1 SCOMA Scientific Computing and Optimization in Multidisciplinary Applications Prof. Pekka Neittaanmäki University of Jyväskylä Deparment of Mathematical Information Technology
2 Main areas of the research Reliable computing Robust methods for data driving Hybrid optimization methods Effective solvers for large linear algebraic systems
3 Milestones Research in scientific computing and optimization has been systematically developed at the University of Jyväskylä during the last 15 years. The activities started 1988 in numerical analysis of partial differential equations, and were later extended to calculus of variations, optimal control, optimization, and their numerical methods and efficient implementations on different computer platforms.
4 SCOMA Center Aims of the SCOMA Center: to combine multi- and interdisciplinary applications in industry, economics and finances to link with worldwide scientific computing centers and collaborate with small and large industries on R&D projects to develop multidisciplinary methods and tools and share knowledge with small and large industries
5 SCOMA Center to host visits by international scientists to the Center to leverage expertise in Jyväskylä for technologies and research into interdisciplinary engineering and societal applications to host master's degree and PhD programs in scientific computing and optimization to arrange the annual international Jyväskylä Summer School together with the Faculty of Mathematics and Science and the Faculty of Information Technology
6 SCOMA (Tekes project) SCOMA - Scientific Computing and Optimization in Multidisciplinary Applications Responsible Leader: Prof. Pekka Neittaanmäki Project manager: M.Sc. Tero Tuovinen Budget: /year Partners: TEKES, JYU, ABB Ab Corporate Research, Comsol Oy, CSC Scientific Computing, Metso Paper Oy, Moventas Oy, Patria Aerostructures Oy, TeliaSonera Oy, City of Jyväskylä, Research Institute of Finnish Economy (ETLA)
7 Work Package 1a Fluid-structure interactions / Simulation of paper web dynamics Goal Mathematical modelling and simulation of coupled, time-dependent moving web-air behaviour in paper making process Group: M.Sc. Tero Tuovinen, Prof. Raino A. E. Mäkinen, Prof. Pekka Neittaanmäki, M.Sc. Sanna Mönkölä Collaboration: Matti Kurki (Metso Paper/Metsä Botnia) Sakari Lukkarinen (Comsol) Johan Hoffman (KTH) Wulf G. Dettmer (Univ. Of Wales Swansea)
8 Work Package 1b Reliable finite element analysis of gearboxes / Design of gearboxes The accuracy of the finite element solutions plays crucial role in the simulation of industrial processes and in this case, especially in the stress-strain analysis of gearboxes parts, especially for the purposes of the fracture and fatique analysis. It is well known that only mesh-adaptive methods based on a posteriori error indicators can provide an efficient modelling able to give reliable solutions with minimal expenditures. Goal update information of new tools for industrial simulation design. resulting testcases: Error checker, new algorithm, methods. Group: Prof. Pekka Neittaanmäki, Prof. Sergey Repin, Petri Lahtinen (Moventas), Pavlo Turchyn, Elena Gorshkova, Maxim Frolov Collaborators: Moventas Oy.
9 Work Package 1b Results: A new posteriori error indicator is developed in order to estimate the accuracy of approximate solutions of fluid flows as well as the construction certain parts of the gearboxes. On the basis of this approach it was developed a computer code. It was tested on a series of model problems related to problems in the mechanics of solids and fluids and on 3D objects offered by the collaborator that present parts of a gearbox. Experiments have shown high efficiency of the method. Theses: A PhD thesis entitled "Adaptive meshes in computer graphics and model--based simulation" by a project member P. Tuchin is presented for the defence procedure December 28, 2006.
