Statistical energy analysis for a compact refrigeration compressor and model improvement

Jimin Lim, Purdue University

Abstract

Traditionally the prediction of the vibrational energy levels of the components in a compressor is accomplished by using a deterministic model such as a finite element model. While a complete dynamic analysis based on a deterministic approach requires much detail and computational time, an analysis performed using statistical energy analysis (SEA) requires much less information and computing time. All of these benefits can be obtained by using data averaged over the frequency and spatial domains instead of the direct use of deterministic data. In this work, SEA was applied to a compact refrigeration compressor for the prediction of dynamic behavior of each subsystem. Since the compressor used in this application was compact and stiff, the modal densities of its various components were low, especially in the low frequency ranges, and much energy in these ranges transmits through indirect coupling paths instead of via direct coupling. Indirect coupling is usually significant in systems that are not well matched with the assumptions on which the development of classical SEA is based. For this reason, experimental SEA (ESEA), a good tool for the consideration of the indirect coupling, was used to derive the SEA formulation in this case. Direct comparison of SEA results and experimental data for an operating compressor will be shown. The power transfer path analysis at certain frequencies made possible by using SEA will also be described to show the advantage of SEA in this application. Even though a system can be modeled without error in terms of being able to reproduce measured results, it does not always guarantee an ideal SEA model for diagnostic purposes. For this reason, model quality indices were introduced to check an experimental SEA model, and random weighting factors were directly applied to the components in the SEA model to obtain an ideal model. The presence of strong coupling in an SEA model can cause the SEA parameters to take undesirable values, and this effect usually prevents clear interpretation of SEA results in a physical sense. To overcome this problem, it was proposed that strongly coupled subsystems should be combined and modeled as one in an SEA model. Applications and verifications of the latter point were demonstrated with the help of finite element analyses, and the concept was applied to the experimental SEA model of an actual compressor in the final stage.

Degree

Ph.D.

Advisors

Bolton, Purdue University.

Subject Area

Mechanical engineering

COinS