Data fusion increases the reliability and reduces the difficulty of gear-damage diagnosis.

system for real-time detection of surface- fatigue-pitting damage to gears for use in a helicopter transmission is based on fuzzy-logic used to fuse data from sensors that measure oil-borne debris, referred to as “oil debris” in the article, and vibration signatures. A system to detect helicopter-transmission gear damage is beneficial because the power train of a helicopter is essential for propulsion, lift, and maneuvering, hence, the integrity of the transmission is critical to helicopter safety. To enable detection of an impending transmission failure, an ideal diagnostic system should provide real-time monitoring of the “health” of the transmission, be capable of a high level of reliable detection (with minimization of false alarms), and provide human users with clear information on the health of the system without making it necessary for them to interpret large amounts of sensor data.

One of the main ideas underlying the present development is that by integrating oil-debris and vibration sensor subsystems into a single diagnostic system, wherein the data from the two types of sensors are appropriately fused, it is possible to make the damage-detection and decision-making capabilities of the resulting diagnostic system better than those of a diagnostic system that incorporates only one of the sensor subsystems. This idea was tested in 24 experiments in NASA Glenn Research Center’s Spur Gear Fatigue Rig, wherein vibrations were measured by two accelerometers and oil debris were measured by a commercially available inductance-type oil-debris sensor. Speed and load were also measured. The vibration and speed data were processed by two gear diagnostic algorithms that yielded temporally varying statistical parameters known in the art as “FM4” and “NA4 Reset,” respectively.

These Vibration (FM4 and NA4 Reset) and Oil-Debris Data and the corresponding output of the datafusion model are the products of one of the experiments performed to test key parts of the developmental diagnostic system.
Multisensor-data-fusion analysis techniques were applied to the FM4, NA4 Reset, and oil-debris data. Data from the different sensors were combined to make inferences that could not be made on the basis of data from a single sensor. Such a process is similar to the process in which a human integrates data from multiple sources and senses to make decisions.

Sensor data can be fused at the raw data level, feature level, or decision level. In this development, the decision level was chosen because it does not limit the fusion process to a specific feature or sensor. The FM4 and NA4 Reset parameters and the accumulated mass of the debris were the features selected for use as input to the data-fusion part of the system. Fuzzy logic was used to identify the damage level indicated by each feature and to perform decision-level fusion on the features. The resulting data-fusion model was capable of discriminating between the stages of pitting wear. The output of the data-fusion model was in the form of parameters indicating which of three discrete conditions represents the current state of damage and the corresponding action recommended to end users. The three condition/action combinations were denoted “OK” (no damage and no action necessary), “inspect” (initial pitting), and “shutdown” (severe pitting).

The upper part of the figure depicts the FM4, NA4 Reset, and oil-debris data from one experiment during which pitting damage occurred. The lower part of the figure shows the corresponding output of the data-fusion model. Readings were taken once per minute. The triangles indicate when the gear was inspected for damage. As shown in the photograph connected to the second triangle, damage began to occur at approximately reading 2,669 during this experiment. Analysis of the data collected during this and the other experiments confirmed the expectation that it is advantageous to fuse features of data obtained through different sensors and that, as desired, the output of the data-fusion model amounts to clear, reliable information that can be used in making decisions about the health of the affected gears.

This work was done by Paula Dempsey of Glenn Research Center. For further information, access the Technical Support Package (TSP) free on-line at www.techbriefs.com/tsp under the Mechanics category.

Inquiries concerning rights for the commercial use of this invention should be addressed to:

NASA Glenn Research Center

Innovative Partnerships Office

Attn: Steve Fedor

Mail Stop 4–8

21000 Brookpark Road

Cleveland, Ohio 44135.

Refer to LEW-17889-1.

The U.S. Government does not endorse any commercial product, process, or activity identified on this web site.