全面研發免疫演化法的設計最佳化(II)

機構典藏 > College of Engineering > Graduate Institute & Department of Mechanical and Electro-Mechanical Engineering > Research Paper > Item 987654321/102894

Please use this identifier to cite or link to this item: https://tkuir.lib.tku.edu.tw/dspace/handle/987654321/102894

Title:	全面研發免疫演化法的設計最佳化(II)
Other Titles:	Comprehensive Immunity Based Evolutionary Design Optimization (Ii)
Authors:	史建中
Contributors:	淡江大學機械與機電工程學系
Date:	2012-08
Issue Date:	2015-05-11 14:37:03 (UTC+8)
Abstract:	本研究是摹仿生物之免疫系統為基礎，研發數學模型，再應用於研發全面的最佳化設計方法及軟體程式。整體計畫主要分為兩階段實施，第一階段為建構一般最佳化模型；第二階段發展高階的最佳化。經由探討免疫系統學理推論得知，該系統必須非常準確及非常迅速的對付來犯的病毒，達到最佳的消滅病毒狀態。所以，師法免疫系統原理所發展的數值最佳化，預測應具有相當高精準性及高效率性。又由於病毒多樣及動態變化，因此師法其原理所發展的最佳化方法應是可適當處理多種狀況的問題及應用於變化多端的實際問題。本研究之特色為發揮純免疫系統的精神與性能，根據免疫系統的理論，建構出模擬免疫系統的數值演算模型，再以最佳化數學為基礎，發展模擬免疫系統演算法於最佳化設計及程序技術之全面研發，應用於多樣類型的工程設計問題中，期望在解工程問題時，能得到兼具良好的收斂性，精確性以及穩健性的全域最佳解。本全程研究之主項目的可綜述如下：1. 發展無限制條件的演化法，可解單一全域極值問題；可解全部全域極值問題；可解全部全域極值及區域極值問題；及可解全部全域極值及其附近之部分區域極值問題。 2. 應用本方法技術於多極值的結構設計應用。 3. 根據前述結果繼續發展限制條件的的處理方法，其中含模擬純生物免疫原理求解及結合數值應用求解的方法。 4. 發展求解前述項目含有整數及離散變數之混合變數問題。 5. 研發求解前述項目的多目標設計問題，可解得多個折衷解或唯一平衡解。 6. 發展目設計函數是隱函數之免疫系統最佳化解法。 7. 發展免疫系統最佳化求解法解多變數及多限制隱函數的大型工程設計問題。前四項的研究屬於第一階段，後三項屬於第二階段，第一階段預計持續三年，第二階段預計持續二至三年。在主項目下，尚有其他次項目研究，不但可系統化發展強大能力的無梯度最佳化方法，除了相當具有學術價值外，亦具有實用價值，同時可持續培訓研究生之研發能力。目前在第一階段研究的第一年研究已完成，根據生物免疫學的原理以 C++的程式語言自行研究開發全部類免疫為基的最佳化程式及數值演算法，已完成免疫系統原理探討分析，最佳化模型建構。含有四種類型的考慮，(1) 只有單一全域極值的求解。 (2) 多個全域極值的求解。 (3) 多個全域極值及最靠近的部分區域極值求解。 (4) 全部多個全域極值及多個分區域極值求解。在第一階段研究計畫的第二年研究，預期完成之項目為： 1. 發展類免疫為基的桁架型態及結構最佳化，含二維桁架型態及結構最佳設計及三維桁架型態及結構最佳設計。 2. 探討限制條件處理策略，包括以多目標技術處理限制條件。 3. 雙熱臂電熱式微致動器最佳化設計。順應著免疫學的原理以C++的程式語言自行研究開發全部類免疫為基的最佳化程式及數值演算法，並改善最佳化演算法之精確性，效率性及穩健性。 The proposed research project simulates biological immune system (S) as the core mechanism to develop mathematical models, and further to full potential of non-gradient based optimization. The whole research process primarily divides into two stages. The first stage contains the modeling of general optimization; and the second stage contains the development of advanced optimization methods. From the exploration of the immune theory, it is recognized that the IS must precisely and effectively defeat the virus so as to eliminate them in a perfect (optimization) state. Therefore, if the operation of immune system can be applied to bio-optimization technique via the simulation technique, the optimization result should has high precision and efficiency. Additionally, due to the variety and dynamic characteristics of viruses, the simulation to optimization may deal with vary situation of problems and real dynamically unstable problems. The feature of this proposed research focuses on the original theory of immune system in the simulation of optimization model and software. The main topics in the complete research period contains: 1. the development of immunity based unconstrained evolutionary optimization. This development can solve single model problem, single global optimum in multi-model problems, multiple global optimums in multi-model problems, complete global and local optimums. 2. To apply this development to structural designs. 3. The development of dealing with constraints problems. 4. The development of containing a mix of real, integer and discrete variables. 5. The development for multiple objective optimization problems to obtain Pareto front design or unique balanced design. 6. the development of handling implicit objective and/or constrained functions. 7. the solution to large-scale explicit and implicit engineering design optimization problems. The previous four topics belong to the first stage work and the rest of them are second stage. Eventually, a powerful non-gradient based optimization utilizing immunity theory can be well developed that can win the academic contribution, practical values and perform as a perfect training for graduate training program. The first year in first stage work, some items have been done and finished. 1. A complete study for biological immune theory to build up the solid foundation. 2. The simulation modeling of optimization from immune theory. This is an important field of software engineering that considers four sub-problems research: single global optimum, multiple global optimums, multiple global optimums and some local optimums, and a complete set of global and local optimums. 3. To examine and improve the optimization software by using several bench-mark problems. Using C++ programming develop the whole program of immunity based evolutionary optimization software. In the second year research proposed here contains several main items. To examine and improve the optimization software by applying to topology and size optimization problems of 2-D and 3-D truss structures. To develops constraints handling technique in current study. The developing method will be applied to double hot-arm electro-mechanical actuator for the improvement of output performance. The project can develop the complete program written in C++ of immunity based evolutionary optimization software.
Appears in Collections:	[Graduate Institute & Department of Mechanical and Electro-Mechanical Engineering] Research Paper

Files in This Item:

There are no files associated with this item.

Loading...