A Probabilistic Model for Information and Sensor Validation

摘要: This paper develops a new theory and model for information sensor validation. The represents relationships between variables using Bayesian networks utilizes probabilistic propagation to estimate the expected values of variables. If estimated value variable differs from actual value, an apparent fault is detected. only since it may be that itself based on faulty data. extends our understanding when possible isolate real faults potential supports development algorithm capable isolating without deferring problem use expert provided domain-specific rules. To enable practical adoption real-time processes, any time version developed, that, unlike most other algorithms, returning improving assessments validity sensors as accumulates more evidence with time. developed tested by applying validation temperature during start-up phase gas turbine conditions are not stable; known challenging. concludes discussion applicability scalability model.