Scholarlink

This article describes the design of a control system for a common industrial process, the Reactor-Separator-Recycler (RSR), and the development of two estimators: a full-state observer and an artificial neural network, to estimate the concentration at the output of the Continuous Stirred Tank Reactor (CSTR) in the RSR process. For the controller design, empirical approximate models are obtained from the reaction curve via identification procedures. The estimated concentration is used to close a control loop for the CSTR. Temperature perturbation tests and initial concentration estimations at the mixing point are performed. The results show the efficiency of the neural network compared to the full-state observer.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Full State Observer Versus NARX-Based as Soft Sensor for Reactor–Separator–Recycle Process Control

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Marco Herrera,
  • Luis Caiza,
  • Diego Benítez,
  • Noel Pérez-Pérez,
  • Oscar Camacho,
  • Álvaro Prado

摘要

This article describes the design of a control system for a common industrial process, the Reactor-Separator-Recycler (RSR), and the development of two estimators: a full-state observer and an artificial neural network, to estimate the concentration at the output of the Continuous Stirred Tank Reactor (CSTR) in the RSR process. For the controller design, empirical approximate models are obtained from the reaction curve via identification procedures. The estimated concentration is used to close a control loop for the CSTR. Temperature perturbation tests and initial concentration estimations at the mixing point are performed. The results show the efficiency of the neural network compared to the full-state observer.