Background-oriented schlieren visualization of heating and ventilation flows: HVAC-BOS.
Frequently the need arises to measure HVAC flow patterns to evaluate performance characteristics. Traditionally, a limited set of point measurements is made inside a flowfield, and overall flow characteristics are extrapolated from these limited data. Modern optical techniques, such as particle image velocimetry (PIV), particle streak velocimetry, infrared (IR) imaging, and large-field schlieren imaging, have also been applied to measuring whole-field HVAC airflow characteristics, as recently reviewed by Sandberg (2007). Many of these techniques, however, require complex instrumentation or elaborate physical setups and can directly interfere with or modify the flowfield under study through the injection of particles, smoke, or imaging screens.
The most successful method surveyed by Sandberg (2007) was a simple and ingenious use of ordinary window screen stretched across a room. Temperature variations imposed upon the window screen by the airflow were observed using an IR camera (Anderson et al. 1991). However, this approach is intrusive to the room airflow and has other problems, such as radiation between the screen and the walls, no frequency response, and the need for an expensive camera. A novel approach is suggested here for nonintrusive imaging of ventilation airflows: background-oriented schlieren (BOS).
The BOS technique was introduced almost simultaneously by Meier (Meier 1999) and by Dalziel et al. (1998), who called it "synthetic schlieren.". This technique allows refractive flowfields to be visualized by their distortion of a distant patterned background. By comparing two images of the background, one with a refractive disturbance and one without, the refractive index and density gradients within the disturbance can be determined from the apparent shift in the background pattern (Venkatakrishnan and Meier 2004).
The BOS technique is essentially a modern reinvention of Schardin's simple background-distortion schlieren method (Schardin 1942; Settles 2001), but with the advantage of modern digital technology. The two images required for a BOS visualization can now be digitally recorded with high fidelity using modern consumer-grade single-lens-reflex (SLR) cameras. The image processing, performed photographically by Schardin (Schardin 1942), is now accomplished using readily available computer software algorithms. Neither the camera nor the background is essentially new in BOS, but the addition of modern digital image processing makes this a new and valuable flow visualization technique.
The simplicity of the BOS technique allows it to be applied across a wide range of scales, with the measurement region defined purely by the available camera field-of-view and the physical size of an appropriate background. This technique has been successfully applied to large-scale visualizations outdoors for measurement of helicopter rotor vorticies (Richard and Raffel 2001) and to thermal plumes from automobiles and flames (Hargather and Settles 2010). In the laboratory environment, the BOS technique has been used primarily to visualize and measure density gradients in high-speed compressible flows (Venkatakrishnan and Meier 2004; Elsinga et al. 2004) and density-stratified liquids (Dalziel et al. 2000, 2007).
Presented in this article is the use of the BOS technique for large-field, in situ visualization of HVAC flowfields. The physical principles involved are discussed, adaptations for present purposes are considered, and a variety of HVAC flows visualized by the BOS technique are shown.
Experimental setup: BOS imaging basics
The BOS technique visualizes refractive disturbances via their distortion of a distant background. To perform BOS imaging, two images of a background pattern are required: one with a refractive flow disturbance and one without. These two images, which together compose a BOS image pair, are then analyzed using digital image correlation (DIC) software. The software identifies locations within the background pattern that have changed between the two images and quantifies the change in the background pattern in terms of a "pixel shift" through which the background has been apparently moved (distorted) due to the refractive disturbance. This pixel shift is directly related to the refraction in the schlieren object and to the geometry of the BOS setup.
Schardin's (1942) schlieren method #1, being typical of BOS illumination, is illustrated in Figure 1. Here, for simplicity, a single light-dark boundary is shown in the background (B). Point P in the schlieren object S refracts light through angle [[epsilon].sub.y]. This causes an apparent background distortion or shift of a point on the optical axis to a point at P". This shift is recorded by the camera as a distortion of the background, i.e., a "pixel shift."
The sensitivity of this system, i.e., the smallest refraction angle [[epsilon].sub.y] that can be detected, is a function of the optical geometry shown in Figure 1, the camera capability, and the strength of the schlieren object S. In general, a light ray traveling in the negative z-direction will be refracted through an angle [[epsilon].sub.y] due to the refractive index gradient in the y-direction within the schlieren object S:
[[epsilon].sub.y] = 1/n [integral] [partial derivative]n/[partial derivative]y dz.
