Adaptive ANN Research Articles

Development of an Adaptive Optimization ANN Model for an Automatic Identification and Classification of Stuttering in Speech

Development of an Adaptive Optimization ANN Model for an Automatic Identification and Classification of Stuttering in Speech

Machine Learning Techniques Applied to the Harmonic Analysis of Railway Power Supply

Harmonic generation in power system networks presents significant issues that arise in power utilities. This paper describes a machine learning technique that was used to conduct a research study on the harmonic analysis of railway power stations. The research was an investigation of a time series whose values represented the total harmonic distortion (THD) for the electric current. This study was based on information collected at a railway power station. In an electrified substation, measurements of currents and voltages were made during a certain interval of time. From electric current values, the THD was calculated using a fast Fourier transform analysis (FFT) and the results were used to train an adaptive ANN—GMDH (artificial neural network–group method of data handling) algorithm. Following the training, a prediction model was created, the performance of which was investigated in this study. The model was based on the ANN—GMDH method and was developed for the prediction of the THD. The performance of this model was studied based on its parameters. The model’s performance was evaluated using the regression coefficient (R), root-mean-square error (RMSE), and mean absolute error (MAE). The model’s performance was very good, with an RMSE (root-mean-square error) value of less than 0.01 and a regression coefficient value higher than 0.99. Another conclusion from our research was that the model also performed very well in terms of the training time (calculation speed).

Face recognition with illumination, scale and rotation invariance using multiblock LTP-GLCM descriptor and adaptive ANN

Face recognition with illumination, scale and rotation invariance using multiblock LTP-GLCM descriptor and adaptive ANN

Image processing system for identifying groundnut plant disease

Groundnuts (Arachis hypogaea), also known as peanuts are the edible seeds of a legume plant that grow to maturity in the ground. Diseases and its diagnosis methodology are the main challenges to groundnut production. Manual means of identifying plant disease is the most difficult activity with a high rate of mistake and time taking procedures. Therefore, the main aim of this research was to design an image processing system to identify groundnut plant disease. In general, the major significance of the study is to provide effective and simple groundnut plant disease diagnosis system that supports disease controlling mechanism and experts in the domain area. While the process of developing the system, researchers used purposive sampling techniques for acquiring 320 sample leaves images from four classes of groundnuts plant. Those are Cercospora personatum, Cercospora arachidicola, Puccinia arachidis, and healthy leaf from Bishan-Babile and Gemechu peasant associations in Babile district, and Awdal peasant association in Gursum districts in East Hararghe zone of Oromiya Regional State. Also, the researchers conducted a survey study which helps to identify experts view and deep understanding of the domain area. The design, phase used; adaptive median filtering, K-means clustering, SVM and ANN techniques with MATLAB programming tool. In general, the overall result shows that the developed system achieved a better result with an accuracy of 90.6%.

An adaptive nonlinear autoregressive ANN model for high time resolution traffic noise predictions. Experimental results for a port city waterfront

In this research study an adaptive recurrent artificial nonlinear neural network identification model has been developed and experimentally tested for dynamically predicting the traffic noise level Leq,1′ with a time refinement of 1 min. The model has been successfully applied in three selected positions, representative of the waterfront in a Mediterranean port city. Several maritime cities are exposed to a wide range of road traffic fluctuations that negatively impact liveability in the area concerned. Large volumes of road traffic periodically access the port, dynamically affecting the acoustic scenario in neighbouring areas, especially in seaside towns during the tourist season. A signalized intersection, a roundabout, and a wide entrance to a vehicular underpass have been analyzed in the course of two characteristic periods, during which traffic ranged widely from normal to peak yearly intensity. Detailed traffic data for 15 road lanes and noise sequence regressors have been considered as input data sources. This exploratory investigation reveals a good predictive performance of the model developed, the prediction error of Leq,1′ falling prevalently within the range ±0.5 dB. The experimental profile of Leq,1′ is well reflected by the simulated sequence, and the auto and cross correlation functions confirm how well the identified neural model is able to explain the functional dependence underlying the experimental observations.

