Evaluation of Time and Frequency Condition Indicators from Vibration Signals for Crack Detection in Railway Axles
Abstract
:1. Introduction
2. Background
2.1. Condition Indicators
2.2. Random Forest for Feature Selection
Feature Selection by MDA
2.3. k-Nearest Neighbor Classifier
3. Experimental Set-Up
3.1. Bogie Test Bench
3.2. Signal Acquisition and Experimental Conditions
4. The Proposed Approachs
4.1. Proposed Approach 1
- Data acquisition: two sets of signals are acquired, from the left accelerometer (LL) along the longitudinal direction and from the right accelerometer (RL) along the same direction under different fault, load, and speed conditions, as explained in Section 3.2.
- Feature extraction: the feature extraction was applied as explained in Section 3.2. Figure 4 shows examples of vibration signals in time domain, and signals in the frequency domain using FFT and PSD from both the left and right accelerometers, for normal (healthy) condition and fault level 1. For each signal, different CIs were obtained according to three domains: (i) 30 CIs for time domain (CI_T), (ii) 24 CIs for frequency domain (CI_F), and (iii) 54 CIs from time + frequency domain (CI_TF). Therefore, six data sets were completed. Three data sets were obtained with the left-side accelerometer along the longitudinal direction (LL): the first CI_TL contains CIs extracted from the left side accelerometer and in time domain, the second CI_FL contains the CIs in frequency domain for the same accelerometer, and finally, the third CI_TFL has CIs in the time + frequency domain, also with the same accelerometer. Similarly, with the right-side accelerometer along the longitudinal direction (RL), we obtained other three data sets: CI_TR, CI_FR, and CI_TFR on time, frequency and time + frequency domain, respectively.
- Feature selection: this process starts by removing correlated features. Correlation coefficient value on 0.8–1 between two CIs suggests these two features are highly correlated [30]. In this study, a threshold value equal to or greater than 0.95 (95%) was selected to identify two highly-correlated features, in all data sets. This threshold was stated after obtaining good performance in classification. Next, a normalization process was applied by scaling characteristics between −1 and 1. Normalization can be applied to Normal distributed data, as well as to data with another type of distribution. This process helps KNN to give equal importance to all CIs. After the preprocessing step, a Random Forest model was implemented with 40 trees as the main parameter. MDA metric was used to select the ten most important CIs, and then, the selection of the ten best-ranked features from each dataset.
- Classification: each dataset is organized in a matrix of samples in the rows and CIs in the columns. Each pre-selected dataset (built from 10 ranked CIs) in stage 3 is classified using KNN (a value of k = 3 was chosen as the main parameter after testing different number of neighbors) and the cosine distance metric was used.Usually, when using k-folds cross-validation, values of k = 5 or k = 10 are chosen due to the good results obtained empirically [31]. In our work, the best results were obtained for k = 5. Therefore, a five-fold cross-validation strategy was carried out; the average accuracy and standard deviation () on the cross-validation process were calculated from 5 runs.Finally, many classifications were performed; these starts by using the first feature for a classification, next, the two first features are used for a new classification, and so on, until reaching the ten first features. The purpose is to analyze the contribution of each ranked feature.
4.2. Proposed Approach 2
- Data acquisition: the six vibration signals, three accelerometers mounted on the left side LV, LL, LA, and three accelerometers placed on the right side RV, RL, RA, were acquired under different faults, load and velocity conditions.
- Feature extraction: for each signal, different condition indicators were obtained according to two domains: (i) 30 CIs in time domain and (ii) 24 CIs in the frequency domain. All those condition indicators are combined, so, for each vibration signal, 54 CIs are obtained. Therefore, six data sets were used, three different sets of CIs per each accelerometer.
- Data-Fusion: from the six data sets of CIs obtained in the feature extraction stage, data fusion is performed at the indicator level. Three new sets of indicators are obtained: (i) The sets of CIs extracted from signals of the three accelerometers on the left side (LV, LA, LL) are fused, obtaining the first set of fused CIs called CI_DFL; (ii) the CIs of the three accelerometers on the right side (RV, RA, RL) are fused, obtaining a fused data set called CI_DFR; and finally (iii) the CIs of the six accelerometers are fused, obtaining the third set of fused CIs called CI_DFRL.
