- Email: [email protected]
Assessment of Life Prediction Methods for Elastomeric Seals – A Review
CHAPTER 4
Assessment of Life Prediction Methods for Elastomeric Seals – A Review JR Daley Trelleborg Sealing Solutions, Ashchurch, Tewkesbury, Glos., UK
SYNOPSIS The prediction of the life of a sealing element within a system is required to allow planned maintenance to be carried out at the most efficient time. In the automotive industry there is a trend for seals to last for ever increasing amounts of time, with the latest goal being 15 years (150 000 miles). With greater life times required, methods have to be adopted to predict the life of the seal in application, and accelerate testing to allow this to be completed in an acceptable time period. In this chapter several common methods of prediction for static seals are reviewed and their applicability to the actual sealing application environment assessed.
1 INTRODUCTION Increasingly, the sealing industry is being faced with the question of component service life: how long will components last? This apparently simple question is one which is certainly difficult to answer and impossible to without putting a considered margin around the life prediction figure. Static seals, such as gaskets and O-ring seals, are used in all industries. Restricting ourselves to static seals eliminates the complications of dynamic interactions and factors such as wear, friction and tribological effects (Hertz, 1992), which all add to the degradation of the seal (and mating surface), and hence may affect seal life (Coveney & Menger, 1999, 2000; Menger, 2001). For static seals the primary considerations are the changes in the properties of the elastomer with time, and the effect of these changes have on the sealing ability of the seal. Any of the following, and combinations thereof, can lead to seal failure. • • • •
Relaxation and compression set of the seal Excessive hardening Chemical attack Explosive decompression (ED) 51
52 • •
Elastomers and Components: Service Life Prediction – Progress and Challenges Incompatible thermal expansion Extrusion damage
Service life of elastomer seals is limited by the interaction of the polymer with its environment (temperature, pressure and chemical), and is affected by changes in the characteristics of the elastomer over time, and the design of the seal. General discussions on material selection for seals are given elsewhere (see for example Eason & Barnes, 1994). Here I consider seal life, and the methods of predicting the life expectancy of static seals.
2 STRESS RELAXATION AND SET Compression set and compressive stress relaxation are normally the key issues when considering static seals. [Coveney & Rizk (2005) and Morgan et al (2005) consider other causes of failure.] The loss of the sealing force during the life of a seal under compression is well known (Bunting et al, 1992). There will in the seal’s life come a point when the retained sealing force is inadequate to overcome the leakage pressure of the fluid, and hence the life of the seal is at an end. The actual point at which this occurs is, however, open to debate, and is certainly a function of applied pressure – low pressure sealing is particularly susceptible to leakage, when the material has lost compliance to provide sealing, in comparison with higher pressure applications where the seal is often forced into providing a high enough contact stress to prevent leakage [pressure activated sealing (Coveney & Rizk, 2005)]. Stress relaxation (and set) can be considered to be as a result of “physical” (i.e. viscoelasticity-related) and chemical effects. Generally, physical relaxation proceeds approximately linearly (but perhaps more accurately logarithmically) with logarithmic time and is often overtaken by chemical relaxation which proceeds approximately linearly with linear time (Stevenson & Campion, 1992). In order to predict the failure point in the future, Arrhenius, and WLF (Williams-LandelFerry transform) methods may be used [Huy & Evrard, 2000 and Equation (1)]. Arrhenius plots are mainly used to predict chemical, e.g. oxidative ageing [Albihn, 2005 and Equation (2)] while the WLF transform is often employed to predict viscoelastic effects. In practical applications there is often a combination of both physical and chemical effects occurring in parallel, which can lead to difficulties. The stress relaxation curves for an elastomeric material, are shown in Figure 1. The three curves correspond to the three test temperatures (160°C, 180°C and 200°C), with the higher test temperature resulting in the highest relaxation rates. The data was collected using continuous compression stress relaxation (CSR) measurement as outlined by Spetz (2000). The results below are from tests carried out in air, in application seals are used in numerous fluids and, where exposure is seen, the compression tests must be carried out in the fluid concerned (Stevens, 2000) to ensure any related physico-chemical effects are taken into account. A (WLF based) shift of data is shown in Figure 2 – from 200°C, where the most data was collected, to the lower temperatures of 180°C and 160°C. [See Equation (1), below and Ferry, 1992.] In Figure 2 the measured data is shown in black, and the transposed data in grey. Increasing the test temperature accelerates the viscoelastic relaxation rate of the material. At the lower temperatures, more time is required than at higher temperatures for the material to reach the same state of stress relaxation. It may be seen that to reach a 50% loss in initial force
53
Fn
Assessment of Life Prediction Methods
Fig. 1 Compressive stress relaxation (CSR), plot of force (Fn) normalised to initial value against time (t) in hours of a fluorocarbon elastomer FKM [upper, middle and lower curves are for 160°C, 180°C and 200°C respectively.]
