Loading...

Development of a Real Time System Identifier for Interconnected Systems

Babaei, Mohammad Reza | 2021

345 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 54294 (06)
  4. University: Sharif University of Technology
  5. Department: Chemical and Petroleum Engineering
  6. Advisor(s): Bozorgmehri Boozarjomehri, Ramin
  7. Abstract:
  8. Access to a reduced and fast mathematical model that can predict the input-output behavior of interconnected systems with acceptable accuracy and in a short time is crucial. The purpose of this project is to provide an object-oriented platform for identifying a model based on input-output data of interconnected systems in real-time. On this basis, in the case of developing fast estimation algorithms and updating the model quickly (in case of extensive changes in the system), good predictions of the model can be used for various applications. In this research, machine learning and data analysis methods are used to extract useful features and relations from data. Then, taken advantage of this information to design model structure. In order to let the model learn continuously, universality and fast training, deep learning, image processing, and natural language processing models are used to model time-series in real-time. graph-type neural network, recursive neural network (RNN), long-short term memory (LSTM), and convolutional (CNN) neural networks have been used for this purpose. The use of the mentioned models has challenges such as over-fitting. To prevent these problems, appropriate solutions such as using a random removal layer have been considered. The implemented platform is first used to identify two series reactors and then is used to identify the Tennessee Eastman process. Obtained results from both relation analysis between variables in the system and predicting the system behavior is promising. In addition, the processing time and model identification promise to use this platform for faster processes. The final model obtained can be used in the field of simulation and control of interconnected systems
  9. Keywords:
  10. Object-Oriented Programming ; Machine Learning ; Neural Network ; Real Time System ; Parameters Identification ; Parametric Prediction ; Tennessee-Eastman Process ; Transient Behavior ; Interconnected System Transient Behavior

 Digital Object List

 Bookmark

No TOC