Souven Jawa Steward Gima <br/><br/>
Multivariate statistical process monitoring based on dissimilarity matrix / 
Souven Jawa anak Steward Gima 
 - Kuantan, Pahang :  UMP,  2012 
 - xiii, 49 p. :  ill. ;  30 cm. +  1 CD-ROM 
<br/><br/>Project paper (Bachelor of Chemical Engineering) -- Universiti Malaysia Pahang – 2012 
<br/><br/>Bibliography : p. 35-37 
<br/><br/> This thesis mainly deals with the development of new technique in process monitoring which integrates dissimilarity matrix together with the traditional technique which is the conventional Principal Component Analysis (cPCA). From previous research, it is concluded that the cPCA technique itself is not effective when the process variables of the particular process is non-linearly correlated. Specifically, the cPCA cannot detect a change of correlation among process variables as long as the monitored indexes are inside the control limits. The new technique which is dissimilarity integrated with PCAbased Multivariate Statistical Process Monitoring (MSPM), was developed by using Matlab. By using the software, two runs are executed which are based on the traditional technique and the new technique respectively. The data used is based on the simulation of CSTR system and two faults are tested with both abrupt and incipient types of faults are taken into account. The final results are projected in form of Hotelling’s T2 and Squared Prediction Error (SPE) monitoring statistics to be analyzed. Significantly, the cPCA can detect fault introduced into the system earlier if compared to the new technique specifically for incipient faults. Nevertheless, both methods are able to perform fault detection equally well for abrupt faults. 
<br/><br/><label>ISBN: </label>THE0002018(Local) 
<br/><br/><label>Subjects--Topical Terms: </label><br/>Multivariate anlaysis<br/>Principal component analysis