1 0.8
Fn
0.6 0.4 0.2 0 1
10
100 t (h)
1000
10000
Fig. 2 WLF transform applied to CSR data from FKM – q.v. Figure 1. Plot of force (Fn) normalised to initial value against WLF-shifted time (t in hours). [Upper, middle and lower curves are for 160°C, 180°C and 200°C respectively.]
requires 150 hours at 200°C, and 2000 hours at 160°C. Hence a significant reduction in laboratory test time is achievable, provided that the data may be accurately extrapolated. The data from the shorter term, higher temperature test is used by transposing it onto the lower temperature test results. The curves fit well, and the form of the curves are correct, and this method may be used for extrapolation. Both the WLF [Equation (1)], and the Arrhenius [Equation (2)] plots allow a prediction of long term material behaviour using measurements made over shorter timescales, such as
54
Elastomers and Components: Service Life Prediction – Progress and Challenges
seen in Figure 1. The extrapolations are made on the assumption that the methods are valid and that there that there are no unaccounted for effects impinging on the relationship between the material property and time. Unfortunately, the methods of extrapolation suffer from two major sources of error. One of the sources of error relates to the paucity of data. Testing is time-consuming and in the case of the Arrhenius method only one data point is utilised per test – e.g. the point at which a given property falls to 50% of its initial value. Thus practical limits are imposed on precision. The second source of error concerns the extrapolation of the data points from short term (more accurate) to longer (increasingly less accurate) time frames (see also Albihn, 2005). Moreover, during long tests problems with fluctuating conditions, drift etc can be significant. One form of the WLF transform equation is (Ferry, 1992):
(1) where C1 and C2 are material constants, T is the temperature, and T0 is the reference temperature. The Arrhenius equation is: (2) where k is the rate (e.g. of a chemical reaction), A is a constant, Ea is the activation energy of the mechanism, R is the universal gas constant, T is the absolute temperature. If the failure process conforms to the Arrhenius equation it follows that a plot of log(life) against T -1 is a straight line (see Albihn, 2005). By means of Arrhenius plots, service lives are often extrapolated from the material’s property curve, for the temperature required. However (see above and Albihn, 2005) these extrapolations are not without risk. The data in Figure 1 has been transferred to such a plot in Figure 3. Despite the fact that the stress relaxation process for this material are believed to be primarily “physical” the Arrhenius plot is approximately straight, suggesting the utility of the method. [Mechanisms of stress relaxation are discussed by Fuller (1988), Chapman & Porter (1988), Barnard & Lewis (1988) and Azura & Thomas (2005).] The “failure” criterion is often conservatively set at 50% loss of original force. The graph indicates the expected time it would take, at a given temperature, for 50% of the original compressive force to be lost. In reality, significantly more than 50% of the original value must be lost for leakage to occur. These two methods (WLF transform and Arrhenius plot) give methodologies for the use of reduced time laboratory test data. An indication of the relaxation / set behaviour of the elastomer or seal is thereby obtained but these methods do not give a prediction of whether the seal will operate for a given period of time. Rather, a comparison of materials is given. With this data the choice of one material over another for a given target life may be made. In order to make satisfactory life predictions for the seal an understanding of modes of operation and failure is required. Finite element analysis (FEA) can greatly assist - especially in the case of complex geometries.
55
t (yrs)
Assessment of Life Prediction Methods
Fig. 3 Arrhenius plot applied to CSR data from FKM – q.v. Figure 1. Time (t) to 50% of original force against reciprocal of absolute temperature (T). [Extrapolation to T-1 = 0.0027 K-1 corresponds to 97ºC.]
