Vector Quantization Neural Network Research Articles

Digital Image Processing for Detecting Industrial Machine Work Failure with Quantization Vector Learning Method

Todays, digital image processing is widely used in various fields to facilitate humans in doing work by analyzing videos or images for use in decision making in the industrial world. The use of industrial machine technology is one of the most important factors in efforts to facilitate human work, but an industrial machine is inseparable from work failure that can hinder the production process and cause harm to the industry. This study aims to detect a failure in industrial machinery by using video data of industrial machine movements recorded using a webcam camera. For the preprocessing stage, the image is resized then converted to grayscale imagery and segmented using the thresholding method, then morphological operations are performed with an opening operation, feature extraction is done by changing the binary image into a vector data that is used as input data in the classification process using the Learning Vector Quantization Neural Network Algorithm (LVQ NN) version 1. The results showed the results of the detection of machine working errors can be done well with an accuracy value of 94.24% in training and 92.38% in the testing phase.

Simulation and Experimental Investigation on Active Thermography Test of the Solder Balls

The miniaturized electronic products require not only high density, but also high reliability. Defect inspection for the high-density package becomes challenging gradually. In this article, active thermography technology is used to detect the solder balls in flip-chip (FC) packages. A single-layer FC model is constructed and heat conduction in the FC package is simulated by using the finite-element analysis code of COMSOL. Experimental investigation is also carried out to inspect the solder defects by using the laser excitation thermography test system. Infrared images of the SFA1 package are captured, and the spatial adaptive filtering algorithm is used to remove the thermal noise. The hot spots of the solder balls are segmented from the filtered infrared image, and the representative features are extracted for each hot spot. The modified learning vector quantization (LVQ) neural network is used for the classification of the solder balls. The missing solder balls are identified accurately. The simulation and experimental results prove that the proposed approach using active thermography and LVQ algorithm is effective for defect inspection in high-density electronic devices.

Optimal multi-product supplier selection under stochastic demand with service level and budget constraints using learning vector quantization neural network

In today’s competitive marketplace demand, evaluation and selection of suppliers are pivotal for firms, and therefore decision makers need to select suppliers and the optimal order quantities when outsourcing. However, there is uncertainty and risk due to lack of precise data for supplier selection. Uncertainty can impose shortage or overstocks, because of stochastic demand, to firms; in this case, considering inventory control is essential. In this research, an appropriate spatial model is developed for a multi-product supplier selection model with service level and budget constraints. Learning Vector Quantization Neural Network is used to find the optimal number of decision variables with the goal of maximizing the expected profit of supply chains. By analyzing a practical example and conducting sensitivity analysis, we find that corporate profit will be maximized if the optimal integration of suppliers and the optimal order quantities from each supplier is determined. In addition, budget and service level should be considered in the process of finding the best result.

Traffic injury severity prediction along with identification of contributory factors using learning vector quantization: a case study of the city of London

This study presents learning vector quantization neural network modelling to predict injury severity of driver as well as riders, which applies to the backbone of traffic networks for London’s central business district. The potential associations between injury severity classes and crash related factors that contribute to their generation are discovered. Accordingly, the model is addressed as an identification technique for contributory factors and range of interventions for road safety. Unsurprisingly, approaching a T/staggered junction is detected as an accident hotspot. Injuries caused by going ahead on a bend and turning manoeuvres are ranked as the next most important contributory factors. Likewise, the affect of most junction actions were almost triple compared to the other indexes. All other sensitive predictors approximately were held near as equal; injuries involving a stationary or parked vehicle, factors related to junction control, crossing facilities, alcohol involvement, rush hours, and vehicle type. Following this implication, with the purpose of maximising the likelihood of injury accuracy, the model is predicted through the most sensitive predictors.

Classification and characterization of thermally modified timber using visible and near-infrared spectroscopy and artificial neural networks: a comparative study on the performance of different NDE methods and ANNs

