Sample preparation techniques on dynamic shear rheometer testing: round robin tests on bitumen

Three interlaboratory tests conducted by European laboratories were studied, in which 37 laboratories participated in 2018, rising to 46 in 2019 and 2020. A RTFO-aged neat bitumen 20/30 and PMB 45/80-55 in RTFO-aged and unaged conditions were tested using a DSR at frequency of 1.59 Hz with PP08 and PP25 geometries at temperatures ranging from 10°C to 65°C. To assess variation in numerous areas of test method EN14770:2012 practices and the underlying impact on test outcomes, a statistical analysis on G* and δ is reported. The finding emphasises the need of using a detailed written sample preparation protocol to limit operators’ latitude while conducting tests. The statistical investigation suggests that the sample mounting onto rheometer stage and the sample manufacturing oven setup process require more attention. Furthermore, δ precision outperforms the EN14770:2012 criteria; nevertheless, |G*| precision does not match EN14770:2012 for all the testing combinations.


Introduction
The complex shear modulus (G * ) is used to characterise the rheological properties of the bitumens.It is defined as G * = |G * | e iδ , where |G * | is the dynamic shear modulus, δ is the phase angle and i is the imaginary unit (i 2 = −1).The complex shear modulus can be determined by using a dynamic shear rheometer (DSR) according to the standards (AASHTO, 2010;ASTM, 2008;EN14770, 2012).The Strategic Highway Research Program (SHRP) that was carried out in the U.S. A. between 1988A. between and 1993A. between (Anderson et al., 1994;;Petersen et al., 1994) led to increasing interest in performance-based methods for characterising bituminous binders and to using in numerous new purchase specification and standard (AASHTO, ASTM, EN).The new methods distinguish the rheological properties of bitumens as a function of loading time and temperature to simulate actual behaviour in practice rather than empirical properties such as softening point temperature and penetration value, among others.
The procedure of preparing and conditioning the samples is mainly described in the method EN 14770, 'Bitumen and bituminous binders.Determination of complex shear modulus and phase angle.Dynamic Shear Rheometer (DSR)'.However, in DSR testing of bitumen there are several factors that could affect the measurements of the rheological properties.For example, several studies show the importance of having the same thermal treatment to achieve an acceptable precision on the complex modulus, phase angle, and SHRP parameter (G * /sin δ) when testing in low frequency and for polymer modified bitumens (PMBs) during the sample preparation and testing phase (Eckmann et al., 2012;Mouillet et al., 2004;Soenen et al., 2005).The plate diameter is selected based on the maximum sample stiffness at the specific testing condition.The effect of different plate diameter as well as the gap size between the plates have been presented in earlier studies.It was shown that higher gap size leads to a decreased value of the complex modulus and phase angle at higher temperatures (Carswell et al., 1998;Liu et al., 2020;Singh et al., 2016).Another study presented differing values and lack of consistency of results from DSR tests using three different sample preparation methods (G.Airey et al., 2002b).This shows the importance of a strictly defined and applied procedure for enabling the comparison of observations.Some studies demonstrate that the sensitivity of the equipment to measure torque (Divya & Krishnan, 2018), and the limit of linear viscoelasticity (G.D. Airey et al., 2002), should be considered to obtain a consistent result from DSR testing.The testing of polymer modified bitumens using a DSR normally leads to greater variation in the result compared to testing of unmodified bitumen (Airey & Hunter, 2003).The morphology of the PMB increases the challenge to limit the variation in the testing (Airey, 2003;Soenen et al., 2008;Zhu et al., 2017).Thus, to ensure comparable and accurate results, there is a need for defining the existing procedure more strictly and highlighting the importance of following the standard.
The purpose of this study is to see how different sample preparation affect the results.The aim is also to identify the phases of sample preparation that are most critical to achieve consistent results.Furthermore, the EN14770:2012 method precision is examined using repeatability and reproducibility analyses of reported G * and δ values.The study also contains an overview of prior studies to emphasise the importance of using the same sample preparation and test conditioning techniques, as well as demonstrating their potential impact on outcomes.