3 FINITE ELEMENT ANALYSIS In order to estimate the life of the seal non-linear finite element analysis methods (FEA or FEM) have been used. FEA allows the modelling of both simple and complex crosssections of seals, and hence is useful for assessing the change of stress within complex seal geometry and the effect on its sealing performance. For FEA performed on a seal, when the sealing force per unit area (contact pressure) becomes less than the system pressure this is generally taken to indicate the onset of leakage. It should be noted that in FE models the seal surface and the mating faces are assumed to be perfectly smooth – ignoring the possibility of “seepage” along a scratch on the metal housing, for example. Also the possibility of adhesion is generally ignored. In order to accurately model the life of a static seal, material data is key. Initial data may be precisely tested for, and included in the model (Daley & Mays, 1999), but the material properties change over the life of the seal. The seal may swell due to interaction with the contained fluid, the physical properties can change due to oxidation, and fluid effects, and the sealing force will drop due to stress relaxation. Although many have modelled specific elements of a seal in operation, no one has put a unified FEA model together to fully simulate the life of a seal. The various aspects, which have been modelled, in addition to hyperelastic behaviour, include: permeation (Ho, 2001; Ho & Nau, 1996); explosive decompression (Ho, 2000); fatigue (Busfield, 2001). Modelling of the aspects listed above has been done, with good reason, to assess the most likely failure mode for the application. However, without full interactive account being taken of the chemical and physical changes within the seal, uncertainty is introduced in the precision of the prediction. As discussed above, in the case of a static seal, compressive relaxation is generally taken as the probable cause of failure. Sealing geometries range from the simple – such as the O-ring seal, as shown in Figure 4, to more complicated designs and loadcases – such as the gasket shown in Figure 5. [For
56
Elastomers and Components: Service Life Prediction – Progress and Challenges
Fig. 4 Plots for 2-D hyperelastic finite element analysis (FEA) of an O-ring seal in pre-relaxed state (left) and following stress relaxation (right). Cauchy 22 stresses are shown (MPa). The FE code used was MARC.
Fig. 5 Stress plot for 3-D hyperelastic finite element analysis (FEA) of a gasket. The FE code used was MARC.
a discussion of hyperelastic modelling and FEA see Gough et al, 1999; Cadge & Prior, 1999.] The stress pattern in Figure 4 is simpler, more uniform, and easier to interpret, than the multifacetted loading of the gasket in Figure 5 – nevertheless the advantages of FEA are clear.
4 OTHER FAILURE MODES Other failure modes, such as explosive decompression or extrusion damage of the seal, are features which should be anticipated in design (Morgan et al, 2004). It is however the case
Assessment of Life Prediction Methods
57
that the seal environment predicted and the actual conditions seen are not necessarily the same, and hence such failures do occur. Chemical attack of seals is seen in more aggressive environments, such as occur in the chemical industry, and here a service life of months rather than years is usually accepted. In considering more expensive materials, such as perfluoroelastomers, the cost / life benefit in the application must be weighed up (Daley & Phillips, 2000). Although use of perfluoroelastomers is a fairly extreme example, the type of elastomer must fit the application – and relevant fluids and temperatures – to ensure that chemical degradation of the seal is not accelerated. Further failure modes include mismatch of thermal expansion, especially if the elastomer seal does not recover quickly enough under thermal loading. A simple case is cold start testing, where the components are cooled, usually to -40°C, and then the engine is started. If the housing around the seal expands at a faster rate than the seal can follow, then a leakage path will occur. Low temperature elasticity of elastomers is commonly measured through low temperature retraction (TR) tests (ISO 2921,1997). In addition to the above factors which can be designed against, there can be damage to seals through assembly, transport or storage, which reduces their life span.
5 CONCLUSIONS Methods do exist to assess the lifetime of elastomer materials and seal designs, through accelerated testing and analysis. However in order to utilise the laboratory results there must be consideration of the environment that the seal actually experiences. In use, a seal will be subjected to a range of operating conditions; this range must be applied to the theoretical methods of seal life prediction. Therefore a combination of laboratory material tests, application testing, mathematical modelling, and application knowledge and experience is required for life prediction. Even then, this is a best guess estimate, due to uncertainties in the final application environment. Therefore, in practice, any prediction work carried out can only consider samples which experience typical or representative conditions which will occur in the application environment.
