Computational Engineering

Baker, A.J.

Brief Bio

A. J. Baker moved to Professor Emeritus, UT/Knoxville, in 2010 following 34+ years on the Engineering College faculty. He founded the UT Computational Fluid Dynamics (CFD) Laboratory in 1982 to facilitate collaboration among academics, national laboratory scientists and commercial industry professionals in development/application of CFD technology. The CFD Lab organized graduate academic program has matriculated some three dozen PhD-MSc degrees during his tenure as Director.

His PhD is in Engineering Science, SUNY/Buffalo, 1970, with dissertation topic in atmospheric pressure magnetohydrodynamic plasmas. He joined Bell Aerospace Company as Principal Research Scientist to pursue the fledgling field of finite element CFD. Progress created collaborative research opportunities with NASA, hence 1972-73 Visiting Scientist stints at the NASA Langley Institute for Computational Applications in Science and Engineering (ICASE). These led to a Visiting Professorship at Old Dominion University, 1974-75, whereupon he joined the UT Engineering Science faculty with goal to initiate a graduate degree program in CFD.

With CFD Lab colleagues and graduate students, Baker is widely published including 74 archival journal articles, 138 conference proceedings papers, chapter contributions in 9 monographs and 71 research contract reports. He has authored 5 textbooks, most recently Optimal Modified Continuous Galerkin CFD (2014) and Finite Elements ⇔ Computational Engineering Sciences (2012), both published by John Wiley and Sons, London. He has presented numerous invited lectures nationally and internationally with stints as visiting professor at universities in Japan, Taiwan and South Africa.

Research

Professor Baker’s 4+ decade record in aerospace industry, then academic research leadership in continuum weak formulation focus in laminar/time-averaged/space-filtered Navier-Stokes PDE system CFD algorithms is well documented. The culmination is derivation and validation of advanced performance modified continuous Galerkin finite element discrete implementations amenable to efficient HPC design. He has archived academic course content to Internet outreach spanning baccalaureate CES (www.wiley.com/go/baker/finite) to graduate and advanced level CFD (www.wiley.com/go/baker/GalerkinCFD).

Following are brief descriptions of ongoing research topics:

Continuum mPDE theory for CFD O(h²) dispersion error annihilation: Present generation HPC implemented commercial/production CFD codes for time averaged (RaNS) Navier-Stokes PDE systems are principally based on finite volume second order accurate numerics. These theories are required artificial diffusion augmented due to nonlinear destabilizing discretization-induced O(h²) dispersive error mode. A weak form theorization generates a continuum modified PDE (mPDE) replacement which embeds a dispersive/anti-dispersive operator replacement validated to annihilate this error mechanism.

Analytical space filtered NS tensor/vector quadruple closure: The 5+ decade literature addressing space filtered NS (LES) replaces identified tensor quadruple with resolved scale velocity tensor product plus a subgrid scale (SGS) tensor model. A 2013 UT CFD Lab PhD dissertation derives an analytical closure (arLES theory) for rigorous tensor/vector quadruples absent any Reynolds number assumption. Inclusion of boundary commutation error (BCE) integrals renders resultant mPDE system well-posed on bounded domains. Differential definition weak form-approximate deconvolution (AD) algorithm generates state variable non-homogeneous Dirichlet BCs at boundary nodes whereon the NS BC is no-slip.

HPC efficient radiosity theory replacement for Stefan-Boltzmann: Current CFD codes addressing RaNS hypersonic aerothermodynamics with combustion omit radiation due principally to Stefan-Boltzmann linear algebra jacobian complexity. Radiosity theory replacement generates a Fredholm integral equation of the second kind. Resultant weak formulation eliminates cited jacobian complexity with algebraic hypermatrix statement generating piecewise discontinuous solutions validated to adhere to asymptotic error theory.

PICMSS HPC platform: Via long standing research collaboration between CFD Lab and JICS, validation of cited arLES theory algorithm possessing 27 state variable members was enabled via weak formulation implementation in PICMSS (Parallel Interoperable Computational Mechanics System Simulator), the HPC multi-CPU/GPU brainchild of Dr. Kwai Wong, Research Scientist. Only via this HPC implementation was arLES theory validation enabled including prediction of laminar-turbulent wall jet transition absent any modeling component.