Mixed-integer Programming Representations of Linear Model Decision Tree Surrogates
Explore the source record for details and available documents.
Ammari, Bashar↗
Engineering topics
Kim, Taehun
Publications and source records attributed to Kim, Taehun.
Bond-dependent anisotropy and magnon decay in cobalt-based Kitaev triangular antiferromagnet
Following tremendous success of the graphene-derived fundamental and applied research, the magnetic van der Waals materials that can be cleaved into monolayer two-dimensional atomic crystals have emerged as a new platform in the studies of low-dimensional physics and in the design of artificial heterostructures with novel properties. Because of that, the family of dihalides and trihalides of the 3d transition group receive a strong renewed interest. Together with our experimental collaborators, we have investigated the novel dynamical properties of one of such promising material CoI 2 , and demonstrated that spin excitations in it are prone to substantial breakdown and complex level repulsion, the important quantum effects whose theoretical understanding has been significantly advanced by our group in the last two decades. Both phenomena are dramatically revealed by experiments and verified by the theory in our joint study, published in Nature Physics. Altogether, our results pave the way toward a new research direction for the magnetic van der Waals materials.
36 MATERIALS SCIENCE↗
Linear model decision trees as surrogates in optimization of engineering applications
Machine learning models are promising as surrogates in optimization when replacing difficult to solve equations or black-box type models. This work demonstrates the viability of linear model decision trees as piecewise-linear surrogates in decision-making problems. Linear model decision trees can be represented exactly in mixed-integer linear programming (MILP) and mixed-integer quadratic constrained programming (MIQCP) formulations. Furthermore, they can represent discontinuous functions, bringing advantages over neural networks in some cases. We present several formulations using transformations from Generalized Disjunctive Programming (GDP) formulations and modifications of MILP formulations for gradient boosted decision trees (GBDT). We then compare the computational performance of these different MILP and MIQCP representations in an optimization problem and illustrate their use on engineering applications. Importantly, we observe faster solution times for optimization problems with linear model decision tree surrogates when compared with GBDT surrogates using the Optimization and Machine Learning Toolkit (OMLT).
42 ENGINEERING↗
Optimization in Engineering with Embedded Linear Model Decision Trees
Explore the source record for details and available documents.
Ammari, Bashar↗
Dynamic modeling and simulation of pressure swing adsorption processes using toPSAil
Pressure swing adsorption (PSA) has attracted significant recent interest for chemical process intensification due to its potential for high energy efficiency and amenability to small, modular designs. However, the lack of simulation tools that are readily available, transparent, and trusted, has been identified as a serious impediment to widespread adoption of PSA, as well as to further research on PSA modeling, numerical solution, optimization, and control. This paper presents a complete framework for dynamic modeling and simulation of PSA processes and its implementation in an open-source simulator called toPSAil. Further, the presentation is tutorial and includes many modeling and implementation details often overlooked in existing literature. Novel methods are presented for handling flow reversals and implementing various pressure–flow relationships, along with controlled boundary conditions. Finally, the code contains several innovations designed to improve efficiency and reduce the extensive trial-and-error tuning often required to produce a working PSA cycle.
37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