1 Introduction
2 Theoretical background
2.1 Global luminescent oil-film skin-friction field estimation
2.2 Photogrammetry and determination of the oil-film thickness
2.3 Requirements on the measurement conditions
2.3.1 Critical limit of roughness
2.3.2 Courant–Friedrichs–Lewy condition
2.3.3 Neglection of pressure gradient and body force effects
3 Experimental configurations
3.1 Wind tunnel, test object, and conventional measurement techniques
3.2 Luminescent oil
3.3 Optics
3.4 Flow conditions
4 Measurement procedure
4.1 Measurement parameters
4.1.1 Estimation of the critical limit of roughness
4.1.2 Determination of the oil dynamic viscosity and estimation of the CFL number
4.1.3 Evaluation of the pressure gradient and body force effects on the oil-film thickness development
4.2 Calibration procedure
-
Dark current images [\(I_{\mathrm{dark}}\) in Eq. (9)] were acquired to account for the influence of residual light and camera electronic noise. These images were taken without any lighting (see Fig. 7a) for each \(\mathrm{AoA}\). The exposure time of the camera was the same as that set in the measurement (50 ms). The averages of the dark current images were then subtracted from the luminescent oil images (see Sect. 5.2).
-
To determine the calibration parameters for the projection of the GLOF images into the physical space (see Sects. 5.1 and 5.2), the airfoil model was covered by a flexible plastic sheet with a printed grid pattern, as shown in Fig. 8a. The grid images (Fig. 7b) were taken at each \(\mathrm{AoA}\) with an adjusted camera exposure time (different from the aforementioned exposure time of the dark current and luminescent oil images). The relative position of the grid with respect to a model reference point (the port side of the trailing edge) was also recorded.
-
Excitation light distribution images (\(I_{\mathrm{exc}}\) in Eq. (9)) were acquired by attaching a white paper sheet fully covering the GLOFSFE-evaluation region (see Fig. 8b). In fact, the used white paper included a luminescent dye, and thus allowed the simulation of a homogeneous distribution of luminescent oil film. The \(I_{\mathrm{exc}}\) images were taken at each \(\mathrm{AoA}\) under the excitation light source (see Fig. 7c). As compared to the dark current and luminescent oil images, the exposure time was shortened to 40 ms to avoid image saturation, since the intensity of the light emitted by the white paper was higher than that of the oil film.
-
Oil-droplet images were taken for the determination of the unit thickness \(h_*\), according to the ratioed image film thickness method (Husen et al. 2018)—see Eq. (7). Oil droplets were dropped onto the surface by means of a micro-pipette (Biomaster 4830, from Eppendorf) with a volume of 20 \(\upmu\)L, and then spread out using compressed air. The oil-droplet images (Fig. 7d) at each \(\mathrm{AoA}\) were acquired with the same exposure time as that of the dark current and luminescent oil images.
4.3 Wind-tunnel operation sequence
-
The oil was placed approximately at the upstream limit of the GLOFSFE-evaluation region, i.e., at \(x/c\sim 9\%\), prior to series S1; in this series, the \(\mathrm{AoA}\) was increased from \(-0.4^{\circ }\) to 2.0\(^{\circ }\) (S1–01 to S1–08), and then decreased back to \(-0.4^{\circ }\) (S1–09 to S1–15).
-
After the base coat had been exchanged, before series S2, the oil was applied at approximately the same location as that of series S1. The \(\mathrm{AoA}\) was varied from 1.2\(^{\circ }\) to 2.0\(^{\circ }\) (S2-01 to S2-04) in series S2.
-
After cleaning the oil, the oil application position was then moved to approximately 20% of the chord, and, in series S3, the \(\mathrm{AoA}\) was varied from 2.0\(^{\circ }\) to 1.2\(^{\circ }\) (S3-01 to S3-04), and then from \(-0.4^{\circ }\) to 0.8\(^{\circ }\) (S3-05 to S3-08). The base coat was not further exchanged after series S2.
-
After series S3, the oil was directly added to the investigated surface and, before the wind tunnel was operated again, spread out by means of compressed air. In the final series S4, the \(\mathrm{AoA}\) was varied from 1.2\(^{\circ }\) to 2.0\(^{\circ }\) (S4-01 to S4-04), and then from \(-0.4^{\circ }\) to 0.0\(^{\circ }\) (S4-05 to S4-06).
