Fault detection for iterative learning control of time-varying systems

Abstract

The double-roller casting system comprises a pair of casting rollers rotating in opposite directions, wherein an adjustable roller gap is formed between the casting rollers; a casting roll controller configured to adjust a roll gap between the casting rolls in response to a control signal; a cast strip sensor that measures a parameter of the cast strip and generates a strip measurement signal; and an iterative learning control (ILC) controller that receives the tape measurement signal and provides a control signal to the cast roll controller. The ILC controller includes a fault detection algorithm that receives the control signal and the tape measurement signal and generates a fault detection signal indicating when a fault condition is detected, and an iterative learning control algorithm that generates the control signal. The fault detection algorithm indicates a fault condition when it is detected that the control signal exceeds an upper control saturation threshold or the ILC controller operates in an instable state.

Description

Background technique [0001] Iterative Learning Control (ILC) is a control method that is particularly suitable for processes that are repetitive in nature. When applying ILC in practice, the algorithm should be set up to ensure that 1) the system will remain stable, and 2) the performance specifications will be met. However, these goals may conflict with each other, especially when the plant model contains uncertainty and / or iteratively changing dynamics. [0002] Recent studies have examined the convergence and robustness properties of ILC algorithms applied to iteratively varying or non-repetitive systems. More specifically, previous research developed an optimal design procedure for ILCs that balances system performance with robust stability criteria. However, a limitation of this approach is that satisfying the robust stability criterion requires knowledge of the upper bound on the uncertainty of the input and output operators. In some applications, the upper bound may ...

AI Technical Summary

Problems solved by technology

However, this approach is limited in that satisfying the robust stability criterion requires knowledge of upper bounds on the uncertainty of the input and output operators

In some applications, the upper bound may be too conservative during normal operation, leading to poor performance of conservative learning algorithms

Furthermore, if there is uncertainty in the boundary values ​​used to design the ILC algorithm, the ILC algorithm can become unstable if the device violates the assumed upper bound

Images