[FIGURE 1 OMITTED]
For a two-dimensional schlieren object of length Z along the optical path, this equation can be reduced (Settles 2001) to
[[epsilon].sub.y] = Z/n [infinity] [partial derivative]n/[partial derivative]y,
where the refractive index of the ambient air is given by n [infinity]. The refractive index gradient in the y-direction [partial derivative]n/[partial derivative]y is directly related to the density-gradient distribution of the schlieren object in the y-direction through the Gladstone-Dale relation (Settles 2001), which can be converted to a temperature distribution via integration and application of the ideal gas law. The stronger the variations in density and temperature in the schlieren object, the greater the gradient of refractive index and, thus, the greater the refraction angle [[epsilon].sub.y].
In the BOS technique, the refraction angle is measured via the pixel shift between the two images of a BOS image pair. This pixel shift, represented as e in Figure 1, is related to the refraction angle through
tan [[epsilon].sub.y] = e/L-t.
By measuring the pixel shifts, and thus the refraction angles, throughout an image, a quantitative measurement of the refractive index, temperature, and density fields can be obtained (Goldhahn and Seume 2007; Hargather and Settles 2011). The present article, however, presents purely qualitative visualization of the refractive fields in terms of pixel shifts, since, in HVAC, a qualitative image of the airflow is often more useful than quantitative data.
The smallest refraction angle that can be visualized, i.e., the smallest temperature or density gradient observable, depends on the ability to measure a pixel shift between two images. Current commercial digital-image-correlation algorithms using sub-pixel resolution techniques can accurately resolve pixel shifts on the order of one-tenth of a pixel width. The geometry of a BOS arranged should thus be set up so that the smallest refractive index gradient to be visualized creates a shift of at least 0.1 pixel. With a fixed minimum observable pixel shift, the sensitivity of a BOS system can then be increased by increasing the distance between the subject and the background, i.e., increasing L-t in Figure 1.
Camera lens focal length and camera sensor pixel size also affect the ability to detect small disturbances, but, in general, a distant background, imaged with a long-focal-length lens by a camera of high pixel resolution, results in the greatest sensitivity. This is constrained, however, by depth-of-field, since maintaining both the background and the schlieren object (the subject airflow) in reasonable focus is important for the success of BOS processing. In the present work, the required sensitivity was established by choosing a suitable distance L-t within the constraints of a given room, and then choosing t to provide acceptable depth-of-field. This procedure usually places the schlieren object at about 0.5< t/L <0.75. Ideally, the schlieren object would be halfway between the camera and the background, t/L = 0.5, but this may be limited by the available depth-of-field and is sometimes difficult to obtain in practical indoor situations.
BOS background design
A pixel shift between two BOS images will only be detected if the light refraction caused by the airflow under study is strong enough to cause an apparent change in the background pattern or an effective crossing of a light-dark boundary within the background. The background for BOS imaging thus is composed of random patterns, i.e., many light-dark boundaries on a scale that can be clearly and uniquely imaged by the camera that is focused upon it (Hargather and Settles 2010). In order to effectively measure a schlieren disturbance, the background pattern should have pattern sizes that are approximately "tuned" to the expected magnitude range of the schlieren disturbance being studied. Small disturbances will not be observed, for example, if they do not cause a measurable distortion of the background pattern.
By incorporating different-scale pattern "noise" into the background, it has been found that results are qualitatively better than from a purely random pattern (Figure 2). The incorporation of different background scales facilitates the simultaneous imaging of a wide range of schlieren disturbance scales, e.g., different turbulent structure sizes, and allows the background to be used for a range of distances L. The multi-scale backgrounds, however, can also lead to noisy or grainy schlieren images if the largest scale structures are too prominent, and thus result in regions with no distinct features. These qualitative observations are similar to those seen in related fields, where random patterns are used to resolve image details or pixel deformations (Cook and DeRose 2005; Siebert et al. 2007).
[FIGURE 2 OMITTED]
A consumer-grade Nikon D90 SLR camera with an 18-135-mm zoom lens and 12.3-megapixel sensor resolution was used for recording the present images. A modern digital SLR camera is preferred here for recording the BOS image pair because it provides high pixel resolution (thus high sensitivity), ease of use, portability, and modest cost. Modern high-speed digital cameras can also be used, but these typically have lower pixel resolution, limited portability, and prohibitive cost, which decrease their utility for HVAC-BOS. They are, in general, not required or preferred for most HVAC-BOS scenarios.
When recording each BOS image pair, it is important to minimize camera motion and shutter jitter and to avoid any changes in field-of-view or lens zoom between images. To prevent any camera motion caused by physically pressing the shutter release, the camera here was rigidly mounted to a tripod and triggered using an IR trigger available as a camera accessory.
