Kojo Sarfo Gyamfi
Coventry University, UK
gyamfik '@' uni.coventry.ac.uk

Craig Marshall
National Engineering Laboratory, TUV-NEL, UK
Craig.Marsall '@' tuv-sud.co.uk

Data Set Information:

Meter A contains 87 instances of diagnostic parameters for an 8-path liquid ultrasonic flow meter (USM). It has 37 attributes and 2 classes or health states:
Class '1' - Healthy
Class '2' - Installation effects

Meter B contains 92 instances of diagnostic parameters for a 4-path liquid USM. It has 52 attributes and 3 classes:
Class '1' - Healthy
Class '2' - Gas injection
Class '3' - Waxing

Meter C contains 181 instances of diagnostic parameters for a 4-path liquid USM. It has 44 attributes and 4 classes:
Class '1' - Healthy
Class '2' - Gas injection
Class '3' - Installation effects
Class '4' - Waxing

Meter D contains 180 instances of diagnostic parameters for a 4-path liquid USM. It has 44 attributes and 4 classes:
Class '1' - Healthy
Class '2' - Gas injection
Class '3' - Installation effects
Class '4' - Waxing

Attribute Information:

All attributes are continuous, with the exception of the class attribute.

Meter A
(1)        -- Flatness ratio
(2)        -- Symmetry
(3)        -- Crossflow
(4)-(11)   -- Flow velocity in each of the eight paths
(12)-(19)  -- Speed of sound in each of the eight paths
(20)       -- Average speed of sound in all eight paths
(21)-(36)  -- Gain at both ends of each of the eight paths
(37)       -- Class attribute or health state of meter: 1,2

Meter B
(1)       -- Profile factor
(2)       -- Symmetry
(3)       -- Crossflow
(4)       -- Swirl angle
(5)-(8)   -- Flow velocity in each of the four paths
(9)       -- Average flow velocity in all four paths
(10)-(13)  -- Speed of sound in each of the four paths
(14)       -- Average speed of sound in all four paths
(15)-(22)  -- Signal strength at both ends of each of the four paths
(23)-(26)  -- Turbulence in each of the four paths
(27)      -- Meter performance
(28)-(35) -- Signal quality at both ends of each of the four paths
(36)-(43) -- Gain at both ends of each of the four paths
(44)-51   -- Transit time at both ends of each of the four paths
(52)      -- Class attribute or health state of meter: 1,2,3

Meters C and D
(1)       -- Profile factor
(2)       -- Symmetry
(3)       -- Crossflow
(4)-(7)   -- Flow velocity in each of the four paths
(8)-(11)  -- Speed of sound in each of the four paths
(12)-(19) -- Signal strength at both ends of each of the four paths
(20)-(27) -- Signal quality at both ends of each of the four paths
(28)-(35) -- Gain at both ends of each of the four paths
(36)-(43) -- Transit time at both ends of each of the four paths
(44)      -- Class attribute or health state of meter: 1,2,3,4

Relevant Papers:

K. S. Gyamfi, J. Brusey, A. Hunt, E. Gaura , a€?Linear dimensionality reduction for classification via a sequential Bayes error minimisation with an application to flow meter diagnostics,a€? Expert Systems with Applications (IF: 3.928), September 2017

Citation Request:

K. S. Gyamfi, J. Brusey, A. Hunt, E. Gaura , a€?Linear dimensionality reduction for classification via a sequential Bayes error minimisation with an application to flow meter diagnostics,a€? Expert Systems with Applications (IF: 3.928), September 2017