to these issues, firstly, a feature selection method combining Euclidean distance, classifier criterion, max-relevance min-redundancy and clustering index is proposed, ECMC. This method can filter out weak correlation features and redundant information in the time, frequency and time–frequency domains. The optimal feature data set constructed by ECMC can fully reflect the characteristics of arc faults. Then, a stochastic configuration network (SCN) based on variational Bayesian (VB) optimization, VB-SCN, is proposed. This method uses VB to optimize the output weight and scale function of SCN iteratively. VB-SCN can improve the quality of hidden-layer nodes and the generalization ability of the network while ensuring the adaptive learning of current features, making arc fault detection fast and stable. Finally, the experiments based on the self-built sample set prove that the optimal feature dataset constructed by ECMC has perfect separability. VB-SCN can adaptively realize the accurate and fast detection of arc faults. The detection accuracy is stable between 98.2506% and 100%, which has better robustness and stability.

Response Surface Method for Reliability Analysis Based on Iteratively-Reweighted-Least-Square Extreme Learning Machines

A response surface method for reliability analysis based on iteratively-reweighted-least-square extreme learning machines (IRLS-ELM) is explored in this paper, in which, highly nonlinear implicit performance functions of structures are approximated by the IRLS-ELM. Monte Carlo simulation is then carried out on the approximate IRLS-ELM for structural reliability analysis. Some numerical examples are given to illustrate the proposed method. The effects of parameters involved in the IRLS-ELM on accuracy in reliability analysis are respectively discussed. The results exhibit that a proper number of samples and neurons in hidden layer nodes, an appropriate regularization parameter, and the number of iterations for reweighting are of important assurance to obtain reasonable precision in estimating structural failure probability.

Research on Fault Diagnosis Algorithm of Ship Electric Propulsion Motor

The permanent magnet synchronous motor (PMSM) has been used in electric propulsion and other fields. However, it is prone to the stator winding inter-turn short-circuit, and if no effective measures are taken, the ship’s power system will be paralyzed. To realize intelligent diagnosis of inter-turn short circuits, this paper proposes an intelligent fault diagnosis method based on improved variational mode decomposition (VMD), multi-scale principal component analysis (PCA) feature extraction, and improved Bi-LSTM. Firstly, the stator current simulation dataset is obtained by using the mathematic model of the inter-turn short-circuit of PMSM, and the parameters of VMD are optimized by the grey wolf algorithm. Then, the data is coarse-grained to obtain multi-scale features, and the main features are selected as the sample data for fault classification by PCA. Subsequently, the Bi-LSTM neural network is used for training and analyzing the data of the sample set and the test set. Finally, the learning rate and the number of hidden-layer nodes of the Bi-LSTM are optimized by the whale algorithm to increase the diagnosis accuracy. Experimental results show that the accuracy of the proposed method for inter-turn short-circuited fault diagnosis is as high as 100%, which confirms the effectiveness of the method.

Low cycle fatigue life prediction of titanium alloy using genetic algorithm-optimized BP artificial neural network

This paper proposes a novel approach for estimating titanium alloy structural parts’ low cycle fatigue (LCF) life based on the continuous damage mechanics (CDM) model. Using a genetic algorithm-optimized back-propagation artificial neural network (GABP-ANN), the LCF life of titanium alloy structural parts is accurately predicted. Firstly, experimental data and finite element simulation data are combined to constitute a fatigue life database, and the relative error is taken as the training target for the BP neural network. The relationship is established between the relative error and the initial parameters (such as hidden layer nodes and learning rate). Secondly, the GABP-ANN model improves the shortage of the BP model at randomly selected initial training parameters and further enhances the predictive performance. Then, the accuracy and stability of five titanium alloy structural parts under two machine learning (ML) models are analyzed. The experimental values, finite element model (FEM) simulated values, and predicted values of the two ML models are compared and analyzed. It is found that the GABP-ANN model has a significant advantage in predicting the fatigue life of titanium alloy structural parts. Finally, it is verified that the proposed ML models have better data learning ability. The above results indicate that the GABP-ANN technique provides a highly accurate and stable method for predicting the LCF life of titanium alloy structural parts.