- Feature selection and classification: from the three data sets fused in stage 3, the next stages 4 (feature selection) and 5 (classification) were carried out by following the same procedures, experimental conditions, and CIs analysis of the stages Feature Selection and Classification of the approach 1 detailed in Section 4.1 and presented in Figure 3a.
5. Results and Discussion
5.1. Results of Approach 1
5.2. Results of Approach 2
5.3. Discussion
6. Conclusions
Author Contributions
Funding
Acknowledgments
Conflicts of Interest
References
- Zhang, X.; Zou, Z.; Wang, K.; Hao, Q.; Wang, Y.; Shen, Y.; Hu, H. A new rail crack detection method using LSTM network for actual application based on AE technology. Appl. Acoust. 2018, 142, 78–86. [Google Scholar] [CrossRef]
- Hassan, M.; Bruni, S.; Carboni, M. Crack detection in railway axle using horizontal and vertical vibration measurements. In Proceedings of the 7th IET Conference on Railway Condition Monitoring 2016 (RCM 2016), Birmingham, UK, 27–28 September 2016. [Google Scholar]
- Cavuto, A.; Martarelli, M.; Pandarese, G.; Revel, G.; Tomasini, E. Experimental investigation by laser ultrasonics for high speed train axle diagnostics. Ultrasonics 2015, 55, 48–57. [Google Scholar] [CrossRef] [PubMed]
- Cheng, Y.; Wang, Z.; Zhang, W. A novel condition-monitoring method for axle-box bearings of high-speed trains using temperature sensor signals. IEEE Sens. J. 2018, 19, 205–213. [Google Scholar] [CrossRef]
- Artagan, S.S.; Ciampoli, L.B.; D’Amico, F.; Calvi, A.; Tosti, F. Non-destructive assessment and health monitoring of railway infrastructures. Surv. Geophys. 2019, 41, 1–37. [Google Scholar] [CrossRef]
- Sabnavis, G.; Kirk, R.G.; Kasarda, M.; Quinn, D. Cracked shaft detection and diagnostics: A literature review. Shock Vib. Dig. 2004, 36, 287. [Google Scholar] [CrossRef]
- Gómez, M.J.; Corral, E.; Castejón, C.; García-Prada, J.C. Effective Crack Detection in Railway Axles Using Vibration Signals and WPT Energy. Sensors 2018, 18, 1603. [Google Scholar] [CrossRef] [Green Version]
- Li, Y.; Zhang, W.; Xiong, Q.; Lu, T.; Mei, G. A novel fault diagnosis model for bearing of railway vehicles using vibration signals based on symmetric alpha-stable distribution feature extraction. Shock Vib. 2016, 2016. [Google Scholar] [CrossRef] [Green Version]
- Sánchez, R.V.; Lucero, P.; Vasquez, R.; Cerrada, M.; Macancela, J.; Cabrera, D. Feature ranking for multi-fault diagnosis of rotating machinery by using random forest and KNN. J. Intell. Fuzzy Syst. 2018, 34, 3463–3473. [Google Scholar] [CrossRef]
- Gómez, M.; Castejón, C.; García-Prada, J. Review of recent advances in the application of the wavelet transform to diagnose cracked rotors. Algorithms 2016, 9, 19. [Google Scholar] [CrossRef]
- Wei, J.; Liu, C.; Ren, T.; Liu, H.; Zhou, W. Online condition monitoring of a rail fastening system on high-speed railways based on wavelet packet analysis. Sensors 2017, 17, 318. [Google Scholar] [CrossRef]