Visible and near-infrared (VIS–NIR) spectroscopy was used for classifying and predicting the properties of thermally modified Western hemlock wood. The specimens were treated at 170 °C, 212 °C, and 230 °C. The dimensional reduction was performed using linear discriminant analysis, and the resulted dataset was used for wood classification using the support vector machines and the linear vector quantization neural network. The VIS–NIR dataset was also used to predict the wood moisture content, swelling coefficient, water absorption, density, dynamic modulus of elasticity, and hardness. The “adaptive neuro-fuzzy inference system” (ANFIS), “Group Method of Data Handling” (GMDH), and “multilayer perceptron” (MLP) neural networks were employed for predicting the wood properties. It was shown that regardless of the type of the neural network, NIR dataset provided a robust model with 100% classification accuracy, which can be implemented in industrial scale for in-line timber quality control. The results indicated that the ANFIS and GMDH neural network showed higher performance than the MLP model for predicting the wood properties. While the VIS–NIR data resulted in a promising accuracy for predicting the wood moisture content and dimensional stability parameters, it did not seem suitable for the prediction of wood density and its mechanical properties. The performance of the VIS–NIR spectroscopy method for classification and characterization of heat-treated timber was compared with that obtained using the color measurement and the stress wave method detected by the acoustic emission sensor.

Evolution prediction of unconventional emergencies via neural network: An empirical study of megacities

Megacities occupy an important position in the country they locate and even the whole world. Therefore it is of great significance to investigate the evolution mechanism of unconventional emergencies and to explore the probability of predicting the secondary disasters in megacities. In this paper we study the evolution prediction problem of unconventional emergencies and formulate it as multi-label classification. A novel multi-label learning vector quantization (LVQ) neural network is proposed to construct the prediction model able to forecast the type of sub-events. A real data set of 85 megacities all over the world is collected and used as a case study in the experiments. The prediction performance is measured by the match between the real label set and retrieved label set. The empirical results demonstrate the effectiveness of using LVQ in a multi-label scenario for predicting the type of secondary disasters. The model is beneficial for emergency managers to predict the potential secondary disasters and thus make informed decisions for disaster prevention and mitigation.

Real-Time Control Strategy for CVT-Based Hybrid Electric Vehicles Considering Drivability Constraints

The energy management strategy has a great influence on the fuel economy of hybrid electric vehicles, and the equivalent consumption minimization strategy (ECMS) has proved to be a useful tool for the real-time optimal control of Hybrid Electric Vehicles (HEVs). However, the adaptation of the equivalent factor poses a major challenge in order to obtain optimal fuel consumption as well as robustness to varying driving cycles. In this paper, an adaptive-ECMS based on driving pattern recognition (DPR) is established for hybrid electric vehicles with continuously variable transmission. The learning vector quantization (LVQ) neural network model was adopted for the on-line DPR algorithm. The influence of the battery state of charge (SOC) on the optimal equivalent factor was studied under different driving patterns. On this basis, a method of adaptation of the equivalent factor was proposed by considering the type of driving pattern and the battery SOC. Besides that, in order to enhance drivability, penalty terms were introduced to constrain frequent engine on/off events and large variations of the continuously variable transmission (CVT) speed ratio. Simulation results showed that the proposed method efficiently improved the equivalent fuel consumption with charge-sustaining operations and also took into account driving comfort.

Control Strategy for PHEB Based on Actual Driving Cycle with Driving Style Characteristic

To exert fully the energy economy performance of plug-in hybrid electric buses (PHEBs) and enhance the adaptability to different drivers and driving cycles, a control strategy for PHEB based on actual driving cycle with driving style characteristic is proposed in this paper. Through the actual city bus driving data, collected in real time, 6 actual driving cycles with driving style characteristic are fitted by using Principal Component Analysis (PCA) and Cluster Analysis (CA). Based on the 6 driving cycles, the key parameters of rule-based control strategy are optimized and established by a combinatorial optimization algorithm in Isight. Then, an identification model to recognize the current condition based on the Learning Vector Quantization (LVQ) neural network has been built and trained offline, which is integrated in the control strategy for PHEB to invoke the corresponding optimized key control parameters in real time. A hardware-in-the-loop (HIL) test is conducted, and the result shows that the proposed strategy could improve the energy consumption by 4.94%, compared with the original rule-based control strategy, and its validity and practicability are fully verified.

Learning Vector Quantization Neural Network (LVQNN): Can it be implemented on the Forecasting Electrical Loads?

Short-term load forecasting is an important activity in planning the operation of the electric power system to estimate the load conditions of the following days and the results may help in the decision making. The imbalance in electrical power between the supply side and the demand side may lead blackouts on the consumer side. Consequently, the generating unit must be operated to meet its load requirements. Various techniques and methods are used to short-term load forecasting. Learning vector quantization neural network (LVQNN) is a classification algorithm that is superior in classifying digital images. Based on these considerations this research aimed to develop LVQNN to forecast short-term electricity peak loads. The idea was used as a reference on the discovery of architectural data classification processes that resembled forecasting techniques. LVQNN development was carried out by adjusting the sample data to the LVQNN architecture. First-distinct sample data were used to obtain the weight vector, then the remaining data from the distinct data process were divided into training data and testing data. By developing the new concept of LVQNN into forecasting technique, the preliminary research was conducted on 11 commodities to predict price fluctuations that have relatively precise forecasting values when implemented in the population. Based on these results, this algorithm will be developed to forecast the fluctuation of electrical load to provide better results in the upcoming research. It is expected to perform with better accuracy results to indicate that the mean absolute percentage error (MAPE) of the predicted values of loads (in MW) will close to the actual loads.

