Browsing by Author "Nie, Tiantian"

Now showing 1 - 2 of 2

An enhanced logarithmic method for signomial programming with discrete variables
(2016-12-16)
Li, Han-Lin
;
Fang, Shu-Cherng
;
Huang, Yao-Huei
;
Nie, Tiantian
;
資訊管理與財務金融系註：原資管所+財金所
Signomial programming problems with discrete variables (SPD) appear widely in real-life applications, but they are hard to solve. This paper proposes an enhanced logarithmic method to reformulate the SPD problem as a mixed 0-1 linear program (MILP) with a minimum number of binary variables and inequality constraints. Both of the theoretical analysis and numerical results strongly support its superior performance to other state-of-the-art linearization methods. We also extend the proposed method to linearize some more complicated problems involving product and fractional terms in discrete and continuous variables. (C) 2016 Elsevier B.V. All rights reserved.
Enhanced linear reformulation for engineering optimization models with discrete and bounded continuous variables
(2018-06-01)
An, Qi
;
Fang, Shu-Cherng
;
Li, Han -Lin
;
Nie, Tiantian
;
資訊管理與財務金融系註：原資管所+財金所
In this paper, we significantly extend the applicability of state-of-the-art ELDP (equations for linearizing discrete product terms) method by providing a new linearization to handle more complicated non-linear terms involving both of discrete and bounded continuous variables. A general class of "representable programming problems" is formally proposed for a much wider range of engineering applications. Moreover, by exploiting the logarithmic feature embedded in the discrete structure, we present an enhanced linear reformulation model which requires half an order fewer equations than the original ELDP. Computational experiments on various engineering design problems support the superior computational efficiency of the proposed linearization reformulation in solving engineering optimization problems with discrete and bounded continuous variables. (C) 2017 Elsevier Inc. All rights reserved.