Elman Recurrent Neural Network Research Articles

Development of a Production Machine Maintenance Predictive Model Using the Elman Recurrent Neural Network Algorithm

<table width="593" border="1" cellspacing="0" cellpadding="0"><tbody><tr><td valign="top" width="388"><p>PT Simba Indosnack Makmur is a factory that produces snacks. In the production process the machine has worked very optimally, the problem that is often faced by the Quality Control department is often finding non-standard product weights. This problem is caused by a machine that already requires maintenance. So far, the maintenance process has to get approval from the manager, which sometimes takes quite a long time to be inspected so that the maintenance process is delayed, which results in reduced production targets. By implementing a predictive maintenance model that utilizes time series data in the production process, applying the Elman Recurrent Neural Network will be able to provide notifications for machine maintenance before the machine is inaccurate in snack production. The Elman structure was chosen because it can make iterations much faster, thus facilitating the convergence process. The input vector used uses windows size. The results of the study using a target error of 0.001 show the smallest MSE value of 0.002833 with windows size 11. Then by using 13 neurons in the hidden layer a minimum error value of 0.003725 is obtained.</p></td></tr></tbody></table>

A Dynamic Recurrent Neural Network for Predicting Higher Heating Value of Biomass.

The higher heating value (HHV) is the main property showing the energy amount of biomass samples. Several linear correlations based on either the proximate or the ultimate analysis have already been proposed for predicting biomass HHV. Since the HHV relationship with the proximate and ultimate analyses is not linear, nonlinear models might be a better alternative. Accordingly, this study employed the Elman recurrent neural network (ENN) to anticipate the HHV of different biomass samples from both the ultimate and proximate compositional analyses as the model inputs. The number of hidden neurons and the training algorithm were determined in such a way that the ENN model showed the highest prediction and generalization accuracy. The single hidden layer ENN with only four nodes, trained by the Levenberg-Marquardt algorithm, was identified as the most accurate model. The proposed ENN exhibited reliable prediction and generalization performance for estimating 532 experimental HHVs with a low mean absolute error of 0.67 and a mean square error of 0.96. In addition, the proposed ENN model provides a ground to clearly understand the dependency of the HHV on the fixed carbon, volatile matter, ash, carbon, hydrogen, nitrogen, oxygen, and sulfur content of biomass feedstocks.

A comprehensive investigation of wetting distribution pattern on sloping lands under drip irrigation: A new gradient boosting multi-filtering-based deep learning approach

Accurate estimation of the wetting distribution pattern (WDP) around the emitters of a drip irrigation system in sloping lands can minimize surface runoff losses by determining the placement status of plants and emitters. In this study, both experimental and computational efforts were made to estimate the WDP in sloping lands with drip irrigation. 486 sets of laboratory experiments were conducted and a series of soil characteristics data were collected. Particularly, the upstream wetting radius (R−), downstream wetting radius (R+), and wetting depth (D) were measured and further used as the target variables of three modeling scenarios. In the modeling effort, a new hybrid framework, consisting of a light gradient boosting machine (LightGBM) and best subset regression (BSR) integrated with bidirectional recurrent neural network (Bi-RNN), was developed for precise simulation of WDP. The main model (i.e., Bi-RNN) was compared with the Elman recurrent neural network (ERNN) and bagging regression tree (BGRT) in the advanced multi-filtering framework for all the scenarios. In the first stage, the LightGBM tree-based feature selection filtered the significant predictors in each scenario. In the second stage, the three best possible input combinations using the N predictors selected in the first stage were extracted among 2N possible combinations via the BSR strategy. The performances of the models were evaluated by using different statistical metrics. It was demonstrated that Bi-RNN achieved the highest accuracy in all the hybrid models for the three scenarios, followed by the ERNN and BGRT models. Also, a resampling bootstrap-based uncertainty analysis proved that the developed multistage-filtering strategy before the deep learning model feeding decreased the uncertainty associated with input combination effects. By determining the placement status of plants and emitters, the proposed framework can effectively reduce surface runoff losses.

