Features Of Intrinsic Mode Functions Research Articles

Fault diagnosis method of rolling bearings based on adaptive modified CEEMD and 1DCNN model

The working environment of rolling bearings is highly complex and often the vibration signal of the bearing is mixed with noise, which makes fault diagnosis challenging. As such, it is imperative to denoise the vibration signal of rolling bearings, extract effective vibration features, and improve classification accuracy. In this research, we propose a rolling bearing fault diagnosis model based on adaptive modified complementary ensemble empirical mode decomposition (AMCEEMD) and a one-dimensional convolutional neural network (1DCNN). Firstly, the AMCEEMD method is proposed. This algorithm is an improved signal processing technique based on CEEMD, which introduces fuzzy entropy and kurtosis values to remove noise and identify impulse signals. The purpose of AMCEEMD is to obtain standard Intrinsic Mode Functions (IMFs) while removing noise. Secondly, we introduce the energy ratio, fuzzy entropy, and kurtosis as selection indices for IMFs. The selection of IMFs is adapted, and the selected IMF features are inputted into 1DCNN for fault classification. Finally, it was validated by two bearing experiments and compared with other classification methods. The classification accuracy of AMCEEMD-1DCNN method in this study is higher than other methods. The effectiveness of the AMCEEMD-1DCNN fault diagnosis model was verified.

Feature Extraction based on Empirical Mode Decomposition for Shapes Recognition of Buried Objects by Ground Penetrating Radar

Ground penetrating radar (GPR) is one of the promising non-destructive imaging tools investigations for shallow subsurface exploration such as locating and mapping the buried utilities. In practical applications, GPR images could be noisy due to the system noise, the heterogeneity of the medium, and mutual wave interactions thus, it is a complex task to recognizing the hyperbolic signature of buried objects from GPR images. Therefore, this paper aims to develop nonlinear feature extraction technique of using Empirical Mode Decomposition (EMD) in recognizing the four geometrical shapes (cubic, cylindrical, disc and spherical) from GPR images. A pre-processing step of isolating hyperbolic signature from different background was first employed by mean of Region of Interest (ROI). The hyperbolic signature that describes the shapes was extracted using EMD decomposition to obtain a set of significant features. In this framework, the hyperbolic pattern was decomposed of using EMD, to produce a small set of intrinsic mode functions (IMF) via sifting process. The IMF properties of the signature that exhibit the unique pattern was used as potential features to differentiate the geometrical shapes of buried objects. The extracted IMF features were then fed into machine learning classifier namely Support Vector Machines. To evaluate the effectiveness of the proposed method, a set data collection of GPR-images has been acquired. The experimental results show that the recognition rate of using IMF features was achieved 99.12% accuracy in recognizing the shapes of buried objects whose shows the promising result.

Classification of Body Mass Index Based Facial Images using Empirical Mode Decomposition

Human faces contain rich information. Recent studies found that facial features have relation with human weight or body mass index (BMI). Decoding “facial information” from the face in predicting the BMI could be linked to the various health marker. This paper proposed the classification of body mass index (BMI) based on appearance based features of facial images using empirical mode decomposition (EMD) as feature extraction technique. The facial images that describe the body mass index was extracted using EMD to obtain a set of significant features. In this framework, the facial image was decomposed using EMD to produce a small set of intrinsic mode functions (IMF) via sifting process. The IMF features which exhibit the unique pattern were used to classify the BMI. The obtained features were then fed into machine learning classifier such as k-nearest neighbour and support vector machines (SVM) to classify the three BMI classes namely normal, overweight and obese. The obtained results show that the IMF2 feature using SVM classifier achieved recognition rate of 99.12% which show promising result.

A Transfer Learning Based Unmanned Aerial Vehicle MEMS Inertial Sensors Fault Diagnosis Method

In this paper, we propose a novel transfer learning based micro-electromechanical system (MEMS) inertial sensors fault diagnosis method. First, the MEMS inertial sensors fault diagnosis method is formulated to a deep transfer learning problem in which the offline samples are deemed as source domain and the online samples are set to target domain features. Second, the bidirectional long short-term memory and Hilbert-Huang transformation-based feature transfer model is designed to decrease the discrepancy between SD and TD, that performs the transfer operation using intrinsic mode function features. Then we propose a convolutional neuro network-based transfer learning algorithm to further decrease deep features discrepancy and perform the fault classification tasks on TD. According to the experiments, the proposed FD method has achieved excellent fault classification performance and significantly improvement comparing with the state-of-the art methods.