REFERENCES Albihn P (2005) “The 5-year accelerated ageing project for thermoset and thermoplastic elastomeric materials: A service life prediction tool” (This volume). Azura AR & Thomas AG (2005) “Effect of heat ageing on the strength properties of elastomeric components” (This Volume). Barnard D & Lewis PM (1988) “Oxidative ageing”, Chapter 13 in Natural Rubber Science and Technology, AD Roberts (ed), Oxford University Press, 621-678. Bunting WM, Russell WD, Doin JE, Tate AL, & Slocum GH (1992) “Compression Stress Relaxation I, An Important Test For Evaluation of Sealant Materials”, Society of Automotive Engineers Standard, 920135. Busfield JJC & Thomas AG (2001) “Using FEA techniques to predict failure in elastomers”, Service Life Prediction of Elastomer Components, Institute of Materials, 15 October, London, UK. Cadge D & Prior A (1999) “Finite element modelling of three-dimensional elastomeric components” in
58
Elastomers and Components: Service Life Prediction – Progress and Challenges Finite Element Analysis of Elastomers, D Boast and VA Coveney (eds), Professional Engineering Publishing, London, ISBN 1860581714, 187-205.
Chapman AV & Porter M (1988) “Sulphur vulcanization chemistry”, Chapter 12 in Natural Rubber Science and Technology, AD Roberts (ed), Oxford University Press, 511-620. Coveney VA & Menger C (1999) “Initiation and Development of Wear of an Elastomeric Surface by a Blade Abrader”, Wear 233-235, 702-711. Coveney VA & Menger C (2000) “Behaviour of Model Abrasive Particles Between a Sliding Elastomer surface and a Steel Interface”, Wear 240, 72-79. Coveney VA & Rizk R (2005) “Life prediction of O-rings used to seal gases” (This Volume). Daley JR & Phillips HM (2000) “Cost-effective perfluoroelastomer sealing solutions for aggressive environments”, Proceedings of the 16th International Fluid Sealing Conference, Brugge, 18-20 September, R Flitney (ed), Professional Engineering Publishing, ISBN 1860582540, 135-146. Daley JR & Mays S (1999) “The Complexity of Material Modelling in the Design Optimisation of Elastomeric Seals”, in Finite Element Analysis of Elastomers, D Boast and VA Coveney (eds), Professional Engineering Publishing, London, ISBN 1860581714, 119-128. Eason RJH & Barnes N (1994) “Seals and Sealing”, Chapter 15 Section 2 in Mechanical Engineers Reference Book, EH Smith (ed), Butterworth-Heinemann, Oxford, 15-18 & 15/75. Ferry JD (1992) “Elastomers: dynamic mechanical properties” in Concise Encyclopaedia of Polymer Processing and Applications, PJ Corish (ed), Pergamon, Oxford, 252-256. Fuller KNG (1988) “Rheology of raw rubber”, Chapter 5 in Natural Rubber Science and Technology, AD Roberts (ed), Oxford University Press, 141-176. Gough J, Gregory IH & Muhr AH (1999) “Determination of constitutive equations for vulcanized rubber” in Finite Element Analysis of Elastomers, D Boast and VA Coveney (eds), Professional Engineering Publishing, London, ISBN 1860581714, 5-26. Hertz DL (1992) “Introduction”, Chapter 1 in Engineering with Rubber – How to Design Rubber Components, AN Gent (ed), Hauser, Munich, 1-9. Ho E & Nau BS (1996) “Gas emission by permeation through elastomeric seals”, Tribology Transactions, 39 (1), 180-186. Huy ML & Evrard G (2000) “Methodologies for lifetime predictions of rubber using Arrhenius and WLF models”, Lifetime estimation of rubber materials by ageing and stress relaxation tests, 31 October, Sweden. ISO 2921 (1997) “Rubber, vulcanized – determination of low temperature characteristics – temperatureretraction procedure (TR test)”, International Standards Organisation. Menger C (2001) “Behaviour of sliding seals in abrasive fluids”, PhD Thesis, University of the West of England, Bristol, UK. Morgan GJ, Campion RP & Derham CJ (2005) “Stress-induced phenomena in elastomers and their influence on design and performance” (This Volume). Spetz G (2000) “Stress Relaxation – Test methods, instruments and lifetime estimation”, Proceedings of Elastocon Conference, Sweden, 31 Oct-1 Nov. Stevens RD (2000) “Permeation and Stress Relaxation Testing of Fuel Seal Materials – A Comparison of Fuel System Elastomers”, RAPRA Conference Birmingham, UK, May. Stevenson A & Campion RP (1992) “Durability”, Chapter 7 in Engineering with Rubber – How to Design Rubber Components, AN Gent (ed), Hauser, Munich, 169-207.