An experimental modelling and performance validation study: Top gas pressure tracking system in a blast furnace using soft computing methods

The blast furnace is a master iron-producing plant of iron and steel factories and affected by several process parameters as well as top gas pressure , which is a key process control phenomenon to maintain stability and operational productivity in such plants. Blast furnace operation is not tolerant to any interruption, unbalanced operations, momentary disturbances or loss of control due to its nature of intensive chemical reactions and heat balance requirements. Consequently, it is crucial to monitor and control top gas system components of the furnace with instrumentation measurements to maintain stable, efficient operation and system safety ongoing. In this study, a novel top gas pressure tracking system is developed using the chronologically obtained live process data of Erdemir BF#2 in Turkey. Eight process parameters are considered as input parameters as per the plant maintenance team's recommendations and soft computing methods, artificial neural networks and adaptive neuro fuzzy inference system are employed and a statistical regression tool, autoregressive integrated moving average, is also applied for comparison. Performance and success ratio analysis is carried out using coefficient of determination ( R2), mean absolute percentage error and root mean squared error terms. The best performing model output for the adaptive neuro fuzzy inference system is found to be 0.95, 1.21 and 0.023, and slightly lower performance is obtained for the artificial neural network model with the output values of 0.94, 0.029 and 1.32 against R2, mean absolute percentage error and root mean squared error terms, respectively. The maximum prediction error is found to be 9.85% and 10.2%, and the average prediction error is found to be 1.19% and 1.29% for adaptive neuro fuzzy inference system and ANN models, respectively, for optimum simulations. The proposed neuro-fuzzy-driven top gas pressure prediction system is unique in the literature and should be integrated into existing control systems to improve operational awareness and sustainability or can be used as input guidance for a possible future top gas recovery system.

Hybridization of harmonic search algorithm in training radial basis function with dynamic decay adjustment for condition monitoring

In recent decades, hybridization of superior attributes of few algorithms was proposed to aid in covering more areas of complex application as well as improves performance. Condition monitoring is a major component in predictive maintenance which monitors the condition and identifies significant changes in the machinery parameter to perform early detection and prevent equipment defects that could cause unplanned downtime or incur unnecessary expenditures. An effective condition monitoring model is helpful to reduce the frequency of unexpected breakdown incidents and thus, facilitates in maintenance. ANN has shown effective in various condition monitoring and fault detection applications. ANN is popular due to its capability of identifying the complex nonlinear relationships among features in a large dataset and hence, it can perform with an accurate prediction. However, a drawback is that the performance of ANN is sensitive to the parameters (i.e., number of hidden neurons and the initial values of connection weights) in its architecture where the settings of these parameters are subject to tuning on a trial-and-error basis. Hence, a wide range of studies have been focused on determining the optimal weight values of ANN models and the number of hidden neurons. In this research work, the motivation is to develop an autonomous learning model based on the hybridization of an adaptive ANN and a metaheuristic algorithm for optimizing ANN parameters so that the network could perform learning and adaptation in a more flexible way and handle condition classification tasks more accurately in industries, such as in power systems. This paper presents an intelligent system integrating a Radial Basis Function Network with Dynamic Decay Adjustment (RBFN-DDA) with a Harmony Search (HS) to perform condition monitoring in industrial processes. RBFN-DDA performs incremental learning wherein its structure expands by adding new hidden units to include new information. As such, its training can reach stability in a shorter time compared to the gradient-descent based methods. To achieve optimal RBFN-DDA performance, HS is proposed to optimize the center and the width of each hidden unit in a trained RBFN. By integrating with the HS algorithm, the proposed metaheuristic neural network (RBFN-DDA-HS) can optimize the RBFN-DDA parameters and improve classification performances from the original RBFN-DDA by 2.2% up to 22.5% in two benchmarks datasets, which are numerical records from a bearing and steel plate system and a condition-monitoring system in a power plant (i.e., the circulating water (CW) system). The results also show that the proposed RBFN-DDA-HS is compatible, if not better than, the classification performances of other state-of-the-art machine learning methods.

데이터센터 최적 열환경 제공을 위한 적응형 인공신경망 냉매 유량 예측 모델 개발

데이터센터 최적 열환경 제공을 위한 적응형 인공신경망 냉매 유량 예측 모델 개발

Heterogeneous Sensor Data Analysis Using Efficient Adaptive Artificial Neural Network on FPGA Based Edge Gateway

We propose a FPGA based design that performs real-time power-efficient analysis of heterogeneous sensor data using adaptive ANN on edge gateway of smart military wearables. In this work, four independent ANN classifiers are developed with optimum topologies. Out of which human activity, BP and toxic gas classifier are multiclass and ECG classifier is binary. These classifiers are later integrated into a single adaptive ANN hardware with a select line( s) that switches the hardware architecture as per the sensor type. Five versions of adaptive ANN with different precisions have been synthesized into IP cores. These IP cores are implemented and tested on Xilinx Artix-7 FPGA using Microblaze test system and LabVIEW based sensor simulators. The hardware analysis shows that the adaptive ANN even with 8-bit precision is the most efficient IP core in terms of hardware resource utilization and power consumption without compromising much on classification accuracy. This IP core requires only 31 microseconds for classification by consuming only 12 milliwatts of power. The proposed adaptive ANN design saves 61% to 97% of different FPGA resources and 44% of power as compared with the independent implementations. In addition, 96.87% to 98.75% of data throughput reduction is achieved by this edge gateway.