A novel classification method based on ICGOA-KELM for fault diagnosis of rolling bearing

A novel classification method based on ICGOA-KELM is presented in this paper. In ICGOA-KELM, an improved circle chaotic map with grasshopper optimization algorithm (ICGOA) is designed to optimize the parameters of Kernel extreme learning machine (KELM) to improve the stability and accuracy of fault classification for rolling bearing based on parameter modification of circle chaotic map. Grasshopper optimization algorithm (GOA) is a new heuristic optimization algorithm, which has strong global searching ability. However, it still may fall into local optimization in some cases. In this paper, the vibration signals of rolling bearing are preprocessed by using Variational Modal Decomposition (VMD). Then Multi-scale Permutation Entropy (MPE) is utilized to extracted features of intrinsic mode functions (IMFs) decomposed by VMD. In addition, KPCA is adopted to select the salient features with high contribution rates to remove redundant and irrelevant features. Finally, the salient features are fed into ICGOA-KELM to fulfill fault classification. Therefore, a new fault detection and classification method based on VMD, MPE, KPCA and ICGOA-KELM is proposed. This method is applied to the fault classification of rolling bearing and the identification of different damage fault degrees. Experiments verify that the proposed method is more effective than CGOA-KELM for fault diagnosis of rolling bearing.

An optimized design of seizure detection system using joint feature extraction of multichannel EEG signals.

The detection of seizure onset and events using electroencephalogram (EEG) signals are important tasks in epilepsy research. The literature available on seizure detection has discussed the implementation of advanced signal processing algorithms using tools accessed over the cloud. However, seizure monitoring application needs near sensor processing due to privacy and latency issues. In this paper, a real time seizure detection system has been implemented using an embedded system. The proposed system is based on ensemble empirical mode decomposition (EEMD) and tunable-Q wavelet transform (TQWT) algorithms. The analysis and classification of non-stationary EEG signals require the wavelet transform with high Q-factor. However, direct use of TQWT increases the computational complexity of feature extraction from multivariate EEG signals. In this paper, the first step is to process the signal by using EEMD to obtain 8 intrinsic mode functions (IMFs). The Kraskov (KraEn), sample (SampEn), and permutation (PermEn) entropy features of IMFs are extracted and based on optimum values, and 4 IMFs are decomposed using TQWT. Secondly, centered correntropy (CenCorrEn) features of the 1st and 16th sub-band of TQWT have been used as classifier inputs. The performance of multilayer perceptron neural networks (MLPNN), least squares support vector machine (LSSVM), and random forest (RF) classifiers has been tested on the multichannel EEG data recorded from a local hospital. The RF classifier has produced the highest accuracy of 96.2% in classifying the signals. The proposed scheme has been employed in developing an embedded seizure detection system to assist neurologists in making seizure diagnostic decisions.

Internal leakage identification of hydraulic cylinder based on intrinsic mode functions with random forest

Monitoring for internal leakage of hydraulic cylinders is vital to maintain the efficiency and safety of hydraulic systems. An intelligent classifier is proposed to automatically evaluate internal leakage levels based on the newly extracted features and random forest algorithm. The inlet and outlet pressures as well as the pressure differences of two chambers are chosen as the monitoring parameters for leakage identification. The empirical mode decomposition method is used to decompose the raw pressure signals into a series of intrinsic mode functions to obtain the essence in experimental signals. Then, the features extracted from intrinsic mode functions in terms of statistical analysis are formed the input vector to train the leakage detector. The classifier based on random forest is established to categorize internal leakage into proper levels. The accuracy of the internal leakage evaluator is verified by the experimental pressure signals. Moreover, an internal leakage evaluator is established based on the support vector machine algorithm, in which the wavelet transform is applied for feature extraction. The accuracy and efficiency of different classifiers are compared based on leakage experiments. The results show that the classifier trained by the intrinsic mode function features in terms of random forest algorithm may more effectively and accurately identify internal leakage levels of hydraulic cylinders. The leakage evaluator provides probability for online monitoring of the internal leakage of hydraulic cylinders based on the inherent sensors.

