On-line lower-order modeling via neural networks.

Area of Science:

• Control Engineering
• Artificial Intelligence
• Industrial Automation

Background:

• Accurate modeling of dynamic systems is crucial for effective process control.
• First-order plus dead-time (FOPDT) models are widely used in process industries.
• Traditional parameter estimation methods can be complex and require significant a priori information.

Purpose of the Study:

• To present a novel neural network-based method for determining FOPDT model parameters.
• To develop an adaptive controller using the proposed parameter estimation technique.
• To demonstrate the applicability of neural networks for on-line industrial control.

Main Methods:

• Utilized neural networks to directly output gain, dominant time constant, and apparent time delay of FOPDT models.
• Integrated the neural network parameter estimation with Proportional-Integral (PI) or Proportional-Integral-Derivative (PID) controllers.
• Developed an adaptive control scheme requiring minimal prior knowledge of the plant.

Main Results:

• Neural networks accurately determined the key parameters of FOPDT models.
• The combined adaptive controller demonstrated effective performance.
• Simulation and experimental results validated the feasibility and adaptive capabilities of the proposed scheme.

Conclusions:

• The proposed neural network method offers a simplified approach for FOPDT model parameter identification.
• The adaptive controller is effective and requires little a priori plant knowledge.
• This work presents a viable new approach for implementing neural network-based on-line industrial control.