Performance Measurements of a Full-stage Centrifugal Process Gas Compressor Test Rig

The paper presents the first experimental performance data for a mid-stage centrifugal compressor test rig built at RWTH Aachen University. The results provide an analysis of the operational and loss behavior and will be published in an open test case for validating numerical methods. The test rig was built to investigate the operational and loss behavior of a complete return channel in detail and is broadly introduced. The measurements described in this paper were designed based on numerical calculations. Further, the stage performance is analyzed based on a compressor map for design speed, supplemented by detailed measurements inside the diffusion system. The performance of the diffuser and the return channel are discussed separately. The significant influence of the return channel design on the overall stage efficiency and pressure build-up is emphasized. Additional angle measurements at the stage outlet provide details about the inflow to the potential next compressor stage. NOMENCLATURE avg average b channel width c absolute velocity cD dissipation coefficient cf friction coefficient Cp pressure recovery DP design point exp experiment f error h enthalpy m number of parameters M Mach number n number of samples NC near choke NS near stall p pressure r radius rh degree of reaction Re Reynold’s number tp student factor T temperature u circumferential velocity uΦ uncertainty of Φ V̇ volume flow rate x measured parameter α absolute flow angle β relative flow angle ǫ interpolation threshold ηp polytropic efficiency ν kinematic viscosity Π pressure ratio ρ density φStage flow coefficient Φ defined quantity Ψh work input coefficient Ψy head coefficient Subscripts 1− 5 plane definitions diff diffuser dyn dynamic DP design point i control variable i, j measurement planes imp impeller in inlet plane inf infinite blade number m meridional direction n normalized out outlet plane r random error s systematic error slip impeller slip t total conditions therm thermal ts total to static tt total to total u circumferential direction 1 Proceedings of 11 European Conference on Turbomachinery Fluid dynamics & Thermodynamics ETC11, March 23-27, 2015, Madrid, Spain OPEN ACCESS Downloaded from www.euroturbo.eu Copyright © by the Authors INTRODUCTION The high pressure ratios required for the chemical industry often cannot be realized with only one compressor stage. Therefore, centrifugal compressors are typically single-shaft, multistage compressors of several rotors with a radial outflow. After leaving the rotor, the fluid passes the attached diffusion system consisting of a vaned or vaneless diffuser followed by a return apparatus. The diffuser decelerates the flow and thus leads to an increase in static enthalpy. Then the fluid is redirected radially inward by a U-bend before it flows through a vaned return channel, which aligns the flow for the next stage. A final L-turn directs the flow in the axial direction to the next rotor. The flow inside the diffusion system contains complex 3D-structures and secondary flows. Their characteristics are e.g. discussed based on experimental studies of the velocity distribution inside the U-bend and the return channel by Inoue and Koizumi (1983), Simon and Rothstein (1983), Rothstein (1984) and Rothstein and Gallus (1983) for different geometries. These investigations show that complex flow phenomena and their appearance are not fully understood. With a 5 to 10 percentage point loss in overall compressor efficiency (see Aalburg et al. (2011)), the flow inside the diffusion system remains the topic of numerous investigations. One of the main objectives of these in most cases numerical investigations are the return channel vanes. Since cylindrical profiles are usually used in industrial compressor stages, they have great potential for optimization. The aim is to provide a larger redirection while the losses decrease, so that the outer stage diameter can be decreased for nearly constant efficiency. This leads to smaller friction and pressure losses inside the diffuser. The big difficulty in numerical computations is to choose the right turbulence model to solve the secondary flow phenomena inside the return channel vanes as is shown in Lenke (1999), Lenke and Simon (1999) and Lenke and Simon (1998). Even Reutter et al. (2011) and Hildebrandt (2011) conceded that their results could be improved by validating their turbulence modeling with measurements. This shows that experimental results are essential for optimizing the flow path. Simpson et al. (2008) and Schmitz et al. (2008) therefore described the buildup of a 90◦ cascade rig and the validation of the numerical setup based on the measurement data. Based on this work, smaller diffusion systems for small flow rates were developed, which were measured on a rotating test rig by Aalburg et al. (2008). The results show that a decrease in the outer stage diameter is possible while the efficiency is nearly constant. Since the measurements were performed for small flow rates, they are not applicable to machines with higher flow rates. The flow through narrow channels is dominated by boundary layer effects due to the higher relative boundary layer height that prevents the formation of most complex 3D flow phenomena occuring in higher channels with higher flow rates. The test rig at RWTH Aachen University was built to optimize the diffusion system of a radial compressor stage with flow rates up to φStage = 0.18. The results will be published in an open test case. This paper presents the first performance data and confirms their reliability. EXPERIMENTAL SETUP Compressor Test Rig The Institute of Jet Propulsion and Turbomachinery is investigating the centrifugal stages of single-shaft, multistage compressors for industrial applications on one test rig. The test rig, designed and set up in close collaboration with MAN Diesel & Turbo SE, Oberhausen, consists of a single stage and was completed in January 2014. The rig is intended to operate with various types of closed impellers and is equipped with a high flow rate stage in the current setup as shown in Figure 1. The compressor is powered by a 1600 kW asynchronous motor, which is coupled to a 12.5:1 ratio gear box with rotational speeds up to 18750 rpm. The test rig has a magnetic bearing system that significantly decreases friction losses and thus enables an accurate measurement of the mechanical work using a torque meter. The test section is a closed loop for high repeatability of the experiments and an independent variation of Reynolds and Mach