Method
This research is based on data from three European round robin tests (RR tests) conducted in three different years.The practices, procedures, and varied selections of temperature and time inside and beyond the EN14770:2012 guideline used by different laboratories, as well as their underlying effects on test results, are screened in this paper.The brand of equipment, sample manufacture method, storage time before testing, pre-heating time and temperature conditioning for manufacturing sample, sample placing onto rheometer, equilibrium duration, and testing within the viscoelastic linear range are all investigated.A distinct experiment design is required to study the interaction effect and pinpoint more precisely the degree to which each phase of the sample preparation and conditioning influences the outcome, as it is done in a study (Sheidaei et al., 2023).

Material
The tests were carried out on a penetration graded bitumen of the type 20/30 in 2018, as per EN 12591(2009), and a polymer modified bitumen of the type 45/80-55 in 2019and 2020, as per EN 14023 (2010)).All materials were distributed in fresh state.In 2018 and 2019, the tests were conducted after each individual lab had performed short-term ageing in a rolling thin film oven (RTFO) in accordance with EN 12607-1 (2014), while in 2020, they were performed on an unaged bitumen, i.e. without RTFOT or PAV.
Table 1 shows the material type, date of random distribution, size of sub-samples produced, penetration value and softening point, and G * and δ testing plan at various temperatures with corresponding parallel plate dimensions (PPmm).The needle penetration test (PEN), according to EN 1426(2015), and the softening point test (SP), according to EN 1427(2015), are used by suppliers for checking the homogeneity and stability of the sampling prior to the distribution of material to ensure minimal effects on results.

Information about test project participants and instruction
In 2018, 37 laboratories participated, rising to 46 in 2019 and 2020.The participating laboratories conducted two independent repeats for each test combination at pre-defined temperatures and plates as stated in Table 1, at frequency of 1.59 Hz (10 rad/s).The participating laboratories were instructed to use their own usual practices, including sample preparation and conditioning procedures, when applying EN14770 (2012) on DSR testing.

Method of statistical evaluation of results
The central tendency (means) and variances of G * and δ are compared between the various practices used by participant labs in different step of the DSR testing.Laboratories were divided into groups based on practices they used.is used to quantify the direction and strength of the relationship.A statistical significance test is then performed to determine whether the correlation between variables is significant.
The overall trueness of the reported results by participant labs were measured by applying statistical tests to the relative deviation from the theoretical value of all samples analysed (z-score) according to ISO 13528 (2015) by round robin organiser per year, which is available to participants.The repeatability standard deviation (S r ), reproducibility standard deviation (S R ), repeatability coefficient of variation (CV-r), and reproducibility coefficient of variation (CV-R) are used to express the precision analysis.According to ISO 5725-6 (2001), the difference between results obtained in the same or alternate laboratory is significant, if it is greater than the repeatability limit (r = 1.96 × √ 2 S r ) or the reproducibility limit (R = 1.96 × √ 2 S R ), respectively.

Results
The following sections examine variations in practices used by participant labs when testing according to EN14770:2012 method, as well as how these variations may affect the values of the complex shear modulus G * and phase angle δ.The letters a, b, and ab were used to show that there is a statistically significant difference (P < 0.05) between the groups/categories for the test combination.The repeatability and reproducibility of the results obtained each year will also be shown.It should be noted that there were two phases in the round robin tests: one in which participants were free to use their own conventional practices, such as mounting the sample onto a rheometer, heating procedure, and/or sample storage time as long as they follow the test standard, and another in which organiser gave strict instructions on which temperatures, frequency, and plate geometries the tests should be performed.

