Purdue Uniersity Purdue e-pubs International Compressor Engineering Conference School of Mechanical Engineering 2000 Multi Criteria Analysis of the Supporting System of a Reciprocating Compressor M. Lamantia Electrolux Compressors Companies A. Faraon Electrolux Compressors Companies A. Solari Electrolux Compressors Companies Follow this and additional works at: https://docs.lib.purdue.edu/icec Lamantia, M.; Faraon, A.; and Solari, A., "Multi Criteria Analysis of the Supporting System of a Reciprocating Compressor" (2000). International Compressor Engineering Conference. Paper 1431. https://docs.lib.purdue.edu/icec/1431 This document has been made aailable through Purdue e-pubs, a serice of the Purdue Uniersity Libraries. Please contact epubs@purdue.edu for additional information. Complete proceedings may be acquired in print and on CD-ROM directly from the Ray W. Herrick Laboratories at https://engineering.purdue.edu/ Herrick/Eents/orderlit.html
MULTI CRITERIA ANALYSIS OF THE SUPPORTING SYSTEM OF A RECIPROCATING COMPRESSOR Authors: e-mail: Maurizio Lamantia, Alessio Faraon, Alessandro Solari Electrolux Compressors Companies Via Consorziale 13 Comina 33170 (PN) Tel. 0434393489 Fax : (39)434/393313 maurizio.lamantia@notes.electrolux.it alessio.faraon@notes.electrolux.it alessandro.solari@notes.electrolux.it ABSTRACT The use of finite element analysis with parametric dynamic models gies a powerful tool for the optimization of a irtual compressor. A multi-criteria approach has been applied to the system to reduce noise, ibration, dynamic stress and costs and improe efficiency. The best project is a compromise of all these factors, because these factors generally gie opposite results. This work presents the deelopment of a irtual model for a domestic hermetic compressor and a practical case of balancing optimization, from the numerical analysis to the experimental test. 1.INTRODUCTION A irtual dynamic model of a hermetic compressor has been created. Een if presently the model has been useful to optimize balancing of the crank mechanism reducing low frequency ibrations (about 50 Hz), the model will be used for many other purposes: -Analysis of dynamic forces on mechanical components -Analysis of ibration of the supporting system -Prediction of power consumption -Considering flexible shaft also flexional and torsional ibrations can be analyzed -Analysis ofhydrodynamic lubrication -Optimization ofbalancing -Analysis of ariable speed compressor -Start up analysis of the compressor -Vibro-acoustic analysis The model will be connected to a thermo-fluid dynamic package (written in Fortran or C) or a Simulink model to calculate the correct PV cycle in specific conditions or the correct speed s. torque in ariable conditions. This is useful to study ariable speed compressors. The paper describes the dynamic model, which uses multi-body [1] and flexible elements. So this approach has less simplification than classical analytical approach. Dynamic of a reciprocating hermetic compressor has been described from seeral authors. Equation of the motion of the slider crank mechanism can be soled to compute the force applied on the supporting system. The analysis of a rigid body applied on the system permits to ealuate acceleration. Unfortunately this approach has seeral simplifications and it is generally difficult to know the uncertainty of the result. The parametric model has been deeloped with the use of finite element analysis and dynamic analysis. Some components of the compressor, such as supporting springs and the internal discharge pipe hae been considered as flexible bodies[2], so that they hae a dynamic transfer function which has been calculated through a spectral analysis performed with Ansys [3]. All other elements has been assumed as rigid, een if it's 531