- Lucero, P.; Sánchez, R.V.; Macancela, J.C.; Cabrera, D.; Cerrada, M.; Li, C.; Alonso, H.R. Accelerometer Placement Comparison for Crack Detection in Railway Axles Using Vibration Signals and Machine Learning. In Proceedings of the 2019 Prognostics and System Health Management Conference (PHM-Paris), Paris, France, 2–5 May 2019; pp. 291–296. [Google Scholar]
- Trilla, A.; Gratacòs, P. Maintenance of bogie components through vibration inspection with intelligent wireless sensors: A case study on axle-boxes and wheel-sets using the empirical mode decomposition technique. J. Rail Rapid Trans. 2016, 230, 1408–1414. [Google Scholar] [CrossRef]
- Aravanis, T.C.I.; Sakellariou, J.S.; Fassois, S.D. Spectral analysis of railway vehicle vertical vibration under normal operating conditions. Int. J. Rail Transp. 2016, 4, 193–207. [Google Scholar] [CrossRef]
- Rolek, P.; Bruni, S.; Carboni, M. Condition monitoring of railway axles based on low frequency vibrations. Int. J. Fatigue 2016, 86, 88–97. [Google Scholar] [CrossRef]
- Liu, T.; Chen, J.; Dong, G. Zero crossing and coupled hidden Markov model for a rolling bearing performance degradation assessment. J. Vib. Control 2014, 20, 2487–2500. [Google Scholar] [CrossRef]
- Ninoslav, Z.F.; Rusmir, B.; Cvetkovic, D. Vibration feature extraction methods for gear faults diagnosis—A review. Facta Univ. Ser. Work. Living Environ. Prot. 2015, 12, 63–72. [Google Scholar]
- Cai, G.; Zhao, J.; Song, Q.; Zhou, M. System architecture of a train sensor network for automatic train safety monitoring. Comput. Ind. Eng. 2019, 127, 1183–1192. [Google Scholar] [CrossRef]
- Hua, Z.; Sun, Z.; Jia, H.; Zhao, Z.; Hou, L. Track Occupation Detection Based on a Maximum Posterior Probability Model Using Multisensor Data Fusion. Comput. Sci. Eng. 2018, 21, 40–54. [Google Scholar] [CrossRef]
- Li, H.; Xiao, D.Y. Fault diagnosis based on power spectral density basis transform. J. Vib. Control 2015, 21, 2416–2433. [Google Scholar] [CrossRef]
- Khaleghi, B.; Khamis, A.; Karray, F.O.; Razavi, S.N. Multisensor data fusion: A review of the state-of-the-art. Inf. Fusion 2013, 14, 28–44. [Google Scholar] [CrossRef]
- Mendel, J.M. Tutorial on higher-order statistics (spectra) in signal processing and system theory: Theoretical results and some applications. Proc. IEEE 1991, 79, 278–305. [Google Scholar] [CrossRef]
- Toledo, E.; Pinhas, I.; Aravot, D.; Akselrod, S. Bispectrum and bicoherence for the investigation of very high frequency peaks in heart rate variability. In Proceedings of the Computers in Cardiology 2001, Rotterdam, The Netherlands, 23–26 September 2001; pp. 667–670. [Google Scholar]
- Breiman, L. Random forests. Mach. Learn. 2001, 45, 5–32. [Google Scholar] [CrossRef] [Green Version]
- Biau, G.; Scornet, E. A random forest guided tour. Test 2016, 25, 197–227. [Google Scholar] [CrossRef] [Green Version]
- Duda, R.O.; Hart, P.E. Pattern Classification and Scene Analysis; Wiley: New York, NY, USA, 1973. [Google Scholar]
