This paper proposes a target specialized meta-heuristic optimization algorithm, called Mixture Design Threshold Accepting (MDTA) algorithm, which applies the idea of the Threshold Accepting to generate the corresponding approximate D-optimal designs for a wide range of mixture models, with or without constraints imposed on the components. The MDTA algorithm is tested by many of common mixture models, among which some even have no solutions of the D-optimal design available in the literature. Other tests include 5 models with specific upper bound constraints. These results prove that the MDTA algorithm is very efficient in finding D-optimal designs for mixture models. In some scenarios it even outperforms the state-of-art algorithms, such as the ProjPSO algorithm and the REX algorithm. The source codes of the MDTA algorithm are freely available by writing to the first author.

本文提出了一种目标专用元启发式优化算法，称为混合设计阈值接受 (MDTA) 算法，该算法应用阈值接受的思想为各种混合模型生成相应的近似D最优设计，无论有无对组件施加的约束。MDTA 算法经过了许多常见的混合模型的测试，其中一些甚至没有文献中可用的D最优设计的解决方案。其他测试包括 5 个具有特定上限约束的模型。这些结果证明MDTA算法在寻找D时非常有效- 混合模型的最佳设计。在某些情况下，它甚至优于最先进的算法，例如 ProjPSO 算法和 REX 算法。MDTA 算法的源代码可通过写信给第一作者免费获得。