PROJECT ABSTRACT: HIGH-SPEED FLOWS ARE RELEVANT TO A NUMBER OF NASA APPLICATIONS INCLUDING LAUNCH VEHICLES SCRAMJETS AND SUPERSONIC AIRCRAFT. IN GENERAL SUCH FLOWS ARE COMPRESSIBLE AND INVOLVE ONE OR MORE SHOCK STRUCTURES TURBULENCE AND CHEMICAL REACTIONS. ALTHOUGH NUMERICAL SIMULATIONS THROUGH THE USE OF COMPUTATIONAL MODELS HAVE PROGRESSED TREMENDOUSLY IN THE LAST TWO DECADES DUE TO THE VAST GROWTH IN SUPERCOMPUTING POWER NUMERICAL SIMULATIONS OF THESE FLOWS QUICKLY BECOME INTRACTABLE. IN THE PROPOSED WORK MACHINE LEARNING TOOLS WILL BE USED TO INTEGRATE MODELS INTO EXISTING SIMULATION TOOLS IN TWO DIFFERENT WAYS: A) DEVELOPMENT OF MODELS FOR THE RESIDUAL TERMS IN THE COMMONLY-USED LARGE EDDY SIMULATION (LES) APPROACH WHICH REPLACES PURELY PHYSICS-BASED MODELING BY COMBINING PHYSICS-BASED MODEL FORMULATION AND A DATA-DRIVEN FUNCTIONAL RELATION BETWEEN THE MODEL INPUTS AND OUTPUTS; B) DEVELOPMENT OF REDUCED-FIDELITY MODELS BASED ON FLOW DECOMPOSITION TECHNIQUES THAT PROVIDE A FAST ALGORITHM FOR SIMULATING EVENT TRANSITIONS SUCH AS INLET UNSTART OR FLAME EXTINCTION INSIDE HIGH-SPEED SCRAMJET ENGINES. IF SUCCESSFUL THIS PROGRAM WILL CHANGE THE SCOPE OF COMPUTATIONAL MODELING BY PROVIDING A RAPID MODELING TOOL THAT IS CAPABLE OF LEVERAGING LEGACY DATA TO ADVANCE PREDICTIVE ACCURACY WHILE PROVIDING A FRAMEWORK FOR INTEGRATING EXPERIMENTAL AND HIGH-FIDELITY SIMULATION DATA AS THEY BECOME AVAILABLE