Establishment of multiple diagnosis models for colorectal cancer with artificial neural networks.

The current study aimed to develop multiple diagnosis models for colorectal cancer (CRC) based on data from The Cancer Genome Atlas database and analysis with artificial neural networks in order to enhance CRC diagnosis methods. A genetic algorithm and mean impact value were used to select genes to be used as numerical encoded parameters to reflect cancer metastasis or aggression. Back propagation and learning vector quantization neural networks were used to build four diagnosis models: Cancer/Normal, M0/M1, carcinoembryonic antigen (CEA) <5/≥5 and Clinical stage I–II/III–IV. The performance of each model was evaluated by predictive accuracy (ACC), the area under the receiver operating characteristic curve (AUC) and a 10-fold cross-validation test. The ACC and AUC of the Cancer/Normal, M0/M1, CEA and Clinical stage models were 100%, 1.000; 87.14%, 0.670; 100%, 1.000; and 100%, 1.000, respectively. The 10-fold cross-validation test of the ACC values and sensitivity for each test were 93.75–99.39%, 1.0000; 80.58–88.24%, 0.9286–1.0000; 67.21–92.31%, 0.7091–1.0000; and 59.13–68.85%, 0.6017–0.6585, respectively. The diagnosis models developed in the current study combined gene expression profiling data and artificial intelligence algorithms to create tools for improved diagnosis of CRC.

Risk Assessment of Hypertension in Steel Workers Based on LVQ and Fisher-SVM Deep Excavation

The steel industry is one of the pillar industries in China. The physical and mental health of steel workers is related to the development of China's steel industry. Steel workers have long been working in shifts, high temperatures, noise, highly stressed, and first-line environments. These occupational related factors have an impact on the health of steel workers. At present, the existing hypertension risk scoring models do not include occupational related factors, so they are not applicable to the risk score of hypertension in steel workers. It is necessary to establish a risk scoring model for hypertension in steel workers. In this study, the learning vector quantization (LVQ) neural network algorithm and the FisherSVM coupling algorithm are applied to estimate the hypertension risk of steel workers, and the microscopic laws of the tailing phenomenon of the two algorithms are analyzed by means of graphics analysis, which can describe the influence trend of sample size change in different intervals on the classification effect. The results show that the classification accuracy of the algorithm depends on the size of the sample space. When the sample size n ≤ 30 * (k + 1), the Fisher-SVM coupling intelligent algorithm is more applicable. Because its average accuracy rate is 90.00%, the average accuracy of the LVQ algorithm is only 63.34%. When the sample size is n > 30 * (k + 1), the LVQ algorithm is more applicable. Because its average accuracy rate is 93.33%, the average accuracy of the Fisher-SVM coupling intelligent algorithm is only 76.67%. The sample size of this paper is 4422, and the prediction of LVQ neural network model is more accurate. Therefore, based on the relative importance of each risk factor obtained by this model and to establish a steel worker hypertension risk rating scale, the score greater than 18 is considered as the high risk, 12-18 is considered as the medium risk, and less than 12 is considered as the low risk. Through the example's verification, the accuracy rate of the scale is 90.50% and the effect is very good. It shows that the established scoring system can effectively assess the risk of hypertension in steel workers and provide an effective basis for primary prevention of hypertension in steel workers.

Series Arc Fault Detection of Indoor Power Distribution System Based on LVQ-NN and PSO-SVM

When a series arc fault occurs in indoor power distribution system, current value of circuit is often less than the threshold of the circuit breaker, but the temperature of arc combustion can be as high as thousands of degrees, which can lead to electrical fire. The arc fault experimental platform is used to collect circuit current data of normal work and arc fault. Five types of loads which are commonly used in indoor distribution system, such as resistive and inductive in series load, resistive load, series motor load, switching power load and eddy current load, are chosen. This paper uses four features of current in time domain, i.e. current average, current pole difference, difference current average and current variance. Ten features of current in frequency domain are extracted by Fast Fourier Transform (FFT). The learning vector quantization neural network (LVQ-NN) is designed to judge the load type. The support vector machine optimized by particle swarm optimization (PSO-SVM) is designed to detect the arc fault. The simulation results show the effectiveness of the proposed method.