10 Work Package 1c Simulation of electromagnetic scattering by paper, paper coating and coated paper Goals are to Group: Prof. Timo Tiihonen, Prof. Tuomo Rossi, Kari Saarinen (ABB), Jukka Räbinä Collaboration: ABB, Roland Glowinski (Houston)
11 Work Package 1c Results: We have studied methods to optimize Monte Carlo simulations of light scattering We have done some radiative transfer simulations using Mie scattering We have studied how packing density affects on light scattering Theses: Master s thesis Simulating light scattering with Monte Carlo method University of Jyväskylä, 13 October 2006, Jukka Räbinä
12 Work Package 1d Simulation of quantum waveguides The research is focused on deriving and studying mathematical models of electron or/and phonon transport in two-and three-dimensional quantum waveguides and nanostructures. The models will be investigated by means of numerical simulation and asymptotic analysis. The results are needed for the development of new devices in nano-and microelectronics (quantum switches of new type, high sensitivity bolometers, optical sensors for non-contact high current and voltage measurements, cold cathodes for vacuum microelectronics) Goal Numerical study of electron transport in branching waveguides. Applications to the control of electron beams. Asymptotic and numerical studying resonant tunneling in quantum waveguides (elaboration of methods and implementations). Numerical investigation of phonon transport in deformed nanofibers (elaboration of methods and implementations). Applications to the control of thermal conductance of nanofibers.
13 Work Package 1d Group: Prof. Pekka Neittaanmäki, Panu Koppinen (Dept of Physics, JyU), M.Sc. Elias Markkanen, M.Sc. Tuomas Puurtinen, Prof. Boris Plamenevskii and Prof. Lev Baskin (St. Petersburg) Results: Numerical study of electron beams in quantum branching waveguides. Devising a switch for electron beams (application for patent). Asymptotic theory of resonant tunneling in two dimensional waveguides with narrows (in preparation for publication). Method of numerical analysis of phonon transport in nanofibers. Method of the control of thermal conductance of nanofibers (application for patent)
14 Work Package 2a Data mining Goal: develop efficient, reliable, and automated methods for clustering and classification develop methods for projection and understandable visualization of obtained clusters and classes Group: Prof. Tommi Kärkkäinen PhD Napsu Haarala PhD Kirsi Majava PhD Sami Äyrämö Paavo Nieminen (M.Sc. at end-of-2006) Collaboration: R. Glowinski (Houston) K. Kunisch (Graz)
15 Work Package 2a Results: Working towards prototype implementations of robust and efficient methods for clustering and neural network regression analysis with noisy and incomplete data: An early prototype of the neural network tool gives promising results in preliminary testing (illustrated on the following slide) Documentation and comparative analysis of the prototype implementation is to be produced. Theses: Doctoral thesis Knowledge mining using robust clustering, Sami Äyrämö, 2006 Publications Äyrämö, S. and Kärkkäinen, T. Introduction to partitioning-based clustering methods with a robust example, Reports of the Dept. of Math. Inf. Tech. (Series C. Software and Computational Engineering), University of Jyväskylä, 2006 Kärkkäinen, T. and Äyrämö, S., On Computation of Spatial Median for Robust Data Mining, In Proceedings of Sixth Conference on Evolutionary and Deterministic Methods for Design, Optimisation and Control with Applications to Industrial and Societal Problems (EUROGEN 2005), Munich, September 12-14, 2005 Kärkkäinen, T. and Äyrämö, S., Robust Clustering Methods for Incomplete and Erroneous Data, in Data Mining V: Data Mining, Text Mining and their Business Applications, 2004
16 Work Package 2b Customer Relation Management (CRM) and Profiling using Robust Data Mining Methods Goal: Efficient use of large datasets that describe customer behaviour Reliable and scalable computational methods to reveal intrinsic groups of customers, e.g., for directing marketing campaigns for right customers Convenient visualizations of CRM data Apply the methodology developed as part of strategic application area research to meet the above goals Group: Prof. Tommi Kärkkäinen PhD Sami Äyrämö