Repeatability and reproducibility
The reproducibility of 10% for the complex shear modulus G * and 5% for the phase angle δ are the recommended values in EN 14770:2012 based on the Rilem TC 180-PEB work (Sybilski et al., 2004).The repeatability for G * and δ was determined to be 5% and 1% for paving grade bitumen and 8% and 2% for polymer modified bitumen from a study conducted by Eckmann et al. (2008).Table 2 lists the number of laboratories that carried out the test combination at overlap temperatures.The values of |G * | and δ at the overlap temperatures differ from the mean of both geometries by much lower than the recommended values of 15% for G * and 3°for δ.The stiffness values reported by labs fall outside the recommended range for the predetermined plate geometries used at test temperatures of 25, 30, and35°C in 2018 (20/30 RTFOT) and10, 15, and20°C in 2019 (45/80-55 RTFOT) shown in bold in Table 2.The precision results clearly show the consequences of departing from the operating stiffness ranges advised by the standard.According to the standard, PP25 is generally suitable for stiffness between 1 and 100 kPa, and PP08 is suitable for stiffness between 100 kPa and 10 MPa.
For the G * , the coefficient of variation under repeatability conditions (CV-r) and the coefficient of variation under reproducibility conditions (CV-R) related to geometric mean value (G.M.) are obtained.It is found that for the PP08 and PP25 the repeatability values are within a range of 2-12%, while reproducibility values vary between 7% and 20% (Figure 1).Thus, the G * reproducibility and repeatability do not meet the criteria obtained by previous studies for all the testing conditions.However, when eliminating the extremes (PP08 data at a high temperature of 35°C, and the plates utilised outside of the operational stiffness level), the precision improves significantly with a CV-r of 2-8% and CV-R of 7-16%.These extremes are shown in the column chart with the pattern-fill.
For the δ, the coefficient of variation under reproducibility conditions (CV-R) are much better than the criteria of 5% given in EN14770: 2012 for all test combinations after eliminating the extremes where plates are utilised outside of the operational stiffness level.However, for the δ, result show that value of CV reduces as temperature increases, as also mentioned by Büchner et al. (2022), which is related to greater value for phase angle as temperature increases.To prevent this reliance, for the δ, the repeatability limit (r) and reproducibility limit (R) related to the arithmetic mean value are expressed in   absolute value, which will remain constant regardless of temperature (Figure 2).Precision improves to a limit of r = 2°and R = 3°by eliminating the extremes shown in the column chart with the pattern-fill in Figure 2.
Another interlaboratory study with a smaller number of participants using the same type of bitumen, PMB 45/80-55 but with different Pen and SP values, shows similar result at test temperature 60°C, i.e. a wider spread of data around the mean for phase angles than the complex shear modulus (Błażejowski et al., 2016).From 2018 to 2020, regardless of material type, there is an overall improvement in precision, which may be attributed to round robin instruction experience and the removal of the influence of varied individual practices on RTFOT technique done by each participating laboratory from the last year tests.

Equipment
Figure 3 shows the percentage of different equipment from different manufacturers used by all the participant laboratories.Most laboratories used various models of Anton Paar manufactured rheometers (MCR 101,102,301,302,501,502,702,smart pave 102/301,.Malvern Panalytical rheometers represented six different models (Kinexus DSR, DSR+, Pro, Pro+, Ultra+, and KNX).The Discovery HR 2 and AR1000 models, and the Haake Mars 2 and 3 models were produced by TA Instruments and Thermo Scientific, respectively.The least amount of used DSR devices were produced by Bohlin, representing one model (DSR2).
Figure 4 shows the average G * and δ by equipment brand.The term 'other-brand' refers to the combination of the three least used brands (TA Instruments, Thermo Scientific, and Bohlin).A letter above the column bar denotes test conditions that were shown to have statistically significant differences.For the G * , we found that there are statistically significant differences between Malvern brand and other-brand for the RTFOT 20/30 at test condition PP25-55°C.In terms of δ, Malvern brand was significantly different from other-brand for the RTFOT 20/30 at PP08-20°C, for the RTFOT 45/80-55 at PP08-10°C, and for the unaged 45/80-55 at PP08-15°C, at PP08-10°C, and at PP25-60°C.
When comparing the Malvern and Anton Paar brands, the Malvern resulted in a higher and lower G * , when applying PP025 and PP08, respectively.However, the differences are only significant in the two test conditions listed here: for unaged 45/80-55 at PP08-20°C, and at PP08-15°C.In terms of the δ, Malvern brand resulted in a higher value than Anton Paar for all test conditions except testing at 40°C and 50°C with a PP25 for unaged 45/80-55.The differences, however, are insignificant.