A-DQRBRL: attention based deep Q reinforcement battle royale learning model for sports video classification

ABSTRACT Analyzing video content and classification are the major challenges for computer vision applications. Thus, the proposed research develops a new attention-based deep Q reinforcement battle royale learning (A-DQRBRL) model for effective sports video classification. Initially, the required keyframes from the input sports video are extracted by an action template approach. Then, pre-processing is performed to remove unwanted noises through cross-guided bilateral filter (CGBF). From the pre-processed frames, the most discriminative features are extracted by Elman recurrent neural network (ERNN). Then, the needed portions are segmented from the extracted features by utilizing a cascaded random forest (CRF) approach. Finally, the sports video is classified by proposing a new A-DQRBRL method. The experimental setup is done by python tool, and the proposed work attains an accuracy of 99.29% in the SVW dataset and 98.97% in the UCF101 dataset. Thus, the experimental results show that the proposed techniques perform better than previous methods.

Predicting the performance of thermal, electrical and overall efficiencies of a nanofluid-based photovoltaic/thermal system using Elman recurrent neural network methodology

In this research, the Elman recurrent neural network methodology is used to develop a relationship to predict the performance of a photovoltaic/thermal device with sheet-and-serpentine tube collector. In this system, PV panels are cooled using the water-magnetite nanofluid. The thermal (ηTH), electrical (ηEL) and overall (ηOV) efficiencies of the PV/T device are utilized as objective functions, and the mass flow rate (m˙) and concentration (φ) of nanofluid are considered as input parameters. The range considered for the m˙ and φ is 20-80 kg/hr and 0-2% respectively. To measure the accuracy of the models developed, the mean absolute percentage error (MAPE), root mean square error (RMSE), and correlation coefficient (R) were used. It was found that the best accuracy is obtained for the ηTH in terms of (Training: R=0.9978 and RMSE=0.2045; Testing: R=0.9973 and RMSE=0.1973) followed by ηOV in terms of (Testing: R=0.9959 and RMSE=0.1797), and ηEL in terms of (Testing: R=0.9825 and RMSE=0.0037), respectively.

Survival study on deep learning techniques for IoT enabled smart healthcare system.

The paper is to study a review of the employment of deep learning (DL) techniques inside the healthcare sector, together with the highlight of the strength and shortcomings of existing methods together with several research ultimatums. Our study lays the foundation for healthcare professionals and government with present-day inclinations in DL-based data analytics for smart healthcare. A deep learning-based technique is designed to extract sensor displacement effects and predict abnormalities for activity recognition via Artificial Intelligence (AI). The presented technique minimizes the vanishing gradient issue of Recurrent Neural Networks (RNN), thereby reducing the time for detecting abnormalities with consideration of temporal and spatial factors. Proposed Moran Autocorrelation and Regression-based Elman Recurrent Neural Network (MAR-ERNN) introduced. Experimental results show the feasibility of the proposed method. The results show that the proposed method improves accuracy by 95% and reduces execution time by 18%. MAR-ERNN performs well in the activity recognition of health status. Collectively, this IoT-enabled smart healthcare system is utilized by enhancing accuracy, and minimizing time and overhead reduction.

Simulasi Penggunaan Listrik Tarif Sosial Menggunakan Algoritma ERNN

The use of electricity under the social tariff category has increased significantly each year, both purely social and commercial social use. The use of social tariff electricity is intended for public interest activities for both the lower and upper middle social strata which are oriented towards fulfilling growth and development facilities for the public interest, so that a simulation of social electricity usage is needed to map a picture of the condition of the amount of social electricity usage in the future. The research was conducted to determine the estimation of how much electricity is used by using the Elman Recurrent Neural Network (ERNN) algorithm by reducing the input dimensions. The ERNN algorithm is used to simulate network parameters formed from complex input-output relationships, so that data patterns can be found. The factors of the input dimensions of this study are demographic data, electricity usage, social customers, population, gross regional domestic product (GRDP) and industrial growth. The results showed that the ERNN algorithm is capable of simulating formed network parameters that can be used for training and validation so that the value of the network Mean Square Error (MSE) can be determined, with prediction accuracy using the Mean Absolute Percentage Error (MAPE) for forecast in sample in the forecast period of 5 years obtained an average of 0.77%, and able to know the dominant factors that influence the use of social tariff electricity.

