以遺傳演算法鑑別結構接點參數

Identification of the Structural Joint Parameters by Genetic Algorithm Method

何信宗(Shin-Tzung Ho)

指導教授 ： 楊大中

元智大學/工程學院/機械工程學系/碩士(2004年)

摘要

本研究主要是以遺傳演算法，鑑別結構中的接點參數，藉由遺傳演算法搜尋全域最佳解。鑑別過程結合遺傳演算法，有限元素分析軟體，與模態實驗數據，以自動化方式進行計算，減少人工操作，有效提升鑑別效率。
本文以遺傳演算法來鑑別結構接點之勁度與阻尼參數。在過去的研究中，大多只鑑別單接點或是少數的接點參數。而本文一單邊支撐樑之螺栓接點參數，及以受楔型夾持器拘束之振動平板為例子，進行多邊界接點參數的鑑別。此外，本文除了考慮接點阻尼效應之外，還考慮了結構本身的阻尼效應，使鑑別的有限元素模型與真實結構更為趨近。
本文所建立的有限元素模型在夾持邊界下，各模態共振區利用各模態的材料參數分別修正，使分析的頻率響應函數更接近實驗值。透過足夠資訊的導入目標函數後，可針對單一Hz之多接點參數進行鑑別。由鑑別結果得知，當越多資訊被導入至目標函數中，遺傳演算法鑑別接點參數也就越準確，且鑑別值的落點也較集中。且當鑑別參數對目標函數的靈敏度越高時，鑑別值的落點範圍也就較小。大致而言，FRF條數愈多或其相異性愈大，其鑑別結果較佳。此係FRF之相異性愈大代表其攜帶之訊息愈豐富，故鑑別準確性較好。
接點阻尼參數之鑑別相對誤差，相較於接點勁度參數來的大；旋轉自由度之鑑別相對誤差，相較於平移自由度來的大。但就分析的頻率響應函數分析結果來看，其誤差影響並不顯著。平板的各接點參數的鑑別值，會有著由外向內遞增及對稱的現象。只有 因為敏感度低，此現象較為不明顯，但並不影響FRF分析值之精確度。平板邊界的勁度與阻尼總和值，不論3x5或是6x10網格分割在對稱簡化及未簡化的條件下，總和值相趨近。

關鍵字

接點 ； 參數鑑別 ； 遺傳演算法 ； 頻率響應函數

並列摘要

This research aims at identification of structural joint parameters by using the Genetic Algorithm method in searching the global optimum solutions. The Genetic Algorithm programs, combined with the finite element analysis software and frequency response functions (FRF) of vibration experiments, are automated to reduce the labor efforts and promote the efficiency of the identification processes.
Identification of single joint or a few joint parameters of structures were conducted in previous research works mostly, whereas the stiffness and damping parameters of multiple joints are identified in this study. Examples of a cantilever beam bolted at one of its ends and a plates clamped by a wedge-lock retainer at one of its edges are investigated, and the results show that the identification process proposed in this work is effective. The structural damping is also included in the finite element modeling of the structures to enhance the accuracy of the structural models.
The frequency response functions of the constrained structures from the finite element analyses are improved in each resonant frequency region by using the calibrated material properties for each specific mode. After introducing enough information into the objective functions in the GA programs, multiple joint parameters can be identified at each single frequency by the identification processes. Generally, the results show that more accurate results and less scattering of the identified values can be obtained with more information used. Using more FRF curves or increasing the inherent discrepancy of the FRF curves mean more information is poured into the identification processes and the identified results are better. Also the parameters with higher sensitivity have less scattering of their values.
The errors of the joint damping are larger than those of joint stiffness, and the parameters of rotational degree-of-freedom have larger errors than those of translational degree-of-freedom. However, these errors will not influence the accuracy of the calculated FRFs significantly. The identified values of the joint parameters for the vibrating plate increase from the outer nodes to the inner nodes and have the trends of symmetric distribution along the constrained edge of the plate. Only the rotational damping does not show this trend due to its less sensitivity to the vibration responses of the plate, but this difference has negligible effects on the FRF curves. The contribution sums of the joint damping and stiffness of the nodes along the constrained edge of the plate are close for 3x5 and 6x10 meshing conditions, and for either symmetric simplification of the constrained boundary conditions or not.

並列關鍵字

Joint ； Parameter identification ； Genetic algorithm ； Frequency response function

參考文獻

[10]楊大中,何信宗,嚴啟峻, “振動平板之邊界接點參數鑑別”, 第十二屆中華民國振動與噪音工程學術研討會, 2004.

[2]T. and Schwefel, H. P., “Evolutionary Computation: An Overview,” Proceedings of IEEE International Conference, pp. 20-29, 1996.

[4]Holland, J. H., “Adaptation in Natural and Artificial System”, University of Michigan Press, Ann Arbor, Michigan., 1975.

[5]Goldberg, D. E., Genetic Algorithms in Search, Optimization, and Machine Learning, Addsion Wesley Publishing Company, Inc., 1989.

[6]Kim, T. R., Ehmann, K. F. and Wu, S. M., “Identification of Joint Structural Parameters Between Substructures,” Journal of Engineering for Industry, Vol. 113, p419-424, 1991.