Improving mechanical seal reliability with advanced computational engineering tools, part 1: FEA
K-D. Meck and G. Zhu, John Crane EAA, Manchester, UK. The use of finite element analysis (FEA) and computational fluid dynamics (CFD) to predict the performance of mechanical seals has become well established in recent years. This feature describes the background to the development of FEA with integrated fluid film and tribology model at John Crane, the capabilities of the software and examples of the validation from test work. A second part next month will discuss CFD and a case study solved using both techniques.
would be almost impossible. However, analytical models to simulate or calculate performance parameters of mechanical seals exist and computer based FEA and CFD models have been developed for almost 20 years. These numerical models have different levels of sophistication. The motivation behind the development of the software has been: • • • • Increasing seal reliability. Increasing operational duty conditions for seals. Reduced development time providing a quicker response and reduced time to market. Reduced development costs.
Mechanical seals and mechanical seal technology have come a long way, from the earliest application as part of the cooling system in the legendary Ford model T to the high performance sealing tasks in modern turbo gas compressors. To meet current and future challenges, traditional engineering methods are complemented by modern computer based simulation programmes. These, often customised, software programmes allow engineers to simulate and understand mechanical seals and the application in great detail. This feature is the first of two articles that will discuss advanced computer based engineering tools that support the design and application of mechanical seals. This feature describes CSTEDY/ CTRANS, a propriety finite element analysis (FEA) tool with integrated fluid film and tribology model developed by John Crane. A subsequent feature will discuss a commercial computational fluid dynamics (CFD) software package and its application/ adaptation for mechanicals seals in conjunction with the FEA programmes and an example of the solution of an actual seal problem. In addition to the theoretical elements the correlation of seal performance predictions with test results is discussed.
components and materials of construction. In addition the following are required: • • A thermal model of components including heat conduction and transfer. A thermodynamic model, including multiphase and particles, of the fluids surrounding the seal. A hydrodynamic/ tribological model of the sealing interface.
FEA based mechanical seal simulation software
CSTEDY/CTRANS is a finite element based computer programme designed to analyse the operation and performance of mechanical face seals. The software is used for contacting and non-contacting, often known as ‘wet’ and ‘dry’ seals, and also mechanical seals with and without interface grooves. The seal analysis can be performed for steady state conditions where there is continuous operation at constant application condition, by CSTEDY or transient conditions looking at time dependent conditions, such as start-up, using CTRANS.
As if this is not challenging enough, the models need to be fully integrated in an iterative loop, meaning that one model affects the results of the others until all individual models agree with each other or the global model converges respectively. And then it is necessary to consider transient conditions. With such a task ahead and the aim to satisfy all the requirements at the same time, a practical approach to achieving the optimum solution
Simulating the operation of mechanical seals
Developing an analytical model that simulates the performance of a mechanical seal is a highly complex task. It is necessary to start with a mechanical model of the seal with its different 8
Figure 1: Data Input for a typical wet seal, left and graphic output for a non-contacting gas seal, right.
This specialist software combines many features that support the analysis requirements for mechanical seals. leakage. process fluid details and boundary condition that define the environment of the seal. no universal formulae exist to calculate the hydrodynamics of a contacting sealing gap. as well as pressure increase close to the sealing interface exit under
Figure 2: Generic flow diagram of iterative loop used with CSTEDY. water. lapping angle of the seal face and component waviness. Analysis predictions are generally correlated to seal testing on oil. for water vapour/ liquid fraction and methanol. if present. The results are presented as numerical values or graphical interpretation of these values as shown in Figure 1.
Figure 3: Non-uniform deflection as a function of non-axisymmetric features. To solve both models. Component deflections. fluid film stability and condition. etc.
Figure 4: Example of a numerical model. hot water and liquid propane. pressure/ temperature matrix. features that can not be found within commercial. but current equations are the result of permanent review and refinement and many hours of testing. Although it may appear at first that customising a FEA package by linking it with a fluid film analysis is relatively straight forward. the tribology calculation is linked in an iterative loop with the component FEA. multidiscipline FEA software. mechanical and thermal. The pressure distribution across the sealing interface needs to consider effects such as cooling or heating of the gas when it passes through the narrow sealing gap. component deflections that influence the sealing interface shape and consequently influence the interface lubrication condition will itself be influenced by the changed interface fluid film condition.