Sample manufacturing
According to EN 14770 (2012), sample manufacturing can be performed in 3 different ways, as follows: Pour into moulds or sheets, directly onto plates, and apply vials.Using a vial is not recommended for polymer modified bitumens.However, a previous study shows that repeatability of the hot pour onto plate method is slightly higher than the repeatability of using mould (silicon) and weighing methods, which involves pouring a pre-weighted amount of hot bitumen directly onto a plate (G.Airey et al., 2002a).Figure 5 shows that the most frequently used method during the round robin test between 2018 and 2020 is the silicon mould.The trend for using the silicone mould is increasing, which can be the result of a larger amount of Anton Paar equipment among participant laboratories.Nevertheless,   only a few laboratories mentioned their mould supplier.Each year, one laboratory used hot pouring onto plate method when PP25 was applied.
Statistical tests were used to examine the differences in the G * and δ based on the sample manufacturing method used by participants (Figure 6).The term 'other-mfg.' refers to the combination of manufacturing methods other than sheet and silicone mould.The sheet manufacturing method resulted in a higher G * , and a lower δ on average than mould manufacturing method for all test conditions, but there were no significant differences between the two methods.There are, however, statistically significant differences (p < 0.05) between other-mfg.and sheet, as well as between othermfg.and mould at few testing conditions as it follows: For G * , for the RTFOT 45/80-55 at 15°C and 20°C when P25 applied.For δ, for the RTFOT 20/30 at 35°C, 30°C and 25°C with PP08, for the RTFOT 45/80-55 at 10°C with PP08 and at 50°C with PP25, and for the unaged 45/80-55 at 60°C with PP25.

Waiting time (storage time) before testing the manufactured sample
Previous studies reveal that the waiting time at room temperature between the bitumen sample casting and the start of the test procedure does not have a significant impact on the test result for pure bitumen.However, for polymer modified bitumens different interpretations can be represented, which demonstrates the importance of reaching bitumen's full equilibrium when considering slow isothermal crystallisation during storage for polymer modified bitumens.For instance, for the ethylene-vinyl acetate (EVA) modified bitumen, a slight increase is observed for the phase angle (δ) at low frequencies over a maximum of 24 h time delay (Eckmann et al., 2012).Another study also shows that the complex viscosity (η * = G * (ω)/ ω) changes more than a decade at 0.001 Hz for the styrene-butadiene-styrene (SBS) modified bitumen.This influence is also observed in the phase angle, which decreases at 0.001 Hz over the waiting time (Soenen et al., 2006).
According to EN14770 (2012), the maximum delay recommended is 72 h regardless of the type of bitumen, with a minimum storage duration of 2 and 12 h for pure and polymer-modified bitumen, respectively.The average storage time of 38.7, 25.3, and 23.3 h were used by the participating laboratories for RTFO aged 20/30, RTFO aged 45/80-55, and unaged 45/80-55, respectively.A relatively higher storage time is obtained for the 20/30 bitumen compared to other materials, which is due to two laboratories having a time delay as high as 7 and 14 days.However, both mentioned laboratories have relatively high precision in their reported results among others.Table 3 shows the storage time (h) used by laboratories on two repeats for each test compared to the recommended limits of EN 14770.Note that the storage time for each test has not been reported by all laboratories.
The statistical tests were conducted to examine the differences on results by the three categories of waiting time applied by laboratories (short: ≤ 2 h, ≤ 12 h (PMB), medium: 2 < h < 72, 12 < h < 72 (PMB), and long: ≥ 72 h) (Figure 7).In most test conditions, long waiting time resulted in a higher G * , and a lower δ when compared to short waiting time.Also, as waiting time passes, the ratio of standard deviation to mean for G * and standard deviation for δ drops (reduced variability).However, there  are statistically significant differences only for 45/80-55 in unaged condition as follows: for the G * at test condition PP08-20°C between long waiting time and medium, and at 50°C and 60°C with PP25, between the long and short waiting time.In terms of the δ, significant difference was found only at PP25-65°C.