Research on calibration of a binocular stereo-vision imaging system based on the artificial neural network.

Camera calibration is a key problem for 3D reconstruction in computer vision. Existing calibration methods, such as traditional, active, and self-calibration, all need to solve the internal and external parameters of the imaging system to clarify the image-object mapping relationship. The artificial neural network, which is based on connectionist architecture, provides a novel idea for the calibration of nonlinear mapping vision systems. It can learn the image-object mapping relationship from some sample points without considering too many uncertain factors in the middle. This paper discusses the learning ability. A binocular stereo-vision mapping model is used as the learning model to explore the ability of image-object mapping for artificial neural networks. This paper constructs sample libraries by pixel and world coordinates of checkerboard corners, builds the artificial neural network, and, through the training samples and test samples prediction, verifies the learning performance of the network. Furthermore, by the laser scanning binocular vision device constructed in the authors' laboratory and trained-well network, the 3D point cloud reconstruction of a physical target is performed. The experimental results show that the artificial neural network can learn the image-object mapping relationship well and more effectively avoid the impact of lens distortion and achieve more accurate nonlinear mapping at the edge of the image. When the X and Y coordinates are in the range of 100mm and the Z coordinates are in the range of a 1000mm, the absolute error rarely exceeds 2.5mm, and the relative error is in the level of 10-3; for 1000mm distance measurement, the standard deviation does not exceed 1.5mm. Network parameter selection experiments show that, for image-object mapping, a three-layer network and increasing the number of hidden layer's nodes can improve the training time more significantly.

Extreme learning machine–radial basis function neural network–based state-of-charge estimation of lithium-ion batteries assisted with fiber Bragg grating sensor measurements

Lithium-ion batteries have been widely used in power systems for accepting large-scale renewable energy generations and in the transport sector to facilitate vehicle decarbonization, the accuracy of state-of-charge (SOC) estimation is extremely crucial for the battery systems to operate safely and reliably. Many approaches have been developed for battery SOC estimation based on the measurable electrical signals. However, the electrical measurements are usually corrupted by sensor noise and are also limited by insulation, corrosion, and electromagnetic compatibility. To address these problems, fiber Bragg grating (FBG) sensors are used to replace thermocouple sensors, the battery SOC therefore can be estimated based on the measurement of the strain information. In order to train the SOC estimation model for lithium-ion batteries, an FBG sensor–based battery SOC estimator, namely, ELM-RBFNN, is developed using the radial basis function neural network (RBFNN) identified by extreme learning machine (ELM). According to the ELM principle, the ELM-RBFNN model randomly selects the radial basis function (RBF) centers and determines the parameters of the hidden-layer nodes without any manual adjustment. The ELM-RBFNN estimation model has the advantages of fast learning speed, global optimization, and excellent generalization. Experimental results confirm that compared to the linear estimator and the polynomial estimator, the developed SOC estimator is able to produce more accurate results.

A novel correlation Gaussian process regression-based extreme learning machine.

An obvious defect of extreme learning machine (ELM) is that its prediction performance is sensitive to the random initialization of input-layer weights and hidden-layer biases. To make ELM insensitive to random initialization, GPRELM adopts the simple an effective strategy of integrating Gaussian process regression into ELM. However, there is a serious overfitting problem in kernel-based GPRELM (kGPRELM). In this paper, we investigate the theoretical reasons for the overfitting of kGPRELM and further propose a correlation-based GPRELM (cGPRELM), which uses a correlation coefficient to measure the similarity between two different hidden-layer output vectors. cGPRELM reduces the likelihood that the covariance matrix becomes an identity matrix when the number of hidden-layer nodes is increased, effectively controlling overfitting. Furthermore, cGPRELM works well for improper initialization intervals where ELM and kGPRELM fail to provide good predictions. The experimental results on real classification and regression data sets demonstrate the feasibility and superiority of cGPRELM, as it not only achieves better generalization performance but also has a lower computational complexity.