Feature analysis of dysphonia speech for monitoring Parkinson's disease.

Parkinson's disease (PD) is a progressive neurodegenerative disorder that has no known cure and no known prevention. Early detection is crucial in order to slow down the progress. In the past 10 years, interest in PD analysis has visibly increased. Speech impairment affects the majority of people with Parkinson's (PWP). New features and machine learning algorithms were proposed to help diagnose PD and to measure a patient's progress. Using sustained vowel /a/ recordings, we identified a more prominent set of Mel-Frequency Cepstral Coefficient (MFCC) and Intrinsic Mode Functions (IMF), and other parameters that can best represent the characteristics of Parkinson's dysphonia to assist with the diagnosis process. For higher quality audio signals, there is a visible difference in the higher MFCC coefficients, the wider spectrum bandwidth in the first four IMFs of PWP, and higher power intensity in the healthy subjects. We also found that even when the signals are downsampled into toll-quality, the distinguishable MFCC and IMF features were largely maintained. This enabled a whole possibility of providing telemedicine for PWP.

Ictal EEG classification based on amplitude and frequency contours of IMFs

Electroencephalogram (EEG) signal serves is a powerful tool in epilepsy detection. This study decomposes intrinsic mode functions (IMFs) into amplitude envelope and frequency functions on a time-scale basis using the analytic function of Hilbert transform. IMFs results from the empirical mode decomposition of EEG signals. Features such as energy and entropy parameters were calculated from the amplitude contour of each IMF. Other features, such as interquartile range, mean absolute deviation and standard deviation are also computed for their instantaneous frequencies. Discriminative features were extracted using a large database to classify healthy and ictal EEG signals. Normal EEG segments were differentiated from the seizure attack in individual IMF features, multiple features with individual IMF, and individual features with multiple IMFs. Discriminating capability of three Cases was tested. (i) Artificial neural network and (ii) adaptive neuro-fuzzy inference system classification were used to identify EEG segments with seizure attacks. ANOVA was used to analyze statistical performance. Energy and entropy-based features of instantaneous amplitude and standard deviation of instantaneous frequency of IMF2 and IMF1 have 100% accuracy, sensitivity, and specificity. Good performance with a single feature that represents information of the whole data was obtained. The result involved less complicated computation than other time–frequency analysis techniques.

Hilbert-Huang Transformation Based Analyses of FP1, FP2, and Fz Electroencephalogram Signals in Alcoholism.