Series | AoA [\(^{\circ }\)] | Initial oil position [%] | New base coat | |
---|---|---|---|---|
S1 | \(-0.4\) to 2.0, increase | 1.8 to \(-0.4\), decrease | 9 | Yes |
S2 | 1.2 to 2.0, increase | 9 | Yes | |
S3 | 2.0 to 1.2, decrease | \(-0.4\) to 0.8, increase | 20 | No |
S4 | 1.2 to 2.0, increase | \(-0.4\) to 0.0, increase | 20–40 | No |
5 Data processing
5.1 Camera calibration
5.2 Image projection
5.3 GLOF image analysis and flow-data integration
5.4 Estimation of the uncertainties
-
The uncertainty in the oil dynamic viscosity \(\Delta \mu\) was estimated from the model surface temperature variation. The adiabatic-wall temperature variation in the GLOFSFE-evaluation region with laminar flow (the area where quantitative skin-friction estimations could be carried out, see Sect. 6.1), evaluated on the basis of the measured surface pressure distribution, flow total temperature, and freestream Mach number, was within ±1.1 K for all examined test conditions. The surface temperature variation in the considered evaluation region was expected to be even less than the adiabatic-wall temperature variation; for a conservative estimation, the uncertainty in the oil temperature was assumed to be \(\Delta T~=~\pm\)1.1 K. Incidentally, this is the same uncertainty in the oil temperature as that reported in Lee et al. (2020c), where it corresponded to an uncertainty in the oil dynamic viscosity of ±2.3%. This can be taken as a conservative estimation of the uncertainty in the oil dynamic viscosity in the present work.
-
The uncertainty in the unit length \(\Delta x_*\) depends on the camera pixel density, on the geometry of the surface, on the camera angle \(\theta _P\), on the camera lens distortion, and on the camera calibration and image projection process (see Sects. 5.1 and 5.2). The resulting estimation of the uncertainty in the unit length was within ±0.5%.
-
The uncertainty in the unit time \(\Delta t_*\) is determined by the camera timing accuracy. In the present work, \(\Delta t_*\) was estimated to be even smaller than the values reported in Lee et al. (2020c) (within ±0.02%) and was therefore regarded as negligible.
-
The uncertainty in the unit oil-film thickness \(h_*\) depends on \(\Delta x_*\) (discussed above), on the uncertainty in the calibration (ratioed) image \(\Delta \overline{r_{{\mathrm{cal}}}}\) and on the uncertainty in the oil-droplet volume \(\Delta v_{{\mathrm{droplet}}}\). Since the ratioed image film thickness method was applied in the same manner as in Lee et al. (2020c), the uncertainty in the oil-droplet volume was estimated as \(\Delta v_{{\mathrm{droplet}}}\) = ±1.65%, whereas the uncertainty in the calibration image could be considered as negligible (\(\Delta \overline{r_{{\mathrm{cal}}}}\) within ±0.003%).
-
As discussed in Lee et al. (2018, 2020c), the uncertainty in the normalized skin friction \(\Delta \hat{\tilde{\tau }}\) is determined by the GLOF image noise, the number of GLOF image pairs, and the “image characteristics”. This latter definition comprehends several factors, such as the offset error related to the base coat influence on the GLOF measurements (see Sect. 6.1) and the investigated flow conditions. In the present study, it was not possible to evaluate systematically \(\Delta \hat{{{\tilde{\tau }}}}\) because of the extremely challenging transonic wind-tunnel environment. Challenges for the application of the GLOF measurement technique in the continuously driven DNW-TWG wind tunnel were, among others, the accessibility for the oil application, the time needed to prepare the facility for wind-tunnel operation, and the time necessary to attain stable flow conditions (see Sect. 4.3). The issues that limited the repeatability of the GLOF measurements are described and discussed in Sect. 6.1.An estimation for \(\Delta \hat{{{\tilde{\tau }}}}\) in this work may be ±22.6%, based on the worst-case scenario presented in Lee et al. (2018).
