Table 1 Descriptive statistics of the utilized features for detecting pipeline failures.

From: Semi-supervised learning framework for oil and gas pipeline failure detection

Feature

Mean

Median

Minimum

Maximum

Standard Deviation

Unintentional release (MCF)

24,318.20

4,000.00

0.00

585,457.00

57,576.27

Intentional release (MCF)

4,354.93

205.00

0.00

89,152.00

10,084.19

Cover depth (ft)

51.00

42.00

0.00

336.00

48.89

Pipe diameter (in)

17.40

16.00

2.00

42.00

9.20

Pipe thickness (in)

0.28

0.25

0.13

1.00

0.09

Pipe strength (psi)

45,839.39

46,000.00

29.00

80,000.00

11,802.71

Potential impact radius (ft)

352.18

335.00

23.00

1,155.00

209.27

Operational pressure (psi)

656.91

690.50

1.50

2,689.00

317.75

Max allowable pressure (psi)

814.85

810.00

120.00

3,255.00

358.50

Potential isolated segment (ft)

59,425.05

48,014.50

1.00

801,150.00

79,836.08

Pipe age (years)

52.70

54.00

0.00

96.00

17.66

Impact area (in2)

3,273.60

1.67

0.00

114,022.50

12,195.38

Prompted shutdown (1:yes, 2:no)

1.91

2.00

1.00

2.00

0.29

Ignition (1:yes, 2:no)

1.15

1.00

1.00

2.00

0.36

Crossing (1:yes, 2:no)

1.12

1.00

1.00

2.00

0.32

Pipe coating type (multiple)

Location class (multiple)

Previous inspection (1:yes, 2:no)

1.56

2.00

1.00

2.00

0.50

SCADA anomaly flag (1:yes, 2:no)

1.20

1.00

1.00

2.00

0.40

  1. Observations with partially available information are included in the dataset used to generate the statistics. The training and testing datasets are normalized before proceeding to the proposed CIC framework.