State Prediction in the Presence of Input, State and Output Delays: Application to Compressor Surge Control Using Active Magnetic Bearings

本研究项目的目标是开发新的设计方法，以控制在输入、输出和状态中出现时间延迟的工程系统。这些方法将帮助工程师减轻延迟对稳定性和性能的不利影响。多级离心式压缩机是最广泛的工业设备类型之一，并消耗大量的能源。它们用于驱动天然气管道，是建筑供暖、通风和空调系统的重要组成部分。将项目结果应用于喘振不稳定性将使这些压缩机的运行效率更高，从而降低运行成本并减少总体能源消耗。通过弗吉尼亚大学的旋转机械与控制（ROMAC）联盟，研究团队将与主要压缩机制造商的工程师进行互动，制定和传播解决方案，以应对行业面临的重大技术挑战。在时滞系统中，当输入延迟大于状态延迟时，传统的预测器反馈会导致非因果控制律的产生。初步调查表明，即使在这种情况下，也有可能建立一个因果状态预测器。其基本思想是将预测时间划分为小于最大状态延迟的部分，并在预测时间内递归地进行预测。由此产生的状态预测器是因果的，但却是无限的维度。这个项目的主旨是设计安全实现无限维预测器的方法。研究结果有望为控制同时存在输入、状态和输出延迟的系统提供强大的新工具。该项目的一个组成部分是将理论结果应用于利用主动磁轴承控制多级离心压缩机喘振不稳定。

The objective of this research project is to develop new design methods for controlling engineering systems in the presence of time delays arising in the inputs, outputs, and states. These methods will help engineers alleviate the adverse effects of delays on stability and performance. Multistage centrifugal compressors are one of the most widespread types of industrial equipment, and consume significant amounts of energy. They are used to drive natural gas pipelines and are essential components for building heating, ventilation, and air conditioning systems. Application of the project results to surge instability will enable higher efficiency operation of these compressors, thereby lowering operating costs and reducing overall energy use. Through the Rotating Machinery and Controls (ROMAC) consortium at the University of Virginia, the research team will interact with engineers from major compressor manufacturers, to formulate and disseminate solutions to major technical challenges faced by the industry.Conventional predictor feedback in time-delay systems leads to noncausal control laws when the input delays are larger than the state delays. Initial investigations indicate that it is possible to construct a causal state predictor even for this case. The fundamental idea is to partition the prediction time into sections that are each shorter than the maximum state delay, and carry out the prediction recursively over the prediction time. The resulting state predictor is causal, but infinite dimensional. The thrust of this project is to devise methods to safely implement the infinite-dimensional predictor. The results are expected to lead to powerful new tools for control of systems subject to simultaneous input, state and output delays. An integrated part of the project is to apply the theoretical results to the control of surge instability in multistage centrifugal compressors by use of active magnetic bearings.