possible to analyze a system with all flexible bodies. This is useful to analyze ibration sources in a higher range of frequency. 2.MECHANICAL MODEL The main assumption is that only discharge pipe and supporting springs are flexible bodies, so that other elements can be considered as rigid. All rigid body elements hae ten characteristic parameters: A center of mass (three coordinates) Amass An inertia (six coordinates) Coupling of substructures permits to analyze flexible bodies, which hae a more complex structure because they hae the same ten parameters of a rigid body, but they hae seeral modal frequencies (eigenalues) and modal shapes (eigenectors). A modal analysis can be performed to find these parameters. The result of the analysis is generally normalized to the maximum displacement, so the real displacement can be ealuated only if the correct excitation set is known. A mechanical dynamic software can ealuate the excitation from the solution of differential equations of the motion of crank mechanism. A great adantage of substructures is that it's possible to modify independently each component of the assembly. A flexible body has been described as a substructure. A constrained modal analysis of the body has been performed, to ealuate modal frequencies and then power spectral density analysis permits to ealuate participation factor of each mode. It's possible to ealuate experimentally a damping coefficient for each modal frequency. Normal constrained modes hae been transformed into normal free-free modes plus a set of interface eigenector modes[ 4]. Rigid body modes hae a ery low frequency (zero) and are normally disabled. The motion of a flexible body can be described as a combination of shape ectors or mode shapes: the position of master nodes of the flexible body is [5]: where: x defines the position ector in the global reference system to the local reference frame, B A is the directions cosines of the orientation between B and the global origin si is the undeformed location of the ith node in B ~ defme the contribution of the mode shapes to the ith node. q is the ector of modal amplitude. Motion can be represented using Euler angles. Generalized coordinates are: where: X is a ector of coordinates of the local body reference frame If/iS the ector of Euler angles ofb relatie to the global origin. q is the ector of modal amplitudes of the m contributing mode shapes. 532
Equations of the motion of the flexible body can be deried from Lagrange's equations:!!_(a~ J _ al + af_ + [ a\f ]T A _ Q = O dt a~ a~ a~ a~ \}' = 0 where: Lis the Lagrangian: L=T-V where T and V denote kinetic and potential energy. e is the energy dissipation function 'I' are the constrain equations..; are the generalized coordinates A. are the Lagrange multipliers for the constraints Kinetic energy can be expressed in generalized coordinates.; : T =.!_~T M(~)~ 2 M(.;) is the mass matrix. l Mu Mtr M= Mrr szmm Mtml Mrm Mmm Where t, r, m refer to translational, rotational, and modal coordinates. Mu=m I where m is the mass and I the identity matrix Mtr =- jpa(s p + <D(P) q )B dv the first term is the center of the mass location multiply by the mas~. ct>(p) is the shape function. M tm = jpa<d(p) dv Mrm =- fpbt(sp +<D(P) qy <D(P) dv Mrr = fpbt(sp +<D(P) qy(sp +<D(P) q)b dv Mrm = fp<d(p)t <D(P) dv Potential energy contains graitational energy and elastic energy: where K is the generalized stiffness matrix (only modal coordinates q participate to elastic energy), and Vg is the graitational energy. The damping force is generally dependent on modal elocityw, according to Rayleigh's dissipation function E>. 1 <I>(P) q is the deformation of a point P of a flexible body {<I>(P) is the shape function, sp is the undeformed position of the point P). 533
E> =.!_wr Dw 2 where: D is damping coefficient matrix and w is the modal elocity ector The asynchronous electric motor has a characteristic speed s. torque cure, so the speed and the power consumption depends on the load. The angular elocity is about 308 rad/sec, een if this is not constant and brushless motors hae also a pulsing torque of double frequency, which has a null mean alue. Motor torque characteristic can be determined in regime, if this function is used to study the start up, there is a high degree of uncertainty. 2 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 Speed Vs. torque diagram ---- ~ torque (t m) / \ / \ - - / \ c-/ \ ~ :----.~ _..\... r~"~"'j -t - - -\- 0 500 1000 1500 2000 2500 3000 Fig. I Motor torque characteristic cure Pressure load on the system has been computed throw a thermo-fluid dynamic software for the specific compressor. So the analysis isn't fully coupled, but can gie a good result in stationary conditions. Pressure (MPa) -- -~ - 20 A(\ t-- \ \.. 1\ \ \ ' "" ~ -- 0 2 -- \olume em). c---- Volume Vs. pressure - - - t---1- ' 3 4 5 6 7 8 I Fig. 2 PV cycle for the refrigerant Rl34a for a hermetic compressor. The model also considers friction of the sliding parts of the system. Friction model torque has been computed from the theory of hydrodynamic lubrication of bearings. The steady incompressible Reynolds equation has been employed with Sommerfeld boundary conditions. where: Purdue Uniersity, West Lafayette, IN, USA July 25-28, 2000 534
