NASA Logo, National Aeronautics and Space Administration


There are many challenges in developing robust, accurate, and fast EMA health management systems. Some of these challenges are as follows:

  • Non-invasive prognostic methods need to be developed.
  • Actuators are often under-instrumented.
  • Local (onboard the vehicle) execution of prognostic algorithms often needs to be highly optimized to adjust for the limited computational resources.
  • Effective fault detection and propagation models are difficult to develop.
  • Noisy environment for sensor measurements
  • Uncertainty in future operating conditions
  • Uncertainty due to human interaction


The overall approach for EMA health management research is illustrated on the figure below.

EMA Health Management Approach
EMA Health Management Approach

First, an extensive literature review of prior EMA health management research was conducted, and a Failure Modes, Effects, and Criticality Analysis (FMECA) was done to compile a prioritized list of the most likely and/or most consequential fault modes to focus on. Some of these faults include ballscrew return-channel jams, spalls, and sensor faults of varying magnitudes.

A model-based diagnostic and prognostic approach was adopted for the health management system, with the understanding that it would be complemented by data-driven machine learning techniques when appropriate. To this end, the first step was generation of nominal and fault progression models of the EMA. The modeling efforts ranged from creation of high-level models for EMA operation, to more detailed studies of motion of bearing balls inside ballscrew raceways and return channels and effects of lubricants on the collision of these balls with each other and the metal surfaces of the ballscrew, for example. Models for winding short effects and motor temperature in case of progressive winding shorts were also developed.

Once the models were available, the diagnostic and prognostic system development started. For the diagnoser, a hybrid model-based and data-driven approach was implemented. The prognostic system, on the other hand, uses Gaussian Process Regression to predict the remaining useful life of the EMA. The diagnosis and prognosis algorithms developed are tested in laboratory using Actuator Prognostics Experiment (APE) testbed and both in laboratory and flight conditions using the Flyable Electromechanical Actuator Testbed (FLEA).

First Gov logo
NASA Logo -