Data-driven approach for time-delay estimation of industrial processes

Area of Science:

• Control Engineering
• Signal Processing
• Industrial Automation

Background:

• Accurate time-delay estimation is critical for effective process control, performance assessment, and controller design.
• Industrial processes often operate under routine conditions with background disturbances, requiring methods that utilize available closed-loop output data.
• Existing methods may require system identification or prior process knowledge, limiting their applicability.

Purpose of the Study:

• To develop a novel, data-driven approach for estimating time delays in industrial processes.
• To provide practical solutions for time-delay estimation using only closed-loop output data under routine operating conditions.
• To validate the proposed method's effectiveness on various numerical and industrial examples.

Main Methods:

• A data-driven approach is proposed, estimating the closed-loop impulse response online from output data.
• For large time delays, direct estimation is performed without system identification or prior knowledge.
• For small time delays, estimation utilizes a stationarilized filter, pre-filter, and loop filter.

Main Results:

• The proposed method successfully estimates time delays using only routine closed-loop output data.
• The approach is effective for both large and small time-delayed processes.
• Validation on industrial examples like distillation columns and refinery heating furnaces demonstrates practical applicability.

Conclusions:

• The developed data-driven method offers an accurate and practical solution for time-delay estimation in industrial settings.
• The approach bypasses the need for system identification and prior process knowledge, enhancing its versatility.
• The method's effectiveness is confirmed across diverse industrial applications, highlighting its robustness.