CFD a Viable Engineering Tool for Compressor Valve Design or Just a Toy
暂无分享,去创建一个
The single most important technical feature of a compressor valve is its available effective valve area. This area depends on the primary design of the sealing element (type, spacing ... ) and the lift at which the valve is used. The traditional method of developing a new design is by means of an analysis of the available lift area and an estimation of the effective valve area by compensating for flow losses through loss coefficients which are known from experience. Any such new design has to be evaluated in a prototype development phase and the actual available valve area has to be verified by means of a measurement in a wind tunnel. If the testing reveals undesirable results the design has to be reevaluated, a new prototype has to be built and the flow testing has to be repeated. Today's demands for reduced time to markets of new products triggered a need for a faster method of this development process. One obvious approach for achieving this goal is by means of a numeric experiment thus solving the governing fluid mechanic equations numerically instead of building an actual prototype. The standard equations describing the three-dimensional stationary flow situation through a compressor valve were already derived in the 18th century. However, it is known that for most practical applications these equations show a chaotic behavior, which is usually referred to by the term "turbulence". Several models exist which try to account for this phenomena. Nevertheless none of these models is able to perfectly simulate a flow situation in which turbulence plays a major role. The purpose of this paper is to compare the results of a "state of the art" commercially available CFD software with the results of wind tunnel testing for several common valve types (reed-, poppet-, ringand plate valve). The comparison also takes into account the different available turbulence models and shows how these models change the outcome of the calculations. Therefore a recommendation will be given on the usefulness of this type of tool for the valve design process. Q .. . 6.p .. . ••• p ... Volume flow pressure loss across valve effective valve area gas density NOMENCLATURE INTRODUCTION The process for the market release of a new compressor valve usually consists of the prototype phase, in which different design principles are investigated and the field test phase, in which the life time requirements have to be met. The prototype phase consists of finding a compromise between the available effective valve area Bauer [ 1] and Lin [3], the required spring force for the closing of the sealing element, the allowable valve losses, the allowable clearance volume, the required MTBF and the required operating conditions. In order to define the available effective valve area an iterative process has to be carried out during which this area is defined through initial flow assumptions, prototype manufacturing and design refinements until the required effective area is achieved. Fifteenth International Compressor Engineering Conference at Purdue University, West Lafayette, IN, USAJuly 25-28, 2000 423 The initial expected effeCtive valve area is usually found through the evaluation of the available lift area and the anticipation of loss coefficients which are known from experience and or prior measurements of similar configurations. For every design option a prototype has to be built to evaluate the actual effective valve area. The closer a new design is to an existing one the smaller is the difference between anticipated and actual effective area. For a totally new valve approach the accuracy of the traditional approach is usually +/-10% accurate and therefore it is not uncommon that it takes several production and evaluation cycles for a new prototype design. Cyklis [2] has shown that the CFD method gives accurate results for a flat ring valve design. He solved the steady state two-dimensional equations for a compressible flow. Deschamps, Ferreira, and Prata, [4] have solved the time averaged Navier-Stokes Equations for a radial diffuser with axial feeding and have proved that the results are in good agreement with experimental results. They used the RNG k-e turbulence model Orzag et. al [8] and report that this model improves the prediction of separation regimes. Due to these promising results a project has been undertaken with the goal to evaluate whether a commercially available CFD computer program can be used as an engineering tool to reduce the lead time for a new prototype. The automotive and airplane industry has been using these tools for decades. However the amount of necessary manand computing power has until recently made the CFD technology unattractive for most of the other industries. Primary to this project a comparison of the available low end CFD packages was undertaken. The requirements for the CFD program were defined as follows: 0 Low end CFD package with an interface to standard 3D CAD system (SAT) o Program must run on a High End PC 0 Results of the effective flow area of a simple sample valve has to be achievable within a reasonable amount of time The program which offered the best compromise between price and usability was purchased. VALVE DESIGNASSUMPTIONS The flow through a compressor valve during the suction intake phase or discharge phase can be described as a three dimensional, time dependent, turbulent compressible flow. During the design phase where the effective area of a new valve is defined, the compressibility of the fluid and the instantaneous flow are usually neglected Bauer [1]. (An accurate modeling of these effect is beyond today's standard know-how and standard capabilities.) Thus the main criteria for judging the aerodynamic efficiency of a new design is the effective valve area (
[1] Cesar J. Deschamps,et al. Turbulent Flow Through Valves of Reciprocating Compressors , 1996 .
[2] P. Cyklis. CFD Simulation of the Flow Thr ough Reciprocating Compressor Self-Acting Valves , 1994 .
[3] Friedrich Bauer. Valve Losses in Reciprocating Compressors , 1988 .
[4] F. Boysan,et al. Renormalization Group Modeling and Turbulence Simulations. , 1993 .