- 學位論文
矩形截面分岐管消音器之性能分析與最佳化設計
Acoustic Performance Analysis and Optimum Design of a Silencer with Multiply Connected Tube of Rectangular Cross Section
摘要
本論文整合類神經網路系統(Neural Network System)與遺傳演算法 (Genetic Algorithm, GA)，搭配聲學分析軟體SYSNOISE 使用邊界元素法 (Boundary Element Method, BEM)於矩形截面分岐管干涉型消音器最佳 化設計。 論文主要分三部份：(1) 干涉型消音器之消音性能，利用消音器的管 路分開為二或更多通路，最後再使其管路匯集為一，來減低噪音，並探 討改變彎管半徑以及改變彎管截面積，對干涉型消音性能之影響。(2) 利用遺傳演算法之最佳化方法和聲學分析軟體SYSNOISE 之結合於干 涉型消音器貼覆吸音材料位置最佳化設計，利用遺傳演算法規劃吸音材 料位置，來達到最佳的消音性能。(3)整合類神經網路系統與遺傳演算法 之最佳化方法，並搭配聲學分析軟體SYSNOISE 於方形截面干涉型消音 器之尺寸最佳化設計。 經過實驗的驗證，與模擬的結果相當吻合，對於工業界的應用，可 有效地減少開發時間和降低生產成本，提升產品競爭力。
並列摘要
This article is integrated by Neural Network System and Genetic Algorithm (GA), collected by the acoustic analysis software SYSNOISE with Boundary Element Method on Acoustic Performance Analysis and Optimum Design of a Silencer with Multiply Connected Tube of Rectangular Cross Section. The article included three parts: (1) the performance of the silencer: using the tubes on the silencer which is separated by two or more paths, then combine together to reduce the noise. Discussing the effects to the performance with change the radius of the tubes and changing the cross-section area of the tubes. (2) Integrating by Neural Network System and GA, and collocated acoustic analysis software SYSNOISE on the optimum design of the silencer: aiming to a specific frequency, using Neural Network System to build a network mode, and using Neural Network System and GA to get optimum dimension and the best Sound Transmission Loss (STL). (3) Combining the optimization of GA with acoustic analysis software SYSNOISE on the optimum design of absorbing material position on the silencer, it can reduce the using of absorbing material and promote the performance of the silencer. After experimental verification, it tallies with the simulation result. For industry application, it can save the time of development, decrease the costs and promote the competitiveness of products.