Limits...
Unsupervised classification of surface defects in wire rod production obtained by eddy current sensors.

Saludes-Rodil S, Baeyens E, Rodríguez-Juan CP - Sensors (Basel) (2015)

Bottom Line: An unsupervised approach to classify surface defects in wire rod manufacturing is developed in this paper.The defects are extracted from an eddy current signal and classified using a clustering technique that uses the dynamic time warping distance as the dissimilarity measure.It is shown that it outperforms other classification alternatives, such as the modified Fourier descriptors.

View Article: PubMed Central - PubMed

Affiliation: Centro Tecnológico CARTIF, Parque Tecnológico de Boecillo 205, 47151 Boecillo, Valladolid, Spain. sersal@cartif.es.

ABSTRACT
An unsupervised approach to classify surface defects in wire rod manufacturing is developed in this paper. The defects are extracted from an eddy current signal and classified using a clustering technique that uses the dynamic time warping distance as the dissimilarity measure. The new approach has been successfully tested using industrial data. It is shown that it outperforms other classification alternatives, such as the modified Fourier descriptors.

No MeSH data available.


Related in: MedlinePlus

Confusion matrix for MFD and spectral clustering.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4482004&req=5

f4-sensors-15-10100: Confusion matrix for MFD and spectral clustering.

Mentions: The adjacency matrix used as the starting point for spectral clustering has been computed over a k nearest neighbors (k-NN) similarity graph [22] with k = 15. The algorithm used is the normalized version, and the confusion matrix is presented in Figure 4. The silhouette index [19] is = 0.321. The silhouette index is a quantitative method of evaluating the results of a clustering process. It was proposed by Russeeuw in [23]. The confusion matrix shows that this method is not capable of discriminating defects in Classes C and D. Moreover, Class B is split into two different clusters, and one of the sequences is mixed with Class A, which is rightly assigned to a cluster.


Unsupervised classification of surface defects in wire rod production obtained by eddy current sensors.

Saludes-Rodil S, Baeyens E, Rodríguez-Juan CP - Sensors (Basel) (2015)

Confusion matrix for MFD and spectral clustering.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4482004&req=5

f4-sensors-15-10100: Confusion matrix for MFD and spectral clustering.
Mentions: The adjacency matrix used as the starting point for spectral clustering has been computed over a k nearest neighbors (k-NN) similarity graph [22] with k = 15. The algorithm used is the normalized version, and the confusion matrix is presented in Figure 4. The silhouette index [19] is = 0.321. The silhouette index is a quantitative method of evaluating the results of a clustering process. It was proposed by Russeeuw in [23]. The confusion matrix shows that this method is not capable of discriminating defects in Classes C and D. Moreover, Class B is split into two different clusters, and one of the sequences is mixed with Class A, which is rightly assigned to a cluster.

Bottom Line: An unsupervised approach to classify surface defects in wire rod manufacturing is developed in this paper.The defects are extracted from an eddy current signal and classified using a clustering technique that uses the dynamic time warping distance as the dissimilarity measure.It is shown that it outperforms other classification alternatives, such as the modified Fourier descriptors.

View Article: PubMed Central - PubMed

Affiliation: Centro Tecnológico CARTIF, Parque Tecnológico de Boecillo 205, 47151 Boecillo, Valladolid, Spain. sersal@cartif.es.

ABSTRACT
An unsupervised approach to classify surface defects in wire rod manufacturing is developed in this paper. The defects are extracted from an eddy current signal and classified using a clustering technique that uses the dynamic time warping distance as the dissimilarity measure. The new approach has been successfully tested using industrial data. It is shown that it outperforms other classification alternatives, such as the modified Fourier descriptors.

No MeSH data available.


Related in: MedlinePlus