International conference on high performance scientific computing: Modelling, simulation and optimization of complex processes. Hanoi, 2003

A Step by Step Evolution of Protein Structures

The paper presented the results for two of the well known proteins viz Villin and Crambin. The predicted structures have about 6.5 Angstrom RMSD from the experimental ones and is much better than the sequential genetic algorithms. The overall topology is quite similar to the experimental one. The results showed that, with limited inputs of force field and sequence this is a notable achievement and may be useful for the proteins which cannot be syntheized easily. An improved algorithm is under development by combining the island model and multi-objective functions.

A general-purpose finite element method for 3D line transfer problems with application to galaxies in the early Universe

The research aims at the modelling of three-dimensional radiation fields in gas clouds from the early universe, in particular as to the influence of varying distributions of density and velocity. In observations of high-redshift gas clouds, the Ly-anpha transition from the first excited energy level to the ground state of the hydrogen atom is usually found to be the only prominent emission line in the entire spectrum. It is a well-known assumption that highredshifted hydrogen clouds are the precursors of present-day galaxies. Thus, the investigation of the Ly-anpha line is of paramount importance for the theory of galaxy formation and evolution. The observed Ly-anpha line - or rather, to be precise, its profile - reveals both the complexity of the spatial distribution and of the kinematics of the interstellar gas, and also the nature of the photon source.

A new approach to the simulation of flash floods in tropical humid monsoon catchments

In this paper a new approach to the simulation of flash floods in tropical humid monsoon regions is developed. The model for flash flood simulation is conceived as a deterministic rainfallrunoff model. The determination and development of a theoretical time-area function for the flow concentration in a catchment model, which is set up mainly by means. of topographical information (GIS), is emphasized in this paper. The determination of the time-area function is derived as a mathematical function from the rising limb of a historical flood hydrograph. The inflection point of this part of the hydrograph can be identified topographically with a specific route through the main stream. The temporal and spatial flood routing can be combined. Since the approach is based on topographically defined features. It requires no extensive hydrological measurements and can be verified simply. The deterministic approach planned deterministic is to a large extent scale independent and regional applicable. The method was developed on flash floods in catchments of Vietnam.

A numerical method for optimal error estimates in boundary value problems (BVPs) and robust optimization

The paper introduced a numerical solution method for general semi-infinite optimization problems with convex lower level problems admits a large number of applications, among them the determination of optimal error bounds in the defect minimization method for elliptic boundary value problems, and the solution of robust optimization problems. The method bases upon a reformulation of the semi-infinite problem as a Stackelberg game and the use of regularized NCP functions. This approach leads to central path conditions for the lower level problems, where for a given path parameter a smooth non-linear finite optimization problem has to be solved. The solution of the semi-infinite optimization problem then amounts to driving the path parameter to zero. The author given convergence properties of the method and illustrate it with a number of numerical examples. The method is easy to implement, and the examples show that it works well for high dimensional index sets.

A posteriori error analysis and adaptive computation for wave scattering by periodic structures

The authors developed a finite element adaptive strategy with error control for the wave scattering by periodic structures. The unbounded computational domain is truncated to a bounded one by an extension of the PML technique which attenuates both the outgoing and evanescent waves in the PML region. The PML parameters such as the thickness ofthe layer and the medium property are determined through sharp a posteriori error estimates. Numerical experiments are included to illustrate the competitive behavior of the proposed adaptive method.

A software package for the optimal operation of continuous moving bed chromatographic processes

This software package includes numerical methods for modeling, parameter estimation, design of experiments and optimal operation of continuous moving bed chromatographic processes. Its application to a production of a high fructose corn syrup in a reactive moving bed process will be demonstrated. The SMB-process is optimized using a sequential approach based on a rigorous mathematical model of the plant. This problem requires formidable efforts and can not be achieved within a fixed sampling time. On the other hand, disturbances and plant/model mismatch can occur and lead to an off-spec product. In order to maintain the product purity while injecting a minimal additional amount of eluent, the authors proposed a modified nonlinear model predictive controller, and calculating a suboptimal but feasible solution and which can be applied in real-time.

A two-stage, high-accuracy, finite element technique of the two dimensional horizontal flow model

An algorithm and essential subroutines programs are presented which implement two stage finite element Galerkin method for integrating the complete two dimensional horizontal flow model. In this method high accuracy is obtained by combining the Galerkin product with a high-order difference approximation to derivatives in the nonlinear advection operator. Program includes the use of a weighted selective lumping scheme in the finite element method, use Gauss-Seidel iterative method for solving the resulting systems of linear equations. Small scale noise was eliminated by using a Shulman filter.

