ACOUSTIC PATTERN RECOGNITION IN FAILURE ANALYSIS

THE NATURAL FREQUENCY OF A MECHANICAL SYSTEM PROVIDES INSIGHT INTO ITS CONDITION. THE DEGREE OF STRESS TO WHICH A SYSTEM IS SUBJECTED IS REFLECTED IN SOUNDS PRODUCED BY THE SYSTEM WHEN EXCITED. IN PARTICULAR, THE MOVEMENT OF A MECHANICAL OR BIOLOGICAL CARDIAC VALVE PROSTHESIS GENERATES AUDIBLE SOUND; THE NATURAL FREQUENCY OF A CORRECTLY FUNCTIONING VALVE DIFFERS FROM THAT OF A FAULTY VALVE. THIS RESEARCH EFFORT WILL FOCUS ON DEFINING A PROTOTYPE DATABASE MANAGEMENT SYSTEM AND ASSOCIATED ANALYSISALGORITHMS FOR IDENTIFYING FAILURE MODES OF CARDIAC VALVE PROSTHETIC DEVICES. TASKS INCLUDE DEVELOPING CRITERIA AND PROTOCOLS FOR COLLECTING DATA FROM PROSTHETIC VALVES MOUNTED IN A COMMERCIALLY AVAILABLE PULSE DUPLICATOR SYSTEM, SPECIFYING APPROPRIATE DATA STRUCTURES AND PROCESSING LOGIC FOR STORING AND ANALYZING THE DATA, AND DEFINING DISPLAY TECHNIQUES TO PERMIT VISUALLY AS WELL AS AUDIBLY ASSOCIATING UNAMBIGUOUS SOUND SIGNATURES WITH PROSTHETIC DEVICES IN VARIOUS STAGES OF FATIGUE OR STRESS FAILURE. SUCCESSFUL DEVELOPMENT OF ANALYSIS ALGORITHMS WILLPERMIT VALVE MANUFACTURERS TO MORE CLOSELY MONITOR MANUFACTURING FOR QUALITY CONTROL PURPOSES AS WELL AS PROVIDING A BASIS FOR ALGORITHMS WHICH CAN BE USED TO DETERMINE THE CONDITION OF VALVES IN VIVO. MORE GENERALLY, THIS ALGORITHM RESEARCH WILL HAVE WIDESPREAD IMPACT ON OTHERMANUFACTURING QUALITY CONTROL AND/OR MAINTENANCE PROCESSES.