- Hu, L.Y.; Huang, M.W.; Ke, S.W.; Tsai, C.F. The distance function effect on k-nearest neighbor classification for medical datasets. SpringerPlus 2016, 5, 1304. [Google Scholar] [CrossRef] [PubMed] [Green Version]
- Prasath, V.; Alfeilat, H.A.A.; Lasassmeh, O.; Hassanat, A.; Tarawneh, A.S. Distance and Similarity Measures Effect on the Performance of K-Nearest Neighbor Classifier–A Review. arXiv 2017, arXiv:1708.04321. [Google Scholar]
- Cunningham, P.; Delany, S.J. k-Nearest neighbour classifiers. Mult. Classif. Syst. 2007, 34, 1–17. [Google Scholar]
- Swinscow, T.D.V.; Campbell, M.J. Statistics at Square One; BMJ: London, UK, 2002. [Google Scholar]
- Casella, G.; Fienberg, S.; Olkin, I. An Introduction to statistical learning with Applications in R. In Springer Texts in Statistics; Springer: New York, NY, USA, 2013. [Google Scholar]
Mean of spectrum | |
Variance of spectrum | |
Skewness of spectrum (Skewnessf) | |
Kurtosis of spectrum | |
Central Frequency | |
STD of spectrum | |
RMS of spectrum | |
CP1 | |
CP2 | |
CP3 | |
CP4 | |
CP5 | |
Centroid of Spectrum | |
Spectrum Spread | |
Entropy of spectrum | |
where is the normalized total spectral energy | |
Total power | |
Median Frequency (MDF) | |
Peak frequency (PKF) | |
First Spectral Moment | |
Second Spectral Moment | |
Third Spectral Moment | |
Fourth Spectral Moment (SM4) | |
Spectral moment ratio (VCF) | |
Frequency ratio (FR) |
3 loads | - 4, 10, y 16 tons |
2 speeds conditions | - 20 km/h - 50 km/h |
6 vibration measurement placements | - Left vertical (LV) accelerometer, left longitudinal (LL) accelerometer, left axial (LA) accelerometer - Right vertical (RV) accelerometer, right longitudinal (RL) accelerometer, right axial (RA) accelerometer |
Rotational directions | - Clockwise and counterclockwise |
4 fault conditions, see Figure 2 | - Healthy (Normal) - Crack level 1 (e = 5.7 mm) - Crack level 2 (e = 10.9 mm) - Crack level 3 (e = 15 mm) |
CI_TL | CL_RT | ||||||
---|---|---|---|---|---|---|---|
# | CIs | Accuracy (%) | Std | # | CIs | Accuracy (%) | Std |
1 | Zero crossing | 25.6 | 0.20 | 1 | SSC | 25.85 | 0.26 |
2 | Wave length | 60.45 | 3.26 | 2 | SRAV | 54.74 | 3.63 |
3 | Kurtosis | 76.63 | 2.04 | 3 | Wave length | 85.43 | 3.15 |
4 | WAMP | 85.07 | 1.29 | 4 | WAMP | 87.21 | 1.32 |
5 | SSC | 86.93 | 1.50 | 5 | Shape factor | 89.27 | 1.83 |
6 | CPT4 | 89.69 | 1.12 | 6 | Skewness | 90.43 | 1.85 |
7 | Skewness | 92.89 | 1.45 | 7 | Zero crossing | 93.65 | 2.46 |
8 | Shape factor | 92.89 | 1.18 | 8 | Kurtosis | 95.17 | 1.30 |
9 | CPT3 | 92 | 1.61 | 9 | Energy operator | 94.64 | 0.93 |
10 | Energy operator | 91.64 | 0.74 | 10 | CPT3 | 94.36 | 1.38 |
# | CIs | Classes (Accuracy%) | |||
---|---|---|---|---|---|
1 | 2 | 3 | 4 | ||
1 | Zero crossing | 100 | 0 | 0 | 0 |
2 | Wave length | 65.95 | 55.44 | 48.58 | 71.92 |
3 | Kurtosis | 86.47 | 74.04 | 71.58 | 74.06 |
4 | WAMP | 89.23 | 84.56 | 83.09 | 83.21 |
5 | SSC | 90.97 | 84.56 | 86.34 | 85.76 |
6 | CPT4 | 94.45 | 88.77 | 84.55 | 90.9 |
7 | Skewness | 97.57 | 92.63 | 88.84 | 92.32 |
8 | Shape factor | 98.25 | 94.04 | 87.75 | 91.23 |
9 | CPT3 | 97.57 | 93.68 | 88.15 | 88.32 |