Time-Series Forecasting of Chlorophyll-a in Coastal Areas Using LSTM, GRU and Attention-Based RNN Models

The chlorophyll-a (Chl-a) concentration is commonly considered as the main indicator of phytoplankton biomass in coastal waters. Forecasting and understanding the status of Chl-a is beneficial to coastal ecosystem management and is an important emergency management measure for algae blooms. To obtain accurate predictions, the long short-term memory neural network (LSTM) and gated recurrent unit neural network (GRU) were implemented for Chl-a forecasting, and based on the LSTM and GRU units, two simplified attention-based encoder-decoder recurrent neural network (AEDRNN) models were also developed for time series predictions. The performance of the proposed models was compared with that of the auto-regressive integrated moving average (ARIMA), multilayer perceptron (MLP) and Elman recurrent neural network (ERNN) models by experimentally generating multi-step-ahead predictions using a dataset in the Zhejiang coastal areas of China. The results demonstrated that the LSTM, GRU and AEDRNN models significantly outperformed the ARIMA, MLP and ERNN models according to multiple statistical indicators. Moreover, the AEDRNN models were superior to the LSTM and GRU models, especially for middle-term predictions. In addition, the AEDRNN model with LSTM units was more robust than the AEDRNN model with GRU units in terms of accuracy and stability; therefore, it was considered to be the best model for Chl-a forecasting.

Detection of Human Stress Using Optimized Feature Selection and Classification in ECG Signals

An autonomic nervous system (ANS) of humans is majorly affected by psychological stress. The changes in ANS may cause several chronic diseases in humans. The electrocardiogram (ECG) signal is used to observe the variation in ANS. Numerous techniques are presented for an ECG stress signal handling feature extraction and classification. This work managed a heart rate variability feature acquired from smaller peak waveforms such as P, Q, S, and T waves. Also, the R peak is detected, which is a significant part of the ECG waveform. In this work, the proposed stress classification work has been categorized into two main processes: feature selection (FS) and classification. The main aim of the proposed work is to propose an optimized FS and classifier model for the detection of stress in ECG signals. The Metaheuristics model of the African vulture optimization (AVO) technique is presented to perform an FS. This selection is made to choose the required features and minimize the data for classification. The AVO‐based modified Elman recurrent neural network (MERNN) technique is proposed to perform an efficient classification. The AVO is used for fine‐tuning the weight of the MERNN technique. The experimental result of this technique is evaluated in terms of Recall (91.56%), Accuracy (92.43%), Precision (92.78%), and F1 score (95.86%). Thus, the proposed performance achieved a superior result than the conventional techniques.

An Intelligent Path Planning Algorithm and Control Strategy Design for MultiMobile Robots based on a Modified Elman Recurrent Neural Network

An Intelligent Path Planning Algorithm and Control Strategy Design for MultiMobile Robots based on a Modified Elman Recurrent Neural Network

Use of multilayer recursive model for non-linear dynamic system identification

In practice, the dynamics of the system are uncertain due to nonlinear and dynamic characteristics. It is difficult to establish accurate identification and prediction of the nonlinear plants that require dynamic modelling of the system. Extreme learning machine (ELM) as the recursive model due to its fast training and convergence speed is utilized in this work. However, its limitation is that it has only 1 hidden neuron which tends to make evolution speed low. Further, Multi-layer ELM (ML-ELM) model is applied on a nonlinear Auto-regressive complex benchmark system. The performance of ML-ELM is compared with dynamic recurrent functional link neural network (DRFLNN), functional link neural network (FLNN), nonlinear auto-regressive moving average (NARAX), multi-layer perception (MLP), radial basis function network (RBFN), Elman recurrent neural network (ERNN), and basic ELM models. From the comparison table, it can be seen that ML-ELM has better performance as compared with other models.

