Multi-packet transmission aero-engine DCS neural network sliding mode control based on multi-kernel LS-SVM packet...

Area of Science:

• Control Engineering
• Artificial Intelligence
• Aerospace Systems

Background:

• Aero-engine digital control systems (DCS) exhibit strong nonlinear characteristics.
• Network-induced delays and random packet dropout significantly degrade DCS performance.
• Existing control methods struggle to effectively compensate for these network-induced issues.

Purpose of the Study:

• To propose a robust control strategy for aero-engine DCS facing nonlinearities, delay, and packet dropout.
• To develop an online compensation method for packet dropout using a sliding window multi-kernel LS-SVM.
• To design a neural network-based PID sliding-mode controller for improved tracking control.

Main Methods:

• Equivalent transformation of the time-delay term to establish a discrete system model.
• Optimization of multi-kernel function coefficients using a chaos adaptive artificial fish swarm algorithm.
• Online packet dropout compensation via sliding window multi-kernel LS-SVM.
• Neural network-based PID sliding-mode controller with online parameter adjustment.
• Simulation using Truetime for performance validation.

Main Results:

• The multi-kernel LS-SVM achieved a 29.21% and 44.66% reduction in packet dropout prediction error at 30% and 60% dropout rates, respectively, compared to combined kernel LS-SVM.
• The proposed neural network PID sliding-mode controller significantly reduced chattering amplitude compared to five other methods.
• The controller demonstrated a fast response speed, ensuring effective tracking control of the aero-engine DCS.

Conclusions:

• The proposed method effectively compensates for packet dropout and reduces chattering in aero-engine DCS.
• The neural network PID sliding-mode controller enhances tracking control performance in challenging network conditions.
• This approach offers a viable solution for robust control of networked aero-engine systems.