THERMAL ANALYSIS PREDICTIONS FOR ABLATIVE PHENOLIC COMPOSITES

Transcription

1 THERMAL ANALYSIS PREDICTIONS FOR ABLATIVE PHENOLIC COMPOSITES ABSTRACT J.R. Williamson, L. Levan, C. Semprimoschni, M. van Eesbeek Thermal analysis was performed on a phenolic composite material which could be used for re-entry applications where controlled ablation influences the materials performance. An assessment of a carbonphenolic material was performed to determine char levels suitable for subsequent mechanical test samples. The evaluations utilised thermoravimetric analysis (TGA) combined with lifetime predictions from advanced model free kinetics (AMFK). Correct characterisation and modellin of the thermal deradation process includin the charrin phenomena is crucial to the short-term operation and performance of the material in an ablative role. This approach enables testin of material on a laboratory scale usin conventional heatin rates in the rane of 5-5 Cmin - to simulate more rapid heatin scenarios such as reentry or launch. A comparison has been made between the mass loss estimates derived from existin char data for an ablative application and predictions from model free kinetics. From this data values of charrin were able to be estimated and predictions made for lower temperature charrin cycles which could then be used for preconditionin of test samples. This script presents the data obtained, comparisons of models with data and discusses the effects, trends and predictions made from the thermal analysis data. INTRODUCTION. Backround Recent ESA prorammes such as Aurora, FLPP, Vea, IXV and ARD/ARV have hihlihted the need to investiate and understand ablative effects on composites utilised for launch vehicles, sample return systems and also to develop a European re-entry capability. A critical part of this requirement is the capability to test and evaluate materials such as composites in a suitable condition representative of either their launch or re-entry state. This means that determination of key materials properties such as thermal conductivity, coefficients of thermal expansion, mechanical properties etc. needs to be performed on material that is in a suitable condition i.e. pre-charred. Further the level of charrin may ultimately impact on the materials performance so TEC-QEM, ESTEC, ESA, Postbus 299, Noordwijk zh, The Netherlands, , Jason.williamson@esa.int understandin and controllin this behaviour is key to successfully predictin the materials performance. The TEC-QEM laboratories performed investiations to establish the viability of utilisin conventional thermal analysis techniques to predict the ablative response of a composite material. These techniques are however often limited by relatively slow heatin rates of 5-5 Cmin - compared to values of up to 6 Cmin - for certain launcher applications. The use of model free kinetic simulations allows hih heatin rate predictions to be made based upon low heatin rate experiments and further allows an understandin of the actual mass losses which occur in relatively rapid heatin scenarios. This data can then be utilised to predict and determine suitable, accurate and repeatable sample pre-conditionin for other characterisation tests utilisin relatively low cost experiments..2 Model Free Kinetics Lifetime predictions have been derived from a model free kinetics (MFK) iso-conversional approach developed by Vyazovkin[]. This utilises a model-free kinetic analysis to establish Arrhenius parameters thus enablin kinetic predictions based upon the effect of variables such as temperature, pressure and concentration on the reaction rate. Thermal decomposition of a solid material such as a polymer involves the chemical breakdown of the material, transportation and removal from the specimen. All of these processes will be kinetically driven and therefore have a minimum activation enery (E a ) required to allow the reaction to initiate. Arrhenius developed a relationship for the reaction rate based upon the reaction temperature (T) and activation enery for simple order reactions: Ea k( T ) = Aexp () RT where k(t) is the temperature dependent rate constant, A is a pre-exponential factor, R is the universal as constant, t is time and, is the deree of conversion or extent of reaction. From the Arrhenius equation, it can be seen that the reaction rate and hence time to completion is

