A Comparative Study of Valve Natural Frequency Estimation Using Finite Element Analysis (FEA), Raleigh’s Principles and Laboratory Tests

The natural frequency of valves is an important design requirement to ensure that valves do not go into resonance during operation and consequently fail structurally or fail to perform their design and safety related functions. Besides its impact on operability, valve resonance can initiate piping vibration that could damage pipes and their supports; which is undesirable. As important as equipment natural frequency is to valve operability, one would expect that testing should be the de facto method for confirming its value. Ideally, this should be the case, however, cost considerations limit the extent to which testing is used. On the other hand, testing does have some issues with respect to accuracy such as the effect of supporting structure flexibility resulting in a conservatively lower natural frequency measurement. In addition, the multiplicity of valves in nuclear power plants with different designs, sizes and safety classes limit the use of testing to establish valve natural frequencies except when required in the equipment specifications. Frequently, valve natural frequencies are determined by analysis either using finite element techniques (FEA) or by first principles of beam and mass models; the latter being more frequently used. This paper presents the studies performed to correlate valve natural frequency test results to the results derived from analytical techniques using Raleigh’s energy principle and from finite element analysis (FEA) methods. In a previous paper on valve natural frequency [1], Ezekoye et al. presented a model for estimating valve natural frequency by incorporating mass inertia of the valve structures with the more traditional methods that are based on a lumped mass model to determine displacements. In the process, the flexibility of the extended structure (otherwise referred to as the superstructure) and the valve body itself are considered. Using limited test data, Ezekoye et al. showed that there is merit in using their enhanced analysis model. Their correlation was promising. The finite element analysis, on the other hand, is a well-established technique for solving complex structural mechanics problems and should be expected to provide reasonable results comparable to actual valve tests provided the boundary conditions provide a reasonable representation of the actual valves tested. In this paper, ANSYS Version 12.1 was used to model valve natural frequencies. Additionally, a more extensive testing of valves for natural frequency was performed in this paper than was reported in Reference 1. The results of both the FEA and the Raleigh’s principle model as presented in Ezekoye et al. are compared against the test results. By comparing the three results, strengths and weaknesses of each method become apparent. The choice of whether or not one chooses to test or perform analysis depends on the valve specification requirement and the preference of the designer.