Adaptive FEM for nonlinear problems

This lecture presents an approach towards goal-oriented adaptivity in which the error in a functional of the solution is written in terms of primal and dual residuals. For evaluating the resulting a posteriori error estimate an associated linear dual problem has to be solved. The linearization leaves a remainder term which is cubic in the errors and is neglected in the course of the refinement process. In this way mesh adaptation is oriented at the pre-defined goal of the computation and reaches high computational efficiency. This general approach can be used in the context of boundary value problems, optimal control problems as well as eigenvalue problems.

Adaptive finite elements for output-oriented parameter identification problems

The author considered parameter identification problems involving partial differential equations with finite number of unknown parameters. The quantity of interest (output) depends on the state variable as well as on the parameters. Finite elements on locally refined meshes are employed for discretization of the state equation. He develop an a posteriori error estimator, which aims to control the error in the quantity of interest and is used for the successive improvement of the accuracy by appropriate mesh refinement. Several examples illustrate the behavior of an adaptive mesh refinement algorithm based on error estimator.

Advanced techniques in the column generation method for crew pairing problems

The main idea of the method has two basic steps: (1) solving the linear problem 'for a subset of variables, (2) pricing a subproblem to find out new variables with negative reduced costs. Since the dual variables are not bounded, this leads to instability of the standard form of this method. Therefore, a so-called stabilized column generation is suggested in step (1) to overcome this phenomenon by reducing the number of redundant variables. With a good approach of updating algorithm parameters, the performance could be improved dramatically. The time consuming pricing step (2) is also under consideration with several pricing algorithms which are tuned to adapt to the crew pairing problem. This paper will discuss an approach of using a cluster of computers to solve the problem. With a quite simple implementation, a good speedup is obtained in solving randomly generated crew pairing problems.

An approach to parameter estimation and model selection in differential equations

The estimation problem in parameterized systems of differential equations starts with data acquired through observations of system trajectories made in the presence of noise and seeks to estimate the parameter values by solving an optimization problem which matches solutions of the differential equation to the observed data. Thus, there is an explicit stochastic component to the problem. Typically, two classes of method are considered: 1/ Explicitly computed solution trajectories are compared directly with the observations in an unconstrained optimization procedure; 2/ The system of differential equations is imposed as explicit constraints on the optimization problem and the resulting mathematical program typically is solved by a variant of sequential quadratic programming.

An efficient approach to vague joins in the vague query system

This article presents problems arisen when processing complex vague joins in the Vague Query System (VQS) and introduces a new solution to efficiently solve these problems. Join operation is one of the most expensive ones in database management systems. In context of the VQS, vague joins are prohibitively expensive in terms of IO-cost and CPU-cost because they must undergo intermediate processing steps with the sheer volume of multidimensional data in multiple feature spaces. The new approach not only significantly reduces the costs (IO- and CPU-cost), but also returns to users the best match (or approximate nearest neighbors with a certain tolerant error e) in the query relation/view. Experimental results will show performance of the proposed approach.

An explicit divergence-free ppwinding method for the MHD-equations

Many flows, particularly astrophysical flows, are electrically conducting, and the electromagnetic forces can be of the same order as the hydrodynamic forces. The governing equations of Magnetohydrodynamics (MHD) basically merges the Euler equations of gas dynamics with the Maxwell equations of electromagnetic. A feature of the MHD-equation, in contrast to the Euler system, is the additional constrain div B = 0 on the magnetic field. Based on the method of transport, which will be explained briefly, the author presented an explicit multi-dimensional upwind scheme for the MHD-equations, which conserves divB according to a numerical divergence-operator and whose time step is fully local.

An integral model for calculation of LPG jet development in combustion chamber of spark ignition engine

The present paper introduces some calculation results on distribution of fuel and oxygen concentrations of LPG jet in the combustion chamber of spark ignition engine. The conserved scalar model is developed for species exchange in turbulent diffusion jet. The integral equations system for convection-diffusion process is closed by turbulence k-e model. The model is validated by experimental data of air jet given by LDA method. It is then developed to predict the velocity and concentration fields of LPG jet in combustion chamber in order to develop a new kind of engine: Gas Direct Injection Engine which has double advantages of Gasoline Direct Injection (GDI) engine and gaseous fuel.

An object-oriented approach to specification and composition of web services

In this paper, the authors proposed an objectoriented approach to web services, whereby each service is considered as a class consisting of its functions and all the services are organized into a class hierarchy. With this approach, not only web service functions can be reused, but also developed in a structural way without duplication in different services. Firstly, they presented the main features of our proposed Object-Oriented Web Service Language (OOWSL) for specification of web services, their hierarchy and composition. Then present the operational architecture for the language, regarding processing and monitoring service transaction, logging, timeout, retry, exception and inheritance. An example of OOWSL usage is also implemented and demonstrated.

