Refine
Year of publication
- 2009 (35) (remove)
Document Type
- Report (35) (remove)
Has Fulltext
- yes (35)
Keywords
Faculty / Organisational entity
Radiotherapy is one of the major forms in cancer treatment. The patient is irradiated with high-energetic photons or charged particles with the primary goal of delivering sufficiently high doses to the tumor tissue while simultaneously sparing the surrounding healthy tissue. The inverse search for the treatment plan giving the desired dose distribution is done by means of numerical optimization [11, Chapters 3-5]. For this purpose, the aspects of dose quality in the tissue are modeled as criterion functions, whose mathematical properties also affect the type of the corresponding optimization problem. Clinical practice makes frequent use of criteria that incorporate volumetric and spatial information about the shape of the dose distribution. The resulting optimization problems are of global type by empirical knowledge and typically computed with generic global solver concepts, see for example [16]. The development of good global solvers to compute radiotherapy optimization problems is an important topic of research in this application, however, the structural properties of the underlying criterion functions are typically not taken into account in this context.
Four aspects are important in the design of hydraulic lters. We distinguish between two cost factors and two performance factors. Regarding performance, filter eciencynd lter capacity are of interest. Regarding cost, there are production considerations such as spatial restrictions, material cost and the cost of manufacturing the lter. The second type of cost is the operation cost, namely the pressure drop. Albeit simulations should and will ultimately deal with all 4 aspects, for the moment our work is focused on cost. The PleatGeo Module generates three-dimensional computer models of a single pleat of a hydraulic lter interactively. PleatDict computes the pressure drop that will result for the particular design by direct numerical simulation. The evaluation of a new pleat design takes only a few hours on a standard PC compared to days or weeks used for manufacturing and testing a new prototype of a hydraulic lter. The design parameters are the shape of the pleat, the permeabilities of one or several layers of lter media and the geometry of a supporting netting structure that is used to keep the out ow area open. Besides the underlying structure generation and CFD technology, we present some trends regarding the dependence of pressure drop on design parameters that can serve as guide lines for the design of hydraulic lters. Compared to earlier two-dimensional models, the three-dimensional models can include a support structure.
We prove a general monotonicity result about Nash flows in directed networks and use it for the design of truthful mechanisms in the setting where each edge of the network is controlled by a different selfish agent, who incurs costs when her edge is used. The costs for each edge are assumed to be linear in the load on the edge. To compensate for these costs, the agents impose tolls for the usage of edges. When nonatomic selfish network users choose their paths through the network independently and each user tries to minimize a weighted sum of her latency and the toll she has to pay to the edges, a Nash flow is obtained. Our monotonicity result implies that the load on an edge in this setting can not increase when the toll on the edge is increased, so the assignment of load to the edges by a Nash flow yields a monotone algorithm. By a well-known result, the monotonicity of the algorithm then allows us to design truthful mechanisms based on the load assignment by Nash flows. Moreover, we consider a mechanism design setting with two-parameter agents, which is a generalization of the case of one-parameter agents considered in a seminal paper of Archer and Tardos. While the private data of an agent in the one-parameter case consists of a single nonnegative real number specifying the agent's cost per unit of load assigned to her, the private data of a two-parameter agent consists of a pair of nonnegative real numbers, where the first one specifies the cost of the agent per unit load as in the one-parameter case, and the second one specifies a fixed cost, which the agent incurs independently of the load assignment. We give a complete characterization of the set of output functions that can be turned into truthful mechanisms for two-parameter agents. Namely, we prove that an output function for the two-parameter setting can be turned into a truthful mechanism if and only if the load assigned to every agent is nonincreasing in the agent's bid for her per unit cost and, for almost all fixed bids for the agent's per unit cost, the load assigned to her is independent of the agent's bid for her fixed cost. When the load assigned to an agent is continuous in the agent's bid for her per unit cost, it must be completely independent of the agent's bid for her fixed cost. These results motivate our choice of linear cost functions without fixed costs for the edges in the selfish routing setting, but the results also seem to be interesting in the context of algorithmic mechanism design themselves.
We present a parsimonious multi-asset Heston model. All single-asset submodels follow the well-known Heston dynamics and their parameters are typically calibrated on implied market volatilities. We focus on the calibration of the correlation structure between the single-asset marginals in the absence of sucient liquid cross-asset option price data. The presented model is parsimonious in the sense that d(d􀀀1)=2 asset-asset cross-correlations are required for a d-asset Heston model. In order to calibrate the model, we present two general setups corresponding to relevant practical situations: (1) when the empirical cross-asset correlations in the risk neutral world are given by the user and we need to calibrate the correlations between the driving Brownian motions or (2) when they have to be estimated from the historical time series. The theoretical background, including the ergodicity of the multidimensional CIR process, for the proposed estimators is also studied.