The BOS image pair obtained as described above must then be computer processed to yield a measure of background distortion between the two images. Commercially-available PIV or DIC software packages are useful for this purpose (Atcheson et al. 2009), although in principle, a custom-written cross-correlation algorithm also works (Hargather et al. 2010). "Vic 2D" image-correlation software from Correlated Solutions, Inc., (2009) was used here. Results are ultimately presented using a range of plotting packages, such as TecPlot 360 as used here.
Experimental results: image processing and setup geometry influences
BOS image-pair processing
Traditional BOS imaging compares a background image taken with a refractive flowfield disturbance to a "tare" image taken with no flow present. Processing this image pair reveals the refractive index gradient field that was present at the instant of the flowfield image.
The processing of images begins by identifying a unique structure in the "tare" (no-flow) background image and then locating the same structure in the flowfield image. Any difference in the structure's pixel location between the images is a result of the refractive disturbance between the camera and the background. The change in the structure's pixel location can be measured in terms of a pixel shift in the two-dimensional plane of the background. The pixel shift magnitude and direction indicate the strength and direction, respectively, of the refractive index gradient field. The entire image is thus processed via the identification of unique background structures and their apparent change in location between the tare and flowfield images, providing refractive index gradient data throughout the available image area. The measurement region of a BOS system is thus limited only by the camera field of view and the dimensions of the background with its unique features.
[FIGURE 3 OMITTED]
The BOS refractive index gradient field can be visualized by contour plotting the measured pixel shifts as previously described. Figure 3b shows a contour plot of the measured horizontal pixel shift between a tare and flowfield image of a candle. The horizontal pixel shift contour plot is optically similar to a traditional vertical knife-edge schlieren image, as shown in Figure 3a, and clearly shows the hot candle plume, including its laminar, transitional, and turbulent regions.
A new method of processing BOS images is presented in Figures 3c and 3d; no tare image is used, but instead, two different flowfield images are processed relative to one another. This procedure reveals only the changes in the refractive flowfield between the two images. Figures 3c and 3d both use the same base image but compare it to images recorded 0.02 sec and 0.1 sec later. The laminar region of the candle plume is not revealed because its refraction does not change between the images of the BOS pair. The turbulent part of the plume is visible, however, due to advection of turbulent eddies in the interim. This technique can thus reveal the time-history of refractive disturbances in turbulent flows or flows that are otherwise unsteady. It is also effective for HVAC flowfields where turbulent airflow is the norm but where the flow may not always be easily shut off in order to take a tare image.
Sensitivity changes due to BOS setup geometry
The sensitivity of the BOS technique is directly related to the experimental setup geometry, especially the distance L-t from the schlieren object to the background (Figure 1). Figure 4 shows three BOS images of a hot teakettle with increasing camera-to-subject distance t and fixed camera-to-background distance L = 4.7 m (15.4 ft). Figure 4b represents the ideal setup with the teakettle located halfway between the background and the camera. Overall, the subject is in reasonable focus while the background is sharply focused so that pixel shifts are accurately resolved. Although Figure 4b shows the highest BOS sensitivity and resolution available in this example, it does not rival that of the traditional mirror-schlieren image in Figure 4a. On the other hand, such a large traditional mirror-schlieren setup is impactical for studying HVAC flow patterns in buildings. As the subject approaches the background (t [right arrow] L), in Figures 4b-4d, much of the flowfield detail disappears as pixel shifts are reduced toward the ambient noise level of the system.
[FIGURE 4 OMITTED]
BOS image presentation
The proper presentation of BOS results can be thwarted if default values, especially color contour maps, of the commercial image-processing software are accepted uncritically. The trained eye expects a schlieren image to appear in highlights and shadows upon a gray or neutral background. Different color schemes, such as the ubiquitous spectrum, can result in un-schlieren-like BOS images like the one illustrated in Figure 5b, which lacks contrast. On the other hand, a simple grayscale BOS image, like those shown in Figures 3 and 4, can be post-processed by popular software, such as Adobe Photoshop, to produce a variety of effects including colorization. Figure 5 shows the results of different contour-map choices for rendering the horizontal pixel shifts of the candle-plume image from Figure 3b.
[FIGURE 5 OMITTED]
[FIGURE 6 OMITTED]
Experimental results: HVAC applications and practical considerations
For purposes of illustration, a representative set of HVAC scenarios was visualized within a residential home and an office building using the BOS above-described techniques.
The first scenario, a ceiling-mounted hot-air diffuser vent, is shown in Figure 6. To image this airflow, the camera and background were raised to the ceiling and positioned so that the vent was halfway between them. The overall distance L was maximized within the available room size, thus maximizing sensitivity. Tare images were recorded before the heat was turned on, then several flow images were recorded with hot air being exhausted at approximately 10[degrees]C (50[degrees]F) above ambient.