. also require comprehensive computational solvers. material properties. Some of this work dates back to the 1980’s. Unlike non-contacting seals where the pressure distribution in the sealing gap is calculated using first principle equations. The interface tribology and hydrodynamic fluid film generation is calculated by proprietary equations that include in the calculation parameters such as real fluid properties. sealing surface roughness. are calculated using conventional FEA techniques. Figure 2 shows a generic flow diagram illustrating how the FEA and sealing interface analysis are linked in the programme. the Joule Thompson Effect. Non-contacting gas lubricated seals. Based on this input the software calculates service dependent seal performance values such as power loss.FEATURE
The engineers using this software input a numerical interpretation of the mechanical seal by providing geometric seal component details. specific heat. face gap. in reality this is not the case and it is a particularly challenging task for contacting ‘wet’ seals. particularly at high pressure. The calculation model for contacting seals in this software is based on empirically derived formulae. This means. asperity contact.
Figure 3 shows an example of the axisymmetric compatibility of a primary seal ring for a turbo gas seal at high pressure loading. CFD or correlation with test results.
Figure 7: Test plot of fluid film thickness on OD and ID of the seal in thermo-cycling mode and temperature variation (left). Sealing gap and taper itself are influenced by mechanical and thermal loads. and of course the properties of the fluid sealed. springs. In the case of mechanical seals there are a number of parameters that can be used as performance indicators and measured to validate the analytical model. The seal leakage is influenced by the seal gap. viscosity. including the phase change from water to steam at elevated water temperatures. Leakage. This can be done by 3D single component FEA models.
Figure 5: Face temperature rise for hot water seal. angular alignment. are not modelled but their affect on the seal ring components is acknowledged by applying boundary conditions. The analysis result shows the influence of the non-axisymmetric features on the component flatness and how waviness can be reduced by increasing the number of anti-rotation grooves. seal ring component temperatures are measurements that are intrusive and require additional preparation of the seal. A number of simplifications and assumptions are made when a seal model is constructed: 10
The component deflection. • Adaptive components such as a sleeve. This technique verifies whether the axisymmetric model is
. CTRANS simulation of the test seal (right). power consumption and face component temperature are parameters that can be measured without interfering with the operation of the seal. is treated as axi-symmetric. concentricity.FEATURE
compatible with the ‘real’ component deflection. hydrodynamic grooves such as LaserFace or Spiral Groove etc. grooves etc. Seal leakage is very sensitive to the height and shape of the sealing gap as it is an h3 function in the Reynolds equation:
Figure 6: Leakage prediction vs. For most seal designs this is a successful method in particular when boundary conditions have been validated by additional single component FEA. Other effects considered in the mathematical solver model that influence the sealing gap include waviness and angular misalignment. mechanical and thermal. test results for a large dry gas agitator seal. secondary seal friction. as a function of pressure and temperature as shown in Figure 4.
choked flow conditions. The results illustrate the ability of the model to simulate the actual seal performance closely. secondary seals. surface finish. taper of the sealing interface. seal plate. A very powerful measure to verify the accuracy of the seal simulation is the correlation of seal leakage with the measured results. although. non-axisymmetric distortions. thermal conductivity etc. including waviness. Figure 5 shows an example of the correlation between the numerical predicted face temperature rise and test data for a contacting plain face seal. The fluid is modelled by using its fluid properties such as density.
Simplification of the Analysis Model
As with any comprehensive numerical model simplification is introduced to manage the complexity of the analyses. This is for most seal components a good estimate because geometry is primarily axi-symmetric. etc. Particles carried in the fluid are not considered. test compared with a CSTEDY prediction. which can introduce non-axisymmetric deflection of the components need to be carefully assessed.