10 | Energy operator | 98.26 | 93.68 | 86.33 | 87.95 |
CI_FL | CI_FR | ||||||
---|---|---|---|---|---|---|---|
# | CIs | Accuracy (%) | Std | # | CIs | Accuracy (%) | Std |
1 | PKF | 25.6 | 0.15 | 1 | Skewnessf | 25.85 | 0.17 |
2 | CP2 | 56.8 | 2.93 | 2 | Spectrum spread | 41.77 | 2.44 |
3 | SM4 | 85.51 | 1.35 | 3 | PKF | 71.38 | 2.65 |
4 | FR | 72 | 2.54 | 4 | CP2 | 82.75 | 3.13 |
5 | Skewnessf | 79.03 | 2.57 | 5 | MDF | 83.99 | 2.48 |
6 | Spectrum spread | 87.21 | 2.43 | 6 | FR | 89.36 | 1.58 |
7 | VCF | 91.29 | 2.92 | 7 | CP3 | 94.01 | 1.16 |
8 | Centroid of Spectrum | 90.4 | 2.82 | 8 | SM1 | 93.47 | 1.32 |
9 | Kurtosisf | 90.13 | 1.08 | 9 | VCF | 94.36 | 0.93 |
10 | Entropy of spectrum | 90.23 | 1.97 | 10 | Centroid of Spectrum | 94.27 | 1.70 |
# | CIs | Classes (Accuracy%) | |||
---|---|---|---|---|---|
1 | 2 | 3 | 4 | ||
1 | PKF | 100 | 0 | 0 | 0 |
2 | CP2 | 53.8 | 65.61 | 56.14 | 51.47 |
3 | SM4 | 78.14 | 85.96 | 84.53 | 93.8 |
4 | FR | 71.86 | 68.77 | 68.33 | 79.16 |
5 | Skewnessf | 81.98 | 77.89 | 74.79 | 81.41 |
6 | Spectrum spread | 85.4 | 83.86 | 84.18 | 95.62 |
7 | VCF | 89.24 | 89.82 | 87.77 | 98.54 |
8 | Centroid of Spectrum | 87.86 | 89.82 | 86.32 | 97.81 |
9 | Kurtosisf | 87.15 | 90.88 | 84.86 | 97.81 |
10 | Entropy of spectrum | 87.85 | 89.47 | 85.63 | 98.17 |
CI_TFL | CI_TFR | ||||||||
---|---|---|---|---|---|---|---|---|---|
# | CIs | Domain | Accuracy (%) | Std | # | CIs | Domain | Accuracy (%) | Std |
1 | Zero crossing | Time | 25.6 | 0.15 | 1 | SSC | Time | 25.85 | 0.17 |
2 | PKF | Freq | 52.8 | 1.33 | 2 | FR | Freq | 50.81 | 1.77 |
3 | WAMP | Time | 73.43 | 3.37 | 3 | SM1 | Freq | 71.56 | 1.86 |
4 | CP2 | Freq | 78.85 | 1.91 | 4 | Skewnessf | Freq | 88.56 | 1.88 |
5 | SSC | Time | 82.49 | 1.89 | 5 | Spectrum spread | Freq | 93.83 | 1.67 |
6 | Shape factor | Time | 92.62 | 1.70 | 6 | Zero crossing | Time | 95.26 | 1.19 |
7 | Skewness | Time | 93.69 | 2.08 | 7 | SM4 | Freq | 94.63 | 1.15 |
8 | Kurtosis | Time | 94.49 | 0.89 | 8 | CP3 | Freq | 96.87 | 0.55 |
9 | SM4 | Freq | 94.4 | 2.35 | 9 | CP2 | Freq | 96.6 | 2.46 |
10 | FR | Freq | 96.62 | 0.59 | 10 | PKF | Freq | 97.14 | 1.29 |
# | CIs | Domain | Classes (Accuracy%) | |||
---|---|---|---|---|---|---|
1 | 2 | 3 | 4 | |||
1 | Zero crossing | Time | 100 | 0 | 0 | 0 |
2 | PKF | Freq | 63.91 | 43.51 | 41.36 | 62.4 |
3 | WAMP | Time | 79.89 | 64.21 | 61.16 | 88.69 |
4 | CP2 | Freq | 77.11 | 75.09 | 72.64 | 90.88 |
5 | SSC | Time | 84.06 | 79.3 | 76.63 | 90.15 |
6 | Shape factor | Time | 93.04 | 89.47 | 92.79 | 95.25 |
7 | Skewness | Time | 96.19 | 90.88 | 96.77 | 90.86 |
8 | Kurtosis | Time | 96.88 | 92.28 | 96.39 | 92.32 |
9 | SM4 | Freq | 96.88 | 91.58 | 97.48 | 91.59 |
10 | FR | Freq | 97.21 | 95.09 | 96.4 | 97.81 |
CI_DFL | CI_DFR | ||||||||
---|---|---|---|---|---|---|---|---|---|
# | CIs - Domain | Sensor | Acc(%) | Std | # | CIs - Domain | Sensor | Acc(%) | Std |
1 | CPT6 - Time | LL | 25.5 | 0.13 | 1 | SSC - Time | RA | 25.5 | 0.17 |
2 | PKF - Freq | LA | 62.21 | 3.45 | 2 | Skewnessf - Freq | RV | 46.11 | 1.45 |