Chronic alcoholism may damage the central nervous system, causing imbalance in the excitation-inhibition homeostasis in the cortex, which may lead to hyper-arousal of the central nervous system, and impairments in cognitive function. In this paper, we use the Hilbert-Huang transformation (HHT) method to analyze the electroencephalogram (EEG) signals from control and alcoholic observers who watched two different pictures. We examined the intrinsic mode function (IMF) based energy distribution features of FP1, FP2, and Fz EEG signals in the time and frequency domains for alcoholics. The HHT-based characteristics of the IMFs, the instantaneous frequencies, and the time-frequency-energy distributions of the IMFs of the clinical FP1, FP2, and Fz EEG signals recorded from normal and alcoholic observers who watched two different pictures were analyzed. We observed that the number of peak amplitudes of the alcoholic subjects is larger than that of the control. In addition, the Pearson correlation coefficients of the IMFs, and the energy-IMF distributions of the clinical FP1, FP2, and Fz EEG signals recorded from normal and alcoholic observers were analyzed. The analysis results show that the energy ratios of IMF4, IMF5, and IMF7 waves of the normal observers to the refereed total energy were larger than 10 %, respectively. In addition, the energy ratios of IMF3, IMF4, and IMF5 waves of the alcoholic observers to the refereed total energy were larger than 10 %. The FP1 and FP2 waves of the normal observers, the FP1 and FP2 waves of the alcoholic observers, and the FP1 and Fz waves of the alcoholic observers demonstrated extremely high correlations. On the other hand, the FP1 waves of the normal and alcoholic observers, the FP1 wave of the normal observer and the FP2 wave of the alcoholic observer, the FP1 wave of the normal observer and the Fz wave of the alcoholic observer, the FP2 waves of the normal and alcoholic FP2 observers, and the FP2 wave of the normal observer and the Fz wave of the alcoholic observer demonstrated extremely low correlations. The IMF4 of the FP1 and FP2 signals of the normal observer, and the IMF5 of the FP1 and FP2 signals of the alcoholic observer were correlated. The IMF4 of the FP1 signal of the normal observer and that of the FP2 signal of the alcoholic observer as well as the IMF5 of the FP1 signal of the normal observer and that of the FP2 signal of the alcoholic observer exhibited extremely low correlations. In this manner, our experiment leads to a better understanding of the HHT-based IMFs features of FP1, FP2, and Fz EEG signals in alcoholism. The analysis results show that the energy ratios of the wave of an alcoholic observer to its refereed total energy for IMF4, and IMF5 in the δ band for FP1, FP2, and Fz channels were larger than those of the respective waves of the normal observer. The alcoholic EEG signals constitute more than 1 % of the total energy in the δ wave, and the reaction times were 0_4, 4_8, 8_12, and 12_16 s. For normal EEG signals, more than 1 % of the total energy is distributed in the δ wave, with a reaction time 0 to 4 s. We observed that the alcoholic subject reaction times were slower than those of the normal subjects, and the alcoholic subjects could have experienced a cognitive error. This phenomenon is due to the intoxicated central nervous systems of the alcoholic subjects.

A NEW DATA MINING APPROACH FOR GEAR CRACK LEVEL IDENTIFICATION BASED ON MANIFOLD LEARNING

Gear crack is a common damage model in the gear mechanisms, and an unexpected serious crack may break the transmission system down, leading to significant economic losses. Efficient incipient fault detection and diagnosis are therefore critical to machinery normal running. One of the key points of the fault diagnosis is feature extraction and selection. Literature review indicates that only limited research considered the nonlinear property of the feature space by the use of manifold learning algorithms in the field of mechanic fault diagnosis, and nonlinear feature extraction for gear crack detection are scarce. This paper reports a novel data mining method based on the empirical mode decomposition (EMD) and supervised locally linear embedding (SLLE) applied to gear crack level identification. The EMD was used to decompose the vibration signals into a number of intrinsic mode functions (IMFs) for feature extraction, whilst the SLLE for nonlinear feature selection. The experimental vibration data acquired from the gear fault test-bed were processed for feature reduction and extraction using the proposed method. Study results show that the sensitive characteristics between different gear crack severity vibration signals can be revealed effectively by EMD-SLLE. The energy distribution and the statistic features of IMFs vary with the change of the gear operation conditions, and the most distinguished features can be extracted by nonlinear method of SLLE. In addition, the performance of feature extraction of SLLE is better than that of the linear method of principal component analysis (PCA).DOI: http://dx.doi.org/10.5755/j01.mech.18.1.1276

Classification of parkinsonian and essential tremor using empirical mode decomposition and support vector machine

Taking into account two types of tremor, namely essential tremor (ET) and Parkinson@?s disease (PD), which are often misdiagnosed in clinical practice, a novel approach using singular value decomposition (SVD) to extract the features of intrinsic mode functions (IMFs) and support vector machine (SVM) is proposed to distinguish between them. The hand acceleration signals of 25 voluntary subjects with two conditions were collected and preprocessed. The empirical mode decomposition (EMD) method decomposed the signals into a number of stationary IMFs. The singular value features of the preceding four IMFs were extracted, and the features were inputted to the SVM classifier, which can recognize the two types of tremor. To comparing, we also used the singular value features of discrete wavelet transform (DWT) as input to the SVM. Cross-validation testing results indicated that the accuracy, sensitivity, and specificify of EMD-SVD features extracted are all remarkable higher than that of DWT-SVD method. Due to the accuracy, sensitivity, and specificify could arrive at 98%, 97.5% and 98.33% respectively, thus, practical guiding significance for diagnosing tremor types in clinic is provided.