学术报告第201750期

报告题目：From Mixed-Integer Optimization to Optimal Control: General Introduction and New Results

报告人：Vassilios Vassiliadis（Cambridge University）

报告时间：2017年9月25日 上午8:00—12:00

报告地点：南一楼中311

报告摘要：

Topic1

Title: Constrained NLP via Gradient Flow Penalty Continuation: Towards Self-Tuning Robust Penalty Schemes

Summary: This lecture presents recently published work on the development of a new algorithm for the solution of Nonlinear Programming Problems (constrained optimization). It explores an old idea for minimizing unconstrained functions, that of “Gradient Flow” methods, and generalises it via a penalty-multiplier scheme for generally constrained nonlinear problems.

Topic2

Title: Linear and Quadratic Programming and Interior Point Methods: General Introduction and New Results

Summary: This lecture presents an exploration of high-order sensitivity methods to produce extrapolations of solutions for interior point methods for nonlinear optimization. The methodology based on solving fully one interior point problem for fixed barrier parameter and then introduces an extrapolation scheme to the limit case where the barrier parameter driven to 'zero'. It is shown that without further Jacobian evaluations it is possible to obtain good solutions to select case studies, and then use further refinement to achieve a solution of great accuracy.

Topic3

Title: Cleaning Scheduling of Heat Exchanger Networks: From Mixed-Integer Optimization to Optimal Control

Summary: This lecture presents a novel approach for the solution of Heat Exchanger Cleaning scheduling. The cleaning scheduling of heat exchangers plays an important role in chemical industry, and belongs to the class of Mixed-Integer Nonlinear Programming problems. Through a novel demonstration, it shown that these problems also belong to a class of optimal control problems that allows the calculation of locally optimal integer solutions very efficiently especially when parallelization is used.

Topic4

Title: Beyond Newton: Arbitrary Order Solution Methods– From Systems of Nonlinear Equations to Nonlinear Optimisation

Summary: This lecture presents a theoretical discussion on higher order solution methods for nonlinear systems of equations, via the use of Taylor series expansions, to address difficult to solve large-scale systems. These nonlinear systems can either arise from simulation models, or result from the necessary conditions of optimality of nonlinear optimization problems. In the latter case, particular attention given to the case of interior point methods. Higher order expansions in Taylor series would normally also require calculation of high order derivative tensors, and this is avoided by a suitable reformulation of the original system via function factorisation.

报告人简介：Professor Vassilios Vassiliadis is the University Senior Lecturer (tenured) from Cambridge University, the curriculum under his teaching includes Process Dynamics and Control, Optimization, Process Synthesis, Engineering Mathematics. Received his Ph.D. degree from Imperial College of Science. He is the Founder and director of Cambridge Simulation Solutions LTD from 2014. He has published more than 70 articles in the journals of Computers and Chemical Engineering, Numerical Linear Algebra with Applications, Applied Energy, etc.