Assessment of Life Prediction Methods for Elastomeric Seals – A Review JR Daley Trelleborg Sealing Solutions, Ashchurch, Tewkesbury, Glos., UK
SYNOPSIS The prediction of the life of a sealing element within a system is required to allow planned maintenance to be carried out at the most efficient time. In the automotive industry there is a trend for seals to last for ever increasing amounts of time, with the latest goal being 15 years (150 000 miles). With greater life times required, methods have to be adopted to predict the life of the seal in application, and accelerate testing to allow this to be completed in an acceptable time period. In this chapter several common methods of prediction for static seals are reviewed and their applicability to the actual sealing application environment assessed.
1 INTRODUCTION Increasingly, the sealing industry is being faced with the question of component service life: how long will components last? This apparently simple question is one which is certainly difficult to answer and impossible to without putting a considered margin around the life prediction figure. Static seals, such as gaskets and O-ring seals, are used in all industries. Restricting ourselves to static seals eliminates the complications of dynamic interactions and factors such as wear, friction and tribological effects (Hertz, 1992), which all add to the degradation of the seal (and mating surface), and hence may affect seal life (Coveney & Menger, 1999, 2000; Menger, 2001). For static seals the primary considerations are the changes in the properties of the elastomer with time, and the effect of these changes have on the sealing ability of the seal. Any of the following, and combinations thereof, can lead to seal failure. • • • •
Relaxation and compression set of the seal Excessive hardening Chemical attack Explosive decompression (ED) 51
52 • •
Elastomers and Components: Service Life Prediction – Progress and Challenges Incompatible thermal expansion Extrusion damage
Service life of elastomer seals is limited by the interaction of the polymer with its environment (temperature, pressure and chemical), and is affected by changes in the characteristics of the elastomer over time, and the design of the seal. General discussions on material selection for seals are given elsewhere (see for example Eason & Barnes, 1994). Here I consider seal life, and the methods of predicting the life expectancy of static seals.
2 STRESS RELAXATION AND SET Compression set and compressive stress relaxation are normally the key issues when considering static seals. [Coveney & Rizk (2005) and Morgan et al (2005) consider other causes of failure.] The loss of the sealing force during the life of a seal under compression is well known (Bunting et al, 1992). There will in the seal’s life come a point when the retained sealing force is inadequate to overcome the leakage pressure of the fluid, and hence the life of the seal is at an end. The actual point at which this occurs is, however, open to debate, and is certainly a function of applied pressure – low pressure sealing is particularly susceptible to leakage, when the material has lost compliance to provide sealing, in comparison with higher pressure applications where the seal is often forced into providing a high enough contact stress to prevent leakage [pressure activated sealing (Coveney & Rizk, 2005)]. Stress relaxation (and set) can be considered to be as a result of “physical” (i.e. viscoelasticity-related) and chemical effects. Generally, physical relaxation proceeds approximately linearly (but perhaps more accurately logarithmically) with logarithmic time and is often overtaken by chemical relaxation which proceeds approximately linearly with linear time (Stevenson & Campion, 1992). In order to predict the failure point in the future, Arrhenius, and WLF (Williams-LandelFerry transform) methods may be used [Huy & Evrard, 2000 and Equation (1)]. Arrhenius plots are mainly used to predict chemical, e.g. oxidative ageing [Albihn, 2005 and Equation (2)] while the WLF transform is often employed to predict viscoelastic effects. In practical applications there is often a combination of both physical and chemical effects occurring in parallel, which can lead to difficulties. The stress relaxation curves for an elastomeric material, are shown in Figure 1. The three curves correspond to the three test temperatures (160°C, 180°C and 200°C), with the higher test temperature resulting in the highest relaxation rates. The data was collected using continuous compression stress relaxation (CSR) measurement as outlined by Spetz (2000). The results below are from tests carried out in air, in application seals are used in numerous fluids and, where exposure is seen, the compression tests must be carried out in the fluid concerned (Stevens, 2000) to ensure any related physico-chemical effects are taken into account. A (WLF based) shift of data is shown in Figure 2 – from 200°C, where the most data was collected, to the lower temperatures of 180°C and 160°C. [See Equation (1), below and Ferry, 1992.] In Figure 2 the measured data is shown in black, and the transposed data in grey. Increasing the test temperature accelerates the viscoelastic relaxation rate of the material. At the lower temperatures, more time is required than at higher temperatures for the material to reach the same state of stress relaxation. It may be seen that to reach a 50% loss in initial force
53
Fn
Assessment of Life Prediction Methods
Fig. 1 Compressive stress relaxation (CSR), plot of force (Fn) normalised to initial value against time (t) in hours of a fluorocarbon elastomer FKM [upper, middle and lower curves are for 160°C, 180°C and 200°C respectively.]
