DESIGN OF EXPERIMENT IN OIL & GAS
Oil and gas industry, unlike others, is a very interesting and challenging when it comes to DOE .
In oil and gas the noise factor is very large and to find the optimum setting for the controllable factors is not an easy task. Most of the time, the type of data that we deal in the O&G industry, especially when the project is related to operation area, is a continuous data for both inputs and the outputs.
During the analyze phase, if no data discretion is taking place then the tool that will definitely be in used are:
(1) Regression or
(2) The combination of Regression and Stepwise Regression
Both of this tools have inherent similar purpose: i.e. To find the significant Xs from all the potential Xs that are listed down from team members brainstorming session or the PFD (Process Flow Diagram ) itself .
The method of selecting the significant Xs from either of this tool depends on the p-value outcome for each factor. The same as hypothesis testing, a factor with p-value less than 0.05 indicates that the factor is a significant factor, thus the need for it to be considered for improvement.
As for factor with p-value more than 0.05, you may eliminate this factor from the model as it is considered insignificant, provided the factor does not impact the process safety or other requirement.
Below is the example of Regression result from Minitab:
Predictor Coef SE Coef T P VIF
Constant 15.966 9.522 1.68 0.095
ERC1FYA.PV 0.19557 0.01770 11.05 0.000 1.472
TI10128.PV 1.2658 0.1610 7.86 0.000 1.138
TC106.PV -0.5624 0.1226 -4.59 0.000 1.596
PC103.PV 2.2534 0.7987 2.82 0.005 1.171
AI102A.PV 0.8829 0.2224 3.97 0.000 1.287
FIC164.PV 0.15304 0.05340 2.87 0.004 1.408
FC111.PV -0.09290 0.03862 -2.41 0.017 1.590
In this example, all of the factors are significant due to their low p-value.
More importantly, always remember to look at the VIF (Variance Inflation Factors). The VIF is an indicator of factors multicollinearity.
 |
