A systematic decomposition approach of nonlinear systems by combining gap metric and stability margin

In this paper, in order to control a nonlinear dynamic system via multi-model controller, we propose a systematic approach to determine the nominal local linear models. These models are selected from the local models bank and results in a reduced nominal models set that provides enough information to design a multi-model controller. To determine the initial local models bank, gap metric is used so that the distance between two successive local models is smaller than a threshold value. Then, a systematic approach that aims to get a reduced nominal models bank is developed. Based on this approach, first, a binary gap matrix is defined by combining gap metric and stability information. Then, several rows of this matrix are selected such that the sum of them becomes a non-zero vector. The proposed approach along with a designed robust controller is validated on a pH neutralization regarding to its highly nonlinear behavior.

Multi-Model Identification of HVAC System

Heat, ventilation and air conditioning (HVAC) is a crucial system for maintaining acceptable air quality and keeping the building and its occupants healthy. There are some challenges in controlling and identifying this system as it commonly operates in different operation conditions. Furthermore, various types of un-controlled sources disturb the steady operations. In addition, an HVAC system is an inherently nonlinear system and varies with time. As a result, conventional methods are not successful in identifying and controlling this system. This paper proposes a new multi-model approach in which the clustering and regression steps are performed simultaneously to tackle this problem. Cost functions of clustering and regression steps are combined and optimized using an iterative algorithm. After identifying the local models, a gap metric based approach is used to develop a global model of the process. The proposed approach is tested on a simulated ventilation unit system and real-world dataset. The results show the performance of the proposed method of identifying the ventilation system.