DETECTION OF PROCESS AND SENSOR FAULTS USING MODEL-BASED APPROACHES IN INDUSTRIAL BATCH PROCESSES

摘要

The pressure to develop new approaches for process control with improved performance leads to the application of model based predictive control methods in a variety of processes. However, the main drawback of implementing these approaches lies in the need for expensive model development. In this work, new prospects for current advanced process control projects in the chemical industry incorporating algorithms for fault detection and identification are presented. The starting point in this contribution stems from a research cooperation work with the Evonik Degussa GmbH at Hanau-Wolfang, Germany. The proposed design approach will show that additional benefit can be made accessible by using process models for fault detection and identification in addition to regulatory control. The considered industrial case studies reveal the improvement and the need for robust symptom generation in potential online applications.