Application of K-e model for three dimensional simulation of wind field

A 3D model for simulating air flow and dispersion in the atmospheric boundary layer, which is based on three-dimensional non-hydrostatic flow equations and two turbulence closure equations, is described and analyzed. The closed system of equations is solved for mean flow characteristics, turbulent kinetic energy and its dissipation rate. A method for successful solution of two turbulence closure equations together with flow equations is presented. Some validations of the code are also presented in the paper.

Application of vehicle routing problem with hard time window constraints

This paper proposes a heuristic, Tabu-Perturbation Algorithm (TPA), to efficiently and effectively solve Vehicle Routing Problem with Hard Time Window Constraints (VRPHTW). TPA integrates Tabu Search (TS) and Noising Method (NM). A software system is developed based on it and is used in a company that delivers equipment spare parts to the plants manufacturing semiconductor wafers and TFT-LCDs. VRPHTW is NP-hard due to the NP-hardness of Vehicle Routing Problem (VRP). Previous work on VRPHTW includes both optimization algorithms and heuristic approaches, but current research focuses on heuristic approaches. This research proposes a heuristic TPA, integrating TS and NM, to solve VRPHTW. TPA consists of three phases: route construction, local search improvement, and generic search improvement.

Applied stochastic integer programming: scheduling in the processing industries

The approach is conceptually different from those above as much as the possibility of recourse actions is modeled by means of two-stage stochastic integer programs. Deterministic equivalents can be formulated as large MILPs with a block-angular matrix structure; they are solved by a dual decomposition approach based on Lagrangian relaxation of the nonanticipativity constraints. The paper presented a production process from the polymer industries as a real-world benchmark problem. The scheduling task is dominated by the strong coupling of substances and units and the significant uncertainties of demands and the process itself. The focus of the presentation is on the master scheduling problem as part of a two layer cascade-like feedback structure. It is stated as a large-scale multiperiod model with accurately formulated capacity constraints and realistic production goals.

Automatic analysis and evaluation of scarce Jacobian matrices

The paper addressed the question how Jacobains and Hessians should be analyzed and provided most appropriately for the purposes sketched. As it turns out their explicit accumulation as a rectangular array of numbers may be wasteful and destroy structure even if sparsity is taken into account. The author illustrated the new concept of Jacobian scarcity and present some basic theoretical and numerical results concerning partial accumulation procedures for minimizing the costs of matrix-vector products and Newton-step calculations.

Balance algorithm - A new approach to solving the mapping problem on heterogeneous systems

A fundamental issue affecting the performance of a parallel program is the as signment of tasks to processors in order to get the minimum rompletion time. Most of state-of-the-art approaches consider homogeneous MIMD multiprpcessor systems, in which all communication channels have the same bandwidth and all processors are equally powerful. These algorithms do not run efficiently on heterogeneous systems. In this paper, the authors presented a new approach for the mapping problem on arbitrary systems. The main idea is based on the "global load balancing, local cut size optimization" principle. This approach has achieved encouraged results that are verified by experiments for various random graphs and processor numbers.

Balancing cost versus quality in UMTS radio access networks

The Universal Mobile Telecommunications System (UMTS) is a 3rd generation cellular system for mobile telecommunication. UMTS is much more flexible on the radio interface than GSM, supporting data rates ranging from a few bits upto several hundred kilobit per second. The reuse of the available bandwidth for simultaneous and co-located radio communication is enabled by a wide-band COMA technology. As a consequence, UMTS radio cell coverage and capacity are strongly inversely coupled through system selfinterference. Cell capacity can be increased at the expense of coverage and vice versa. The presentation sketches the basic principles of the UMTS radio interface. Models and simulations for the automatic planning and optimization of UMTS radio networks are currently under development in the project MOMENTUM, supported by the European Union under the Information Society Technologies (IST) programme. The presentation will survey the optimization approach taken within MOMENTUM. The focus will be on a mixed linear integer programming model for optimizing location selection and configuration for the UMTS base stations.

Calculating consistent initial values for structurally singular differential-algebraic equation systems

A method for calculating consistent initial values for such a class of differential-algebraic equations (DAE) F(t,y,y') = 0 is presented. Starting from consistent initial values for y, consistent initial values for the derivative y' are determined. The article derived the hidden constraints by differentiating suitable algebraic constraints and combine the result with the original DAE system to form an equation system, which can be solved to obtain initial consistent values for the original system. Automatic differentiation or finite difference techniques are used to evaluate these hidden constraints. Some practical application examples from the mentioned fields are presented.