1 0.8
Fn
0.6 0.4 0.2 0 1
10
100 t (h)
1000
10000
Fig. 2 WLF transform applied to CSR data from FKM – q.v. Figure 1. Plot of force (Fn) normalised to initial value against WLF-shifted time (t in hours). [Upper, middle and lower curves are for 160°C, 180°C and 200°C respectively.]
requires 150 hours at 200°C, and 2000 hours at 160°C. Hence a significant reduction in laboratory test time is achievable, provided that the data may be accurately extrapolated. The data from the shorter term, higher temperature test is used by transposing it onto the lower temperature test results. The curves fit well, and the form of the curves are correct, and this method may be used for extrapolation. Both the WLF [Equation (1)], and the Arrhenius [Equation (2)] plots allow a prediction of long term material behaviour using measurements made over shorter timescales, such as
54
Elastomers and Components: Service Life Prediction – Progress and Challenges
seen in Figure 1. The extrapolations are made on the assumption that the methods are valid and that there that there are no unaccounted for effects impinging on the relationship between the material property and time. Unfortunately, the methods of extrapolation suffer from two major sources of error. One of the sources of error relates to the paucity of data. Testing is time-consuming and in the case of the Arrhenius method only one data point is utilised per test – e.g. the point at which a given property falls to 50% of its initial value. Thus practical limits are imposed on precision. The second source of error concerns the extrapolation of the data points from short term (more accurate) to longer (increasingly less accurate) time frames (see also Albihn, 2005). Moreover, during long tests problems with fluctuating conditions, drift etc can be significant. One form of the WLF transform equation is (Ferry, 1992):
(1) where C1 and C2 are material constants, T is the temperature, and T0 is the reference temperature. The Arrhenius equation is: (2) where k is the rate (e.g. of a chemical reaction), A is a constant, Ea is the activation energy of the mechanism, R is the universal gas constant, T is the absolute temperature. If the failure process conforms to the Arrhenius equation it follows that a plot of log(life) against T -1 is a straight line (see Albihn, 2005). By means of Arrhenius plots, service lives are often extrapolated from the material’s property curve, for the temperature required. However (see above and Albihn, 2005) these extrapolations are not without risk. The data in Figure 1 has been transferred to such a plot in Figure 3. Despite the fact that the stress relaxation process for this material are believed to be primarily “physical” the Arrhenius plot is approximately straight, suggesting the utility of the method. [Mechanisms of stress relaxation are discussed by Fuller (1988), Chapman & Porter (1988), Barnard & Lewis (1988) and Azura & Thomas (2005).] The “failure” criterion is often conservatively set at 50% loss of original force. The graph indicates the expected time it would take, at a given temperature, for 50% of the original compressive force to be lost. In reality, significantly more than 50% of the original value must be lost for leakage to occur. These two methods (WLF transform and Arrhenius plot) give methodologies for the use of reduced time laboratory test data. An indication of the relaxation / set behaviour of the elastomer or seal is thereby obtained but these methods do not give a prediction of whether the seal will operate for a given period of time. Rather, a comparison of materials is given. With this data the choice of one material over another for a given target life may be made. In order to make satisfactory life predictions for the seal an understanding of modes of operation and failure is required. Finite element analysis (FEA) can greatly assist - especially in the case of complex geometries.
55
t (yrs)
Assessment of Life Prediction Methods
Fig. 3 Arrhenius plot applied to CSR data from FKM – q.v. Figure 1. Time (t) to 50% of original force against reciprocal of absolute temperature (T). [Extrapolation to T-1 = 0.0027 K-1 corresponds to 97ºC.]