6 Results and discussion
6.1 Repeatability of the results and evaluation of the GLOFSFE estimations
-
Influence of the initial oil position and of the time history of the oil filmWhen the measurements are started, the skin friction increases along with the flow direction, \({\partial \tau }/{\partial x} > 0\), and the oil film becomes thinner, \({\partial h}/{\partial t}= -h^2/2\mu \cdot {\partial \tau }/{\partial x} <0\). In series S1 and S2, the oil film positioned at 9% of the chord was affected by the stream-wise pressure gradient (see the pressure distribution in Fig. 4), which led to the formation of an oil ‘bump’. This can be observed in Figs. 9b and 12 for the series S2, which was terminated after only four data points. The oil ‘bump’ formed at \(10\%<x/c<20\%\) induced a ‘bump’ also in the estimated skin-friction distributions of series S2, which can be observed in Fig. 14 and, even more clearly, in Fig. 15 (green lines). In addition to this local effect, the oil ‘bump’ also induced premature transition at \(\mathrm{AoA} = 1.8^{\circ }\) and \(2.0^{\circ }\) (data points S2-03 and S2-04). (As introduced in Sect. 4.1.1, the position of transition onset was identified in the present work at the location where \(C_f\) became larger than the reference value for laminar flow \(C_{f,\mathrm{L}}=4 \cdot 10^{-4}\).) At \(\mathrm{AoA} = 1.2^{\circ }\) and \(1.5^{\circ }\) (data points S2-01 and S2-02), the bump effect on boundary-layer transition was negligible, since the chord-wise distance between the oil ‘bump’ and the location of natural transition onset was small. In series S1, the oil ‘bump’ formed at \(x/c\sim 10\%\), and its height was reduced during the conduction of this first series. Nevertheless, a difference in the transition position between the data points at \(\mathrm{AoA} = -0.4^{\circ }\), \(0.0^{\circ }\) and \(0.4^{\circ }\) (S1-01 to S1-03 vs. S1-15 to S1-13) can be seen. At these angles of attack, the boundary-layer stability situations were more sensitive to the effect of the bump (see Costantini et al. (2019)), and transition occurred more upstream for the data points S1-01 to S1-03, as compared to the cases S1-15 to S1-13. This effect of the oil ‘bump’ can be clearly seen in Fig. 15 (red vs. magenta lines). At \(\mathrm{AoA} \ge 0.8^{\circ }\), the influence of the oil ‘bump’ on the transition weakened with increasing angle of attack (until it essentially vanished at \(\mathrm{AoA}\ge 1.5^{\circ }\)), because the boundary-layer stability situations became less sensitive to the bump effect. However, a higher oil ‘bump’ did induce premature transition at \(\mathrm{AoA} = 1.8^{\circ }\) and \(2.0^{\circ }\) in series S3, where the ‘bump’ formed at \(25\%<x/c<30\%\). As can be seen in Figs. 14 and 15, transition occurred at a more upstream location for the data points S3-01 and S3-02, as compared to the data points from series S1. As the angle of attack was further reduced to \(\mathrm{AoA}=1.5^{\circ }\) and \(1.2^{\circ }\) (data points S3-03 and S3-04), the oil ‘bump’ moved into a more downstream location (see Fig. 12), i.e., downstream of the location of natural transition onset. Therefore, the transition location at these two angles of attack in series S3 was essentially the same as that found in series S1 and S2 (see Fig. 15).As an example of the influence of the time history of the oil film on the skin-friction estimations in the turbulent-flow regions, it can be seen that a oil ‘valley’ formed in series S1 at \(x/c>25\%\) starting from data point S1-03, probably induced by the change in the transition location between the data points S1-02 and S1-03. The oil ‘valley’ led to a ‘valley’ also in the skin-friction distributions estimated in the turbulent-flow regions at \(0.8^{\circ }\le \mathrm{AoA}\le 1.5^{\circ }\), which can be observed in Figs. 14 and 15.