17 Work Package 3a Weight optimized composite structures for aerospace applications Goal design composite structures (or combined composite/metal structures) in aerospace applications minimize weight with performance constraints augment the FE analysis with shape optimization as well as integration of FE and optimization tools with CAD system Group: Prof. Pekka Neittaanmäki, Prof. Raino A. E. Mäkinen, Petri Hepola (Patria Aerostructures), Prof. Jacques Périaux (Paris), Lic. Timo Aittokoski, B.Sc. Pasi Saari, Prof. Arthur Rizzi (KTH) Collaboration: Patria Aerostructures
18 Work Package 3a Results: New CAD/FEM interface software is available Theses: Ph. Lic. thesis made by Timo Aittokoski: On optimization of simulation based design (supported by project of Academy of Finland)
19 Work Package 3b Using Catia and Capri in modeling process Goal Research on the improvements that CAPRI offers for the analysis and optimal design processes that are currently in use Implementation of CAPRI to these processes using the Catia CAD-system integrated with FEM and optimization packages Stydy of an open source and "free" software that can be used in conjunction with CAPRI Group: Prof. Pekka Neittaanmäki, Prof. Jacques Periaux, Prof. Dan Tiba, B.Sc. Santtu Salmi, B.Sc. Pasi Saari Collaboration: Robert Haimes, MIT (CAPrI), Prof. Arthur Rizzi (KTH)
20 Work Package 4a Control and management of product life cycle The goal of research is to construct mathematical methods for optimization of investment planning in process industry (paper, energy) and service planning based on game theory approach.
21 Work Package 4a Group: M.Dementieva, P.Neittaanmäki, V. Zakharov Collaboration: Metso Paper Results: New basic model of 3-player conflict with two leaders and one follower. Two leaders move simultaneously and independently, and then the follower chooses its action as an answer to the leaders strategies The model includes: extended form game with stochastically defined payoffs
22 Work Package 4b Optimization and control of nonlinear equilibrium models in economics Goal: trace the numerical difficulties related to FOG (Finnish Overlapping Generations) model developed by ETLA for analyzing the effects of the used economical and social politics in Finland Group: Sauli Ruuska, prof. Marko M. Mäkelä (Univ. of Turku) Collaboration: Eija Kauppi (Etla), Jukka Lassila (Etla), Tarmo Valkonen (Etla)
23 Work Package 4b Results: system has been split into subsystems subsytems has been analyzed in more details possible sources of numerical instabilities has been determined operatios for improvement has been proposed Theses: B.Sc. theses by Sauli Ruuska will be completed soon
24 Work Package 4c New models, Methods, and Simulations for QoS and Pricing in Next Generation Internet Goal Evaluation of market segments and their testing with real data from customers. Developing of suitable parameters for segmentation of CRM system. Developing of criteria for segmentation regognation. Group: Isto Kannisto, prof. Jyrki Joutsensalo, prof. Timo Hämäläinen, Tapio Väärämäki Collaboration: TeliaSonera
25 Work Package 4d Transport Management Technology for Large-Scale Applications Goal: Optimization of public service logistics, e.g. Group: City of Jyväskylä's meal service for elderly Home care services Other transportation problems O. Bräysy, P. Nakari
26 Work Package 4d Results: Meal service routes optimized Over 50% shorter total route length in test optimization
27 Work Package 4d Future plans: City of Jyväskylä's home care Transportation of elderly Possibly material transports, street maintenance, internal post, school transportation etc.
28 Work Package 4d
29 Work Package 4e SCOMA Semantic Web portal Goal To establish an ICT infrastructure for collaborative work of SCOMA partners, sharing and discovering resources, searching people, organizations and groups with needed scientific profiles Apply Semantic Web technology for enhanced structuring, reuse and integration of information and knowledge management Group:Industrial Ontologies Group (IOG) Vagan Terziyan Sergiy Nikitin
30 Work Package 4e Results: Alpha-version of SCOMA web-based application with semantics-enabled search engine Classification of mathematical terms transformed into ontology 3 men/months of resources spent so far (all in 2005) See more in:
31 Work Package 4e Future plans: Wiki-based implementation of a collaborative SCOMA environment Extended with Semantic Web technology for efficient searching and interlinking of the information
32 Work Package 4e SCOMA community portal