SAP PURCHASE REQUISITION: 4200265291 STATEMENT OF WORK FOR DEVELOPMENT OF A NOVEL METHODOLOGY FOR PROGNOSTI…

Public award description

SAP PURCHASE REQUISITION: 4200265291 STATEMENT OF WORK FOR DEVELOPMENT OF A NOVEL METHODOLOGY FOR PROGNOSTICS, UNCERTAINTY REPRESENTATION AND UNCERTAINTY MANAGEMENT 1.0 SCOPE OF WORK IMPACT TECHNOLOGIES WILL PURSUE A THREE-PRONG APPROACH TO DERIVE PROGNOSTIC ALGORITHMS THAT ARE MATHEMATICALLY RIGOROUS, GENERIC, ROBUST AND VERIFIABLE. THEY WILL BE ACCOMPANIED BY RELEVANT UNCERTAINTY REPRESENTATION AND MANAGEMENT METHODS. THE COMPANY WILL EMPLOY A BAYESIAN ESTIMATION METHOD, CALLED RISK SENSITIVE PARTICLE FILTERING, AS THE BASELINE FAILURE PROGNOSIS SCHEME WITH ITS ATTENDANT UNCERTAINTY REPRESENTATION/MANAGEMENT TOOLS. THE SECOND APPROACH WILL RELY ON DATA-DRIVEN TECHNIQUES AND, SPECIFICALLY, A CONSTRUCT CALLED CONFIDENCE PREDICTION NEURAL NETWORK FOR LONG-TERM PREDICTION AND A PARZEN WINDOW FOR UNCERTAINTY REPRESENTATION, WHILE A LAZY-LEARNING TECHNIQUE IS USED FOR UNCERTAINTY MANAGEMENT. THE FINAL APPROACH WILL TAKE ADVANTAGE OF A MARKOV CHAIN MONTE CARLO SIMULATION TOOL AND HYPER-PARAMETER CORRECTIONS TO MANAGE UNCERTAINTY IN A FEEDBACK CONFIGURATION. THE PROPOSED SYSTEMS APPROACH WILL ENTAIL THE DEFINITION OF PERFORMANCE METRICS, A THOROUGH DESIGN OF EXPERIMENTS PROCEDURE ACCOMPANIED BY TOOLS TO EVALUATE DATA QUALITY, TESTING ON A SUITABLE ELECTRO-MECHANICAL ACTUATOR (EMA)TEST STAND AND A GENERIC VERIFICATION AND VALIDATION (V&V) METHODOLOGY. IMPACT TECHNOLOGIES WILL DELIVER TO NASA FULL DOCUMENTATION OF THE ALGORITHMS AND A TOOLSET CONSISTING OF COMPUTER CODE FOR ALGORITHMS/MODELS AS WELL AS A DETAILED DESCRIPTION OF THE EXPERIMENTAL PROCEDURES, PERFORMANCE METRICS, TEST RESULTS, AND THE V&V METHODOLOGY EMPLOYED IN THE PROGRAM. 2.0 APPLICABLE DOCUMENTS/BACKGROUND SELECT BACKGROUND REFERENCES: VACHTSEVANOS, G., LEWIS, F., ROEMER, M., HESS, A. AND WU, B., INTELLIGENT FAULT DIAGNOSIS AND PROGNOSIS FOR ENGINEERING SYSTEMS, JOHN WILEY&SONS, INC. 2006 ORCHARD, M., AND VACHTSEVANOS G., "A PARTICLE FILTERING-BASED FRAMEWORK FOR REAL-TIME FAULT DIAGNOSIS AND FAILURE PROGNOSIS IN A TURBINE ENGINE," 26TH AMERICAN CONTROL CONFERENCE, ACC2007, NEW YORK, USA, 2007. BYINGTON, C. AND VACHTSEVANOS, G., VERIFICATION AND VALIDATION OF DIAGNOSTIC/PROGNOSTIC ALGORITHMS , PROCEEDINGS OF MFPT 59, MACHINERY FAILURE PREVENTION TECHNOLOGY CONFERENCE, VIRGINIA BEACH VA, APRIL 18-21, 2005. VACHTSEVANOS, G. AND WU, B., ADDRESSING UNCERTAINTY AND CONFIDENCE IN PROGNOSIS , PROCEEDINGS OF MATERIALS SCIENCE AND TECHNOLOGY CONFERENCE 2004, NEW ORLEANS, LA, SEPTEMBER 27 28, 2004. 3.0 DESCRIPTION OF TASKS/TECHNICAL REQUIREMENTS 3.1. FY09 TASKS (PROGNOSTIC ALGORITHMS): 3.1.1 DEVELOP AND DOCUMENT THEORETICAL FRAMEWORK FOR 3.1.1.1. MATHEMATICALLY RIGOROUS PROGNOSTIC ALGORITHMS, I.E. THE CORRESPONDING MATHEMATICAL/PHYSICAL FUNDAMENTALS AND MODELS SHALL BE PRESENTED IN A COMPLETE AND VERIFIABLE FORM 3.1.1.2. UNCERTAINTY REPRESENTATION AND MANAGEMENT 3.1.1.3. PARTICLE FILTERING 3.1.1.4. CONFIDENCE PREDICTION NEURAL NETWORK (CPNN) 3.1.1.5. MARKOV CHAIN MONTE CARLO SIMULATION WITH BAYESIAN UPDATE 3.1.2 DEFINE PERFORMANCE METRICS FOR THE PROGNOSTIC ALGORITHMS 3.1.2 DESIGN VALIDATION&VERIFICATION EXPERIMENTS: 3.1.2.1 DEFINE THE APPROPRIATE TEST STAND CONFIGURATION AND INITIAL DESIGN OF EXPERIMENTS 3.1.2.2 DESCRIBE FAULT MODEL AND SENSING STRATEGIES 3.1.2.3 DEVELOP DATA ANALYSIS TOOLS 3.1.3 DEVELOP A PRELIMINARY DESIGN OF SIMULATION AND VISUALIZATION PLATFORM. 3.2. FY10 TASKS (TESTING / SOFTWARE IN THE LOOP SIMULATION): 3.2.1 COMPLETE DEVELOPMENT OF ALL PROGNOSTIC AND UNCERTAINTY REPRESENTATION AND MANAGEMENT ALGORITHMS. 3.2.2 CODE ALGORITHMS AND INSTALL IN TOOLSET 3.2.3 CONDUCT TESTING ON EMA STAND AND ACQUIRE / ANALYZE DATA. 3.2.4 CARRY OUT SOFTWARE IN THE LOOP SIMULATIONS OF ALL ALGORITHMS USING EXPERIMENTAL DATA. 3.2.5. DEFINE V&V FRAMEWORK AND CONDUCT TOLL VALIDATION. 3.2.6 POPULATE TOOLSET WITH ALL ALGORITHMS FOR PROGNOSIS AND UNCERTAINTY MANAGEMENT. 3.3. FY11 TASKS (HARDWARE IN THE LOOP