-
Separation and reattachment linesAs discussed above, the gray-masked stripes visible in Figs. 13 and 14 at approximately \(x/c<30\%\) (data series S1 and S2), which are essentially oriented in the span-wise direction, were very likely related to the zero skin friction at the critical lines and to the large skin friction upstream and downstream of flow reattachment. In particular, this latter effect was expected to be the reason for the efficient oil-film removal around the reattachment location, which most probably led to the very low oil-film thickness in the span-wise-oriented stripes (see Fig. 12), and thus to the impossibility to estimate the skin friction using the GLOFSFE. The appearance of multiple stripes and the “smearing” of the gray-masked stripes with varying angle of attack in the same data series were also influences of the aforementioned time history of the oil film, but are discussed here because of their specificity. In fact, in series S1, the stripes related to the critical lines started to appear in Figs. 12, 13 and 14 from the data point S1-04 (\(\mathrm{AoA}=0.8^{\circ }\)), and the number of stripes increased with increasing angle of attack up to \(\mathrm{AoA}=1.5^{\circ }\) (data point S1-06). In practice, the stripes observed for a certain data point remained “imprinted” also in the subsequent data points. The approximate chord-wise locations of the stripes were in agreement with the expectations for the chord-wise locations of the critical lines, based on the measured pressure distributions (see Fig. 4). It should be noted here that, at \(0.8^{\circ } \le \mathrm{AoA} \le 1.5^{\circ }\), the reattachment location expected for a certain angle of attack approximately overlapped the separation location expected for the subsequent, larger \(\mathrm AoA\). It is therefore difficult to unequivocally discern between stripes related to flow separation or reattachment in the presence of multiple stripes; however, for the reasons discussed above, these most probably originated from the efficient oil-film removal around the reattachment location. The gray-masked stripes observed in series S2 appear to confirm this conjecture. In this second data series, a marked low-\({\bar{h}}\)/gray-masked stripe was seen at \(x/c \sim 19\%\), starting from the first data point of the series (S2-01). An additional stripe was observed at \(x/c \sim 22\%\), starting from the data point S2-02. The chord-wise locations of these two stripes were in agreement with the expected reattachment locations at \(\mathrm{AoA}=1.2^{\circ }\) and \(\mathrm{AoA}=1.5^{\circ }\). In both data series, as the angle of attack was further increased, the low-\({\bar{h}}\)/gray-masked stripes were found to remain “imprinted” also in the subsequent data points, i.e., S1-07 and S1-08 in series S1, and S2-03 and S2-04 in series S2.In series S1, as the angle of attack was reduced again, the stripes were observed to “smear”: this is especially visible for the gray-masked regions of the data points S1-11 to S1-15. The different behavior observed for increasing and decreasing angle of attack was very likely related to the location of flow reattachment at a certain data point, with respect to that at the preceding data point. As the angle of attack was increased, flow separation and reattachment occurred at a more downstream location; because of the still low skin friction in the region around the previous location of flow reattachment, the “imprinted” distributions remained essentially unaffected by the change in \(\mathrm{AoA}\). In contrast, as the angle of attack was reduced, flow separation and reattachment occurred at a more upstream location; the now larger, positive skin friction in the region around the previous location of flow reattachment was capable of modifying the “imprinted” distributions, thus leading to the “smearing” of the low-\({\bar{h}}\)/gray-masked regions.Because of these issues, it was not feasible to extract the locations of flow separation and reattachment from the gray-masked stripes in the GLOFSFE estimations.