Validation of the model
Every numerical simulation model is only as valid as its ability to produce results that correlate with the real performance of the component.
Non-axisymmetric features such as drive lugs.
The analysis included the fluid film variation due to the introduced component waviness. hysteresis. 1992. Parmar. not taking into account other parameters that influence the differential pressure. Volume 213.FEATURE
This means that to predict the leakage of a mechanical seal to within an accuracy of 10%. N. M17 1SS. Similar good correlation is achieved for contacting ‘wet’ seals and non-contacting high performance ‘wet’ and turbo gas seals. 11
. UK. BHRA. Tel: +44 161 886 5993. contacting
Figure 8: Test data of circumferential fluid film variation of a turbo gas seal during coast down with introduced seal face waviness and Talyrond measurement of the prepared seal component. Web: firstname.lastname@example.org. Zhu ‘Computer Prediction of Mechanical Seal Performance and Experimental’. Figure 6 shows an example of the performance correlation between leakage prediction by this software and test data for a large diameter dry gas agitator seal. 433-449. Doust and A.johncrane. Such information is however essential to develop or improve an analytical hydrodynamic model of the interface fluid film.
CSTEDY and CTRANS are registered trade names of John Crane LaserFace is a registered trade mark of John Crane. 11th International Conference on Fluid Sealing’. Manchester. nor can the leakage information distinguish whether the sealing interface is constant and stable at a given operating condition or transiently changing. 2003. EDF-LMS. the approximation of the gap height has to be within an accuracy of about 3%. Figure 8 shows an example of the fluid film measurements with a turbo gas seal. 4.
1. G. Workshop: High Performance Rotary Shaft Seals. Figure 9 shows the correlation with the measured mean film thickness of the above seal and the predicted performance. Fax: +44 161 872 1654. Although a powerful performance indicator such as leakage can be used to validate an analytical seal model. ‘Thermal Cycling In Mechanical Seals – Causes. was first introduced to develop the hydrodynamic equations for wet seals and later adapted to turbo gas seals. gradient and the viscosity.
Contact: John Crane EAA. UK. Meck. The small difference in leakage. leakage as a single performance value can not give enough information to derive the exact shape of the sealing gap during operation. Belgium. where the fluid film stability at low pressure was studied using a seal with introduced seal component waviness. Nash Road.
The presented propriety software is a FEA based computer tool with integrated fluid film and tribology models developed and refined over the past 20 years.
p. ISSN 1350-6501. A. 2. It serves as important tool in design and development. Figure 7 shows the changing dynamic sealing gap of a seal operating in a critical thermo-cycling mode measured by the capacitance probes and the simulation of the seal with CTRANS. Parmar. during pressurisation and depressurisation is explained by friction effects from the secondary balance diameter seal.uk. Experiment and Modelisation. This technology. High Performance Sealing’. Both test examples and the correlation to the predicted performance underlines the ability of the software to simulate the operation of any type of mechanical seal. using capacity probes inserted into the stationary seal face.
Figure 9: ID sealing gap correlation between test and software prediction. seal reliability and gives engineers the capability to study seals and their performance under simulated service conditions. BHR Group Ltd. µ. to a very high degree of accuracy. T. Fluid film measurements were taken throughout the speed range of the seal from 0 to 9000 rpm.M. 3. 1987. wet or gas. Proceedings of the I MECH E Part J Journal of Engineering Tribology. For this purpose Doust and Parmar introduced a technology that allows the direct measurement of the fluid film height and shape of the sealing interface during operation of a mechanical seal. Prediction. France.uk. 1999. Wallace and K-D. ‘Transient Thermoelastic Effect in a Mechanical Face Seal’. This applies to steady state as well as transient seal performance simulations. Prevention’ 13th International Conference on Fluid Sealing. Email: louise. The computational engineering tool discussed has been shown to simulate mechanical seal operation to a good degree of accuracy.co. This is confirmed by the validation test results. ‘LaserFace. The knowledge how the interface shape is changing as a function of time is essential to compute and predict transient seal behaviour. Trafford Park.G.