2 dependent on both the activation enery of the reaction and the temperature. This fact forms the basis for analysis and prediction of isothermal reaction rates utilisin non-isothermal linear heatin prorams. Solid state kinetics can be eneralised to an equation form which takes into account the dependence of the reaction rate on the extent of reaction: d = k dt ( T ) f ( ) where.f() is the reaction model a function describin the dependence of reaction rate on the extent of reaction. The reaction model f() for solid state reactions such as mass loss is often empirically based. For isothermal conditions this equation can be expressed in its interal form; (): = f ( ) ( ) d k( T ) t (2) (3) This equation provides the lobal reaction kinetics from three pieces of information; the reaction model and the two Arrhenius parameters. These three items are sometimes referred to as the kinetic triplet[]. The time required to achieve any particular extent of reaction can be determined for any chosen value of temperature under isothermal conditions by rearranin equation (3) or, for any chosen heatin rate β by rearranin equation (7). Further these values can be obtained as a function of conversion without any knowlede of the specific reactions or the various kinetic processes takin place hence its model-free title. Kinetic predictions such as time to reach a specified conversion or reaction rate at some reference temperature cannot be directly made by equations (2) and (3)[] since activation enery will vary with the extent of reaction. This can be resolved by assumin the partial kinetic triplets remain the same when chanin temperature. Usin this assumption equation (3) and equation (7) can be equated with a iven conversion. Simultaneously solvin these two equations yields[]: For non-isothermal conditions, time dependence can be eliminated from the equation by introducin a constant heatin rate: t E =. E exp RT β exp RT T dt (8) where: d d = dt β dt dt β = dt (4) (5) where T is an experimental value of temperature which corresponds to a iven conversion at the heatin rate β. From equation (8) the time at which a iven conversion will be reached at an arbitrary temperature T can be determined. Similarly other temperatureconversion-heatin rate functions can be established without a knowlede of either the reaction model or the pre-exponential factor. thus the interal can now take the form of: f ( ) ( ) d = k( T ) and introducin the Arrhenius equation by substitutin for k(t) ives: A β β E RT T a ( ) = exp T dt dt 2 PROCEDURE (6) Specimens were analysed usin thermoravimetric analysis (TGA). TGA was carried out with a Mettler- Toledo TGA/SDTA85 e operatin with the STAR e software usin a heatin proramme from 25 C up to 6 C at several different ramp rates ranin from C.min - up to C.min -. The furnace was continuously pured with nitroen as at a flow rate of 7ml.min -. Buoyancy and baseline corrections were made usin an equivalent empty pan run usin the (7) same method. All tests were performed in accordance with ISO specification 358[2].

3 At least four different heatin rates were selected to enable kinetic modellin and lifetime prediction for the materials analysed usin the STAR e model free kinetics analysis module. Where appropriate the thermoram obtained from one of the heatin rates was discarded if necessary to facilitate analysis. All of the resultant thermorams were normalised with respect to sample mass and plotted aainst sample temperature. Analysis of the results was conducted on the resultant normalised mass loss-temperature curves. The derivative of each curve was first obtained to identify suitable analysis limits, the curves were cut to this frame (or no frame was utilised), then the conversion curves were derived for each heatin rate across these limits. Usin the conversion curves the activation enery was determined as a function of the conversion usin Vyazovkin s model free kinetics approach. From the activation enery function time-temperature-mass loss curves were determined. These curves were subsequently converted into time-reciprocal temperature curves. An advanced MFK (AMFK) technique was used. AMFK utilises a differential approach for prediction of the activation enery function compared to an interal approach for the MFK analysis. The AMFK technique enerally results in improved lifetime predictions particularly at low conversion values. A number of isothermal runs were conducted at 45 C. 3 RESULTS & DISCUSSION Initial experiments indicated concerns with the buoyancy and corrections were made first by usin an equivalent empty pan. However for small heatin rates (5 Cmin - and Cmin - ) the diffusion process may be different due to the instrument and/or the sample. This resulted in the samples showin on-oin mass loss above 9 C. Moreover it appears that for low heatin rates, the first decomposition step (around 3 C) is not distinct and may not be present. After extensive investiations, the experiments were redone with the same methodoloy but with a second run of the already tested sample as a baseline. This enabled temperature scan curves to be more reproducible and the plateau of the total mass loss to convere even at low heatin rates. (see Fi. ). Converence of the results is important for the MFK modellin to be consistent. Fi.. Temperature scan curves of CFRP from ambient to 6 C at various heatin rates