Figure 6b was obtained from processing one flowfield image with one tare image, revealing a high turbulent "noise" level. To improve the visualization and reduce the noise, a MATLAB program was used to average five tare images and five flowfield images. These two averaged images were then processed as a BOS pair, resulting in Figure 6c. Through this averaging process, the turbulence "noise" is reduced, and the mean vent flowfield becomes qualitatively more visible.
[FIGURE 7 OMITTED]
The same averaging technique was applied to the visualization of flow from the floor-level forced-air grille shown in Figure 7. The hot air jet exiting the grille attaches to the floor via the Coanda effect, forming a "wall jet" across the field of view. The vertical BOS pixel shift is used here (horizontal knife edge schlieren equivalent) to best visualize the refractive-index gradients that are expected to be vertically oriented.
Figure 8 shows BOS images of (a) a passive wall-mounted heating register and (b)-(c) a fan-driven portable space heater. In Figure 8a, the overall pixel shift is significantly larger than other cases shown here because the hot-air plume is more than 20[degrees]C (36[degrees]F) hotter than the ambient air. Figures 8b and 8c show a good example of a commercial product with a design flaw that is easily detected by BOS visualization; the one-sided and two-sided exhaust control settings produce virtually identical flowfields.
The human thermal plume is the buoyant convective plume of air produced by every human due to skin warmth (Craven and Settles 2006), and is an important flowfield in HVAC applications, including personal ventilation, disease spread, and indoor-air quality. It is visualized in Figure 9 using the BOS technique. The human thermal plume is a weak thermal disturbance in a room, with a temperature variation of only a few degrees from the ambient temperature. In order to visualize these weak thermal gradients, the geometry of the BOS system was optimized for maximum schlieren sensitivity. The background was approximately 10.7 m (35.1 ft) from the camera and 3.7 m (12.1 ft) from the subject. Note that the subject is slightly out of focus at this location and the detected pixel shifts within the human thermal plume are near the detection limit of 0.1 pixels.
[FIGURE 8 OMITTED]
[FIGURE 9 OMITTED]
A cough was also visualized using the BOS technique, as shown in Figure 10. The cough was initially found to be too weak to be visualized with BOS given the present geometry and sensitivity limits. To improve the schlieren visibility of the cough, it was "seeded" by having the subject pre-breathe some helium gas. The use of helium here does not significantly change the flow characteristics of the cough, but it improves the schlieren visibility due to the large refractive index difference between helium and air. In general, this approach of "seeding" an HVAC flow with a thermal or species difference can be useful in some applications to improve BOS visibility as long as the flowfield is not materially altered.
[FIGURE 10 OMITTED]
The BOS technique allows the visualization of a range of HVAC scenarios where traditional schlieren optics cannot be practically implemented. It is easily portable, durable, and inexpensive; it can image large fields-of-view without mirrors or other large optical elements; and it can be applied in close quarters or large interior spaces for the visualization of HVAC flowfields.
Effective BOS image capturing, processing, and post-processing are required to produce high-quality schlieren-like visualizations of HVAC flowfields. The image capture process can be done with a consumer-grade digital SLR camera that is rigidly tripod-mounted, level with the main flowfield, and focused upon a suitable background with the schlieren object (airflow) of interest halfway between. Digital processing of BOS images can be enhanced by first separately averaging several flowfield and several tare images to reduce turbulent "noise" and camera-motion effects. A BOS flow-on image can be processed relative to a tare image or to another flowfield image in order to reveal either the total flowfield or the time-rate-of-change of the refractive-index field. Image post-processing with a grayscale contour map or a color gradient map yields schlieren-like visualizations and enhanced visibility of flow features.
Only a few representative examples of HVAC airflows imaged by the BOS technique are shown here. Much broader applications are possible, including the diagnosis of airflow patterns in new buildings, checking the proper function of air-handling equipment, and the determination of effluent capture and containment in hoods for commercial kitchen ventilation (CKV). Diagnosis of hospital HVAC flows to avoid airborne disease spread is another potential application. For these applications, earlier large-scale fixed schlieren systems of the lens-and-grid type (Settles 1997,2001) are superceded by the simpler BOS approach.
Only qualitative visualizations have been shown here, but quantitative BOS temperature-field measurements are also possible in certain circumstances. For examples of this, see Hargather and Settles (2011).
Anderson, R., V. Hassani, A. Kirkpatrick, K. Knappmiller, and D. Hittle. 1991. Visualizing the airflow from cold air ceiling jets. ASHRAE Journal 33(5):30-5.
