Mixed Data Fusion for Turbine Propulsion PHM

The project develops engine fault estimation from parametric sensor data and discrete BIT (built-in-test) data. The approach addresses the vulnerabilities of existing PHM systems and allows achieving low (the smallest theoretically possible) false alarm rates while simplifying system engineering and sustainment. The fault states are estimated (detected) from the pre-processed sensor data and pre-processed BIT data using fault signature models. Existing PHM approaches use heuristics for determining engine fault alarm conditions from a mix of parametric and discrete data. This project relies on recent breakthroughs in estimation and optimization that allow integrating discrete and parametric data as well as discrete and continuous fault states in a multivariate mixed estimation formulation. Phase II will develop an embedded functional application implementing the approach in the avionics of the target engine. It will also maturate tools supporting the system analysis and design of the mixed data fusion. Phase I demonstrated the feasibility of the approach for a linearized model of F110 engine. Limited functionality and partial fault coverage were implemented. Phase II will develop full functionality for a target engine and integrate it with engine on-board. Its objective is to prepare the developed mixed data fusion function for productization and operational deployment. BENEFIT: The proposed Mixed Data Fusion approach can substantially improve the accuracy of diagnosing root cause engine faults from a mixture of parametric gas path data and discrete (BIT/BITE) data. The accuracy is improved both by reducing the error rate and by allowing discrimination between more different fault conditions. A feasibility study for the F110 engine demonstrated the reduction of the false error rate by a factor of 1.5-2 while discriminating between 15 faults (compared to 4 faults for the baseline approach). The Mixed Data Fusion software can be deployed in on-board engine avionics or in ground maintenance support systems. The ability of the software to reduce false alarms and fault ambiguity groups could help to reduce the maintenance costs significantly.