4 Table : Equivalent conversion rate and relative mass loss Relative mass loss (%) Equivalent conversion (%) Fi. 2. Conversion curves and first derivatives of CFRP materials On the temperature scan curves and their first derivative (Fi. and Fi. 2) three mass loss steps can be seen; the first is a water desorption (from moisture trapped within the material) which occurs from room temperature to around 5 C with an averae mass loss of.5%. Followin this the first decomposition step occurs at around 3 C. A difficulty is observed in that this step is only seen with hih heatin rates (above 2 Cmin - ). This deradation step results in a mass loss of 3.7%. This mass loss step could correspond to either a release of water or other volatiles formed by the deradation of the phenolic resin compound. Note that if oxyen is released by the deradation of the phenolic resin, it may further contribute to the internal oxidation of the carbon phenolic. The main decomposition step occurs from 4 C to 9 C with a mass loss of 2.3%. This step is probably associated with the breakdown of the CFRP resin into small monomers such as phenol and/or the release of volatile compounds (methane, hydroen, carbon monoxide/dioxide etc.). Durin the depolymerisation step above 55 C multiple phenomena are occurrin as part of the charrin of the material. The averae total mass loss can be estimated to be around 6.4 %. This total mass loss appears consistent in all intermediate heatin rate temperature scans. Conversion curves for the selected temperature scans are shown in Fi. 2. It should be noted that a number of iterations were performed to establish suitable curves for analysis. In addition to the baseline correction it was necessary to identify a set of selfconsistent curves. To expedite the analysis the low mass loss reions of the temperature scans were excluded from the evaluation. This was done on the basis that these seem likely to involve volatiles rather than deradation of the phenolic resin and, that the time spent at these temperatures is relatively short. Fi. 3. AMFK analysis usin optimised conversion curves Fi. 2 also presents the first derivative curves for the mass loss steps. It is apparent from these curves that multiple effects are takin place which seem to be dependent on the heatin rates. This is particularly true for the mass loss occurrin at around 3 C which is only seen at hih heatin rates althouh it seems likely that this step also occurs at lower rates but is subsumed into adjacent mass loss steps. The AMFK analysis performed on the data from Fi. 2 is iven in Fi. 3. The conversion curves were derived for each heatin rate across the main decomposition and charrin process (from 377 C to 95 C). From the conversion curves the activation enery was determined as a function of conversion usin Vyazovkin s model free kinetics approach. From the activation enery function time-temperature conversion curves were determined. The isoconversion curves obtained from the AMFK analysis are iven in Fi. 4. These were subsequently converted back to mass loss values utilisin the % conversion value. The conversion rate is related to the mass loss such that the total mass loss is equivalent to % of conversion (See Table and Fi. 4).

5 25 2 Inner Middle Outer Temperature (K) 5 5 Fi. 4. Lifetime prediction curves as a function of conversion rate, time and temperature Time (s) Fi. 5. Composite panel application temperatures The analysis assumes that at % conversion charrin has fully occurred. The temperatures for a typical ablative application are iven in Fi. 5. This data can be combined with the data iven in Fi. 4 where the intersect points with each iso-mass loss curves define the charrin state of the material versus time and temperature (see Fi. 6). Takin the intersect values from Fi. 6 a plot of temperature-time can be constructed with each point associated with a specific mass loss value. This can also be done for the halfdepth temperature. Fi. 7 and Fi. 8 show these values for the inner position and the half depth position respectively. It is apparent from both fiures that the full conversion (i.e. 6% mass loss) is quickly achieved in both cases based upon the likely temperature predictions (Fi. 5). For the inner position this occurs in around 36 seconds and at the half depth at around 8 seconds from onset. To accurately determine the appropriate amount of charrin it is necessary to identify which is the critical time point durin the composites lifetime. This timepoint would typically be where the hihest level of forces are observed durin either a launch or re-entry fliht profile. Further it should be identified if the most critical position is the outer surface or at the half depth position. Once the critical position/time has been established this can then be equated to a massloss level from Fi. 7 and Fi. 8. This value can then be utilised in a suitable isothermal curve (e.. Fi. 9) to ensure the correct level of charrin is performed on any characterisation samples. It should be noted however that firstly the charrin time should incorporate any dwell times observed durin testin and, secondly a number of effects have been observed which could impact on the prediction. These are discussed further in the remainder of the script. time [s] Temperature [ C].3% Mass loss ( 2% conversion).5% Mass loss ( 3% conversion).8% Mass loss ( 5% conversion) % Mass loss ( 6% conversion).5% Mass loss ( 9% conversion) 2% Mass loss ( 2% conversion) 2.5% Mass loss ( 5% conversion) 3% Mass loss ( 9% conversion) 3.5% Mass loss ( 22% conversion) 4% Mass loss ( 25% conversion) 4.5% Mass loss (28% conversion) 5% Mass loss ( 3% conversion) 5.5% Mass loss ( 34% conversion) 6% Mass loss ( 37% conversion) 8% Mass loss ( 49% conversion) % Mass loss ( 62% conversion) 2% Mass loss ( 74% conversion) 4% Mass loss ( 86% conversion) 6% Mass loss ( 99% conversion) Throat inner diameter Throat half depth Throat outer diameter Fi. 6. Lifetime prediction of CFRP material with the inner, middle and outer plate temperature curves. Fi. 7. Mass loss simulation as a function of time and temperature for the inner throat diameter

