Design optimization of the suction manifold of a reciprocating compressor using uncertainty and sensitivity analysis

Nasir Bilal, Purdue University

Abstract

Gas pulsations are the most common problem in reciprocating compressors, resulting in unwanted and annoying noise and adding energy requirements to the compressor. Extensive research has been conducted on developing models of reciprocating compressors to enable design optimization studies, reduce compressor noise, and improve energy efficiency. Uncertainties and sensitivities are inherent in any compressor model and very few studies have focused on how these uncertainties and sensitivities affect the validity of the model when the model is perturbed with respect to a baseline set of parameters. This work focuses on reviewing the approaches taken so far to understand the effects of compressor model uncertainties and to quantify the effects of modeling uncertainties and input variations on a compressor model using uncertainty and sensitivity analysis tools. The objective is to develop better simulation models resulting in lower gas pulsations in the suction manifold and to lower energy requirements to the compressor. The compressor model developed by Park (Park, 2004) was used as the baseline compressor model for demonstrating the application of uncertainty and sensitivity analysis tools for use in the compressor models. Local sensitivity analysis methods, using the derivative as a measure of parameter sensitivity, were implemented to quantify the uncertainty in the output, and global sensitivity analysis methods, based on the decomposition of variance, were implemented to determine the output sensitivity. Sobol's method of global sensitivity analysis was used to calculate the first order effects and total effects of the suction manifold radius, width and depth on the manifold pressure response. A new sensitivity analysis method based on the concept of total pulsation energy in the suction manifold was developed to perform a sensitivity analysis of the compressor suction manifold. This method has the advantage that it not only determined the input parameters that produced the lowest gas pulsations, but it also the minimized the energy requirement to the system. A new systematic approach was developed using sensitivity analysis methods to aid the compressor suction manifold design using uncertainty bounds on the input and specifying the output in terms of a probability distribution. The use of uncertainty and sensitivity analysis methods in this work helped in accomplishing the following: (a) it was determined that the compressor suction valve is the most critical component in the model to which the output response is most sensitive; and based on these findings, improvements were made in the valve model by transitioning from a 1D model to a 2D finite element valve model; (b) the global sensitivity analysis was used to indicate that the manifold pressure response is most sensitive to changes in manifold radius, followed by manifold width and depth; (c) interactions or higher order effects among the input parameters were also determined; (d) global sensitivity analysis methods were used to determine the set of parameters that produce the maximum and minimum gas pulsations in the suction manifold, which is a useful information when conducting design studies; (e) sensitivity analysis methods were used to identify critical frequencies in the pressure response which were later used in the design of suction manifold; (f) a new method of sensitivity analysis was developed and applied to the simulation model to determine the sensitivity of the input parameters; and (g) the new approach for the compressor manifold design was developed based on the sensitivity analysis procedure developed earlier to select the design parameter of the suction manifold.

Degree

Ph.D.

Advisors

Adams, Purdue University.

Subject Area

Mechanical engineering

COinS