傳統上群樁基礎均採用試誤法的程序來設計，雖能符合規範，但其造價的經濟性卻是無法保證。本研究的目的便是最佳化的演算法則來進行群樁基礎的低價化設計，所設計之群樁基礎不僅可滿足規範的要求，亦可符合經濟性的考量。 在設計荷重、地質條件與施工方法已知的條件下，預鑄群樁基礎的設計變數主要包括樁徑、樁數、樁長、基樁間距和樁帽的尺寸，其中預鑄基樁的尺寸和樁帽內的鋼筋，乃由國內廠商已生產之尺寸所建立的資料庫來選取。最佳設計的目標函數則是總造價，包含土方開挖費用、樁帽費用和基樁費用等；而束制條件則包含強度需求、樁頂位移、樁距及樁帽的尺寸等。 由於設計變數為實數與整數之不連續變數，故本研究嘗試利用遺傳演算法來搜尋最佳解。演算法則的效率將透過數個設計例來說明，而影響群樁基礎造價的主要設計參數，亦將透過數值演算結果來探討，以供工程設計之參考。 Conventional design of pile groups is based on the trial-and-error procedures. Although the design results can satisfy strength and displacement requirements that stipulated in code provisions, it is not a minimum cost design. The purpose of this study is to apply the genetic algorithm (GA) for searching the minimum cost design of precast concrete pile groups. The objective function of the problem includes the costs of soil excavation, cap and piles. The design variables are the pile diameter, pile length, spacing of piles, and dimensions of cap, which are all considered as discrete design variables. The size of precast concrete piles and rebars in the pile cap are all selected from the available sections in the engineering market. The strengths and displacement constraints for the minimum cost design of precast concrete pile groups are formulated according to the foundation design code provisions. Size constrains, such as the length of piles, the diameter of piles, and spacing of piles are also considered in the formulation. The application of GA in the minimum cost design of pile groups is shown by a number of design examples. The efficiency of GA and sensitivity analyses of design variables on the cost of pile groups are also discussed.
