高性能混凝土的強度與坍度是混凝土品質的重要因子,由於缺少數理模型,強度、坍度與配比的關係必須透過實驗收集數據,再以迴歸分析或類神經網路建立模型。一般土木材料的實驗設計缺少系統化的方法,因此本研究嘗試以實驗設計(Design of Experiment)來設計實驗。本研究採用傳統的D-Optimal設計方法,以五種實驗數目各自以類神經網路建立強度、坍度預測模型,並與隨機法所建立的模型作比較。本研究結果顯示:(1)要建立準確的預測模型,強度模型需要100個以上的配比實驗;坍度模型只需50個以上的配比實驗。(2) D-Optimal產生的模型相對於隨機法所產生的模型要來得好。
Strength and slump are the important measures of high performance concrete. Because there are no mathematical models, the relationships between strength and slump and proportion must be deduced from collecting experimental data, then establishing models by regression analysis or artificial neural networks. Generally, construction material experiment designs lack systematic methodology. Therefore, this research attempts to use design of experiments (DOE) to design the experiments. This study used the traditional D-Optimal design method, and five kinds of experimental numbers to establish strength and slump models by artificial neural networks, respectively. The results showed that (1) to establish an accurate forecast model, the strength model needs more than 100 mix proportion experiments; the slump model only needs 50 mix proportion experiments, and (2) the models produced by D-Optimal design method are much more accurate than those produced by random design.