## Optimization in Engineering

The EDOM group has a rich history of collaborations with engineers. Projects involving all kinds of engineering questions. We are specifically interested in to bring methods from discrete optimization to application in fields where they are not well established: Most engineering models are modeled with a well-established simulation model in the background. Discrete parameters are then often introduced via trial-and-error or heuristic methods. We have shown in various applications that it is possible to produce guaranteed globally optimal solutions in such cases.

For further information, contact Lars Schewe (lars.schewe[at]math.uni-erlangen.de).

## Current Projects

## Operational management of water supply networksWater supply networks are located in almost every inhabited place around the world. An operational management of such networks could be an extremely costly task. Finding good or even optimal (in the sense of costs and/or w.r.t environmental aspects) settings for the operational management of water supply networks is unavoidable in the near future. Contact: Antonio Morsi | |

## Thermal Sewage Sludge Recovery - Process Efficency Information and Optimization SystemWe develop an open loop optimization for several timesteps and a state estimation of the process to utilize it in a model predictive control loop for a wastewater treatment plant. Contact: Thorsten Gellermann |

## Finished Projects

## LeOpIn - Life-cycle oriented optimization for a resource- and energy-efficient infrastructureThe goal of the project LeOpIn is to devise methods for life-cycle oriented planning and evaluation of buildings and related infrastructure. To this end we develop simulation and optimization tools which can cope with this task. A concrete application is the planning of a building and pipes undergoing high pressure scenarios for which a software solution will be prototypically developed. The development of the planning and evaluation procedures demands a tight interaction between mathematical and engineering techniques. We plan to employ methods of numerical simulation and of discrete and nonlinear optimization. The main focus lies on integration of these techniques since only by this the high complexity of the treated problems can be handled appropriately. Contact: Alexander Martin and Stefan Schmieder (Project A) or Lars Schewe and Jakob Schelbert (Project B) | |

## Optimal combination of active and passive parts in load carrying systemsMechanical trusses are found in many applications (undercarriages of airplanes, bicycles, electrical towers, etc.). Those trusses are often overdimensioned to withstand given forces under several uncertainties in loadings, material, production processes, etc. Active parts (e.g. piezo-elements) can react on these uncertain effects and reduce the dimension of trusses. CRC 805 introduces new technologies to handle uncertainty in load carrying systems. Mathematically, this leads to a mixed-integer semi-definite programming problem. Contact: Sonja Mars or Kai Habermehl (TU Darmstadt)
| |

## Commanding Uncertainties in Systems of Mechanical EngineeringDepending on the level of abstraction, various processing chains with uncertainties occur within this project. Here, we start with a mathematical model for process chains and their inherent uncertainties with the help of a distribution over a set of random envents or scenarios. We follow two approaches. One deals with very large structured mixed integer programs as known from stochastic programming. The other one deals with PSPACE-complete problems like Quantified Integer Programs or some special classes of the Dynamic Graph Reliability problem. Contact: Alexander Martin, Ulf Lorenz (TU Darmstadt), Nicole Ziems
| |

## Collaborative Research Centre 666: Integral sheet metal design with higher order bifurcations - Development, Production, EvaluationBranches are found in a lot of sheet metal products. The CRC666 studies a very new technique, called "Linear flow splitting", to create integral sheet metal products with branches. Mathematical Optimization is the emphasis of two subprojects. The aim of the first one is the optimization of the topological structure and the geometry of the products, using Discrete and Nonlinear Optimization. The second one is engaged to find an optimal procedure to produce the product subject to manufacturing restrictions. Here Discrete Optimization and Graph Theory plays an important role Contact: Wolfgang Hess (TU Darmstadt) or Ute Günther
| |

## Layout and Process Optimization in Recovered Paper ProductionRecovered paper nowerdays is the most important resource in the paper production. One trouble in the paper manufactoring process are production losses due to tacky particles called stickies. The aim of this project is the simultaneous optimization of the layout and setting of a machine cleaning the waste paper suspension from these stickies. The separation process itself can be descibed by nonconvex functions. Including the layout decision then results in a mixed-integer nonlinear problem (MINLP). We solve the model using piecewise linear approximations of the nonlinear functions. Contact: Alexander Martin, Christine Hayn, Armin Fügenschuh (ZIB)
| |

## Examination of the rearrangeable blocking behaviour of Clos-Networks with respect to multicast trafficClos networks have been invented by Charles Clos in 1953 in order to reduce network size and cost. A Clos network is composed from smaller crossbars, which are arranged in three stages. A multicast connection request is an arbitrary partial function from the set of outputs to the set of inputs of a network. To this day, no formula is known in order to determine the minimal number of middle stage crossbars such that in the associated Clos network any multicast connection request is routable. We have computationally determined this number for a set of small Clos networks, yielding significant cost reductions compared to trivial switching networks. Contact: Peter Lietz, Alexander Martin | |

## Optimal Grid Partitioning for Block Structured GridsIn order for mesh-based scientific simulations to be efficiently executed on parallel hardware it is inevitable to distribute the mesh elements among the available processors such that the computational load is balanced and the time required for interprocessor communication is minimized. In this project we develop a new way to reflect hardware restrictions in communication schedules by setting up (and solving) a new integer programming formulation of the graph partitioning problem. Contact: Alexander Martin |