Integrally geared compressors offer a wide range of applications for probabilistic analysis. The combination of a multi shaft compressor with an integral gear under changing operating conditions creates a lot of design challenges.
The gear design needs to meet the requirements of the compression process, like the rotating speed of the pinions or the pinion power. These requirements lead to a specific gear with its specific properties. The examination and verification of the internal correlations between thermodynamics and gear design is one significant objective of the project as a high efficiency of the compression process may be connected with high gear losses, or the other way around.
These design challenges ought to be investigated with probabilistic methods, such as the Monte-Carlo-Simulation. With such methods, it is possible to explore a wide design space automatically in order to learn about correlations between probabilistic input and output parameters as well as in order to choose a better design.
In a first step of this project, all process steps relevant for designing an integrally geared compressor have been combined to form one single automated algorithm. This algorithm is used for Monte-Carlo-Simulations (MCS) with optimal Latin-Hypercube as the sampling method. On the basis of the MCS results, response surfaces can be created to describe the scatter and the behaviour of the result parameters. Furthermore, response surfaces can be used as meta models for optimization and prediction.
This paper seeks to address the use and the performance of response surfaces.