Automatic feature extraction and selection for condition monitoring and related datasets

Schneider, Tizian; Helwig, Nikolai; Schütze, Andreas

doi:10.1109/i2mtc.2018.8409763

Cited by 20 publications

(18 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The features that are traditionally used for CM (RMS, crest factor, kurtosis, …, in the time domain; amplitude of power spectra, band power, envelope, …, in the frequency domain) [4,12,13,14,19,20,21,22,23,24,25], and that are considered in this work, are useful in most applications to maintain the relevant information about the process or tool conditions [4].…”

Section: Methodsmentioning

confidence: 99%

“…In [24,25], authors provide a useful analysis of the main techniques for feature extraction and selection, for CM applications of industrial interest. They propose a set of algorithms which automatically extract and select features, independently of both the type of the specific applications of interest and the classifier used, as a basis for the tasks that the maintenance-on-condition requires.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Physical and Metrological Approach for Feature’s Definition and Selection in Condition Monitoring

D’Emilia

Gaspari

Natale

2019

Sensors

View full text Add to dashboard Cite

In this paper, a methodology is described aiming at emphasizing physical and metrological criteria in feature selection for condition monitoring of a real scale mechatronic system. The device is used for packaging applications according to the movements of its end effector, driven by a couple of brushless servomotors and a kinematic mechanical linkage. The approach is hybrid, meaning that the starting feature set is built with reference to both experimental data from different sensors and to the indication of a simplified kinematic and dynamic model of the mechanical linkage itself. A critical comparison and mixing of theoretical and experimental data, based also on a physical interpretation of differences, suggests some more features, with respect to the classical ones, of hybrid type, which could be mostly correlated to the effects of statuses and defects of the system to be identified. The whole procedure is step by step validated, in order to evaluate the variability of features, throughout the whole procedure. The variability is analyzed depending on the actions that are realized in order to define, select, and use the proposed features for data processing by advanced algorithms, like the most typically used classifiers and artificial neural networks. A comparison with the state-of-the-art automatic feature’s selection procedure is also presented. Experimental results show that the proposed methodology is able to classify with high accuracy many statuses of the mechatronic system, which are only slightly different as for set-up settings and/or mechanical wear and lubrication conditions of mechanical parts of the mechatronic system. Issues to be pursued to a more effective generalization of the method are also discussed.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Physical and Metrological Approach for Feature’s Definition and Selection in Condition Monitoring

D’Emilia

Gaspari

Natale

2019

Sensors

View full text Add to dashboard Cite

show abstract

“…These lines of research, also known as dimensionality reduction, are based on the following assumption: depending on the nature of the time series, its features are usually correlated and redundant. Usually, approaches in this line are categorized into two groups [6], [18], [19]: feature selection and feature extraction. The former focuses on selecting the most significant features from the original data set by applying different techniques, such as univariate statistical tests, variance thresholds, or Principal Component Analysis (PCA) [20], [21], [22], [23].…”

Section: A Time Series Dimensionality Reductionmentioning

confidence: 99%

KPIs-Based Clustering and Visualization of HPC Jobs: A Feature Reduction Approach

2021

View full text Add to dashboard Cite

High-Performance Computing (HPC) systems need to be constantly monitored to ensure their stability. The monitoring systems collect a tremendous amount of data about different parameters or Key Performance Indicators (KPIs), such as resource usage, IO waiting time, etc. A proper analysis of this data, usually stored as time series, can provide insight in choosing the right management strategies as well as the early detection of issues. In this paper, we introduce a methodology to cluster HPC jobs according to their KPI indicators. Our approach reduces the inherent high dimensionality of the collected data by applying two techniques to the time series: literature-based and variance-based feature extraction. We also define a procedure to visualize the obtained clusters by combining the two previous approaches and the Principal Component Analysis (PCA). Finally, we have validated our contributions on a real data set to conclude that those KPIs related to CPU usage provide the best cohesion and separation for clustering analysis and the good results of our visualization methodology. INDEX TERMS Clustering, feature extraction, high-performance computing, time series analysis.

show abstract

“…Some other methods can also be applied to dedicated applications and obtain good results. Schneider et al [22] proposed an automatic extraction and selection method of highly relevant features, the method was tested on eight datasets and obtained a general accuracy over 90%. Rezaie et al [23] proposed a feedback controller framework to adapt the sampling rate for better efficiency and higher accuracy.…”

Section: Related Workmentioning

confidence: 99%

Lower Limb Locomotion Activity Recognition of Healthy Individuals Using Semi-Markov Model and Single Wearable Inertial Sensor

Derrode

Pieczynski

2019

Sensors

View full text Add to dashboard Cite

Lower limb locomotion activity is of great interest in the field of human activity recognition. In this work, a triplet semi-Markov model-based method is proposed to recognize the locomotion activities of healthy individuals when lower limbs move periodically. In the proposed algorithm, the gait phases (or leg phases) are introduced into the hidden states, and Gaussian mixture density is introduced to represent the complex conditioned observation density. The introduced sojourn state forms the semi-Markov structure, which naturally replicates the real transition of activity and gait during motion. Then, batch mode and on-line Expectation-Maximization (EM) algorithms are proposed, respectively, for model training and adaptive on-line recognition. The algorithm is tested on two datasets collected from wearable inertial sensors. The batch mode recognition accuracy reaches up to 95.16%, whereas the adaptive on-line recognition gradually obtains high accuracy after the time required for model updating. Experimental results show an improvement in performance compared to the other competitive algorithms.

show abstract

Automatic feature extraction and selection for condition monitoring and related datasets

Cited by 20 publications

References 14 publications

Physical and Metrological Approach for Feature’s Definition and Selection in Condition Monitoring

Physical and Metrological Approach for Feature’s Definition and Selection in Condition Monitoring

KPIs-Based Clustering and Visualization of HPC Jobs: A Feature Reduction Approach

Lower Limb Locomotion Activity Recognition of Healthy Individuals Using Semi-Markov Model and Single Wearable Inertial Sensor

Contact Info

Product

Resources

About