Detecting Mismapping in Cross-Directional Control Systems

Abstract

Cross-directional (CD) control systems regulate product properties in industrial web processes by manipulating the inputs to arrays of actuators. In many of these processes, the product wanders or undergoes transformations such as stretching that introduce lateral shifts in the location of the actuator responses. The spatial position of the center of the response of each actuator is known as the actuator mapping and for CD controllers to operate effectively, it is important that this is known accurately. Actuator mapping is one of the main sources of uncertainty in the model that is used to design CD controllers and mismapping can lead to closed-loop instability, so CD controllers are often designed conservatively to provide robustness to this uncertainty. Because CD controllers can operate with a significant level of mismapping, both the dynamic and steady-state responses of the closed-loop system are often degraded. In practice, it is difficult to detect the presence of mismapping, but this brief describes an online diagnostic technique based on a vector auto regression (VAR) representation of the closed-loop response of the actuator inputs, which derives a statistic from existing plant data that can be used in a hypothesis test to detect the presence of mismapping. The performance of the technique is demonstrated by results from an implementation on a plastic film process.