h is the film thickness (normally this is h =hmin +hterm cos(q7) + xsin(q7) where h 1 er is due to thermal expansion). w is the speed Than hydrodynamic friction force can be ealuated on shear stress. 3.NUMERICAL AND EXPERIMENTAL RESULTS The first application of the irtual compressor model has been the optimization of balancing. It is possible to optimize the magnitude of a goal function in the frequency domain, for example the fundamental frequency (50Hz) of the goal function. A domestic hermetic compressor has been analyzed arying the counterbalance weight. It is possible to analyze the ariation of the effectie acceleration in some reference points. Defined as coordinate system on this plane x as the piston Numerical and experimental effectie accelerations 8,------.-----.------.------.-----,------.-----.------.------,-----, 7~~~--~--r---r---~~~--~--~--r-~! ~-----:.-- - _, numerical X - - sperimental X I I -~--x -::_:: ~-.. -+-numerical Y -..-- sperimental Y!,: -~-e-,:.c_-~~," - C.:::~ f4 ::~<: I l -----~... ---.. ---.:----------J i 3 ------ --------- --.,. T ~ 2 J---:~----- ---- I 0+-----~----~--~--+-----~-----4------+-----~-----4------+-----~ 350 370 390 410 430 450 470 490 510 530 550 unbalancoment {g mm) Fig. 3 Effectie acceleration: Numerical and experimental data. ~ 0 0 3 0 2 Fig. 4 reference system 0 1 motion direction and y direction orthogonal to x, and defined as origin the center of the crankshaft, maximum effectie acceleration on x and y direction hae been computed and measured. This alue corresponds to the effectie alue of the FFT magnitude diagram at the current frequency (about 50 Hz). Numerical analysis has been performed to find the best counterbalance weight; een if it is always required an experimental alidation of the results and the 2 This compressor uses Rl34a and it has a cooling capacity of 170 kcallh 535
experimental determination of the goal function. Another alidation of the model has been the measure of effectie acceleration of the top of the four supporting springs. These accelerations hae been measured with a particular setup[6] and they hae been calculated numerically. It is possible to notice in tables 1-2 that there is a good correlation. Amplitude (m/sec 2 ) Table 1 Effectie acceleration (measured alues) X y z Phase (deg) Amplitude Phase (deg) Amplitude Phase (deg) (m/sec 2 ) (m/sec 2 ) Spring 1 0.5 0 0.6-140 1.3-130 Spring 2 0.5 0 0.6 106 1.1-50 Spring 3 0.7-80 0.5 130 1.3 50 Spring 4 0.7-80 0.7-130 1.1 130 Table 2 Effectie acceleration (computed alues) X y z Amplitude Phase (deg) Amplitude Phase (deg) Amplitude Phase (deg) (m/sec 2 ) (m/sec 2 ) (m/sec 2 ) Spring 1 0.4 0 0.6-130 1.2-120 Spring 2 0.4 0 0.5 110 1.2-60 Spring 3 0.6-70 0.5 125 1.2 55 Spring 4 0.6-70 0.6-120 1.1 110 The model can calculate FFT diagram of acceleration and the force transmitted to the shell. These informations can be useful for a ibro-acoustic analysis of the system for the ealuation of emitted noise. It is possible to notice in fig. 5 that there is a good correlation for low frequency ibration (under 200Hz). For higher frequency it is important to consider other flexible elements. This will be done in future enhancements of the model. Vibration Analysis 1.E..()7.L...----'---L----'------'---'----'------''-----'----'-- j 0 50 100 150 200 250 Frequency (Hz) 300 350 400 450 500 Fig.S - FFT diagram of acceleration of crankcase 536
3.CONCLUSIONS The first use of the irtual compressor model has a good agreement with experimental results, so this approach can gie an answer to many questions. Similar results for a single model could be achieed with an experimental approach, but a irtual model remarkably improes design capabilities. Further work is under progress to improe and alidate the irtual model of compressor. 4.REFERENCES [1] Craig, R.R. and Bampton, C.C.,Coupling of substructures for Dynamic Analyses, AIAA journal o1.6 no.7 pp. 1313-1319 [2] Adams flex guide Mechanical Dynamics re. 10.0 [3] Ansys theory reference, 8th Edition SAS IP, Inc. [ 4] Mark Baker, Component Mode Synthesis Method for Test-Based, Rigidly connected flexible components Journal of spacecraft ol. 23 No.3 [5] J. T. Spanos, W. S. Tsuha Selection of component modes for flexible multibody simulation AIAAjoumal o1.14 no.2 pp. 278-286 [6] "Running Mode Manual3.5B" LMS Leuen, Belgium 537