Abstract: A method is presented for the control of chemical spills. The approach is based on real-time information provided by microsensors capable of monitoring instantaneous changes in the concentration of a chemical in solution or suspension. The method also utilizes current flow and transport data provided by a simulation model. Once a chemical cloud requiring control action is detected by the sensors, the model provides optimal directions to pre-installed boundary actuators capable of modifying the flow conditions in the system. The technique requires assimilation of data from the sensors to steer the model so the error between its current state and sensor measurements is minimized. The model also performs prediction simulations to determine the optimum set of actuator commands necessary to control the chemical plumes. Results of model control applications are shown to be capable of removing a chemical cloud from a flow through channel.

Plenary Lecture 2

Title: Functional and physical extrapolation relative to evolution of the mammalian erythrocyte

Professor Charles A. Long

Department of Biology

University of Wisconsin

Stevens Point, Wisconsin

54481 U.S.A.

Abstract: The erythrocyte in mammals follows the phylogeny of early lungfish, amphibian, mammal-like reptile line. Although birds are similarly warm blooded and metabolically active, and the reptiles are considered ancestral to the mammalian grade, macroevolutionary specialization of a mammal erythrocyte without marginal microtubule bands, biconcave form, and lacking nucleus and other organelles shows no affinity to extant vertebrate groups. The erythrocyte following loss of the marginal bands and nucleus attains a high concentration of Hb, and “kiting” biconcave shape with high surface to volume and low inertia, but deforms flexibly to enter tight confines. It regains its form even though the endoplasm is purely viscous, by means of tensile elasticity of the cytoskeleton, probably by surface tension and pressure, and has minimal energy costs in bending and minimal “wear and tear” for the life of the cell.

Plenary Lecture 3

Title: MULTI-DISCIPLINARY RESEARCH ACTIVITIES: Mathematics/ High Performance Computing / Biology / Fluid Mechanics / Structural /
Acoustics Applications

Professor Duc Nguyen

Civil & Environmental Engineering Department

Old Dominion University

1319 ECSB

Norfolk, VA 23529

(USA)

Abstract: Fundamental and numerical intensive equations arise naturally from Computation Fluid Dynamics (CFD), Molecular Biology, Structural Analysis, Design Sensitivity Analysis (DSA), Optimal Design, Aerospace/Automobile Design, Mathematics, Operation Research, Ground Water Flows, Electromagnetics, Acoustics etc... are identified.

Effective numerical algorithms to solve such (large-scale) numerical intensive equations have recently been developed. The developed algorithms take full advantages of parallel and/or vector/cache capabilities provided by high-performance computing (HPC) platforms, such as SUN-10000, SGI etc...

Based upon the developed numerical algorithms (for solving systems of SPARSE, linear/nonlinear, symmetric/unsymmetric, positive/negative/INDEFINITE matrix genvalue equations, linear/nonlinear constrained/unconstrained optimization, design sensitivity analysis, 2-nd order P.D.E.), several major (numerical intensive) subroutines have been coded and tested in a parallel and/or vector/ cache computer environments.

Practical NASA engineering problems, such as stress analysis of the Solid Rocket Booster (SRB), High Speed Civil Transport (HSCT) aircraft, NASA LaRC Acoustic, automobile models etc... have been solved to evaluate the performance of the developed algorithms. For example, using our developed SPARSE L*D*L solver, a 250,000 degree-of-freedom (or equations) automobile finite element model can be solved on a "single" Cray-C90 processor in 81 seconds (including approx. 40 seconds re-ordering time, and 41 seconds factorization time).

Fast equation & eigen-solutions for the 640,332 degree-of-freedom (equations) for the NASTRAN structural model have also been obtained on inexpensive/"old" Sun/Ultra 1) workstation in approx. 20 minutes. For this model, the number of non-zero coefficient matrix before & after factorization are 14,790,661 and 88,563,006, respectively. Solutions for 1-6 million unknowns (complex numbers, unsymmetrical systems of linear equations) for acoustics finite element models have recently been solved in inexpensive HPC clusters.

Finally, discussion will be focused on the needs and benefits by incorporating state-of-the-art solvers (in-core and out-of-core parallel-vector equation DENSE solvers),parallel-vector TRI-DIAGONAL (to solve system of 38.4 "million" equations on the "OLD" 128 Intel IPSC/860 "Gamma" processors, in LESS THAN 1 second), eigen solvers, Computational Fluids Dynamics (CFD), etc... into general finite element engineering application codes, within the framework of sub-domains formulation. Large-scale computation for molecular biology in parallel computer environments is also reported.

Plenary Lecture 4

Title: Challenges in Real-Time and Individualized Patient Monitoring, Diagnosis and Decision Assistance: New Paradigms of System Identification

Professor Le Yi Wang

Wayne State University

Department of Electrical and Computer Engineering

5050 Anthony Wayne Dr.

Detroit, MI 48202

U.S.A.

Abstract: Real-time patient monitoring and medical decisions are broadly exemplified by respiratory function monitoring for asthma patient, vital sign monitoring of soldiers in battlefields, anesthesia drug infusion control, fluid resuscitation strategies, pain management, sedation control in intensive care units, automated drug rates for diabetics, etc. The characteristics of patient responses to treatment and drugs in these problems demonstrate significant nonlinearity and time variation, and depend critically on patient medical conditions, surgical procedures, and drug interactions; and hence they are not repeatable. Diagnosis, control and decision assistance in such problems demand individualized and real-time patient models, rendering a central role of system identification in these medical applications. When integrated with internet or wireless networks for telemedicine, these real-time information processing problems are further constrained by data power limitation, noise corruption, and data transmission speed.

In this talk, several new paradigms of system identification will be summarized, beyond traditional identification problems. These will include nonlinear patient models for anesthesia control, reconfigured channel identification for signal separation, and integrated identification and communications in telemedicine. Recent advances in these areas will be presented.

Plenary Lecture 5

Title: Advanced Computational Methods in Bio-imaging and Bio-data Analysis

Assc. Professor Tuan Pham

Bioinformatics Applications Research Centre

School of Information Technology

James Cook University

Townsville, QLD 4811

AUSTRALIA

Abstract: Computational methodology and approaches play a key role in computational life sciences including bioinformatics, computational biology, and biomedical informatics. Some current important research areas of computational life sciences are automated image analysis and identification of cell phases with microscopic time-lapsed imaging sequences, neuronal imaging, classification of genomic and protein sequences, and analysis of gene expression microarray data. In this talk, I will present several recently developed novel computational methods for solving such described problems, and suggest other computational issues and directions for future research in computational life sciences.