AI System for Valve Stiction Identification

Dow is applying AI to plant automation and control systems with the goal of increasing productivity and reducing labor. AI systems are installed in close collaboration with experts, employees and developers to ensure acceptance, reliability and overall improvement.

Valve stiction is a common equipment problem caused by various factors, such as dirty valve stems or actuator problems. When a control valve suffers from stiction, it does not respond promptly to changes in the controller output signal, resulting in poor control performance. This can affect the overall efficiency and performance of a plant unit.

To mitigate the problem of valve stiction, regular maintenance is necessary. Proper diagnosis of con-trol valves helps detect stiction problems early so that preventive maintenance can be performed. This prevents unplanned unit shutdowns, saves money by avoiding unnecessary maintenance, and results in smoother overall plant operation.

Dow uses an AI-based method to identify underperforming control valves based on trended process data and valve position data. This method requires no additional instrumentation. It uses a deep con-volutional neural network (CNN) to extract important patterns and structures from the data to identify valve stiction.

The neural network has been trained on historical data from many control loops. It is now capable of detecting valve stiction problems with a high degree of reliability. Deteriorating control valve perfor-mance is indicated early. The results are displayed on continuously updated dashboards that are easy for maintenance personnel to read and use to make proactive maintenance decisions.


Dow Olefinverbund GmbH

Straße B 13

06258 Schkopau