Experimental investigation on load mitigation of a 5 MW wind turbine in digital fluid power pitch system test rig based on predictive load mitigation controller

ABSTRACT The pitch control system plays a vital role in generating rated power and mitigates mechanical loads during strong wind gusts. Generally, the hydraulic pitch system deployed in a large-scale wind turbine uses a proportional/servo valve to adjust the pitch angle. However, these valves suffer from low operating efficiency which leads to poor pitching performance. Thus, a high-speed digital hydraulic valve-based Digital Fluid Power Pitch System (DFPPS) test rig was proposed in this paper. Moreover, the conventional controller was not able to produce adequate pitching performance during high wind flux. Therefore, a machine learning-based pitch controller was implemented in the developed test rig. The developed DFPPS test rig comprises simulation loop (Wind Generator System (WGS) model, pitch gear model, Predictive Load Mitigation Controller (PLMC), and pitch load model), and DFPPS-hardware interfaced through an I/O interface. Here, PLMC was developed to improve the load mitigation performance without compromising the generator performance at high wind speeds. The proposed PLMC is a cascade arrangement of Recurrent Elman Neural Network (RENN) tuned Fractional-Order Proportional Integral Derivative (FOPID). Further, for collecting the optimal datasets to train the RENNs in PLMC, a Customized Particle Swarm Optimization (CPSO) algorithm was utilized. Two experiments are performed considering the developed WGS model and benchmark Fatigue Aero-elastic Structure Turbulence (FAST) simulator deployed in the DFPPS test rig. The experimental results of PLMC (experiment) were compared to simulation results PLMC (simulation). Further, the experimental output of the PLMC was compared to established controllers in the literature (Radial Basis Function tuned Fractional-Order Proportional Integral Derivative (RBF-FOPID) and baseline Proportional Integral (PI)). In the experiment using FAST, relative to RBF-FOPID, the PLMC (experiment) decreased the Standard Deviation (STD) of blade and tower moment by 81.5% and 81.4%, respectively. Consequently, compared to baseline PI, the PLMC (experiment) reduced the STD of blade and tower moment by 85.6%, and 84.4%, respectively. The proposed controller effectively mitigated the load and also the generator performance was close to the rated value than its counterparts.

ARIMA and ARIMA-ERNN models for prediction of pertussis incidence in mainland China from 2004 to 2021

ObjectiveTo compare an autoregressive integrated moving average (ARIMA) model with a model that combines ARIMA with the Elman recurrent neural network (ARIMA-ERNN) in predicting the incidence of pertussis in mainland China.BackgroundThe incidence of pertussis has increased rapidly in mainland China since 2016, making the disease an increasing public health threat. There is a pressing need for models capable of accurately predicting the incidence of pertussis in order to guide prevention and control measures. We developed and compared two models for predicting pertussis incidence in mainland China.MethodsData on the incidence of pertussis in mainland China from 2004 to 2019 were obtained from the official website of the Chinese Center for Disease Control and Prevention. An ARIMA model was established using SAS (ver. 9.4) software and an ARIMA-ERNN model was established using MATLAB (ver. R2019a) software. The performances of these models were compared.ResultsFrom 2004 to 2019, there were 104,837 reported cases of pertussis in mainland China, with an increasing incidence over time. The incidence of pertussis showed obvious seasonal characteristics, with the peak lasting from March to September every year. Compared with the mean squared error (MSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) of the ARIMA model, those of the ARIMA-ERNN model were 81.43%, 95.97% and 80.86% lower, respectively, in fitting performance. In terms of prediction performance, the MAE, MSE and MAPE were 37.75%, 56.88% and 43.75% lower, respectively.ConclusionThe fitting and prediction performances of the ARIMA-ERNN model were better than those of the ARIMA model. This provides theoretical support for the prediction of infectious diseases and should be beneficial to public health decision making.

