Singular Spectrum Analysis for Bearing Defect Detection

Kilundu, Bovic; Chiementin, Xavier; Dehombreux, Pierre

doi:10.1115/1.4003938

Cited by 45 publications

(30 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recently, it was used for the purposes of engineering application such as fault diagnosis of rolling element bearings [15][16][17][18][19], tool wear health monitoring [20,21] and delamination in composite materials [22].…”

Section: Methodsmentioning

confidence: 99%

Vibration-based damage detection of structural joints in presence of uncertainty

2018

View full text Add to dashboard Cite

Abstract. Early damage detection of structure's joints is essential in order to ensure the integrity of structures. Vibration-based methods are the most popular way of diagnosing damage in machinery joints. Any technique that is used for such a purpose requires dealing with the variability inherent to the system due to manufacturing tolerances, environmental conditions or aging. The level of variability in vibrational response can be very high for mass-produced complex structures that possess a large number of components. In this study, a simple and efficient time frequency method is proposed for detection of damage in connecting joints. The method suggests using singular spectrum analysis for building a reference space from the signals measured on a healthy structure and then compares all other signals to that reference space in order to detect the presence of faults. A model of two plates connected by a series of mounts is used to examine the effectiveness of the method where the uncertainty in the mount properties is taken into account to model the variability in the built-up structure. The motivation behind the simplified model is to identify the faulty mounts in trim-structure joints of an automotive vehicle where a large number of simple plastic clips are used to connect the trims to the vehicle structure.

show abstract

Section: Methodsmentioning

confidence: 99%

Vibration-based damage detection of structural joints in presence of uncertainty

2018

View full text Add to dashboard Cite

show abstract

“…Difference Histogram was used for feature extraction in bearing time series data [6]. Singular Spectrum Analysis (SSA) is introduced for inner race bearing fault detection [7] and multilevel SSA [8] is used for bearing degradation performance assessment. Recently, methods are introduced using features from zero crossing intervals [9] and Euclidean distance [10].…”

Section: Introductionmentioning

confidence: 99%

Symbolic dynamics based bearing fault detection

Sanjith

Sujatha

Jayakumar

2012

2012 IEEE 5th India International Conference on Power Electronics (IICPE)

View full text Add to dashboard Cite

Most of the time domain methods for bearing condition monitoring are machine and load dependent or involves complex mathematical calculations or need of training the algorithm. To overcome the above issues, symbolic dynamics based method is proposed. The time series vibration data is converted into symbolic series from which dictionary of the signal is constructed. Common Signal Index (CSI) a parameter is computed based on dictionary constructed from the reference signal and the test signal. Deviations of the computed CSI value from the CSI value of the healthy state serve as an indicator for the presence of bearing fault. No-load healthy vibration data is used as a reference signal to detect the bearing in healthy or faulty condition. The algorithm is tested with the experimental data obtained for different bearing fault of various sizes and at varying loads. Comparisons of the proposed method with existing time-domain and data based methods are made.

show abstract

“…In this study, only the first three PCs are considered in forming the FVs. The variance proportion accounted for in the first three PCs is at least 75% of the variance of the original signal, which means that the selection of the threshold percentage for PCs meets with one of the guiding criteria mentioned in [14] . Figure 10 shows a 3D visualisation of the 30 feature vectors corresponding to the baseline training sample, which are used to form the baseline features.…”

Section: Case Studymentioning

confidence: 99%

“…For the selection of PCs, there are several guiding criteria mentioned in the literature [14] . One of these criteria is that the percentage of variance portion described in these PCs should be equal to or greater than 75% of the variance of the original signal.…”

Section: Singular Spectrum Analysis | Featurementioning

confidence: 99%

“…From the trend component of each bearing signal a number of features, such as peak value and standard deviation, are used to form the feature vectors and then presented as input to a neural network classifier. In [14] , SSA is used for multi-decomposition analysis. The number of singular values that cumulatively preserve a predetermined variance percentage is used as the indicator for fault presence.…”

Section: Singular Spectrum Analysis | Featurementioning

confidence: 99%

See 1 more Smart Citation