使用類神經網路建立遲滯系統模型

Abstract

遲滯是自然界普遍存在的現象，但也是工程應用上極不易處理而時常 將之忽略的一個課題。本文探討這個現象，試圖獲得更多的了解，以使那 些具有遲滯性質的系統發揮更大效益。 在本文中，我們不再追隨以前 數學工作者的方法，轉而設計一個具有假設意義的模型，將它命名為同步 延遲網路。繼而，我們以前傳式類神經網路將它架構起來，使其具有調適 能力。在文中並將證明它具有某些重要性質，以及展示它的學習效果。 Hysteresis is a frequently observed phenomenon in the realm ofnature. However, in engineering applications, hysteresis is usuallydifficult to deal with, and thus neglected. The purpose of this thesisis trying to understand more of it, such that we may achieve betterperformance from the systems which are hysteresis-embedded. In this thesis, we do not follow the traditional way in whichmathematicians have been working hard to develop mathematical models.Instead, we offer a hypothesis-based neural network model. We name itthe Synchronous Delay Network (SDN) model. And then, we realize it asa feed-forwardneural network and support it with capability of adaptation. Severalimportant properties of this proposed model and network arederived and proved, and the adaptive performance is shown and discussedin this thesis.