NASA Logo, National Aeronautics and Space Administration


This page describes the experiments conducted with the FLEA testbed over the course of the last few years. A typical experiment starts with the nominal actuator active for certain period of time. The control system will then switch operation to the faulty actuator, with the switch transparent to the health management system. From the point of view of the latter, a fault occurs in the system. The diagnostic part of the health management system is then expected to detect and identify the fault, then invoke the prognostic system to predict the remaining useful life for the component and the overall system. The following sections describe the experiments aimed at validating performance of the FLEA diagnoser and prognoser modules. Detailed information on design and implementation of these modules can be found in the papers listed on the References page.

Diagnostic Experiments

To validate performance of the diagnoser, a set of pre-defined scenarios with varying position (sine, trapezoidal, triangular, sine sweep) profiles and load (constant load between -70 and +70 pounds) profiles was executed. Scenarios within the set are described in detail the Laboratory Scenario Description document.

A subset of the scenarios was used for baseline nominal runs, while the rest was used for runs with hardware- (jam, motor failure, and spall) and software-injected faults (sensor faults). For a detailed discussion of the results please refer to Combining Model-Based and Feature-Driven Diagnosis Approaches – A Case Study on Electromechanical Actuators.

Prognostic Experiments

Prognostic experiments up to this point featured jam in the return channel of the ballscrew mechanism as the primary fault mode. As determined through literature reviews and conversations with actuator and aircraft manufacturers, this fault mode is of a significant concern in the field. The fault scenario picked as motivation for this series of experiments was the following: assuming that the jam occurs in-flight and the actuator is still needed to land the aircraft safely, can we estimate the remaining useful life given that the motion and load profiles remain the same as for a healthy actuator? If a helicopter is taken as an example and this particular actuator controls the pitch angle of the main rotor blades (a critical function to safety of flight), how much time would there be available before the vehicle needed to land?

To set up this series of experiments, jam was injected into the return channel of an actuator. A region then was picked from the manufacturer’s performance diagram (please see Experimental Validation of a Prognostic Health Management System for Electro-Mechanical Actuators) where a healthy actuator can operate continuously for prolonged periods of time (i.e. rated for a 100% duty cycle). Increased friction from the jam in the ballscrew nut resulted in additional current directed by the controller into the test actuator motor in order to attempt performing the same motion profile under the same load as a nominal actuator. This above-nominal current resulted in gradual heat build-up inside motor housing - despite the cool-down periods between motion intervals. Excessive heat eventually caused damage of winding insulation, short circuit, and failure of the motor. This sequence of events is illustrated in the figure below.

Fault progression during the prognostic experiments
Fault progression during the prognostic experiments

Initial experiments demonstrated that motor failure would typically occur when the temperature, as measured on the surface of the motor housing, reached approximately 88 degrees Celsius. Next figure illustrates fault progression for three of the experimental runs:

EMA run-to-failure data EMA run-to-failure data

Prognostic algorithm was executed on collected data, using motor housing temperature measurement as the feature, and its End-Of-Life (EOL) predictions were then compared to the actual failure times for the test articles. Predictions were made at an approximately half-way point between the onset of detectable damage and EOL. The next three figures illustrate remaining useful life predictions done by the prognoser:

Life Prediction for EMA under Load Level 1 (+40lbs sine 36) Life Prediction for EMA under +40lbs (compressive) load

Failure for the motor in the next experiment, with a higher, +50 lbs (compressive) load, occurs faster, in only about 640 seconds:

Life Prediction for EMA under Load Level 2 (+50lbs sine 50) Life Prediction for EMA under +50lbs (compressive) load

Life Prediction for EMA under Load Level 3 (-50lbs sine 51) Life Prediction for EMA under -50lbs (tensile) load

The table below summarizes the high-level prediction results for the above experiments:

Prognosis Results

Prognosis Results

Flight Experiments

Flight experiments with the FLEA are being performed on US Army UH-60 helicopter. During the experiments, the test stand executed rigorous motion sequences, matching those of the target UH-60 actuator (forward primary servo, an actuator responsible for pitch control of the main rotor blades). Load profiles executed by the FLEA’s load actuator were derived using the flight condition information (obtained from the aircraft data bus), as well as some of the models developed by NASA's Subsonic Rotary Wing Project. The graphs below illustrate the typical motion and load profiles seen in flight:

Test actuator motion profile during a UH-60 flight segment
Test actuator motion profile during a UH-60 flight segment

Desired Load profile during a UH-60 flight segment
Desired load profile during a UH-60 flight segment

The diagnostic system has been tested in flight experiments on several fault types. The prognostic system has been tested with the ballscrew nut jam scenario similar to the one described above.

First Gov logo
NASA Logo - nasa.gov