KNOWLEDGE BASED SEISMIC EVENT IDENTIFICATION USING SEISMIC ANALYSIS CASE KNOWLEDGE AND EXPLANATIONS

THIS STUDY WILL INVESTIGATE THE APPLICABILITY OF CASE-BASED REASONING TO THE DOMAIN OF SEISMIC EVENT IDENTIFICATION FOR MONITORING TESTBAN TREATIES. A SEISMIC IDENTIFICATION DECISION SUPPORT TOOL WILL BE DEVELOPED TO UTILIZE CASE KNOWLEDGE FROM PREVIOUS EVENTS WHICH HAVE BEEN IDENTIFIED TO IDENTIFY NEW EVENTS WHICH MATCH THE CASE EVENTS. THE CASES, TAKEN FROM THE LITERATURE OR FROM ANALYSIS OF ACTUAL DATA, WILL DESCRIBE HOW A SEISMOLOGIST IDENTIFIED A PARTICULAR EVENT OR GROUP OF EVENTS, INCLUDING WHAT DATA WAS USED, WHAT ANALYSES WERE PERFORMED, AND HOW THE FINAL DECISION WAS MADE. OLD CASES WILL BE SELECTED FOR MATCHING TO NEW CASES THROUGH INDEXED FEATURES, AND THE NEW EVENTS WILL BE CHARACTERIZED AND IDENTIFIED ACCORDING TO THE SIMILARITY OF SIGNAL FEATURES TO THOSE OF THE CASE EVENTS. CONCEPTUAL/DYNAMIC MEMORY STRUCTURES WILL BE DESIGNED TO REPRESENT SEISMOLOGICAL CONCEPTS AND TO ORGANIZE CASE DATA AROUND THE CONCEPTS. EXPLANATION-BASED LEARNING METHODS WILL BE EXPLORED FOR AUTOMATIC INDEXING AND GENERALIZATION OF CASE KNOWLEDGE. WHEN IMPLEMENTED ON A COMPUTER, THE SYSTEM WILL AUTOMATICALLY ANALYZE SEISMIC EVENTS, IN THE SAME MANNER AS SEISMOLOGISTS WHO INVESTIGATE UNIDENTIFIED AND SPECIAL EVENTS, AND BE ABLE TO EXPLAIN ITS RESULTS ABOUT THE NEW EVENT WITH THE CONTEXT OF PREVIOUS CASES.