-
Influence of the base coatThe interaction between the oil film and the base coat applied to the measurement surface affected the GLOF measurements. One result of this interaction can be seen for series S1 in Fig. 12, where fixed cyan circular traces appeared at \(y/c \sim 0.4\) and 0.6. These were traces of the oil droplets used in the calibration process (see Sect. 4.2), which can be observed in the oil-droplet image of Fig. 10c. The observed pattern was the result of the absorption of the remaining luminescent oil droplets by the base coat, and corresponded to approximately 3 \(\upmu\)m offset error in the oil-film thickness. The oil-droplet calibration was not repeated after the base coat had been changed (i.e., from series S2, see Sect. 4.3). Therefore, the oil patterns did not appear in series S2–S4.Another result of the interaction between the oil film and the base coat can be seen in Fig. 12 for series S3 and S4. In these two latter data series, the oil had not been applied upstream of \(x/c \sim 20\%\), but appreciable luminescent intensity was still measurable even at \(x/c<18\%\). This luminescent intensity was due to the dye absorbed in this region during the preceding series S2. In this region, the corresponding offset error in the oil-film thickness of series S3 and S4 was approximately 5 \(\upmu\)m.The offset error affects the estimation of the normalized skin-friction field (see Eqs. (5) and (6) in Sect. 2.1) as:and$$\begin{aligned}&\mathbf{{C}}=\frac{1}{4}\sum \limits _k {{\left\{ {\mathrm{{diag}}\left( {{\mathbf{{M}}_{{\mathrm{c2f}}}} \left( {{{{\underline{\varvec{r}}}}}_k} + \epsilon _r \right) } \right) } \right\} }^2}{\varvec{\Delta }}_{\mathbf{x}}^{{\mathrm{T}}}{\varvec{\Delta }}_{\mathbf{x}}\nonumber \\&\quad {{\left\{ {\mathrm{{diag}}\left( {{\mathbf{{M}}_{{\mathrm{c2f}}}} \left( {{{{\underline{\varvec{r}}}}}_k} + \epsilon _r \right) } \right) } \right\} }^2} , \end{aligned}$$(22)In the above equations, the offset error \(\epsilon _r\) (assumed as a constant) is added to the ratioed image \({{{{\underline{\varvec{r}}}}}_k}\).$$\begin{aligned} \mathbf{{d}}=-\frac{1}{2}\sum \limits _k {{{\left\{ {\mathrm{{diag}}\left( {{\mathbf{{M}}_{{\mathrm{c2f}}}} \left( {{{{\underline{\varvec{r}}}}}_k} + \epsilon _r \right) } \right) } \right\} }^2}{\varvec{\Delta }}_{\mathbf{x}}^{{\mathrm{T}}}{\varvec{\Delta }}_{\mathbf{t}} \left( {{{{\underline{\varvec{r}}}}}_k} + \epsilon _r \right) }. \end{aligned}$$(23)This leads to an offset-induced error \({\varvec{\epsilon _{{\hat{\tau }}}}}\) in the estimated skin friction (see Eq. (3)) that can be presented as:where \({\varvec{\epsilon _{\hat{{{\tilde{\tau }}}}}}}\) is the offset-induced error in the normalized skin friction. In general, the relation between \({\varvec{\epsilon _{\hat{{{\tilde{\tau }}}}}}}\) and \(\epsilon _r\) is non-linear.$$\begin{aligned} {{\varvec{{\hat{\tau }} }}} + {\varvec{\epsilon _{{\hat{\tau }}}}} = \tau _* \left( { {\varvec{\hat{{{\tilde{\tau }}}}}} + {\varvec{\epsilon _{\hat{{{\tilde{\tau }}}}}}}} \right) , \end{aligned}$$(24)The offset error is particularly critical when the oil film is thin. As discussed in the beginning of this Subsection, the regions with low values of the coefficient of determination in Fig. 13 correspond to the areas with low oil-film thickness. These low values of \(R^2\) were induced by the low signal-to-noise ratio of the GLOF images in these regions and by the offset error. In the present work, the offset error could not be compensated, because the pigmentation was neither uniform nor time-independent. The conduction of the GLOF measurements with a comparably thick oil film enabled the impact of the offset error in the GLOFSFE estimations to be mitigated; at the same time, the higher oil-film thickness induced the mutual interactions between the oil film and the boundary layer discussed above.