Atcheson, B., W. Heidrich, and I. Ihrke. 2009. An evaluation of optical flow algorithms for background oriented schlieren imaging. Experiments in Fluids 46(3):467-76.
Cook, R.L., and T. DeRose. 2005. Wavelet noise. ACM Transactions on Graphics 24(3):803-11.
Correlated Solutions, Inc. 2009. Vic-2D user manual. http://www.correlatedsolutions.com/.
Craven, B.A., and G.S. Settles. 2006. A computational and experimental investigation of the human thermal plume. Journal of Fluids Engineering-Transactions of the ASME 128(6):1251-58.
Dalziel, S.B., G.O. Hughes, and B.R. Sutherland. 1998. Synthetic schlieren. Proceedings of the 8th International Symposium on Flow Visualization, September 1-4, 1998, Sorrento, Italy. G.M. Carlomagno, ed. Paper no. 62.
Dalziel, S.B., G.O. Hughes, and B.R. Sutherland. 2000. Whole field density measurements by 'synthetic schlieren'. Experiments in Fluids 28(4):322-35.
Dalziel, S.B., M. Carr, J.K. Sveen, and P.A. Davies. 2007. Simultaneous synthetic schlieren and PIV measurements for internal solitary waves. Measurement Science and Technology 18(3):533-47.
Elsinga, G.E., B.W.V. Oudheusden, F. Scarano, and D.W. Watt. 2004. Assessment and application of quantitative schlieren methods: Calibrated color schlieren and background oriented schlieren. Experiments in Fluids 36(2):309-25.
Goldhahn, E., and J. Seume. 2007. The background oriented schlieren technique: Sensitivity, accuracy, resolution, and application to a three-dimensional density field. Experiments in Fluids 43:241-9.
Hargather, M.J., and G.S. Settles. 2010. Natural-background-oriented schlieren. Experiments in Fluids 48(1):59-68.
Hargather, M.J., and G.S. Settles. 2011. A comparison of three quantitative schlieren techniques. Optics and Lasers in Engineering, Special issue on flow visualization DOI: 10.1016/j.optlaseng.2011.05.012.
Iffa, E.D., A.R.A. Aziz, and A.S. Malik. 2010. Concentration measurement of injected gaseous fuel using quantitative schlieren and optical tomography. Journal of the European Optical Society--Rapid Publications 5:10029.
Meier, G.E.A. 1999. Background Schlieren Method. Deutsche Patentanmeldung Report DE 199 42 856 A1, Gottingen, Germany. [in German]
Richard, H., and M. Raffel. 2001. Principle and applications of the background oriented schlieren (BOS) method. Measurement Science andTechnology12(9):1576-85.
Sandberg, M. 2007. Whole-field measuring methods in ventilated rooms. HVAC&R Research 13(6):951-70.
Schardin, H. 1942. Schlieren methods and their applications. Ergebnisse der Exakten Naturwissenschaften 20:303-439. [In German]
Siebert, T., T. Becker, K. Spiltthof, I. Neumann, and R. Krupka. 2007. High-speed digital image correlation: Error estimations and applications. Optical Engineering 46(5):1 7.
Settles, G.S. 1997. Visualizing full-scale ventilation airflows. ASHRAE Journal 39(7):19-26.
Settles, G.S. 2001. Schlieren and Shadowgraph Techniques. Berlin: Springer-Verlag.
Venkatakrishnan, L., and G.E.A. Meier. 2004. Density measurements using the background oriented schlieren technique. Experiments in Fluids 37(2):237-47.
Michael J. Hargather (1), * and Gary S. Settles (2)
(1) Gas Dynamics Laboratory, Department of Mechanical and Nuclear Engineering, Pennsylvania State University, 208 Reber Building, University Park, PA 16802, USA
(2) Gas Dynamics Laboratory, Pennsylvania State University, 301D Reber Building, University Park, PA 16802, USA
* Corresponding author e-mail: email@example.com
Received December 1, 2010; accepted April 21, 2011
Michael J. Hargather, PhD, is Research Associate. Gary S. Settles, PhD, ASHRAE Member, is Distinguished Professor of Mechanical Engineering and Director of the Gas Dynamics Laboratory.
|Printer friendly Cite/link Email Feedback|
|Author:||Hargather, Michael J.; Settles, Gary S.|
|Publication:||HVAC & R Research|
|Date:||Sep 1, 2011|
|Previous Article:||Methodology for determining the optimal operating strategies for a chilled-water-storage system--part II: project application.|
|Next Article:||Numerical predictions of the broadband turbulent noise for axial flow cooling fans.|