Aero-engine components analysis is quite complex in nature and same is the case with thermal analysis. Building an analytical model to simulate actual working conditions of a jet engine is challenging. Many approximations creep into analytical model in the process of simulating real time working environment. Thermal data match is the process to validate analytical model temperatures against experimental data i.e. temperatures measured by the thermocouples instrumented on various components of test engine.

Traditionally, Thermal boundary conditions are varied within permissible limits as mentioned in standard works to account the inability of empirical correlations in capturing physics and assumptions during modeling. Conventional thermal data match requires large number of iterations to match analytical model temperatures against engine data. This demands huge manual efforts and consumes more cycle time.

A methodology is developed using regression based ML algorithm which uses legacy thermal analysis data to identify scale factor to each thermal boundary condition (BC) that influences temperature at a thermocouple location. Initially ML algorithm identifies the weightage that each BC contributes towards data match status at thermocouple location. Once weights are acquired, objective function is formulated using predictor variables (Delta Temperature), response variables (BC Scale factors) along with the weights. Linear programming techniques are employed to solve the objective function subjected to constraints provided in the standard work. Thus, the output of the program is unique optimal scale factors which satisfy prescribed data match criteria.

The proposed methodology in this paper is validated by data matching compressor case analysis results against test engine thermocouple temperature data. With this approach, user intervention is mostly eliminated. 75% of cycle time reduction is achieved.

This content is only available via PDF.
You do not currently have access to this content.