Anomaly detection to engine performance data sets

Background: The USAF is seeking development of a unique, multivariate, statistics-based approach for applying anomaly detection to engine performance data sets. This analysis methodology shall automate the assumption checks, baseline development, and anomaly detection steps required for multivariate statistical analysis. Requirement: It shall continuously monitor engine parameter readings as they are added and their relation to other engine parameter readings for shifts and trends from normal. Combining the expert knowledge of project engineers, the features extracted from the performance data shall be able to be refined to provide a holistic analysis, detecting changes in gas-path performance without the need for predefined fault codes. In addition, anomaly detection accuracy shall be greatly improved over traditional, population-based methods by automatically developing and utilizing a baseline for each ESN and allowing for rapid decomposition of indications to isolate anomalous parameters. All interested parties should respond with company contact information and provide a brief description of similar work completed along with references. Company contact information shall be in the following format: Company Name Point of Contact Address Phone Number\E-mail Web Address For additional information or questions please contact in writing to: Susan Cockrell Contracting Officer Susan.cockrell@tinker.af.mil (405) 734-6995 Tim Roberts Contract Negotiator Timothy.roberts2@tinker.af.mil (405) 734-2022