Identification of Disease on Leaves Soybean using Modified Otsu and Learning Vector Quantization Neural Networks

Disease of the soybean crop is one of the obstacles to increase soybean production in Indonesia. Some of these diseases usually are found in the leaves and resulted to the crop become unhealthy. This study aims to identify disease on soybean leaf through leaves image by applying the Learning Vector Quantization (LVQ) algorithm. The identification begins with preprocessing using modified Otsu method to get part of the diseases on the leaves with a certain threshold value. The next process is to identify the type of disease using LVQ. This process uses the minimum value, the maximum value and the average value of the red, green and blue color of the image. The testing conducted in this study is to identify two diseases called Peronospora manshurica (Downy Mildew) and phakopsora pachyrhizi (Karat). The result of testing by using 60 training data and the value of all recommendations parameters obtained the highest accuracy of identification is 95% %, but the more stable accuracy is 90%. This result shows that the method perform quite well identification of two mentioned disease.

Analysis of cosmic rays variations on the basis of neural networks

The work is devoted to the study of the dynamics of variations in cosmic rays and the identification of periods of sporadic effects. The data of the world network of neutron monitors have been studied, and signs of sporadic effects have been identified. Various architectures of neural networks were used and the quality of their work was assessed. The effectiveness of the application of vector quantization neural networks for the task of classification of cosmic ray data is shown. A method for improving the quality of the regression neural network operation based on the application of fast wavelet decompositions is proposed.

A Genetic-LVQ neural networks approach for handwritten Arabic character recognition

Handwritten recognition systems are a dynamic field of research in areas of artificial intelligence. Many smart devices available in the market such as pen-based computers, tablets, mobiles with handwritten recognition technology need to rely on efficient handwritten recognition systems. In this paper we present a novel Arabic character handwritten recognition system based on a hybrid method consisting of a genetic algorithm and a Learning vector quantization (LVQ) neural network. Sixty different handwritten Arabic character datasets are used for training the neural network. Each character dataset contains 28 letters written twice with 15 distinct shaped alphabets, and each handwritten Arabic letter is represented by a binary matrix that is used as an input to a genetic algorithm for feature selection and dimension reduction to include only the most effective features to be fed to the LVQ classifier. The recognition process in the system involves several essential steps such as: handwritten letter acquisition, dataset preparation, feature selection, training, and recognition. Comparing our results to those acquired by the whole feature dataset without selection, and to the results using other classification algorithms confirms the effectiveness of our proposed handwritten recognition system with an accuracy of 95.4%, hence, showing a promising potential for improving future handwritten Arabic recognition devices in the market.

Stress wave evaluation by accelerometer and acoustic emission sensor for thermally modified wood classification using three types of neural networks

Classification of thermally modified wood (TMW) allowing the distinction between different processing temperatures and the corresponding changes in wood properties is a crucial task in TMW grading. In this study, stress wave evaluation technique was used to classify the heat treatment level. Accordingly, an acoustic emission (AE) sensor and a pair of accelerometers captured stress waves generated by pendulum impact, and the data was used to classify the heat treatment level of thermally modified Western hemlock wood samples. Sensory features were extracted from time, frequency, and wavelet domain analysis. The extracted features were then used to train multilayer perceptron (MLP), group method of data handling (GMDH), and linear vector quantization (LVQ) neural networks for TMW classification. The results showed that while the features extracted from the accelerometers such as stress wave velocity and wood dynamic modulus of elasticity showed poor classification performance, acoustic emission sensory features were effective for classification of TMW. Wavelet domain features lead to better classification than those extracted from time and frequency domains. Feature fusion approach comprising the features from all the signal domains showed the best classification performance that was further improved by using a dimensionality reduction approach. The linear discriminant analysis was conducted on all acoustic emission features and resulted in 91.1% and 89.1% accuracy obtained from the LVQ and GMDH network, respectively. This performance was further increased to 98% and 97% using the LVQ and GMDH models when the input was combined with wood moisture content. The MLP neural network did not seem as suitable as the other two models. Neural network modeling using the captured stress wave data from an AE sensor could therefore be a promising nondestructive evaluation method for TMW classification.