Forecasting the performance of a diffusion absorption refrigeration system using optimal fluid blend—A neural network approach

AbstractIn recent years, the diffusion absorption system has gained special attention in several types of research due to its ability to promote sufficient cooling effect from the use of low‐grade heat. Normally, in the diffusion absorption refrigeration (DAR) system, ammonia works as a refrigerant, helium as an inert gas, and water as an absorbent. These refrigerant combinations cause certain environmental issues such as global warming, ozone layer depletion, toxic gas formation, etc. This work aims to formulate an environment‐friendly DAR system from among six different alternative refrigerant blends. The refrigeration performances such as coefficient of performance (COP), exergy analysis, heat supplied in the generator, mass flow rate, refrigeration effect, and cooling capacity are determined for different evaporator temperatures. Harris hawk optimization (HHO) screened the optimal refrigerant blend and validated using hybrid Elman recurrent neural network (ERNN)‐based HHO methods. It is observed that the performance of the DAR is optimal when isobutane + dimethylformamide + helium is used as the working fluid.

Predicting Velocity and Direction of Ocean Surface Currents using Elman Recurrent Neural Network Method

Background: Ocean surface currents need to be monitored to minimize accidents at ship crossings. One way to predict ocean currents—and estimate the danger level of the sea—is by finding out the currents’ velocity and their future direction. Objective: This study aims to predict the velocity and direction of ocean surface currents. Methods: This research uses the Elman recurrent neural network (ERNN). This study used 3,750 long-term data and 72 short-term data. Results: The evaluation with Mean Absolute Percentage Error (MAPE) achieved the best results in short-term predictions. The best MAPE of the U currents (east to west) was 14.0279% with five inputs; the first and second hidden layers were 50 and 100, and the learning rate was 0.3. While the best MAPE of the V currents (north to south) was 3.1253% with five inputs, the first and second hidden layers were 20 and 50, and the learning rate was 0.1. The ocean surface currents’ prediction indicates that the current state is from east to south with a magnitude of around 169,5773°-175,7127° resulting in a MAPE of 0.0668%. Conclusion: ERNN is more effective than single exponential smoothing and RBFNN in ocean current prediction studies because it produces a smaller error value. In addition, the ERNN method is good for short-term ocean surface currents but is not optimal for long-term current predictions. Keywords: MAPE, ERNN, ocean currents, ocean currents’ velocity, ocean currents’ directions

A Lyapunov-stability-based context-layered recurrent pi-sigma neural network for the identification of nonlinear systems

A novel higher-order context-layered recurrent pi-sigma neural network (CLRPSNN) is presented for the identification of nonlinear dynamical systems. The proposed model is the modified form of the classical pi-sigma neural network (PSNN) and contains an additional layer (known as the context layer) of the context nodes. Pi-sigma networks involve a product operator/unit in their output layer which indirectly incorporates in them the capability of higher-order networks and also reduces their network complexity. For tuning the weights of the proposed CLRPSNN model, a learning procedure is developed by combining the Back-Propagation (BP) and Lyapunov-stability method. The performance of the proposed model is compared with other models such as PSNN, Feed-forward neural network (FFNN) (containing single hidden layer), and various popular recurrent neural network (RNN) like Elman recurrent neural network (ERNN), Jordan recurrent neural network (JRNN), Diagonal recurrent neural network (DRNN), and a deep neural network (DNN). The simulation study showed that the proposed model has given better results as compared to the other models.

Optimization of RFID reading performance based on YOLOv3 and Elman neural network

As a non-contact automatic identification technology, Radio Frequency Identification (RFID) is of great significance to improve the simultaneous identification of multi-target. This paper designs a more efficient and accurate multi-tag reading performance measurement system based on the fusion of YOLOv3 and Elman neural network. In the machine vision subsystem, multi-tag images are collected by dual CCD and detected by neural network algorithm. The reading distance of 3D distributed multi-tag is measured by laser ranging to evaluate the reading performance of RFID system. Firstly, the multi-tag are detected by YOLOv3, which realizes the measurement of 3D coordinates, improves the prediction accuracy, enhances the recognition ability of small targets, and improves the accuracy of 3D coordinate detection. Secondly, the relationship between the 3D coordinates and the corresponding reading distance of RFID multi-tag are modelled by Elman recurrent neural network. Finally, the reading performance of RFID multi-tag is optimized. Compared with the state-of-the-arts, the multi-tag detection rate of YOLOv3 is 17.4% higher and the time is 3.27 times higher than that of the previous template matching algorithm. In terms of reading performance, the MAPE of Elman neural network is 1.46 %, which is at least 21.43 % higher than other methods. In running time, Elman only needs 1.69s, which is at least 28.40% higher than others. Thus, the system not only improves the accuracy, but also improves the speed, which provides a new insight for the measurement and optimization of RFID performance.

Prediction Active Case of Covid-19 with ERNN

SARS-CoV-2 is known as Covid-19 has been spread in all world since end of 2019. Indonesia, including South Kalimantan has detected first Covid-19 in March 2020. This pandemic has affected in all entirely live in Indonesia. This makes Covid-19 be the main focus of the government. The government has provided aid and imposed restrictions on activities. These policies require planning that can be a solution. Careful planning requires an overview of the data on active cases that are positive for Covid-19. This overview can be obtained through prediction. In this research, Elman Recurrent Neural Network (ERNN) was used to predict active cases of Covid-19. Architecture of ERNN was used ERNN with 3 input nodes, 2 hidden nodes, and 2 context nodes. The data used is 277 data, which is then divided into training data and testing data, respectively 90%-10%, 80%-20%, and 70%-30%. ERNN with a learning rate of 0.1 until 0.9 is applied to data on active cases of Covid-19, then Mean Absolute Percentage Error (MAPE) is calculated to find out performance of model generated by ERNN. The results showed that all of MAPE were below 10% with the smallest MAPE as 3.21% for scenario 90:10 and learning rate 0.6. MAPE value which is less than 10% indicates that ERNN has very good predictive ability.

Slime Mould Algorithm Training Neural Network in Automatic Voltage Regulator

The research is proposed a new method of artificial intelligence (AI) to control automatic voltage regulators. A neural network has improved using a metaheuristic method, namely the slime mould algorithm (SMA). SMA has an algorithm based on the mode of slime mold in nature. SMA has characteristics that use adaptive weights to simulate the process to generate feedback from the movement of bio-oscillator-based slime molds in foraging, exploring, and exploiting areas. The performance of the proposed method is focused on speed and rotor angle. To know the competence and potency of the proposed method, a comparison with feed-forward backpropagation neural networks (FFBNN), cascade-forward backpropagation neural networks (CFBNN), Elman-recurrent neural networks (E-RNN), focused time delay neural network (FTDNN), and Distributed Time Delay Neural Network (DTDNN) method are applied. It can be concluded that the proposed method has the best ability. The Proposed method has ability to reduce the overshoot speed with an average value of 0.78 % and the overshoot rotor angle with an average value of 2.134 %.

Comparison of backpropagation and ERNN methods in predicting corn production

East Java is one of the producers of food crops in Indonesia. Some food crop commodities in East Java Province are corn, soybeans, peanuts, sweet potatoes, and cassava. These food crops have many benefits to make the demand for production increase. The uncertain amount of food crop production will be a problem for the Department of Agriculture and Food Security of East Java Province in determining a policy. To overcome this problem, a system is needed to predict the production of food crops in East Java. This study compares the Backpropagation algorithm and Elman Recurrent Neural Networks (ERNN). The data in this study were obtained from the Department of Agriculture and Food Security of East Java Province starting from 2007-2020 per quarter. The result of this research is that trial scenario 1 produces the best MSE value of 0.00000063 on the Backpropagation algorithm compared to ERNN which only gets an MSE value of 0.00000627. Trial scenario 2 produces the best MSE value, which is 0.000000003 in the Backpropagation algorithm with gradient descent momentum, this is also better when compared to ERNN which gets an MSE value of 0.00000407. It can be concluded that the best algorithm in this study is Backpropagation with gradient descent momentum because it produces MSE values ​​with good prediction results from all algorithms compared.