Utilizing mlr technique in mspm system
The main purpose of this research is to propose a new MSPM technique, where the original variables are modelled into linear composites in order to reduce the number of variables in monitoring, where eventually it may also monitoring the performances. Hence, the objectives are to develop the conventi...
| Main Author: | |
|---|---|
| Format: | Undergraduates Project Papers |
| Language: | English |
| Published: |
2014
|
| Subjects: | |
| Online Access: | http://umpir.ump.edu.my/id/eprint/9095/ http://umpir.ump.edu.my/id/eprint/9095/ http://umpir.ump.edu.my/id/eprint/9095/1/CD8540%20%40%2047.pdf |
| Summary: | The main purpose of this research is to propose a new MSPM technique, where the original variables are modelled into linear composites in order to reduce the number of variables in monitoring, where eventually it may also monitoring the performances. Hence, the objectives are to develop the conventional MSPM method for the original set of variables (System A), to develop the conventional MSPM method, which applies Multiple Linear Regression (MLR) technique (System B) and last but not least is to analyse the monitoring performances between System A and System B. By doing this research, it was proved that, it is able to justify that the developed and upgraded MSPM method is comparatively better than conventional PCA-based MSPM method in monitoring the multivariate of non-linear process. The research also show the development of advanced multivariate way of process monitoring in terms of variables points besides of samples scores. The complete procedures of fault detection and identification comprise of two main phases namely as off-line modelling and monitoring (Phase I) and on-line monitoring (Phase II) where MLR technique is applies in phase I. All the mathematical model were developed into coding of matlab platform of version 7. The results were presented in the form of graphs or Shewart Chart of 95% and 99% confident limit with the Hotelling T-Squared Distribution and Squared Prediction Errors was considered has the statistical tools for observing. Both method which are PCA-based MSPM System and MLR-based MSPM System were compared it performance and had been analysed. From the results, MLR-based MSPM Sytem gives better performance interm of faster detection of fault compared with PCA-based MSPM System. As the conclusion, all the three objectives were achieved successfully |
|---|