Forderungen nach kürzeren Entwicklungszyklen bei gleichzeitig höherer Produktqualität führen in allen Bereichen der Nutzfahrzeugtechnik und insbesondere auch bei Baumaschinen zum verstärkten Einsatz von Simulationssoftware. Um in diesem Sinne Lebensdauerberechnungen durchführen zu können, sind jedoch genaue Kenntnisse über die im Kundeneinsatz auftretenden Betriebslasten und Beanspruchungen erforderlich. Für deren Ermittlung hat der Baumaschinenhersteller VOLVO Construction Equipment einen Mobilbagger umfassend mit Messtechnik ausgestattet, die neben den mechanischen Belastungen an der Arbeitsausrüstung auch wesentliche Kenndaten des Hydrauliksystems und des Fahrantriebs erfasst. Dieser Messbagger wurde bereits bei unterschiedlichen Kunden in Europa eingesetzt. Der Artikel beschreibt die methodische Vorgehensweise zur Verarbeitung der erfassten Daten und zur Generierung von repräsentativen Nutzungsprofilen am Beispiel der mechanischen Belastungen an der Arbeitseinrichtung, die im Wesentlichen vom Fraunhofer Institut für Techno- und Wirtschaftsmathematik (ITWM) erarbeitet wurde.
Bei der Erprobung sicherheitsrelevanter Bauteile von Nutzfahrzeugen steht man vor der Aufgabe, die sehr vielfältige Belastung durch die Kunden abschätzen zu müssen und daraus ein Prüfprogramm für die Bauteile abzuleiten, das mehreren gegenläufigen Anforderungen gerecht werden muss: Das Programm muss scharf genug sein, damit bei erfolgreicher Prüfung ein Ausfall im Feld im Rahmen eines bestimmungsgemäßen Gebrauchs ausgeschlossen werden kann, es soll aber nicht zu einer Überdimensionierung der Bauteile führen, und es soll mit relativ wenigen Bauteilversuchen eine ausreichende Aussagesicherheit erreicht werden. Wegen der hohen Anforderungen bzgl. Sicherheit müssen bei der klassischen statistischen Vorgehensweise – Schätzen der Verteilung der Kundenbeanspruchung aus Messdaten, Schätzen der Verteilung der Bauteilfestigkeit aus Versuchsergebnissen und Ableiten einer Ausfallwahrscheinlichkeit – die Verteilungen in den extremen Rändern bekannt sein. Dazu reicht aber das Datenmaterial in der Regel bei weitem nicht aus. Bei der klassischen „empirischen“ Vorgehensweise werden Kennwerte der Beanspruchung und der Festigkeit verglichen und ein ausreichender Sicherheitsabstand gefordert. Das hier vorgeschlagene Verfahren kombiniert beide Methoden, setzt dabei die Möglichkeiten der statistischen Modellierung soweit aufgrund der Datenlage vertretbar ein und ergänzt die Ergebnisse durch empirisch begründete Sicherheitsfaktoren. Dabei werden bei der Lastfestlegung die im Versuch vorhandenen Möglichkeiten berücksichtigt. Hauptvorteile dieses Verfahrens sind a) die Transparenz bzgl. der mit statistischen Mitteln erreichbaren Aussagen und des Zusammenspiels zwischen Lastermittlung und Versuch und b) die Möglichkeit durch entsprechenden Aufwand bei Messungen und Erprobung die empirischen zugunsten der statistischen Anteile zu reduzieren.
In the ground vehicle industry it is often an important task to simulate full vehicle models based on the wheel forces and moments, which have been measured during driving over certain roads with a prototype vehicle. The models are described by a system of differential algebraic equations (DAE) or ordinary differential equations (ODE). The goal of the simulation is to derive section forces at certain components for a durability assessment. In contrast to handling simulations, which are performed including more or less complex tyre models, a driver model, and a digital road profile, the models we use here usually do not contain the tyres or a driver model. Instead, the measured wheel forces are used for excitation of the unconstrained model. This can be difficult due to noise in the input data, which leads to an undesired drift of the vehicle model in the simulation.
One approach to multi-criteria IMRT planning is to automatically calculate a data set of Pareto-optimal plans for a given planning problem in a first phase, and then interactively explore the solution space and decide for the clinically best treatment plan in a second phase. The challenge of computing the plan data set is to assure that all clinically meaningful plans are covered and that as many as possible clinically irrelevant plans are excluded to keep computation times within reasonable limits. In this work, we focus on the approximation of the clinically relevant part of the Pareto surface, the process that consititutes the first phase. It is possible that two plans on the Parteto surface have a very small, clinically insignificant difference in one criterion and a significant difference in one other criterion. For such cases, only the plan that is clinically clearly superior should be included into the data set. To achieve this during the Pareto surface approximation, we propose to introduce bounds that restrict the relative quality between plans, so called tradeoff bounds. We show how to integrate these trade-off bounds into the approximation scheme and study their effects.