The evaluation of propulsion system test and flight performance data involves reviewing an extremely large volume of sensor data generated by each test. An automated system that screens large volumes of data and identifies propulsion system parameters which appear unusual or anomalous will increase the productivity of data analysis. Data analysts may then focus on a smaller subset of anomalous data for further evaluation of propulsion system tests. Such an automated data screening system would give NASA the benefit of a reduction in the manpower and time required to complete a propulsion system data evaluation. A phase 1 effort to develop a prototype data screening system is reported. Neural networks will detect anomalies based on nominal propulsion system data only. It appears that a reasonable goal for an operational system would be to screen out 95 pct. of the nominal data, leaving less than 5 pct. needing further analysis by human experts. Hoyt, W. Andes and Whitehead, Bruce A. Unspecified Center NAS8-39184; SBA-4-91-2-0357...