6 A comparison of the total mass losses observed under different isothermal durations is iven in Table 2 below. These are also shown schematically in Fi. 9 alon with a prediction from the AMFK analysis. The predictive values are limited to an upper total mass loss of around 6% as exhibited by the TGA temperature scans which indicate a plateau above 9 C in an inert nitroen atmosphere. For this particular material however this assumption appears flawed since isothermal data at 445 C indicates that onoin mass loss occurs above the 6% (see Fi. 9and Table 2). Fi. 8. Mass loss simulation as a function of time and temperature for the half depth throat diameter Relative mass loss [%] C for 7 hours 445 C C for 55 hours Modellin C time [min] Fi. 9 Comparison of isothermal curves at 445 c with their model prediction curves Table 2: Isothermal mass losses observed from phenolic Composite Environment 2 hours at 445 C, 7µl pure of Nitroen as 55 hours at 445 C, 7µl pure of Nitroen as Total Mass Loss 2.4% 24.2% The onoin mass loss effect can possibly be explained by reference to some of the early analysis performed usin lower heatin rates and consideration of the charrin phenomena. Fi. shows temperature scans from some of the initial analysis. This showed continued mass loss in low heatin rate samples above 9 C. Initially it was assumed that this was solely associated with the test rather than the material. In particular it was assumed that the buoyancy and adsorption effects of the sample pan were influencin this observation. This effect has been previously observed and is well established. However when taken into context of the isothermal data (Fi. 9) it seems likely that additional mass loss contributions were made from the samples themselves. This means that durin hih heatin rates the conditions are not present to facilitate continued mass loss. This could be throuh some chemical or physical effect (e.. diffusion of residual oxyen from the pure as and/or the sample material). Indeed it seems likely that the charrin phenomena which is known to protect underlyin material is responsible for protection from the onoin mass loss. Nevertheless in the case considered operational heatin rates are very hih and therefore it seems likely that this phenomena is unlikely to occur durin service. However for pre-charrin of test samples purposes it needs to be ensured that the maximum mass loss does not exceed 6%. In the case of a 445 C pre-charrin temperature this would mean a maximum duration of around 33 hours. At hiher temperatures or where the test temperature is hiher a reduction in the precharrin time will be necessary. It should be noted that no testin was performed on the material in the presence of an oxidative environment such as air. The presence of an oxidative environment could have sinificant implications on the charrin behaviour. Further the source of oxidation miht not necessarily be from the surroundin atmosphere but could be by the eneration of oxidative species internally.

7 % 95 Normalised Tscan - 5 C/min - C/min - 5 C/min - 2 C/min - C C 2 C C 5 C /min First derivative / C C Fi. Initial low heatin rate temperature scans of phenolic composite C Other areas which need to be evaluated include the level of mass loss at the upper test temperatures. It appears from the temperature scans that most of the stable mass loss will occur whilst the samples are bein heated to the test temperature (particularly at or above 8 C). Therefore proloned dwellin at these test temperatures to enable stabilisation of the test temperature may cause a deviation in the level of charrin observed which is unrepresentative of the service environment. It seems likely that some compromise will have to be obtained between a necessary dwell for testin and minimisin the level of non-stable mass loss. 4 CONCLUSIONS The carbon-phenolic composite was successfully investiated usin TGA and AMFK. The derived AMFK data was then able to be utilised in conjunction with the predicted temperature profile to predict the likely mass losses which would occur throuh ablation. TGA indicated that a number of phenomena and overlappin time-temperature effects result in the eneration of a complex analysis. A number of assumptions have had to be made to enable successful analysis. TGA temperature scans indicate that the phenolic composite investiated underoes around 6% mass loss when charred under an inert nitroen atmosphere. This atmosphere is in line with the assumption that durin operation the material is under inert conditions. The analysis indicates that onoin mass loss will occur under isothermal or slow heatin conditions however this is not representative of the service conditions and should be avoided. Only a sinle isothermal temperature has been investiated which indicates onoin mass loss up to 24% after 55 hours at 445 C. 5 REFERENCES [] Vyazovkin S., Wiht C.A., Kinetics in Solids, Annu. Rev.Phys. Chem : [2] ISO Specification 3 58, Plastics Thermoravimetry (TG) of Polymers General Principles, International Standards Oranisation, 997