Pre-heating time and temperature for manufacturing sample
One study, among others, shows that for the SBS modified bitumen, the pouring or homogenisation temperature of the bitumen affects rheological tests more than the storage time.The result of a test conducted at two temperatures of 120°C and 200°C shows that for lower pouring temperatures viscosity η * stabilises as frequency decreases (Soenen et al., 2005).Also, another study shows the influence of the preheating temperature and duration prior to testing on rheological properties of bitumen.It investigates a pure bitumen PG 64-22 and two PMBs PG 70-22 and PG 76-22 in original and after aging processes at temperatures of 143°C and 185°C for a duration of 1/2, 2 and 4 h.The study suggests 2 h of heating at lower limit pouring temperatures (Dessouky et al., 2011).Furthermore, Büchner et al. (2022) reveals that most of the 20 European road asphalt laboratories that participated in an interlaboratory study have heated the pure (50/70) and modified bitumen (25/55-55) around 150°C for sample manufacturing and DSR testing.Current bitumen sample preparation in EN12594 (2014) states that heating unaged bitumen to 100°C above softening point temperatures should be avoided; this is declared as a temperature of (85 ± 5) °C above softening point but not higher than 180°C in EN14770 (2012).However, for modified bitumen obtained under an ageing test, heating within 180-200°C may be needed, but it should not be heated above 200°C (EN14023, 2010).According to EN 14770, the duration of melting, homogenising, and moulding should not exceed 135 min, but reheating time differs between a maximum of 30, 60, or 120 min depend on the mass of sub-samples, for instance, in case there is RTFOT prior to manufacturing the sample.
In this study only in the 2020 round test, samples are directly manufactured after heating, while 2018 and 2019s round tests undergo short term aging and samples manufactured in the three following different ways: • Without additional heating (case 1) • Cooled slightly and reheating to a defined temperature and duration (case 2) • Cooled and stored for later reheating in defined temperature and duration (case 3).
Table 4 shows the oven heating temperature and time for sample manufacturing used by laboratories compared to the recommended limits of EN 14770 and EN12594.An oven sitting duration between 60-90 min for pure bitumen, and no longer than 60 min for PMB regardless of aging condition, are chosen by most laboratories.Case 1 and 3 are the two most used approaches, whereas case 2 is the least favoured option, used by no more than 21% of laboratories.The G * in case 2 and the δ in case 3 have the lowest values in almost all the test conditions.However, only a statistically significant difference was indicated between case 1 and 3 for δ at PP08-20°C for RTFOT 45/80-55.
Table 5 shows the direction of the correlation coefficient r (+, −) between the G * and δ with the heating temperature and heating duration.In general, a negative sign signifies a reduction in G * or δ when heating temperature or duration increases.These factors appear to be positively correlated with each other, meaning that if one increases, the other one also tends to increase.For RTFOT 20/30, the G * shows a statistically significant moderately negative relationship with heating temperature, which is shown in bold in Table 6.While the correlation for δ is positive.However, for RTFOT 45/55-80, G * and δ were found to be insignificantly negatively correlated with heating temperature for sample manufacturing.The unaged 45/55-80, on the other hand, has a different tendency to aged ones at higher test temperatures (PP25), which may be attributed to its stronger relationship (greater r) to heating duration rather than temperature in comparison to other materials.Tables 7-9 display the differences in G * and δ values by the two categories of heating temperature ( ≤ SP + 100°C and > SP + 100°C) used for manufacturing samples by laboratories.For the PMB in both aged and unaged conditions, approximately 30% of laboratories used temperatures lower than SP + 100°C (160°C), yielding more accurate results in terms of coefficient of variation and standard deviation for G * and d.These comparisons show that the upper limit for pure bitumen (160°C) is likewise appropriate for the PMB.It should be noted that the softening points of all tested materials are near to one another.The results are in accordance with the results from correlation test carried out on applied heating temperature shown in Table 5.As contrary to aged bitumens, for unaged 45/55-80 the heating duration decreases from 58 min to 50 min when heating temperature increases from lower than 160°C to higher than 160°C.Nevertheless, the differences are insignificant for all test conditions and materials, except for the d at PP25-60°C for unaged 45/55-80 shown in bold.