Model Hybrid Menggunakan Dekomposisi-Neural Network Untuk Data Indeks Harga Saham Gabungan

 The development of Covid-19 has worsened the economy not only nationally but also globally. Since its spread, the price movement of the Jakarta Composite Index (IHSG) has continued to be volatile. JCI price volatility shows risk and uncertainty in investing. Volatility is used as a barometer to determine portfolio management strategies for financial actors. Therefore, financial actors should find a strategy to be able to predict JCI price movements to reduce risks and gain profits. One way that can be done is to predict the JCI price as a reference in investing. This study uses a hybrid model between the decomposition model and the Neural Network (NN) in predicting JCI price volatility. The decomposition uses two approaches, namely additive and multiplicative, the two approaches will then be combined with NN and the NN algorithm used is Feed Forward Neural Network (FFNN) where the results of the decomposition in the form of seasonal, trend, and random data are used as input in the FFNN architecture. The FFNN architecture in this study differs from the hidden layer nodes and the epochs used. Furthermore, the prediction results from the model are compared with a single NN. The performance of each architecture will be measured using the Mean Absolute Error (MAE) and Mean Square Error (MSE). The results show that the hidden layer with more nodes can provide good performance while the epoch used provides good performance depending on the learning process carried out. The prediction results using the hybrid model can outperform the performance of a single NN.Keywords : time series, volatilitas, studi perbandingan, kecerdasan buatan, statistik.

Evolving convolution neural network by optimal regularization random vector functional link for computational color constancy

To automatically eliminate the influence of external illumination on the acquired image, restore the true color information of the object, and provide true and stable color features for computer vision tasks, our work is based on the convolutional neural network (CNN) VGG19 for feature extraction of image information. Moreover, it proposes a model based on atom search optimization improved by chaotic-logistic maps to optimize the regularization random vector functional link (RRVFL). The optimized RRVFL replaces CNN’s regression layer to estimate the image illumination and then restore the image. First, we used chaotic-logistic maps to optimize the initial value of search agent of the atom search optimization (CASO) algorithm and then used the search agent segment mapping method proposed in our study to simultaneously optimize the number of nodes in hidden layer, regularization factors, input weight, and the bias of RRVFL hidden layer. This avoids the problem of serious fluctuation of prediction results and low accuracy caused by the randomness of the RRVFL parameters. After the predicted illumination information was obtained with the optimized RRVFL, the image was restored using the diagonal transformation method. Comparative experiments showed that the average angular errors of the illumination estimation of the V19-CASO-RRVFL algorithm proposed in our study were 1.73, 0.30, 0.3406, and 0.0976 lower than those of the gray-edge-2, GE-CASO-RRVFL, DS-Net, and color constancy on deep residual learning algorithms, respectively.

Influence Factor Analysis and Prediction Model of End-Point Carbon Content Based on Artificial Neural Network in Electric Arc Furnace Steelmaking Process

In this manuscript, we consider the accuracy of end-point carbon content prediction affected by oxygen injection in multiple stages of electric arc furnace (EAF) melting process. Such a prediction would help to further evaluate process control strategies and optimize overall operation of the electric arc furnace. Principal component analysis (PCA) was used to normalize the 13 input variables affecting the endpoint carbon content. log-sigmoid and tan-sigmoid functions were used to verify the same sample, and it was found that the Mean squared error(MSE) of the model under logsig + logsig function was smaller, indicating that the model was more stable. At the same time, different hidden layer nodes were tried, and finally the structure of the model was determined as 13 × 10 × 8 × 1, and the activation function was logsig + logsig. Using historical smelting data to train and test the neural network model, the correlation coefficient (R) of the verified model is 0.7632, the model prediction is in the range of ±0.03%, the hit rate of the model is 64.5%, and the hit rate of the model is 42% in the range of ±0.02%. Combining the verification basis of the model with the metallurgical reaction principle of the EAF steelmaking process, a pretreatment method of phased input of total oxygen is proposed. The oxygen is divided into three stages, which are the oxygen consumption volume of 0–5 min, 5–30 min and more than 30 min and other variables are kept unchanged. The same neural network is used to train and verify the same data. After verification, the R of the oxygen staged model is 0.8274. The model prediction is in the range of ±0.03%, the hit rate of the model is 78.5%, and the hit rate of the model is 58% in the range of ±0.02%. Finally, an on-line carbon content prediction system based on artificial neural network model is developed and applied to actual production. Running results illustrated that the hit rate of end-point carbon content is 96.67%, 93.33% and 86.67%, respectively when the prediction errors are within ±0.05%, ±0.03% and ±0.01%, the improved neural network model can effectively predict the end-point carbon content, which provides a good basis for the carbon content at the end point of EAF steelmaking process.

