|
|
Analysis of an industrial process
Task: regression
Number of instances: 60
Number of attributes: 7 (6 numerical + 1 discrete)
Type of attribute to be predicted: numerical
Download the data: DataProcess
By analysing data from production, the objective is to determine the really influential parameters, to understand their interactions and thus to optimise them. The explanatory variables are 3 pressures Pa, Pb and Pc (in MPa), 3 temperatures Ta, Tb and Tc (in °C) and the type of machine. We want to determine the influence of these parameters on the quality of the end product which is defined by a note between 0 (very bad) in 4
(excellent).
Sources: BLIASOLUTIONS
Model with 1 variable
* If (Pa is lower than 1,7) then (Quality decreases)
|
Mean Error: 0,49
|
Mean Square Error: 0,72
|
Maximal Error: 1,97
|
|
Maximum : 3 (Pa is higher than 1,7 MPa)
|
|
Minimum : 0 (Pa is lower than 1,57 MPa)
|
This model that directly connects the quality of the parts to pressure Pa is the most precise with only one variable (in other words, pressure Pa is the most influential variable). Beyond 1,7 MPa the quality of the parts is maximal. The following graph represents the model (in red) and the experimental values (points in green):
Model with 2 variables
* If (Pa is higher than 1,55) and (Machine is rather not M3) then (Quality increases)
|
Mean Error: 0,37
|
Mean Square Error: 0,58
|
Maximal Error: 1,87
|
|
Maximum : 4 (Pa is higher than 1,8 MPa and Machine is different from M3)
|
|
Minimum : 0 (Pa is lower than 1,53 MPa and Machine is M3)
|
In this model, in addition to pressure Pa, the quality of the parts is linked to the type of machine: using the machine M3 degrades the quality of the parts. This model enables to obtain all the possible values of quality (from 0 to 4). The following curves represent the model with the machines M1 and M2 (in red) and with the machine M3 (in green):
Model with 3 variables
* If (Tb is close to 80) and (Pa is higher than 1,5) then (Quality decreases)
* If (Pa is higher than 1,5) and (Machine is rather not M3) then (Quality increases)
|
Mean Error: 0,16
|
Mean Square Error: 0,23
|
Maximal Error: 0,65
|
|
Maximum : 4 (Pa is higher than 1,81 MPa, Machine is different from M3 and Tb is lower than 73,1°C or higher than 86,4°C)
|
|
Minimum : 0 (Pa is lower than 1,5 MPa, Machine is M3 and Tb is equal to 79,7°C)
|
In addition to pressure Pa and the type of machine, this model adds the temperature Tb which decreases the quality of the parts when it is close to 80°C (this phenomenon is very localised). The following graph represents the model for the machine M1:
|
|