Clustering algorithms for parallel car-crash simulation analysis

Buckling and certain contact situations cause scattering results of numerical crash simulation: For a BMW model differences between the position of a node in two simulation runs of up to 10 cm were observed, just as a result of round-off differences in the case of parallel computing. An engineer has to measure this scatter, to check whether important parts of the car show such indeterministic behavior and to find the origins. The tool DIFFCRASH compares simulation results and uses data mining technology to cluster those nodes of the car model, which show similar scatter among the simulation runs. For the BMW model the indeterministic behavior could be traced back to a certain part of the motor carrier and was removed by a redesign. In this paper, the authors presented the clustering algorithm and illustrate its usage in car crash simulation analysis.

Comparison of parallel programming models on clusters of SMP nodes

This programming model implied an additional message transfer overhead between the CPUs within a shared memory system (SMP) node. To prohibit this overhead, hybrid programming models are proposed, combining MPI with OpenMP. MPI is used only for the inter-node communication and OpenMP (or automatic parallelization) is used for the parallelization inside of an SMP node. In the most commonly used hybrid model, the MPI routines are called only outside of OpenMP parallel regions, i.e., they are called only from the master threads and only while all other threads in the calling node are sleeping. Although this model avoids massaging overheads inside of the SMP nodes, there are even so serious drawbacks: First, on most platforms, the master thread is not able to use the full bandwidth of the inter-node network. This is shown with different benchmarks on several platforms. Second, if all threads except the master thread are sleeping and dedicated CPUs are used, then all MPI communication must be weighted by the number of threads used in each MPI process.

Computational drug design in the virtual lab

The paper presented the approach of Perron cluster analysis (PCCA) as developed up to now by the author, SCHUTTE, and their co-workers. The key idea of this approach is to directly identify almost invariant sets of the Hamiltonian system as metastable conformations together with their life spans and transition patterns. This leads to the numerical solution of a Perron cluster eigenvalue problem for some Markov operator. Discretization of that operator via hybrid Monte Carlo methods generates transition matrices for nearly uncoupled Markov chains. The discretization of the Markov operator requires careful consideration to avoid the curse of dimension. Three variants have been developed: 1/Temperature embedding via Boltzmann distribution; 2/PCCA combined with Kohonen's self-organized neural networks; 3/Successive PCCA of dihedrals.

Computational methods for large distributed parameter estimation problems in 3D

In this article, the author described the results which have been doing in the context of inverting electromagnetic data in frequency and time domains for geophysical mining applications with the objective of making such computations practically feasible. The techniques was applicable in a wider context, though. The necessary conditions for the inverse optimization problem yield a large, nonlinear, tightly coupled set of PDEs. The author devised multigrid methods as well as preconditioned Krylov solved for the rapid solution of such systems.

Consecutive ones problems

The standard Consecutive Ones Problem is the problem of converting a given 0/1-matrix by column permutations to a matrix having the consecutive ones property for rows, i.e., in every row all 1 entries occur consecutively. The article considered the Weighted Consecutive Ones Problem and its solution with a branch-and-cut algorithm. Furthermore the authors adressed some complexity issues for a fixed number of rows or columns and also the extension where matrices have to be found which have the consecutive ones property simultaneously for rows as well as for columns. Possible applications are discussed.

Constraint retraction in dynamic CSPs over disjoint real intervals

In a dynamic CSP, the author add a new constraint to the constraint network (a restriction) or delete an old constraint (a relaxation) at any time. Therefore, incrementality is very crucial for solving a dynamic CSP since he do not want to resolve the whole constraint system from scratch whenever a restriction or a relaxation occurs. The paper investigated dynamic CSPs over disjoint real intervals, and propose an algorithm that can handle incremental removal of a constraint. Basing on the hierarchical arc-consistency technique for disjoint real intervals developed by G. Sidebottom and W.S. However, the proposed algorithm follows the chain of dependencies among constraints and by doing so it updates only the part of the constraint network which is afFected by the deletion, but maintains the rest of the network untouched. This algorithm makes constraint deletion in a dynamic CSP over disjoint real intervals a feasible task that can be efficiently implemented.

Control of a CPAP-device with a partially observable Markov decision model

The authors studied the performance of such a control in comparison with other, more traditional control strategies. In commercial devices the applied pressure is held on a constant value, carefully determined in a sleep laboratory. More recently, the pressure is controlled by a classification of the airflow during breathing and then choosing the control variable in dependence on the classification features. Within the POMDP approach we introduce the unobservable state of the airway as the hidden state and regard the features of the airflow during breathing as the observations. They also considered models with two and four hidden states, three observation states and three types of actions (values of applied pressure). The parameters of the hidden models are chosen with help of experience from laboratories, where such devices are developed.

Current-voltage characteristics of quantum hydrodynamic model for semiconductors

A numerical study of the stationary isothermal quantum hydrodynamic model for semiconductors is presented. The stationary model consists of nonlinear third-order elliptic equations. The convergence of a Newton iteration is studied. Numerical simulations of a n+ - n - n+ diode are presented.