Improving the maintenance process and enabling prognostics for control actuators using CAHM software

Actuator health monitoring of military systems is currently performed both on-board the aircraft (at the operational-level) and at ground-based maintenance installations (i.e., intermediate or depot levels). Current onboard health monitoring employs built-in-tests (BITs), which apply conservative thresholds to flight control data to identify problems early and avoid in-flight failures. However, the extreme operation and interdependent nature of these military systems often cause flight control parameters to exceed these thresholds even though the component is healthy. Intermediate or depot testing may fail to recreate the in-flight problems experienced or not substantiate the on-board assessment. This has led to high incidences of can not duplicate (CND) errors, high sparing requirements, and has affected maintenance costs and operational readiness. The authors are developing a control actuator health management (CAHMtrade) software package that employs both data-driven and model-based algorithms to provide diagnostic and prognostic indicators for hydraulic servovalves and mechanical actuators. These software modules use actuator command/response data, physical actuator models, signal processing techniques, neural network modeling, advanced feature and knowledge fusion strategies, classification algorithms, and an evolving array of prognostics methods. Advanced reasoners are also being developed to interpret BIT results for better fault isolation. Each class of diagnostic algorithm has specific requirements and advantages, including the low computational burden of data-driven algorithms (when compared with other diagnostic techniques) and the ability of model-based algorithms to relate faults back to physically meaningful parameters. These diagnostic algorithms will not only help isolate faults, but will also provide a gray-scale health assessment of components, which will be much more useful to maintenance technicians than a simple pass/fail BIT designation. In the presence of sufficient historical health information, diagnosis can be extended to prognosis using a suite of trending algorithms and novel tracking methods to provide an estimate of remaining useful life (RUL). This paper highlights the current state of the innovative software tool, which is initially targeted toward implementation in an intermediate test data environment