Active control of hydropneumatic suspension parameters of wheel loaders based on road condition identification

Wheel loader is off-road vehicle and works on uneven terrain, unexpected banks or steep slopes. In order to improve the ride and stability of the vehicle, this study mainly focuses on how to adjust the parameters of hydropneumatic suspension through the identification of road conditions. Firstly, the multibody model of a wheel loader with hydropneumatic suspension is developed by RecurDyn in a co-simulation with MATLAB/Simulink. Secondly, a method of road level recognition based on learning vector quantization neural network is proposed to accurately identify the level of roads on which the wheel loader travels. Then, the hydropneumatic suspension parameters are optimized by using the particle swarm algorithm. A fuzzy controller is established based on the optimized parameters of the hydropneumatic suspension to realize the active adjustment of the suspension parameters under different road levels and driving speeds. Finally, a virtual prototyping model is used to analyse the influence of the active adjustment of suspension parameters on the vertical vibration under different driving conditions. Results show that the fuzzy controller can reasonably adjust the parameters of hydropneumatic suspension according to the identified road condition and effectively reduce the vertical vibration of the wheel loader.

A strategy of electrical energy management for internal combustion engine vehicle based on driving cycle recognition and electrical load perception

Electrical energy consumption is an important component of energy consumption for internal combustion engine vehicle, which directly affects the fuel economy. A bus-based electrical energy management system is built, and an electrical energy management strategy based on driving cycle recognition and electrical load perception is presented to achieve the refined management of vehicle energy. Six typical driving cycles are selected to establish an improved learning vector quantization neural network model for driving cycle recognition. The corresponding model training algorithm is designed by utilizing a similar driving cycle classification and the gradient optimization so that the better recognition accuracy and the less computation intensity can be obtained. An online recognition mechanism based on sliding time window is devised, and the optimal time window length is determined. To minimize fuel consumption, a dynamic optimal regulation strategy for the output power of the alternator and battery, which considers driving cycle recognition and electrical load perception, is proposed. Experimental results show that the strategy can effectually improve the vehicle fuel economy according to the driving cycle and the electrical load change and decrease the fuel consumption per 100 miles of vehicle.

Evolving neuro-fuzzy network for real-time high impedance fault detection and classification

This paper concerns the application of a neuro-fuzzy learning method based on data streams for high impedance fault (HIF) detection in medium-voltage power lines. A wavelet-packet-transform-based feature extraction method combined with a variation of evolving neuro-fuzzy network with fluctuating thresholds is considered for recognition of spatial–temporal patterns in the data. Wavelet families such as Haar, Symlet, Daubechie, Coiflet and Biorthogonal were investigated as a way to provide the most discriminative features for fault detection. The proposed evolving neuro-fuzzy classification model has shown to be particularly suitable for the problem because the HIF environment is subject to concept changes. Different from other statistical and intelligent approaches to the problem, the developed neuro-fuzzy model for HIF classification is not only parametrically, but also structurally adaptive to cope with nonstationarities and novelties. New neurons and connections are incrementally added to the neuro-fuzzy network when necessary for the identification of new patterns, such as faults and usual transients including sag, swell and spikes due to the switching of 3-phase capacitors and energization of transformers. Experimental evaluations compare the proposed classifier with other well-established computational intelligence methods, viz. multilayer perceptron neural network, learning vector quantization neural network and a support vector machine model. Results have shown that the evolving neuro-fuzzy system is effective and robust to changes. The system is able to maintain its detection and classification accuracy even in situations in which other classifiers exhibit a significant drop in accuracy due to gradual and abrupt changes of the fault patterns. Fuzzy rules are useful for interpretability purposes and help to enhance model credibility for decision making.

A Modular and Reconfigurable Pipeline Architecture for Learning Vector Quantization

Learning vector quantization (LVQ) neural networks have already been successfully applied for image compression and object recognition. In this paper, we propose a modular and reconfigurable pipeline architecture (MRPA) for LVQ. The MRPA consists of dynamically reconfigurable modules and realizes a run-time and on-chip configuration for recognition and learning. Prototype fabrication in 65-nm CMOS technology verifies high integration density and memory-utilization efficiency, good performance, and considerable flexibility in vector dimensionality, number of weight-vectors, and adaption strategies. Compared with the embedded microprocessors, which rely on single-instruction-multiple-data processing, the developed prototype increases the performance of both recognition and learning operations. The MRPA prototype shows improvements by factors of approximately 40 and 101 on the well-established performance metrics million connections per second for recognition and million connection updates per second for learning, respectively.