기계학습을 이용한 3D CAD 모델 형상과 경계조건의 유형 식별

In this paper, we propose a machine learning approach for classifying 3D CAD models with boundary conditions. We adopted an extended voxel model to represent a CAD model with its boundary conditions, and to construct the training data. By considering 7 types of part families and 3 types of boundary conditions, we generated 320 similar CAD models for each part family and assigned 3 different boundary conditions to each CAD model, which produces 960 datasets for the part family. We used multi-layer perceptron (MLP) and convolutional neural network (CNN) as machine learning models, which classify the combination type of a given CAD model with its boundary conditions. Using TensorFlow, we trained and tested the models, and compared their performance. We considered the MLP models made of three hidden layers and the CNN models made of two convolutional, two pooling, and three hidden layers. We also conducted a grid search to find the proper number of nodes in hidden layers. From experimental results, we found that the CNN models are better in accuracy than the MLP models. If further enhanced, the proposed approach is expected to become a useful tool for similar case search from archive CAE models.

Application of a New Architecture Neural Network in Determination of Flocculant Dosing for Better Controlling Drinking Water Quality

In drinking water plants, accurate control of flocculation dosing not only improves the level of operation automation, thus reducing the chemical cost, but also strengthens the monitoring of pollutants in the whole water system. In this study, we used feedforward signal and feedback signal data to establish a back-propagation (BP) model for the prediction of flocculant dosing. We examined the effect of the particle swarm optimization (PSO) algorithm and data type on the simulation performance of the model. The results showed that the parameters, such as the learning factor, population size, and number of generations, significantly affected the simulation. The best optimization conditions were attained at a learning factor of 1.4, population size of 20, 20 generations, 8 feedforward signals and 1 feedback signal as input data, 6 hidden layer nodes, and 1 output node. The coefficient of determination (R2) between the predicted and measured values was 0.68, and the root mean square error (RMSE) was lower than 20%, showing a good prediction result. Weak time-delay data enhanced the model accuracy, which increased the R2 to 0.73. Overall, with the hybridized data, PSO, and weak time-delay data, the new architecture neural network was able to predict flocculant dosing.

Design and Implementation of Fault Diagnosis System for Power Internet of Things Equipment Based on Neural Network

The design and application of the equipment fault diagnosis system have been improved and upgraded, allowing it to effectively detect the equipment’s operation status and promptly eliminate hidden faults, reducing the occurrence of unexpected accidents and improving the safety index of people’s lives. The purpose of this essay is to design and apply neural network (NN) fault diagnosis system model in power Internet of things (IOT) equipment and explore its accuracy and effectiveness. The BP neural network (BPNN) algorithm was used to construct model of a fault monitoring testing of the power IOT equipment. Neural network is an algorithmic mathematical model that imitates the behavioral characteristics of animal neural network and performs distributed parallel information processing. The network parameters were as follows: there were four input layer nodes, seven hidden layer nodes, and five output layer nodes, the training times were 10000, and the allowable error was 0.002. In this paper, we use the IOT to detect model of a fault monitoring testing of power equipment designed in each sample, the success rate is as high as 97.5%, and the designed network structure and network parameters are reasonable. The trained loss is less than 0.001, and the nontraining set samples may be appropriately identified. It is clear that the NN has a high application for power equipment fault diagnosis in the IOT value.