Heat Exchanger Modeling by Neural Network Optimization for PETRONAS Penapisan Melaka Sdn. Bhd (PPMSB) Crude Preheat Train

• Main Author: Md. Tahir, Norazliza
• Format: Final Year Project
• Language: English
• Published: Universiti Teknologi Petronas 2005
• Subjects: TP Chemical technology
• Online Access: http://utpedia.utp.edu.my/7654/1/2005%20-%20Heat%20Exchanger%20Modeling%20by%20Neural%20Network%20Optimization%20for%20PETRONAS%20Penapisan%20Melaka%20Sdn.%20.pdf; http://utpedia.utp.edu.my/7654/

The title of this Final Year Research Project is 'Heat Exchanger Modeling by Neural Network Optimization for PETRONAS Penapisan Melaka Sdn. Bhd (PPMSB) Crude Preheat Train'. This project involves the post modeling of heat exchanger sensitivity analysis matcovers neural network based model and implication of statistical analysis to predict the heat exchanger efficiency for maintenance scheduling strategy of Crude Preheat Train (CPT). Themainobjectives of this study are to minimize the error in the predicted values andenhance therobustness ofthe previous model to predict in future. This Final Report consists of five major sections. The first section describes the introduction to Neural Networkbased PredictiveModel, backgroundof the CPT, fouling activity and Heat Exchanger Maintenance in PP(M)SB, problem statement that defined the significant ofthepost modeling heat exchanger sensitivity analysis, project objectives and scope of works done throughout the study. The next section consists of literature review and theory extracted from well established journals and web sites to provide relevant information for the project as references. The third section entails the project methodology comprising series of stages for the project to be carried out. It follows by the fourth section that serves as the gist of the report that presents the findings and includes discussion on the results obtained and significance behind any failure occurs at each stage of the completed optimization strategies. The results are discussed in term of statistical analysis, comparison ofresults between different transfer functions configurations used and graphs of actual denormalized versus predicted outlet temperature for both tube side and shell side. The final section of the report consists of the conclusion corresponds to the objectives set earlier and some recommendations for future improvement of the Neural Network model. The FinalReport ends with a listof references andappendices.