Soft sensor based composition estimation and controller design for an ideal reactive distillation column | JoVE Visualize

Area of Science:

Chemical Engineering
Process Control
Artificial Intelligence in Chemical Processes

Background:

Reactive distillation columns are complex chemical processes requiring accurate state estimation for effective control.
Traditional state estimation methods may face challenges in accurately inferring key process variables like composition from indirect measurements.
Recurrent Neural Networks (RNNs) offer a potential solution for advanced inferential modeling in dynamic systems.

Purpose of the Study:

To design and implement a recurrent neural network (RNN) based inferential state estimation scheme for an ideal reactive distillation column.
To evaluate the performance of the proposed RNN scheme against established methods like Extended Kalman Filter (EKF) and Feedforward Neural Network (FNN).
To control the reactive distillation process by estimating composition from temperature measurements using a Time Delayed Neural Network (TDNN).

Main Methods:

Implementation of a recurrent neural network (RNN), specifically a Time Delayed Neural Network (TDNN), for inferential composition estimation.
Design and implementation of decentralized Proportional-Integral (PI) controllers for process regulation.
Comparative performance analysis of the RNN scheme against EKF and FNN under open-loop and closed-loop conditions, with online training.

Main Results:

The RNN-based state estimation scheme exhibited superior performance compared to the Extended Kalman Filter (EKF) and Feedforward Neural Network (FNN).
Qualitative and quantitative performance indices confirmed the enhanced state estimation capabilities of the RNN.
Online training of RNN and FNN models improved their adaptability to new, un-trained measurements.

Conclusions:

Recurrent Neural Networks, particularly TDNNs, provide a robust and effective approach for inferential state estimation in reactive distillation.
The proposed RNN-based scheme offers significant advantages over traditional methods for controlling complex chemical processes.
Accurate composition estimation from temperature measurements using RNNs enables improved process control and optimization.