-
Turbulent wedgesTurbulent wedges were observed superimposed on the quasi-2D skin-friction distributions for some of the data points (see Fig. 14). In the case of the data points S2-03 and S2-04, a large turbulent wedge was seen in the region around \(y/c=0.60\). After the wind tunnel had been stopped, a particle was found on the oil film during model inspection. The position of the particle corresponded to the location of the apex of the wedge visible in Fig. 9b. Particles passing through the wind tunnel may be captured by the oil film and act as 3D roughness elements. A roughness element of critical size at a critical position could induce a three-dimensional flow and trigger bypass transition (see Sect. 2.3.1), as it was the case for the aforementioned particle in series S2. The small turbulent wedge observed in the data points S1-02 to S1-06 (around \(y/c=0.57\)) probably occurred in a similar manner.The origin of the turbulent wedge seen in the data points of series S4 (S4-03 to S4-06) around \(y/c=0.51\) was different. The span-wise location of the wedge apex matched that of a channel for the installation of the tubes for the pressure taps, which was filled by resin before the experiments were started (see Fig. 1). However, the filling resin contracted over time and formed a shallow depression, which also acted as a 3D surface imperfection and induced the occurrence of the turbulent wedge seen in series S4. Interestingly, a small turbulent wedge at approximately the same span-wise location was observed also in the initial data points of series S1 and S3 (i.e., S1-01 and S3-01). It is possible that, because of the presence of the oil film on the model surface, the boundary layer had not reached stable conditions during the acquisition of the initial data points of these two series; the otherwise small three-dimensional disturbances originating from the filled channel may have had larger initial amplitude or may have undergone stronger amplification, as compared to the other data points, thus leading to the formation of a turbulent wedge at a more downstream location.The sources of the further turbulent wedges observed in Fig. 14 at the span-wise ends of the evaluation region (data points S1-07 to S1-08, S3-01 to S3-02, S3-05, and S4-03 to S4-04) were the edges of the base coat. These turbulent wedges can be also seen in the oil-film patterns shown in Figs. 2 and 9b. The base-coat edges were span-wise steps of approximately 100 \(\upmu\)m height, running from the leading edge to the trailing edge in the stream-wise direction. The step height at the leading edge was sufficiently large so as to affect the laminar flow (see Sect. 4.1.1), and bypass transition occurred.
-
S1-01 to S1-03, S2-03 to S2-04, and S3-01 to S3-02, because of the premature transition induced by the oil ‘bump’;
-
S2-01 to S2-02, because of the ‘bump’ in the skin-friction distribution induced by the oil ‘bump’ upstream of the natural transition onset;
-
S3-03 to S3-04, and S4-01 to S4-02, because the stream-wise extent of the laminar-flow region measurable via GLOF was only a few percent of the chord, before the (shock-induced) transition process was initiated. (The luminescent oil was applied at \(x/c\sim 20\%\) prior to series S3.)
6.2 Aerodynamic coefficients
6.3 Discussion of the GLOFSFE estimations in the laminar-flow region and of the detected locations of transition onset
-
At \(\mathrm{AoA}=1.8^{\circ }\) and \(2.0^{\circ }\), the reasonable agreement of the estimated and computed \(C_{fx}\) profiles was observed in a significant stream-wise region between \(x/c \sim 18\%\) and the location corresponding to the transition onset in the experiments, except for a small region around \(x/c \sim 24\%\), where a minimum of the estimated \(C_{fx}\) can be seen. The low values of estimated \(C_{fx}\) in this small region were due to the locally thinner oil film, as compared to the upstream and downstream regions (see Figs. 12 and 17a). The locally thinner oil film was probably induced by the influences of the time history of the oil film and of the presence of critical lines in the preceding data points (see Sect. 6.1); in particular, the low values of estimated \(C_{fx}\) in the region around \(x/c \sim 24\%\) were likely related to the oil ‘valley’ formed from data point S1-03 and to the region of apparently nearly-zero skin friction formed from data point S1-06.In the remaining portions of the aforementioned stream-wise region of reasonable agreement, the numerical predictions were within the \(C_{fx}\) range given by the GLOFSFE estimations with the representative measurement uncertainty. Moreover, at \(\mathrm{AoA}=1.8^{\circ }\), the estimated location of transition onset was very close to the predicted location of laminar separation (\(x_\mathrm{T,onset}/c\) was indeed slightly downstream of \(x_{\mathrm{S}}/c \sim 31\%\)). At these flow conditions, transition was expected to start shortly downstream of the boundary-layer separation induced by the adverse pressure gradient at \(x/c>30\%\) (see Fig. 4). The estimated location of transition onset was also slightly upstream of the transition location from the TSP experiments, as can be seen in Fig. 19. The observed agreement of the locations of predicted laminar separation, estimated transition onset, and transition detected via TSP was a notable result. At \(\mathrm{AoA}=2.0^{\circ }\), transition occurred in a region significantly upstream of the location of the shock visible in Fig. 4. It appears that the adverse pressure gradient in the region at approximately \(30\% \le x/c \le 37\%\) led to marked amplification of the boundary-layer disturbances and thus induced transition, even though the laminar boundary layer was not predicted to separate (see Figs. 17a and 18). The location of transition onset estimated via GLOFSFE was found to be upstream of the transition location measured via TSP, with a small distance between these two locations, in agreement with the expectations.At approximately \(x/c < 18\%\), the values of skin-friction coefficient obtained for both angles of attack via GLOFSFE were larger than the predicted \(C_{fx}\), since the oil film was too thick in this region close to the leading edge (see Fig. 12). As can be seen in Fig. 17a, \({{{\overline{h}}}}^+ > 7\) seems to provide an indication for the regions where the skin-friction coefficient was overestimated.