3 FINITE ELEMENT ANALYSIS In order to estimate the life of the seal non-linear finite element analysis methods (FEA or FEM) have been used. FEA allows the modelling of both simple and complex crosssections of seals, and hence is useful for assessing the change of stress within complex seal geometry and the effect on its sealing performance. For FEA performed on a seal, when the sealing force per unit area (contact pressure) becomes less than the system pressure this is generally taken to indicate the onset of leakage. It should be noted that in FE models the seal surface and the mating faces are assumed to be perfectly smooth – ignoring the possibility of “seepage” along a scratch on the metal housing, for example. Also the possibility of adhesion is generally ignored. In order to accurately model the life of a static seal, material data is key. Initial data may be precisely tested for, and included in the model (Daley & Mays, 1999), but the material properties change over the life of the seal. The seal may swell due to interaction with the contained fluid, the physical properties can change due to oxidation, and fluid effects, and the sealing force will drop due to stress relaxation. Although many have modelled specific elements of a seal in operation, no one has put a unified FEA model together to fully simulate the life of a seal. The various aspects, which have been modelled, in addition to hyperelastic behaviour, include: permeation (Ho, 2001; Ho & Nau, 1996); explosive decompression (Ho, 2000); fatigue (Busfield, 2001). Modelling of the aspects listed above has been done, with good reason, to assess the most likely failure mode for the application. However, without full interactive account being taken of the chemical and physical changes within the seal, uncertainty is introduced in the precision of the prediction. As discussed above, in the case of a static seal, compressive relaxation is generally taken as the probable cause of failure. Sealing geometries range from the simple – such as the O-ring seal, as shown in Figure 4, to more complicated designs and loadcases – such as the gasket shown in Figure 5. [For
56
Elastomers and Components: Service Life Prediction – Progress and Challenges
Fig. 4 Plots for 2-D hyperelastic finite element analysis (FEA) of an O-ring seal in pre-relaxed state (left) and following stress relaxation (right). Cauchy 22 stresses are shown (MPa). The FE code used was MARC.
Fig. 5 Stress plot for 3-D hyperelastic finite element analysis (FEA) of a gasket. The FE code used was MARC.
a discussion of hyperelastic modelling and FEA see Gough et al, 1999; Cadge & Prior, 1999.] The stress pattern in Figure 4 is simpler, more uniform, and easier to interpret, than the multifacetted loading of the gasket in Figure 5 – nevertheless the advantages of FEA are clear.
4 OTHER FAILURE MODES Other failure modes, such as explosive decompression or extrusion damage of the seal, are features which should be anticipated in design (Morgan et al, 2004). It is however the case
Assessment of Life Prediction Methods
57
that the seal environment predicted and the actual conditions seen are not necessarily the same, and hence such failures do occur. Chemical attack of seals is seen in more aggressive environments, such as occur in the chemical industry, and here a service life of months rather than years is usually accepted. In considering more expensive materials, such as perfluoroelastomers, the cost / life benefit in the application must be weighed up (Daley & Phillips, 2000). Although use of perfluoroelastomers is a fairly extreme example, the type of elastomer must fit the application – and relevant fluids and temperatures – to ensure that chemical degradation of the seal is not accelerated. Further failure modes include mismatch of thermal expansion, especially if the elastomer seal does not recover quickly enough under thermal loading. A simple case is cold start testing, where the components are cooled, usually to -40°C, and then the engine is started. If the housing around the seal expands at a faster rate than the seal can follow, then a leakage path will occur. Low temperature elasticity of elastomers is commonly measured through low temperature retraction (TR) tests (ISO 2921,1997). In addition to the above factors which can be designed against, there can be damage to seals through assembly, transport or storage, which reduces their life span.
5 CONCLUSIONS Methods do exist to assess the lifetime of elastomer materials and seal designs, through accelerated testing and analysis. However in order to utilise the laboratory results there must be consideration of the environment that the seal actually experiences. In use, a seal will be subjected to a range of operating conditions; this range must be applied to the theoretical methods of seal life prediction. Therefore a combination of laboratory material tests, application testing, mathematical modelling, and application knowledge and experience is required for life prediction. Even then, this is a best guess estimate, due to uncertainties in the final application environment. Therefore, in practice, any prediction work carried out can only consider samples which experience typical or representative conditions which will occur in the application environment.
