Dataset Delft pump 3x2

Basic characteristics Delft pump 3x2

137

target objects

The Delft pump dataset, with some description in A. Ypma, Learning methods for machine vibration analysis and health monitoring, thesis Delft university of Technology, 2001. From a submersible pump 5 vibration measurements are taken under different normal and abnormal conditions. From the time signals an 64D envelope spectrum is derived. The 5 measurements are used as independent objects. Some loads with some speeds situations are used in the target class. Download mat-file with Prtools dataset.

463

outlier objects

64

features

Unsupervised PCA Delft pump 3x2

On the left, the PCA scatterplot is shown, on the right the retained variance for varying number of features.
On the left, the PCA scatterplot is shown of data rescaled to unit variance, on the right the retained variance.

Supervised Fisher Delft pump 3x2

On the left, the Fisher scatterplot is shown, on the right the ROC curve along this direction.

Results Delft pump 3x2

The experiments are performed using dd_tools. A rudimentary explanation of the classifiers is given in the classifier section.

544, 0 outliers, AUC (x100) 5x strat. 10-fold
Classifiers Preproc
none unit var PCA 95\%
Gauss 91.1 (0.0) 90.7 (0.0) 57.2 (0.0)
Min.Cov.Determinant NaN (0.0) NaN (0.0) 61.3 (0.0)
Mixture of Gaussians 87.4 (2.8) 86.3 (0.7) 68.8 (3.3)
Naive Parzen 75.2 (0.0) 75.2 (0.0) 67.0 (0.0)
Parzen 81.8 (0.0) 78.1 (0.0) 75.7 (0.0)
k-means 71.4 (7.1) 72.8 (0.4) 56.2 (6.7)
1-Nearest Neighbors 81.8 (0.0) 78.1 (0.0) 75.7 (0.0)
k-Nearest Neighbors 81.8 (0.0) 78.1 (0.0) 75.7 (0.0)
Nearest-neighbor dist 81.0 (0.0) 82.4 (0.0) 73.9 (0.0)
Principal comp. 87.0 (0.0) 87.8 (0.0) 58.3 (0.0)
Self-Organ. Map 83.2 (1.0) 78.9 (0.4) 70.9 (2.0)
Auto-enc network 82.9 (4.1) 74.6 (0.0) 58.3 (0.0)
MST 85.5 (0.0) 79.6 (0.0) 79.7 (0.0)
L_1-ball 53.7 (0.0) 66.4 (0.0) 60.9 (0.0)
k-center 79.2 (1.2) 74.5 (0.9) 69.6 (0.8)
Support vector DD 81.1 (0.3) 58.9 (0.6) 75.8 (0.6)
Minimax Prob. DD 81.9 (0.0) 78.1 (0.0) 76.2 (0.0)
LinProg DD 83.7 (0.0) 74.2 (0.0) 74.9 (0.0)
Lof DD 81.1 (0.0) 87.1 (0.0) 75.1 (0.0)
Lof range DD 78.9 (0.0) 86.6 (0.0) 72.3 (0.0)
Loci DD 78.1 (0.0) 74.5 (0.0) 65.6 (0.0)

Classifier projection spaces The first classifier projection spaces are obtained by computing the classifier label disagreements (setting the threshold on 10% target error) and applying an MDS on the resulting distance matrix between classifiers:



Original



Unit variance



PCA mapped

Classifier projection spaces The second versions of the classifier projection spaces are obtained by computing the classifier ranking disagreements and applying an MDS on the resulting distance matrix between classifiers:



Original



Unit variance



PCA mapped