Sample bonding (mounting) temperatures
When placing a test sample onto the DSR a high temperature of the parallel plates is, to some extent, necessary for ensuring bonding between bitumen and plate, especially when using modified bitumens.However, an upper limit needs to be chosen carefully to avoid aging the sample before testing.
A study shows that, for the tested plastomer modified bitumen, the bonding at either high (80°C) or low (30°C) temperatures does not have a significant impact on the measured stiffness modulus (|G * |).
However, for the sample bonded at a higher temperature, lower phase angles have been measured (Eckmann et al., 2012).
According to the standard EN 14770, the temperature of both the upper and lower plates of the rheometer need to be set to a maximum of the bitumen's softening point, namely plus (20 ± 5)°C, or at (90 ± 5)°C, whichever is the lower, for at least 30 min to facilitate acceptable bonding of the test sample to the plates.Also, it is understood that a manufactured sample can be placed in a refrigerator for a maximum of 30 min prior to de-moulding and bonding the sample material to the DSR device.
Table 10 shows the bonding temperature onto the rheometer and the practice used by laboratories in comparison to the EN recommendation.The duration of the bonding sample is wildly varied for all sample materials (1-30 min); one laboratory for bitumen 20/30, is delayed by 45 min.The Pearson's correlation test result shows a positive relationship between the G * and bonding duration for 20/30, and a negative relationship for 45/55-80 regardless of aging condition.However, this is only significant at PP25-40°C for RTFO aged 45/55-80.
Based on result of the Pearson's correlation test, there is a negative relationship between the bonding temperature used by participants and the G * for RTFO aged bitumens (Table 11).The significant values are shown in bold.However, for unaged bitumen, the bonding temperature positively correlated with the G * across test conditions, although this impact peaked statistically at PP25-40°C.This effect might be brought on by aging of fresh bitumen at higher bonding temperatures.The result of correlation test between the δ and bonding temperature indicates a positive relationship at test temperature between 40°C and 65°C for all the materials.However, the relationship between these variables is negative, for 45/55-80 at test temperature between 10°C and 20°C.For the δ similarly to G * , for most test conditions, the larger coefficient of correlation is indicated with bonding temperature rather than bonding duration.
Additionally, a statistical test was conducted to examine the differences on G * by the three bonding temperature spans used by participant laboratories.The temperature spans are chosen based on the softening point of the materials as follows: lower than SP: 25-55°C, around SP: 55-75°C, and higher than SP: 80-90°C.For both RTFO aged bitumens, the G * value is higher for a bonding temperature lower than the softening point (25-55°C), compared to the higher softening point (80-90°C) for almost all test conditions (Figure 8).Significant differences (p-value < 0.05) were found among the half of the test conditions, indicating the need of maintaining unambiguous upper and maybe lower bonding temperature limits.

Duration of equilibrium time
Determining time for the temperature of the bitumen to reach thermal and mechanical equilibrium is outlined in EN 14770; however, there is no data if participants have followed the procedure.For all the studied materials, the correlation test results show a negative association between G * and equilibrium duration at test temperatures ranging from 40°C to 65°C.Only at 50°C the association was significant, but marginal.The δ, on the other hand, increases as the equilibrium time increases at most of the test conditions.A statistical test was also performed to assess the differences in G * driven by the three durations of equilibrium timespans chosen by participant laboratories.The following timeframes have been chosen: 5-15 min is utilised by 6-11% of labs, 15 min is used by 67-75% of labs, and 15-30 min is used by 14-27% of labs.Result in Table 12 shows that, at test temperatures ranging from 40°C to 65°C and PP25, the average G * is lower for equilibrium duration of 15-30 min compared to other timespans for all the studied bitumens.However, significant variations were detected only for unaged 45/80-55 at 40°C.Nevertheless, when PP08 is used, the converse effect is observed for practically all measured temperatures ranging from 10°C to 35°C (Table 13); a lower average G * for applied equilibrium duration of 5-15 min, with no significant effect.
Further analysis demonstrates that the same model and brand of DSR used varied equilibrium time durations.Only three-five laboratory measurements in increasing and decreasing testing temperature trends were performed using PP08 and PP25, respectively.Nonetheless, for the most recent year-round tests, all PP08 measurements were taken with a decreasing trend, and PP25 measurements were taken with an increasing trend.