-
At \(0.8^{\circ } \le \mathrm{AoA} \le 1.5^{\circ }\), the agreement of the computed and estimated \(C_{fx}\) profiles (considering also the representative measurement uncertainty) was reasonable for small stream-wise regions up to the predicted location of laminar separation: from \(x/c \sim 9\%\) to \(x_{\mathrm{S}}/c \sim 12\%\) at \(\mathrm{AoA}=0.8^{\circ }\), from \(x/c \sim 13\%\) to \(x_{\mathrm{S}}/c \sim 16\%\) at \(\mathrm{AoA}=1.2^{\circ }\), and from \(x/c \sim 13\%\) to \(x_{\mathrm{S}}/c \sim 18\%\) at \(\mathrm{AoA}=1.5^{\circ }\). Similarly to the cases presented above, \(C_{fx}\) at \(\mathrm{AoA}=1.2^{\circ }\) and \(1.5^{\circ }\) was overestimated upstream of \(x/c \sim 13\%\), i.e., in the region close to the leading edge where the oil film was too thick. As compared to \(\mathrm{AoA}=1.8^{\circ }\) and \(2.0^{\circ }\), however, the overestimation region was of smaller stream-wise extent for these lower angles of attack; in particular, at \(\mathrm{AoA}=0.8^{\circ }\), \(\overline{h^+}\) remained close to or below 5 for the whole GLOFSFE-evaluation region up to the estimated location of transition onset. This difference with the two cases at the largest \(\mathrm{AoA}\) was due to the local increase in the boundary-layer thickness at approximately \(x/c > 9\%\) induced by the adverse pressure gradient observed in this region for \(0.8^{\circ } \le \mathrm{AoA} \le 1.5^{\circ }\), as compared to the nearly-zero pressure gradient for \(\mathrm{AoA}=1.8^{\circ }\) and \(2.0^{\circ }\) (see Fig. 4). Also in the three cases at \(\mathrm{AoA}=0.8^{\circ }\), \(1.2^{\circ }\), and \(1.5^{\circ }\), boundary-layer transition was expected to start shortly downstream of the laminar separation induced by the strong adverse pressure gradient (very marked in the case of the shock wave at \(\mathrm{AoA}=1.2^{\circ }\) and \(1.5^{\circ }\)). The estimated location of transition onset was slightly upstream of the predicted location of laminar separation at \(\mathrm{AoA}=1.2^{\circ }\), whereas it was slightly downstream at \(\mathrm{AoA}=0.8^{\circ }\) and \(1.5^{\circ }\). Considering also the reported deviations \(\Delta x_{\mathrm{T,onset}}/c\) in the GLOFSFE estimations, the results were in agreement with the expectations. Moreover, the locations of transition onset estimated via GLOFSFE were in agreement with the transition locations measured via TSP (see Fig. 19).