Real-time damage detection of bridges using adaptive time-frequency analysis and ANN

Real-time damage detection of bridges using adaptive time-frequency analysis and ANN

Computer-aided diagnosis applied to MRI images of brain tumor using cognition based modified level set and optimized ANN classifier

MRI image segmentation and classification is one of the important tasks in medical image analysis and visualization, despite occurrence of noise makes it tough to segment the region of interest. In this paper, the MRI images are pre-processed and segmentation is carried out using modified Level set method for the tumor segmentation. Also, it is important to extract the useful features to predict the image class accurately. The proposed method operates Multi-Level wavelet decomposition features and for the wavelet coefficients modified chief descriptions like Grey Level Co-Occurrence Matrix (GLCM), Gabor and moment invariant features are extracted. The classification is carried out using the Adaptive Artificial Neural Network (AANN) methodology. In the adaptive ANN, the layer neurons are optimized using Whale Optimization Algorithm (WOA). The adaptive neural network optimizes the network structure to increase the classification accuracy and thus gives better classification results of tumors based on the segmented images. The proposed method will be executed in the working platform of MATLAB and the results are compared with the previous state of the art techniques. Finally, the proposed method results in classification accuracy of 98%.

A hybrid wavelet–APSO–ANN-based protection scheme for six-phase transmission line with real-time validation

The ever-increasing load demand motivates the adoption of the six-phase transmission system as a viable alternative to the conventional three-phase double-circuit line because of its enhanced power transfer capability. However, the acceptance of six-phase transmission is highly dependent on its protection scheme, which is a challenging task due to the large number of possible faults. In this regard, this paper describes a novel concept based on the hybridization of wavelet transform, adaptive particle swarm optimization (APSO) and ANN for fault classification and location in the six-phase transmission line. The proposed scheme overcomes the difficulties associated with existing ANN-based protection schemes related to the selection of architecture and convergence to the non-global optimal set of weight vectors and biases post-training. From the features/attributes extracted using wavelet transform, the hybrid algorithm utilizes the strong exploration and mapping capability of APSO and ANN, respectively, for input–output mapping. It significantly increases the probability of global convergence for ANN training, thereby improving the performance of the protection scheme in terms of fault detection, classification and location estimation accuracy. The proposed scheme has been extensively tested for wide variation in different fault attributes. In addition to the offline simulations, the effectiveness of the proposed scheme has been validated in real-time environment by carrying out hardware in loop simulations using OPAL-RT digital simulator.

An Adaptive Activated ANN Equalizer Applied in Millimeter-Wave RoF Transmission System

We experimentally demonstrate a 60-GHz radio-over-fiber (RoF) system using an adaptive activated artificial neural network nonlinear equalizer (ANN-NLE) to improve the bit error rate (BER) performance. We propose a variable- $\alpha $ activation function ANN-NLE with a faster convergence speed and a stronger equalization ability to solve the nonlinear problems generated from the RoF link. To the best of our knowledge, this letter is the first to apply an adaptive activation function ANN to solve a classification problem and equalize the signals in a 60-GHz RoF system. Using the proposed ANN-NLE, a higher convergence speed and better BER performances can be achieved. Comparisons between our proposed ANN-NLE and nonlinear Volterra equalizer in polynomial structure are reported. Using the proposed ANN-NLE, the system successfully transmits a 5-Gbps BPSK signal through a 10-km fiber and 1.2-m wireless link under a forward error correction limit of $10^{-3}$ with much higher power sensitivity than the Volterra equalizer.

Nonlinear control strategies based on Adaptive ANN models: Multi-product semi-batch polymerization reactor case study

In this paper, two versions of Adaptive Artificial Neural Network (AANN) model based Generic Model Control (GMC) [AANNGMC1 & AANNGMC2] are proposed for nonlinear processes. The proposed controllers consist of online parameter estimation of a purely data driven ANN model based on past measurements using Extended Kalman Filter, and control computation to track the predicted output derivative along its desired reference trajectory in a GMC frame-work. A industrial multi-product semi-batch polymerization reactor case study posed as a challenge problem for temperature control has been considered to apply the proposed AANNGMC1 and AANNGMC2 strategies at supervisory level, as cascade control configurations along with Proportional Integral (PI) controller i.e. AANNGMC1–PI and AANNGMC2–PI. The proposed configurations have shown better controller performance in terms of temperature tracking as well as smoother input profiles compared to the exact model based GMC–P1 and standard PI–PI control schemes for two different products, and among the two configurations, AANNGMC2–PI has exhibited better performance. The proposed schemes are found to be versatile although they are based on a purely data-driven models with online parameter estimation.