Testing within the viscoelastic linear range
Rheological properties can be obtained from the conventional frequency sweep at different temperatures according to EN 14770 or as suggested in several studies, with temperature ramping at a fixed frequency, which takes less time (Porot et al., 2020).For each geometry and test temperature, suitable shear strains or stresses should be selected to ensure testing within the viscoelastic linear range (LVErange).Nevertheless, it has been observed that testing within the strain range of 0.005 and 0.100 lies within the linear range for most of the binders, while for PMBs the range is much less (EN14770, 2012).
Only two to three laboratories in the 2019-2020 round test acknowledged that they have not studied the viscoelastic linear range of the sample material.They may have chosen the suitable shear strains or stresses based on their experiences with the material and the device used.Stress-controlled mode was used by five laboratories in 2018 and two laboratories in 2020, while most of the laboratories applied strain-controlled mode.Figure 9 illustrates strain values chosen for the tests by laboratories at different testing conditions.

Conclusions
This research presents different practices and their impacts on lab results in three interlaboratory tests on RTFO-aged neat bitumen 20/30 and PMB 45/80-55 in RTFO-aged and unaged conditions using a DSR at frequency of 1.59 Hz and temperatures ranging from 10°C to 65°C.Despite EN14770:2012 having the technique for preparing and conditioning the sample being primarily detailed, labs selected a wide range of temperatures and/or times in the various steps of sample preparation and testing.However, it appears that the bonding temperature has a small but significant impact on more test conditions than other steps when comparing the variation of obtained rheological parameters.Also, the heating temperature used to manufacture samples should be given additional consideration since it has been demonstrated to have a statistically significant relationship with the acquired results.The main findings of the study are summarised as follows: For the complex shear modulus under the investigated test conditions, repeatability values fall between 2 and 12%, whereas reproducibility values fall between 7 and 20%.However, the precision improves significantly with a CV-r of 2-8% and CV-R of 7-16% after removing data from plates utilised outside of the operational stiffness level according to the EN.For the phase angle, the repeatability limitations (r) are found to be 1-2°, while the reproducibility limits (R) are found to be 1-3°.
To prevent potential differences in distributed materials from significantly affecting the results, the round robin test organiser ensured the homogeneity and stability of the sampling of materials.Moreover, in most test combinations, the equipment brands and manufacturing methods had a negligible impact on the results; the G * was higher when Anton Paar and sheet were used, while Malvern and mould had a higher δ.Therefore, a much better repeatability than reproducibility suggests that, rather than the degree of expertise of the laboratory personnel, differences in distributed material, and equipment brands, the dispersion of results between laboratories may be caused by how strictly the EN test procedure is defined, as well as how closely participant laboratories follow it.
When the waiting time for testing the manufactured sample by participating laboratories differs from the EN limits without influencing the results in a statistically significant way, it is likely that the upper and lower bounds will need to be revised.The long waiting time had a higher G * and a lower δ value in comparison to the short waiting time.However, none of the studied bitumens would be significantly affected by waiting times of less than 2 h or longer than 72 h.
Three methods of oven setting for sample manufacturing were used.Only the δ at test temperature 20°C for RTFO aged 45/80-55 showed a statistically significant difference between cases.For RTFO aged 20/30, the G * and δ were shown to have a statistically significant association with sample manufacturing temperature, indicating that G * increases with temperature while δ decreases.For PMB 45/55-80, both G * and δ tends to decrease with sample manufacturing temperature, except for unaged 45/55-80, which appears to be affected more by heating duration than temperature at test temperatures of 40°C, 50°C, and 60°C.A comparison of laboratory results demonstrates that, the upper limit of heating temperature for pure bitumen (SP + 100°C) is also appropriate for PMB, yielding more accurate results in terms of coefficient of variation and standard deviation for G * and δ, respectively.
Correlation test indicated a significant negative relationship between the bonding temperature and the G * for both RTFO aged bitumens.However, for unaged bitumen, the bonding temperature positively correlated with the G * .All materials examined were shown a positive correlation between the δ and the bonding temperature, except for 45/55-80 at test temperatures of 10°C, 15°C, and 20°C.Similar to G * , a larger coefficient of correlation for the δ is indicated with bonding temperature rather than bonding duration for most of the test conditions.The result shows a positive relationship between the G * and bonding duration for 20/30, and a negative relationship for 45/55-80.In addition, the bonding temperature that is lower than the softening point has a higher G * values than the bonding temperature that is higher for RTFO aged bitumens.The importance of establishing a clear upper and maybe lower limit of temperature is underlined by the statistically significant variations that were found in half of the test conditions.
PP08 at temperatures ranging from 10°C to 35°C revealed a lower average G * for applied equilibrium durations of 5-15 min than 15-30 min when compared to PP25 at temperatures ranging from 40°C to 65°C.However, statistically significant variations between different timespan were detected only for unaged 45/80-55 at 40°C.On the other hand, in most test conditions, the value of δ increases along with the equilibrium time.15 min appears to be a suitable equilibrium duration, with results frequently falling between the upper and lower reported values.

