Bong-Hwan Koh

Fault detection of a roller-bearing system through the EMD of a wavelet denoised signal.

This paper investigates fault detection of a roller bearing system using a wavelet denoising scheme and proper orthogonal value (POV) of an intrinsic mode function (IMF) covariance matrix. The IMF of the bearing vibration signal is obtained through empirical mode decomposition (EMD). The signal screening process in the wavelet domain eliminates noise-corrupted portions that may lead to inaccurate prognosis of bearing conditions. We segmented the denoised bearing signal into several intervals, and decomposed each of them into IMFs. The first IMF of each segment is collected to become a covariance matrix for calculating the POV. We show that covariance matrices from healthy and damaged bearings exhibit different POV profiles, which can be a damage-sensitive feature. We also illustrate the conventional approach of feature extraction, of observing the kurtosis value of the measured signal, to compare the functionality of the proposed technique. The study demonstrates the feasibility of wavelet-based de-noising, and shows through laboratory experiments that tracking the proper orthogonal values of the covariance matrix of the IMF can be an effective and reliable measure for monitoring bearing fault.

Fault Detection of Roller-Bearings Using Signal Processing and Optimization Algorithms

This study presents a fault detection of roller bearings through signal processing and optimization techniques. After the occurrence of scratch-type defects on the inner race of bearings, variations of kurtosis values are investigated in terms of two different data processing techniques: minimum entropy deconvolution (MED), and the Teager-Kaiser Energy Operator (TKEO). MED and the TKEO are employed to qualitatively enhance the discrimination of defect-induced repeating peaks on bearing vibration data with measurement noise. Given the perspective of the execution sequence of MED and the TKEO, the study found that the kurtosis sensitivity towards a defect on bearings could be highly improved. Also, the vibration signal from both healthy and damaged bearings is decomposed into multiple intrinsic mode functions (IMFs), through empirical mode decomposition (EMD). The weight vectors of IMFs become design variables for a genetic algorithm (GA). The weights of each IMF can be optimized through the genetic algorithm, to enhance the sensitivity of kurtosis on damaged bearing signals. Experimental results show that the EMD-GA approach successfully improved the resolution of detectability between a roller bearing with defect, and an intact system.

Fault Detection of Bearing Systems through EEMD and Optimization Algorithm

This study proposes a fault detection and diagnosis method for bearing systems using ensemble empirical mode decomposition (EEMD) based feature extraction, in conjunction with particle swarm optimization (PSO), principal component analysis (PCA), and Isomap. First, a mathematical model is assumed to generate vibration signals from damaged bearing components, such as the inner-race, outer-race, and rolling elements. The process of decomposing vibration signals into intrinsic mode functions (IMFs) and extracting statistical features is introduced to develop a damage-sensitive parameter vector. Finally, PCA and Isomap algorithm are used to classify and visualize this parameter vector, to separate damage characteristics from healthy bearing components. Moreover, the PSO-based optimization algorithm improves the classification performance by selecting proper weightings for the parameter vector, to maximize the visualization effect of separating and grouping of parameter vectors in three-dimensional space.

Compressive Strength Properties of Natural Gas Hydrate Pellet by Continuous Extrusion from a Twin-Roll System

This study investigates the compressive strength of natural gas hydrate (NGH) pellet strip extruded from die holes of a twin-roll press for continuous pelletizing (TPCP). The lab-scale TPCP was newly developed, where NGH powder was continuously fed and extruded into strip-type pellet between twin rolls. The system was specifically designed for future expansion towards mass production of solid form NGH. It is shown that the compressive strength of NGH pellet strip heavily depends on parameters in the extrusion process, such as feeding pressure, pressure ratio, and rotational speed. The mechanism of TPCP, along with the compressive strength and density of pellets, is discussed in terms of its feasibility for producing NGH pellets in the future.

