ABSTRACT: THIS WORK DEALS WITH THE PROBLEM OF FAULT DETECTION AND LOCATION IN MECHANICAL SYSTEMS WITH STATIONARY INPUTS. A KNOWN STATE SPACE MODEL OF MECHANICAL SYSTEM IS CONSIDERED IN THIS APPROACH. THE FAULT IS DEFINED AS AN ABRUPT AND PERMANENT VARIATION OF SOME SYSTEM PARAMETERS. THE PROPOSED METHOD USES A PARAMETER ESTIMATOR BASED ON CORRELATION FUNCTIONS. TWO ACTUAL CONDITIONS ARE MATHEMATICALLY AND NUMERICALLY STUDIED: 1 - RESTRICTIONS ON MEASUREMENTS - SOME SYSTEM STATES CAN NOT BE MEASURED BY INSTRUMENTATION, 2 -RANDOM NOISE- THE SYSTEM SIGNALS ARE CORRUPTED WITH STATIONARY RANDOM NOISE. THE PROCEDURE IS DEMONSTRATED WITH A SIMPLE ONE-DEGREE-OF-FREEDOM MODEL OF MECHANICAL SYSTEM.
KEY-WORDS : PARAMETER ESTIMATION, FAULT DETECTION, DIAGNOSIS. |