Transient Management in Reconfigurable Control Systems

This report investigates the transient phenomena due to run-time reconfigurations in controlled systems. It is shown that in the case of dynamic systems these reconfiguration transients might be reduced to a considerable extent using dedicated design-time and/or run-time methods. The methodology behind applies the so-called model-integrated approach [SK97], which tries to utilize the available knowledge about the system and its environment by constructing " dynamically " consistent computer-based representations. These representations integrate logic, dynamics and constraints [BeMo99], and provide hierarchical system descriptions for monitoring and control applications. Such descriptions can handle evolution governed by on/off switches, if-then-else rules, system degradation due to faults, etc. There are very many approaches that deal with such and similar systems, often called hybrid systems [GNRR93], however, very few studies address the problem of transient phenomena due to the interactions of the higher-level logical/discrete mechanisms and the lower-level continuous dynamics. Since these transients might cause intolerable disturbances both in the value and the time domain, the reduction of such side effects becomes an additional design issue. Throughout this report the interactions causing transients are considered as reconfigurations " implemented " as instantaneous parameter and/or structural changes, and the introduced additional design considerations aim at reducing these transients. The actual course of the reconfiguration transients depends on one hand on the structure of implementation, while on the other on the actual interactions with the environment. For this very reason transient reduction can be achieved partly by selecting proper structure for implementation and partly by techniques utilizing multi-objective, run-time optimization. The term " transient management " tries to reflect the simple fact that to reduce the transients several constraints and control rules are to be applied simultaneously, and therefore a management-like activity is needed. The first part of Section 2 shows how and why transient phenomena might appear after reconfiguration. The second part includes a literature survey covering studies related to different aspects of transient behavior, summarizes the already available results concerning " transient management " , and outlines the content and the structure of the report. Section 3 is devoted to introduce the so-called passive transient management methods, which utilize the fact that the different structures of implementation produce quite different transients in the case of otherwise identical reconfigurations. These passive methods are design-time techniques, and do not influence the transient phenomena directly. This section also contains some considerations concerning multiple-step reconfigurations and provides examples illustrating …