Utilizing multiple linear regression technique for interential measure of continuous-based process monitoring / Muhammad Ridzuan Mamat

Project paper (Bachelor of Chemical Engineering) -- Universiti Malaysia Pahang - 2013

Bibliography : p. 29-31

The present conventional MSPC has several weaknesses in process fault detection and diagnosis. Some researchers in this filed had commented that the MSPC is a powerful tool for data complexity reduction and fault detection in the significant fault appearance data. The current fault detection and diagnosis method via MSPC is limited to significant faults and does not point put the insignificant ones accurately. In the real time, all variable will be used in monitoring. However in this case only a few of them are truly important. By developed modeling based on multiple linear regressions the relationship between these variables can be figure out. Multiple linear regressions (MLR) is a method used to model the linear relationship between a dependent variable and one or more independent variables. Some assumption should be made in order to obtain an accurate data analysis. The assumptions are variables should normally distribute, a linear relationship between the independent and dependent variables must exist and also the variable should be measure without an error. MLR is probably the most widely used in dendroclimatology for developing models to reconstruct climate variables. Besides they also proposed for control charting methods for lumber manufacturing and profile monitoring applied in public health surveillance. The methods to perform this modeling involve two phases which are Phase I: offline modeling and monitoring and Phase II: online monitoring. As a conclusion, the MLR method is success introduced as a significant improvement compared to the conventional method. On top of those objectives, the original goals of SPC are also been considered as well as carried together, such a way that the productivity of multivariate process monitoring is improved