IJISA Vol. 11, No. 10, Oct. 2019
Cover page and Table of Contents: PDF (size: 184KB)
This research aims to test the feasibility of Programmable Logic Controller implementation of an Artificial Neural Network based bearing fault diagnosis using vibration datasets. The main drawback of using a Programmable Logic Controller along with an Artificial Neural Network is that it does not support the parallel nature of neural networks. This drawback is not significant for relatively small applications like bearing diagnosis that involve very short execution time. In this paper, a three layer multilayer perceptron backpropagation neural network is trained using Levenberg-Marquardt training algorithm with vibration dataset consisting of four bearing status classes: normal, outer race way fault, inner race way fault and rolling element (ball) fault. Time-frequency domain and time domain input features were considered in this research. Both approaches have performed well during simulation phase. But the time-frequency feature extraction approach was observed to take too long scan cycle time to be implemented in real-time. This is due to the computationally intensive nature of Fast Fourier Transform algorithm involved during feature extraction. The time domain approach is proved to be feasible for Programmable Logic Controller implementation. The time domain input features used for neural network training were root mean square, variance, kurtosis and negative log likelihood values. The average performance obtained during simulation with 10-fold cross validation performance estimator was an error of 7.9 x10-4. The performance tests of Programmable Logic Controller implementation resulted in 100% bearing fault detection rate.[...] Read more.
The disturbance cancellation techniques are investigated in this paper for Passive Bistatic Radars. The conventional procedure is to compute a clean signal by iteratively constructing an error vector from the residual of the surveillance samples after subtraction of a linear combination of clutters samples. A weight vector is eventually extracted in pure block algorithms, while a weight matrix is computed in iterative schemes. It is illustrated in this paper that the computed weight matrix in the latter case contains valuable information describing the clutters properties. The weight matrix-based disturbance attenuation technique is then innovated and its effectiveness is compared to the conventional error-based procedure in the test bed of several available iterative algorithms. Moreover, a revision of the FBLMS algorithm is presented to cover the case of complex input signals.[...] Read more.
Feature selection is a technique of selecting the most important features for predictive model construction. It is a key component in machine learning for many pattern recognition applications. The primary objective of this paper is to create a more precise system for Human Activity Recognition (HAR) by identifying the most appropriate features. We propose a Cyclic Attribution Technique (CAT) feature selection technique for recognition of human activity based on group theory and the fundamental properties of the cyclic group. We tested our model on UCI-HAR dataset focusing on six activities. With the proposed method, 561 features could be reduced to 63. Using an Artificial Neural Network (ANN), we compared performances of our new dataset with selected features and the original dataset classifier. Results showed that the model could provide an excellent overall accuracy of 96.7%. The proposed CAT technique can specify high-quality features to the success of HAR with ANN classifier. Two benefits support this technique by reducing classification overfitting and training time.[...] Read more.
At present, financial fraud detection is interested by many machine learning researchers. This is because of existing a big ratio between normal transactions and abnormal ones in data set. Therefore, a good result of prediction rate does not mean that there is a good detection result. This is explained that the experimental result might be effected by the imbalance in the dataset. Resampling a dataset before putting to classification process can be seen as the required task for researching in financial fraud detection area. An algorithm, so-called as MASI, is proposed in this paper in order to improve the classification results. This algorithm breaks the imbalance in the data set by re-labelling the major class samples (normal transactions) to the minor class ones basing the nearest neighbor’s samples. This algorithm has been validated with UCI machine learning repository data domain. Then, the algorithm is also used with data domain, which is taken from a Vietnamese financial company. The results show the better in sensitivity, specificity, and G-mean values compared to other publication control methods (Random Over-sampling, Random Under-sampling, SMOTE and Borderline SMOTE). The MASI also remains the training dataset whereas other methods do not. Moreover, the classifiers using MASI resampling training dataset have detected better number of abnormal transactions compared to the one using no resampling algorithm (normal training data).[...] Read more.
One of the main steps in the performance based seismic analysis and design of structures is determination of performance point where the nonlinear static analysis approach is used. The aim of this paper is to predict the performance point of semi-rigid steel frames using Artificial Neural Networks. As such, to generate data required for the prediction, several semi-rigid steel frames were modeled and their performance point was determined then. Ten input variables including number of bays, number of stories, bays width, moment of inertia of beams, cross sectional area of columns, cross sectional area of braces, rigidity degree of connections and soft story (existence or nonexistence) were considered in the prediction. In addition, the actual results were obtained at the presence of different earthquake intensity levels and soil types. Back Propagation with eleven different algorithms and Radial Basis Function Artificial Neural Networks were used in the prediction. The prediction process was carried out in two steps. In the first step, all samples were used for the prediction and the performance metrics were computed. In the second step, three of the best networks were selected, and the optimum number of samples was found considering a very slight reduction in the accuracy of the networks used. Finally, it was shown that, despite using rather limited number of samples, the generated Artificial Neural Networks accurately predict the performance point of semi-rigid steel frames.[...] Read more.
Indian practical rural distribution systems are very long and spread over a wide range of area. The nodes far away from the distribution substation are suffering from low voltage. In India, total distribution system losses are around 20% to 25%. From the past few years, penetration of distributed generation (DG) in to the distribution network/system is increasing expeditiously. DG allocation with appropriate location and size can provide numerous benefits to the distribution companies as well as to the society. In this regard, a new technique called combined sensitivity index (CSI), to find the optimal DG unit location, based on voltage sensitivity and network load magnitude is proposed. To assess the effectiveness of the proposed technique, it is tested on Indian practical 52-bus rural distribution system. The results obtained with the proposed CSI technique is compared with the results obtained with the combined power loss sensitivity (CPLS) technique. Here, the optimal DG unit size is calculated using Bird Swarm Algorithm (BSA). The results show that the proposed CSI technique performs better in minimizing power losses and voltage profile augmentation when compared to existing CPLS technique.[...] Read more.