臺灣博碩士論文加值系統

本文主要是藉由模糊控制理論及神經網路學習法則,其易於處理非線
性系統的優點,來達到對F-100渦輪扇引擎做主燃油控制之目的.文中首先
使用一模糊邏輯控制器來對轉速做控制,再針對不同的測試點利用試誤法
先找出其較適當的參數後,接著使用遺傳演算法來找出各測試點的最適參
數;最後再利用神經網路來做全域操作之控制,經由電腦的模擬結果顯示,
此控制架構均能在引擎的安全限制下達到性能的要求,亦驗證此控制系統
在引擎全域範圍操作之可行性.而當高壓渦輪轉速命令為一時變軌跡時,原
有的控制架構會產生較大的追蹤誤差,因此本文嘗試使用兩種策略來解決
此一問題,一為模糊邏輯控制策略,一為自我修正策略,經由電腦的模擬結
果顯示,所設計的兩種方法皆能在有效的時間範圍內達到軌跡追蹤的效果,
亦驗證了此二種策略的可行性.

The goal of this thesis is to control the main fuel of a
F-100 turbon-fanfan engine by means of fuzzy control and neural
network which are knowns tobe effective methods to deal with
nonlinear systems. A fuzzy logic controleris designed to control
the speed of a F-100 turbon-fan engine. The parametersof the
fuzzy logic controler for different test points are found by
trial-and -error first. The genetic algorithm is then used to
find the most suitable parameters. Finally, neural networks are
used to schedule the parameters forfull flight envelope
operation. Simulation results show that the proposed control
system has satisfactory performance within full engine operation
envelope. When the speed command is a time varying signal, the
original controler will result in large tracking error. Two
methods are proposed toovercome this difficulty: 1)fuzzy logic
control and 2)self-rectification method. Simulation results
show that both methods have satisfactory performancefor time
varying input signals.