Modeling of faults for chemical batch reactor using artificial neural network and fuzzy logic
| dc.contributor.author | Syed Ab. Rahman, Syed Azhar | |
| dc.date.accessioned | 2016-01-22T08:04:24Z | |
| dc.date.available | 2016-01-22T08:04:24Z | |
| dc.date.issued | 2009-06 | |
| dc.description.abstract | Every chemical processes prones to failure. This situation enforces the researchers and industrial to find the appropriate techniques to detect a process failure as early as possible. The best solution is by implementing fault detection and diagnosis system (FDD). In these studies, ten process faults have been designed for the experimental work. The temperature and conductivity data were collected during the experiment and the conversion and concentration of the products were calculated. These data were then acted as an input into the modeling system. In the other hand, the normal and ten faulty situations acted as an output for the modeling system. In the modeling by using Artificial Neural Network (ANN), Multilayer Perceptron (MLP) with single hiddl:(n layer was implemented. For the feature extraction study, the correlation of concentration-conversion-past temperature produced the best result with sum square error (SSE) of 133.38 and r-value of0.999. The optimum number of the hidden layer was found to be 21 neurons with the lowest SSE value of 117.65 and r value of 0.99. The developed ANN was successfully detected and isolated the 10 prescribed faults during the detection and isolation session. This developed modeling then has been further optimized and validated with another set of experimental data which were not used during the training and testing. Again, the developed ANN was successfully produced fault patterns and isolated the faults. The application of an advanced warning and diagnosis threshold with the limit of 0.2 and 0.8 could give an advanced warning and diagnosis on the training and testing data. The 10 designed faults were also successfully detected by using Fuzzy Logic (FL) approach. Comparison on the I different Membership Function (MF) indicated that 5 MFs have better ability to detect the faults compared to 3MFs. The result also shows that Triangular and Gaussian shape MF produced similar the results. However, the Gaussian has the ability to detect and isolate 40% more single fault compared than the Triangular. After eliminating some redundancies rules, the detection and isolation of single fault detected increased about 12% and paired-fault reduces about 76%. Finally, a new Fuzzy Inference System (FIS) has been proposed in the present study to replace the existing FIS in the MA TLABĀ® or proposed from previous researchers. As for the conclusion, the ANN and FL have potential methods in FDD studies. Both these methods were able to detect and isolate various faults considered in the study. It shows that ANN and FL can be implemented for monitoring any process faults in chemical batch reactor. | en_US |
| dc.identifier.uri | http://hdl.handle.net/123456789/1692 | |
| dc.language.iso | en | en_US |
| dc.subject | Chemical batch reactor using | en_US |
| dc.subject | Artificial neural network | en_US |
| dc.title | Modeling of faults for chemical batch reactor using artificial neural network and fuzzy logic | en_US |
| dc.type | Thesis | en_US |
Files
License bundle
1 - 1 of 1
Loading...
- Name:
- license.txt
- Size:
- 1.71 KB
- Format:
- Item-specific license agreed upon to submission
- Description: