60020 - Adjustment and Validation of Monitoring System Algorithms on the Simulated Historical Data of an Aircraft Engine Fleet 

In the field of gas turbine diagnostics, a significant gap is observed between the variety of proposed and investigated diagnostic solutions and a narrow group of the algorithms employed in real monitoring systems. One of the explanations is that particular diagnostic algorithms of a monitoring system are usually developed and verified separately from each other. An additional explanation is related to simplified simulated data used to verify the algorithms.

The present paper aims to adjust and validate joint operation of the algorithms of a gas turbine monitoring system during a whole lifetime. The software tool called the Propulsion Diagnostic Method Evaluation Strategy (ProDiMES) provides the input data. This tool realistically simulates steady-state measurements of an aircraft engine fleet during a total engine life. The simulation is characterized by varying operating conditions, individual engine deterioration profiles, and various scenarios of engine, actuator, and sensor faults.

Using the mentioned software, some diagnostic solutions have been verified so far. However, they do not include all necessary monitoring system components, use a short time interval of input data, and do not allow for a long-term deterioration change. In contrast, this paper considers the operation of a whole monitoring system during an engine life. The need to monitor gradual engine deterioration and diagnose more rapid faults against the background of the deterioration makes the system considerably complex. Hence, a proper adjustment of the rules of interaction between multiple system algorithms becomes a challenging problem.

The present paper proposes and analyzes a system prototype that embraces five main data-driven algorithms aimed at healthy engine model formation, deteriorated engine model actualization, deterioration monitoring, fault detection and identification, and deterioration and faults prediction. Along with engine life usage, the operation of the algorithms and their interactions change, for example, because of models adaption, models replacement, fault appearance, and final diagnosis decision making. Hence, in addition to tuning the algorithms themselves, the paper focuses on optimizing algorithms interactions. This process of monitoring system optimization employs the indicators of model adequacy and diagnostic decision reliability as criteria.