Wavelet energy-based visualization and classification of high-dimensional signal for bearing fault detection

This study investigates a methodology for interpretable visualizing and classifying high-dimensional data such as vibration signals in machine fault detection application. Although principal component analysis is one of the most widely used dimension reduction methods, it does not clearly explain the source of signal variations (i.e., statistical characteristics such as mean and variance), but just locate signals on low-dimensional space which maximizing data dispersion. This deficiency restricts its interpretability to specific problems of process control and thus limits their broader usefulness. To overcome this deficiency, this study exploits the multiscale energy analysis of discrete wavelet transformation, so-called wavelet scalogram, in unsupervised manner. Wavelet scalogram allows us to first obtain a very low-dimensional feature subset of our data, which is strongly correlated with the characteristics of the data without considering the classification method used, although each of these features is uncorrelated with each other. In supervised learning scheme, it can be eventually combined with silhouette statistics for the purpose of more effective visualization of the main sources of different classes and classifying signals into different classes. Finally, nonparametric multi-class classifiers such as classification and regression tree and k-nearest neighbors quantitatively evaluate the performance of our approach for machine fault classification problem in terms of the 10-fold misclassification error rate.

The influence of enhanced closed-loop sensitivity towards breathing-type structural damage

This paper investigates the performance of a nonlinear damage detection method using sensitivity enhancing control (SEC). Damage nonlinearity due to the cyclic behavior of crack breathing could provide valuable evidence of structural damage without information of the structure’s original healthy condition. Not having such information is considered a major challenge in vibration-based damage detection. In this study, two different categories of damage detection methods are investigated: frequency and time-domain techniques focusing on the benefit of SEC for breathing-type nonlinear damage in a structure. Numerical simulations using a cantilevered beam and spring-mass-damper system demonstrated that the level of nonlinear dynamic behavior heavily depends on the closed-loop pole placement through feedback control. According to SEC theory, the characteristic of the feedback gain defines the sensitivity of modal frequency to the change of stiffness or mass of the system. The sensitivity enhancement by properly designed closedloop pole location more visually clarifies the evidence of crack nonlinearity than the open-loop case where no sensitivity is enhanced. A damage detection filter that uses time series data could directly benefit from implementing SEC. The amplitude of damage-evident error signal of the closed-loop case significantly increases more than that of the open-loop case if feedback control or SEC properly modifies the dynamics of the system.

Data Decomposition Techniques with Multi-Scale Permutation Entropy Calculations for Bearing Fault Diagnosis

This paper presents the local mean decomposition (LMD) integrated with multi-scale permutation entropy (MPE), also known as LMD-MPE, to investigate the rolling element bearing (REB) fault diagnosis from measured vibration signals. First, the LMD decomposed the vibration data or acceleration measurement into separate product functions that are composed of both amplitude and frequency modulation. MPE then calculated the statistical permutation entropy from the product functions to extract the nonlinear features to assess and classify the condition of the healthy and damaged REB system. The comparative experimental results of the conventional LMD-based multi-scale entropy and MPE were presented to verify the authenticity of the proposed technique. The study found that LMD-MPE’s integrated approach provides reliable, damage-sensitive features when analyzing the bearing condition. The results of REB experimental datasets show that the proposed approach yields more vigorous outcomes than existing methods.

Identification of structural damage using wavelet-based data classification

Predicted time-history responses from a finite-element (FE) model provide a baseline map where damage locations are clustered and classified by extracted damage-sensitive wavelet coefficients such as vertical energy threshold (VET) positions having large silhouette statistics. Likewise, the measured data from damaged structure are also decomposed and rearranged according to the most dominant positions of wavelet coefficients. Having projected the coefficients to the baseline map, the true localization of damage can be identified by investigating the level of closeness between the measurement and predictions. The statistical confidence of baseline map improves as the number of prediction cases increases. The simulation results of damage detection in a truss structure show that the approach proposed in this study can be successfully applied for locating structural damage even in the presence of a considerable amount of process and measurement noise.