3 | FR -Freq | LV | 84.72 | 1.72 | 3 | SSC - Time | RL | 69.43 | 2.58 |
4 | MDF - Freq | LV | 93.67 | 0.85 | 4 | Skewnessf - Freq | RL | 87.16 | 1.2 |
5 | PKF - Freq | LV | 94.75 | 0.69 | 5 | CP2 - Freq | RL | 93.67 | 1.85 |
6 | Kurtosis - Time | LL | 96.38 | 1.4 | 6 | PKF - Freq | RV | 93.76 | 1.55 |
7 | Skewnessf - Freq | LV | 96.47 | 1.26 | 7 | ShapeFactor - TIime | RA | 96.48 | 1.78 |
8 | SSC - Time | LV | 96.56 | 1.59 | 8 | MDF - Freq | RL | 96.65 | 1.22 |
9 | SM4 - Freq | LA | 97.2 | 0.92 | 9 | VCF - Freq | RL | 97.92 | 1.05 |
10 | VCF - Freq | LV | 97.56 | 0.88 | 10 | ZC - Time | RL | 98.01 | 0.88 |
CI_DFRL | |||||
---|---|---|---|---|---|
# | CIs | Domain | Sensor | Accuracy (%) | Std |
1 | MDF | Freq | LV | 25.5 | 0.19 |
2 | VCF | Freq | RL | 63.47 | 3.96 |
3 | CP2 | Freq | RA | 88.24 | 1.12 |
4 | SSC | Time | LV | 92.68 | 1.67 |
5 | CP2 | Freq | LL | 94.21 | 1.87 |
6 | PKF | Freq | LV | 96.38 | 1.24 |
7 | PKF | Freq | RL | 97.02 | 0.82 |
8 | CP2 | Freq | RL | 97.2 | 1.14 |
9 | ZC | Time | LL | 97.38 | 0.38 |
10 | Skewnessf | Freq | LV | 98.37 | 0.76 |
# | CIs | Domain | Sensor | Classes (Accuracy%) | |||
---|---|---|---|---|---|---|---|
1 | 2 | 3 | 4 | ||||
1 | MDF | Freq | LV | 100 | 0 | 0 | 0 |
2 | VCF | Freq | RL | 70.94 | 46.94 | 25.82 | 39.88 |
3 | CP2 | Freq | RA | 84.4 | 64.75 | 66.18 | 61.97 |
4 | SSC | Time | LV | 90.76 | 84.72 | 79.64 | 93.65 |
5 | CP2 | Freq | LL | 96.46 | 94.29 | 85.82 | 98.13 |
6 | PKF | Freq | LV | 96.8 | 93.93 | 86.55 | 97.77 |
7 | PKF | Freq | RL | 97.53 | 96.44 | 93.09 | 98.88 |
8 | CP2 | Freq | RL | 96.83 | 96.09 | 94.18 | 99.62 |
9 | ZC | Time | LL | 97.88 | 97.51 | 96.36 | 100 |
10 | Skewnessf | Freq | LV | 98.58 | 97.86 | 96.73 | 98.88 |
© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
Share and Cite
Sánchez, R.-V.; Lucero, P.; Macancela, J.-C.; Rubio Alonso, H.; Cerrada, M.; Cabrera, D.; Castejón, C. Evaluation of Time and Frequency Condition Indicators from Vibration Signals for Crack Detection in Railway Axles. Appl. Sci. 2020, 10, 4367. https://doi.org/10.3390/app10124367
Sánchez R-V, Lucero P, Macancela J-C, Rubio Alonso H, Cerrada M, Cabrera D, Castejón C. Evaluation of Time and Frequency Condition Indicators from Vibration Signals for Crack Detection in Railway Axles. Applied Sciences. 2020; 10(12):4367. https://doi.org/10.3390/app10124367
Chicago/Turabian StyleSánchez, Réne-Vinicio, Pablo Lucero, Jean-Carlo Macancela, Higinio Rubio Alonso, Mariela Cerrada, Diego Cabrera, and Cristina Castejón. 2020. "Evaluation of Time and Frequency Condition Indicators from Vibration Signals for Crack Detection in Railway Axles" Applied Sciences 10, no. 12: 4367. https://doi.org/10.3390/app10124367
APA StyleSánchez, R.-V., Lucero, P., Macancela, J.-C., Rubio Alonso, H., Cerrada, M., Cabrera, D., & Castejón, C. (2020). Evaluation of Time and Frequency Condition Indicators from Vibration Signals for Crack Detection in Railway Axles. Applied Sciences, 10(12), 4367. https://doi.org/10.3390/app10124367