-
At \(\mathrm{AoA} \le 0.4^{\circ }\), the comparison of computed and estimated \(C_{fx}\) profiles for the selected data points was possible only for a laminar-flow region of small stream-wise extent, since the GLOFSFE estimations for the data points S3-05 to S3-07 were available at approximately \(x/c \ge 20\%\). With consideration of the representative measurement uncertainty, the numerical and estimated \(C_{fx}\) values at \(\mathrm{AoA} = 0.0^{\circ }\) and \(-0.4^{\circ }\) were in reasonable agreement in the small stream-wise regions between \(x/c \sim 0.22\) and the estimated location of transition onset, i.e., for approximately \(\Delta x/c=4-5\%\). The comparison for the case at \(\mathrm{AoA} = 0.4^{\circ }\) is presented here only for the sake of completeness, since transition was observed to start in the experiments at an even more upstream location (\(x_{\mathrm{T,onset}}/c \sim 23\%\)) than that at \(\mathrm{AoA} = 0.0^{\circ }\) and \(-0.4^{\circ }\). Therefore, reasonable agreement of numerical and estimated \(C_{fx}\) values could be observed only in the proximity of the estimated location of transition onset.At \(\mathrm{AoA} = -0.4^{\circ }\), the location of transition onset estimated via GLOFSFE was in agreement with the transition location measured via TSP (see Fig. 19). However, at \(\mathrm{AoA} = 0.0^{\circ }\) and \(0.4^{\circ }\), \(x_{\mathrm{T,onset}}/c\) was downstream of the location \(x_{\mathrm{T}}/c\) detected in the TSP measurements. An analysis of the TSP results revealed that transition occurred prematurely at these two latter angles of attack, very likely because of an imperfection in the contour of the model when it was equipped with TSP. Therefore, the difference from the expectations observed at \(\mathrm{AoA} = 0.0^{\circ }\) and \(0.4^{\circ }\) seems to be ascribable to the TSP measurements, rather than to the GLOFSFE estimations.
7 Conclusions and recommendations for future work
-
The base coat should be considered for the prevention of the offset intensity error and for the elimination of the span-wise discontinuities at the base-coat edges, which would otherwise induce turbulent wedges. An ideal surface would be a mirror-finished metallic surface that cannot absorb the luminescent dye. In this case, the excitation light reflection should be controlled so as not to be directed toward the camera.If the oil film has to be applied onto a model surface that cannot be modified, and especially when the investigated model surface is already instrumented with pressure taps or other sensors, a thin, reflective metallic coat would be the most appropriate choice as a base coat.
-
The oil-film thickness should be adjusted to a level such that the minimum thickness would be below the critical limit of roughness in both laminar- and turbulent-flow regions, while at the same time, the signal-to-noise ratio would remain sufficiently high. In the case of a still present offset intensity error, the measurement feasibility should be determined by the ratio of the oil luminescent intensity to the offset intensity.According to the present observations and to the finding of Lee et al. (2020c), the optimal oil-film thickness in wall units \(h^+\) should be indicatively slightly below 5. Note here that the coincidence of the recommendations about the optimal \(h^+\) for laminar- and turbulent-flow regions is probably incidental, since the oil-film thickness estimation in the laminar-flow regions examined in the present work was affected by an offset error. Nevertheless, even with consideration of the offset error, the oil-film thickness should be generally kept below \(h^+ \sim 5\) in order to avoid the risk of overestimating the skin friction.Obviously, it is very challenging to guarantee the recommended \(h^+\) condition simultaneously in both laminar- and turbulent-flow regions. Therefore, the recommended oil-film thickness may be achieved, during the same wind-tunnel operation, first in the laminar boundary layer, and then—after running the wind tunnel for a long time—in the turbulent boundary layer. Alternatively, the oil-film thickness may be optimized for laminar- and turbulent-flow regions in different wind-tunnel runs.
-
Efforts in the development of the oil-dye mixture should aim to the achievement of a brighter oil film. This would improve the signal-to-noise ratio of GLOF data, especially in measurements with a thin oil film.
-
To improve repeatability, the measurements should indicatively focus on the 3–5 data points recorded after the first 3–4 different angles of attack, independently of the initial position and distribution of the oil film. In fact, the present observations showed that the first data points were still significantly affected by the starting oil conditions, whereas the signal-to-noise ratio was significantly reduced in the late data points.
-
If laminar separation bubbles are expected to occur for a certain range of \(\mathrm{AoA}\), it is recommended to start the acquisitions from the angle of attack at which flow separation and reattachment are expected at the most downstream locations, and then increase/decrease the angle of attack to induce a progressive, upstream shift of the position of the laminar separation bubble. In this manner, the “imprinting” of regions of very low oil-film thickness in the subsequent data points should be avoided.
-
A reduction of the oil viscosity, as compared to that considered in this work, would be a major advantage for future GLOF measurements with a thin oil film. However, it should be accepted that the oil would relatively quickly reach the trailing edge and then separate from the model, thus possibly damaging the wake-rake probes and contaminating the wind tunnel.