Comparison of gene regulatory networks using adaptive neural network and self-organising map approaches over Huh7 hepatoma cell microarray data matrix

Construction of gene regulatory network GRN is very important as it governs the expression levels of biomolecules in microarray data. In this article, we have developed GRNs by adaptive neural network ANN and self-organising map SOM approaches over Hepatitis C virus infection effect on Huh7 hepatoma cell microarray time series data. We then compared GRNs for the best performance analysis. We used fuzzy C-means clustering method to cluster the normalised dataset and then cluster centres are identified. After constructing GRNs within cluster centres, we analysed that SOM topology results a better performance providing minimum error to construct the GRN from sample data.

Solubility prediction of supercritical carbon dioxide in 10 polymers using radial basis function artificial neural network based on chaotic self-adaptive particle swarm optimization and K-harmonic means

Excellent prediction modeling of CO2solubility in polymers using hybrid computation algorithm.

Estimation and optimisation of right-censored data in survival analysis by neural network

The main topic in medical statistics is survival time. It is difficult to obtain complete data in studies of survival time because of several reasons. One aspect of such difficulty is related to death of all patients. A study is often completed before the death of all patients. Moreover, incomplete information is kept with respect to death of all patients. Usually, exponential and Weibull distribution are used to estimate actual time of the censored data. In general, censored data models describe situations where there are variables of interest that cannot always be observed directly and may deviate from certain values. This study presents an adaptive neural network ANN approach for improved estimation of right censored data. To show the superiority and advantages of the ANN approach, it has been compared with fuzzy mathematical programming and conventional approaches. It is shown that ANN provides better estimation for right censored data in comparison to fuzzy mathematical programming and exponential and Weibull distributions. This is the first study that utilises ANN for improved estimation of right censored data.

Artificial Neural Networks based UPFC for Damping Low Frequency Oscillations

Low Frequency Oscillations(LFO) are a frequent adverse phenomenon which increase the risk of instability for the power system and thus reduce the total and availability transfer capability.LFO occur in power systems because of lack of the damping torque in order to dominance to power system disturbances as change in mechanical input power. In the recent past Power System Stabilizer (PSS) was used to damp LFO. FACTs devices, such as Unified Power Flow Controller (UPFC), can control power flow and increase transient stability. So UPFC may be used to damp LFO instead of PSS. In this research the linearized model of synchronous machine (Heffron-Philips) connected to infinite bus (Single Machine-Infinite Bus: SMIB) with UPFC is used and also in order to damp LFO, adaptive ANN damping controller for UPFC is designed and simulated. Simulation is performed for various types of loads and for different disturbances. Simulation results demonstrate that the developed ANN damping controller would be more effective in damping electromechanical oscillations in comparison with the conventional lead-lag controller.

Non-linear model predictive control based on neural network model with modified differential evolution adapting weights

A modified differential evolution (MDE) optimisation approach is proposed to retrain the network weights of the multi-input multi-output artificial neural network (MIMO-ANN) process model. This is particularly useful for controlling the cases involving changing operating condition as well as highly non-linear processes. The utility of online retraining the network weights using MDE can further improve the predictive performances of the process model including both the possible control accuracy and the computational load reduction. A case study on a distillation column, which is a chemical non-linear process, is used to illustrate the effectiveness of the adaptive ANN based on MDE modelling and control method proposed in this paper. Significant improvements of the proposed strategy were obtained especially when assessing from the perspective of model generalisation.

Adaptive Neural Networks for Flood Routing in River Systems

A methodology based on adaptive ANN models is proposed for flood routing in river systems. The proposed methodology is capable of modeling both converging and diverging river networks. A Multilayer Perceptron Network (MLP), a Recurrent Neural Network (RNN), a Time Delay Neural Network (TDNN) and a Time Delay Recurrent Neural Network (TDRNN) are applied in this study. An Adaptive training procedure based on the Forgetting Factor (FF) approach is used to train ANNs models. The methodology provides a lead time equal to travel time for the flood estimation downstream of the river. The performances of the models are tested within the two distinctive parts of the Karoon River in south-west Iran. The first case study uses synthetic floods generated by the HEC-RAS hydraulic model; the second one uses observed floods. Besides, the Muskingum routing method is used in the second case study to be compared with the results of ANN models. Overall, the results demonstrated that the proposed methodology performs well considering goodness-of-fit criteria. Moreover, the dynamic neural networks outperform the static MLP and the Muskingum model.