REFERENCES Albihn P (2005) “The 5-year accelerated ageing project for thermoset and thermoplastic elastomeric materials: A service life prediction tool” (This volume). Azura AR & Thomas AG (2005) “Effect of heat ageing on the strength properties of elastomeric components” (This Volume). Barnard D & Lewis PM (1988) “Oxidative ageing”, Chapter 13 in Natural Rubber Science and Technology, AD Roberts (ed), Oxford University Press, 621-678. Bunting WM, Russell WD, Doin JE, Tate AL, & Slocum GH (1992) “Compression Stress Relaxation I, An Important Test For Evaluation of Sealant Materials”, Society of Automotive Engineers Standard, 920135. Busfield JJC & Thomas AG (2001) “Using FEA techniques to predict failure in elastomers”, Service Life Prediction of Elastomer Components, Institute of Materials, 15 October, London, UK. Cadge D & Prior A (1999) “Finite element modelling of three-dimensional elastomeric components” in
58
Elastomers and Components: Service Life Prediction – Progress and Challenges Finite Element Analysis of Elastomers, D Boast and VA Coveney (eds), Professional Engineering Publishing, London, ISBN 1860581714, 187-205.
Chapman AV & Porter M (1988) “Sulphur vulcanization chemistry”, Chapter 12 in Natural Rubber Science and Technology, AD Roberts (ed), Oxford University Press, 511-620. Coveney VA & Menger C (1999) “Initiation and Development of Wear of an Elastomeric Surface by a Blade Abrader”, Wear 233-235, 702-711. Coveney VA & Menger C (2000) “Behaviour of Model Abrasive Particles Between a Sliding Elastomer surface and a Steel Interface”, Wear 240, 72-79. Coveney VA & Rizk R (2005) “Life prediction of O-rings used to seal gases” (This Volume). Daley JR & Phillips HM (2000) “Cost-effective perfluoroelastomer sealing solutions for aggressive environments”, Proceedings of the 16th International Fluid Sealing Conference, Brugge, 18-20 September, R Flitney (ed), Professional Engineering Publishing, ISBN 1860582540, 135-146. Daley JR & Mays S (1999) “The Complexity of Material Modelling in the Design Optimisation of Elastomeric Seals”, in Finite Element Analysis of Elastomers, D Boast and VA Coveney (eds), Professional Engineering Publishing, London, ISBN 1860581714, 119-128. Eason RJH & Barnes N (1994) “Seals and Sealing”, Chapter 15 Section 2 in Mechanical Engineers Reference Book, EH Smith (ed), Butterworth-Heinemann, Oxford, 15-18 & 15/75. Ferry JD (1992) “Elastomers: dynamic mechanical properties” in Concise Encyclopaedia of Polymer Processing and Applications, PJ Corish (ed), Pergamon, Oxford, 252-256. Fuller KNG (1988) “Rheology of raw rubber”, Chapter 5 in Natural Rubber Science and Technology, AD Roberts (ed), Oxford University Press, 141-176. Gough J, Gregory IH & Muhr AH (1999) “Determination of constitutive equations for vulcanized rubber” in Finite Element Analysis of Elastomers, D Boast and VA Coveney (eds), Professional Engineering Publishing, London, ISBN 1860581714, 5-26. Hertz DL (1992) “Introduction”, Chapter 1 in Engineering with Rubber – How to Design Rubber Components, AN Gent (ed), Hauser, Munich, 1-9. Ho E & Nau BS (1996) “Gas emission by permeation through elastomeric seals”, Tribology Transactions, 39 (1), 180-186. Huy ML & Evrard G (2000) “Methodologies for lifetime predictions of rubber using Arrhenius and WLF models”, Lifetime estimation of rubber materials by ageing and stress relaxation tests, 31 October, Sweden. ISO 2921 (1997) “Rubber, vulcanized – determination of low temperature characteristics – temperatureretraction procedure (TR test)”, International Standards Organisation. Menger C (2001) “Behaviour of sliding seals in abrasive fluids”, PhD Thesis, University of the West of England, Bristol, UK. Morgan GJ, Campion RP & Derham CJ (2005) “Stress-induced phenomena in elastomers and their influence on design and performance” (This Volume). Spetz G (2000) “Stress Relaxation – Test methods, instruments and lifetime estimation”, Proceedings of Elastocon Conference, Sweden, 31 Oct-1 Nov. Stevens RD (2000) “Permeation and Stress Relaxation Testing of Fuel Seal Materials – A Comparison of Fuel System Elastomers”, RAPRA Conference Birmingham, UK, May. Stevenson A & Campion RP (1992) “Durability”, Chapter 7 in Engineering with Rubber – How to Design Rubber Components, AN Gent (ed), Hauser, Munich, 169-207.