Figure 1 .
Figure 1.Coefficient of variation related to geometric mean value under repeatability conditions (CVr), and under reproducibility conditions (CVR) for G * .The pattern-fill bars show the extreme test conditions considering the plate's operational stiffness level.

Figure 2 .
Figure 2. Repeatability limit (r) and reproducibility limit (R) for δ.The pattern-fill bars show the plates utilised outside of the operational stiffness level according to the EN.

Figure 3 .
Figure 3. Brand of DSR equipment used by participating laboratories.

Figure 4 .
Figure 4. Average values of G * (4.a.) and δ (4.b.) when considering only measurements from specific equipment brands, while letters a and b above the columns indicate a significant difference between the groups at the test combination.

Figure 5 .
Figure 5. Sample manufacturing methods used by participating laboratories.

Figure 6 .
Figure 6.Average values of G * (6.a.) and δ (6.b.) when considering only measurements from specific manufacturing methods.The error bars are equal to the standard deviation, while letters a and b above the columns indicate a significant difference between the groups at the test combination.

Figure 7 .
Figure 7. Average values of G * (7.a.) and δ (7.b.) when considering only measurements from specific waiting times.The error bars are equal to the standard deviation, while letters a and b above the columns indicate a significant difference between the groups at the test combination.

Figure 8 .
Figure 8.Average values of G * when considering only measurements from specific bonding temperature spans.The error bars are equal to the standard deviation, while letters a and b above the columns indicate a significant difference between the groups at the test combination.

Figure 9 .
Figure 9. Strain values applied by participating laboratories.

Table 1 .
Testing plan for bitumens distributed in fresh state for measuring G * and δ with two repeats.

Table 2 .
Examine the measured values at the overlap temperatures.The bold denotes that the plate utilised outside of the operational stiffness level according to the EN.The G * and δ deviations from the mean of both geometries for 20/30 RTFOT

Table 3 .
Storage time used by labs on two repeats for each test in comparison to the EN recommendation.Not all labs have provided storage time for every test.

Table 4 .
Heating temperature and time used by labs for sample manufacturing in comparison to the EN recommendation.

Table 5 .
The sign of the correlation coefficient r (+, −) between G * and δ, with the heating temperature and duration at different test conditions.

Table 6 .
Correlation coefficients (r) and p-values obtained for G * and δ, with heating temperature used by labs at different test conditions for 20/30 RTFOT.

Table 7 .
The average G * and δ by sample manufacturing in an upper and lower temperature and its duration for 20/30 RTFOT.

Table 8 .
The average G * and δ by sample manufacturing in an upper and lower temperature and its duration for 45/80-55 RTFOT.

Table 9 .
The average G * and δ by sample manufacturing in an upper and lower temperature and its duration for 45/80-55 unaged.

Table 10 .
Bonding temperature onto the rheometer and practice used by labs in comparison to the EN recommendation.

Table 11 .
Correlation coefficients (r) and p-values obtained between G * and the bonding temperatures used by labs at different test conditions.

Table 12 .
Average G * by the equilibrium duration timespan applied by labs using PP25.Avg.G * (kPa) at different test temperatures and PP25

Table 13 .
Average G * by the equilibrium duration timespan applied by labs using PP08.