An introduction to CSPSV - CRH 研究小组

1 Background of indoor airflow measurement

Indoor ventilation and air distribution systems play a vital part in heating, ventilation, and air-conditioning systems, which can not only affect the energy efficiency in buildings but also directly influence indoor air quality. However, it is not an easy task to measure the indoor airflow accurately, because it is large-scale, three-dimensional (3D), turbulent, and unsteady.

Modern global-wise technologies for indoor airflow measurements include particle image velocimetry (PIV), particle streak velocimetry (PSV), and particle tracking velocimetry (PTV). PIV is the most successful commercial zonal optical velocimetry method, and it has been used for indoor airflow measurements. Although PIV can achieve high spatial resolutions by micrometer-scale seeding particles, it is only suitable for small zone tests, not room-scale 3D airflow measurements, because of the limitations of its powerful illumination system.

Using helium-filled soap bubbles (HFSBs) as seeding particles, PSV and PTV can provide a wider measurement space. PSV records the bubble’s trajectory with a long exposure time and calculates the velocity by the length of each streak. However, without information of the flow direction, the scheme could result in 3D pair mismatch. PTV takes pictures at a very high speed, normally up to 100 Hz, and tracks the spots frame by frame for the entire trajectory. However, the large amount of data, counted in gigabyte/s, limits each sample time to several seconds. As the frame-rate and resolution are trade-off parameters for cameras, the low resolution of high-speed cameras normally limits the PTV test zone to within 1 m³.

Point wise instruments like HWA is the most widely used sensor for air velocity test with known drawbacks. For example, the measuring sensor must be placed in the flow field, rendering itself an obstacle for airflow, and the released heat limits its use for low-speed airflow measurement. Further, the lack of flow direction information leads to some missing flow characteristics. More importantly, the point-by-point HWA measuring approach is labor intensive and time consuming. UA and LDA are both point-wise solutions that can deliver precise three-component velocity information at one point or on a grid using a mobile system. However, it is difficult to guarantee that the flow field is stable during the testing process, and the test may be time-consuming, which make them only suitable for a stationary laboratory test.

The indoor air flow can be affected by many factors such as partitions, heat sources, movement of objects and radiation, which make it difficult or even impossible to be simulated in a small-scale model. In addition, the velocity of indoor air flow is in a wide range; therefore, the normal intrusive measurement method cannot be sufficiently accurate for both normal and low speed. Therefore, a room-size, high-dynamic-range and non-intrusive flow measurement system is crucial for the laboratory test of thermal comfort, pollution transport and the field test of diffuser design and air flow pattern design.

2 Introduction to CSPSV

2.1 Principle of CSPSV

The basic idea of CSPSV is to capture more information of the seeding bubbles’ trajectory by a novel imaging method for image-based velocimetry with a colour sequence illumination system (CSIS). Synchronized with cameras, the CSIS changes the bubbles’ appearance according to the time sequence and helps to record the bubbles’ entire trajectories during exposure time with a start spot marked in red, a middle spot marked in green, and an end spot marked in blue, all of which contain substantial time sequence information, as shown Fig. 1. Such colour-coded airflow movement information can be extracted by a digital image processing algorithm and bring in advantages in two aspects: firstly, with the known movement direction and displacement of individual seeding particle, the extrapolation tracking is not required; secondly, three colour spots with strict time sequence can further help the reconstruction of the 3D trajectory with improved streak pair match algorithm.

Fig. 1 Operating model of high performance CSPSV

The whole operation process is illustrated below. A commercially available bubble generator seeds the test space with helium-filled soap bubbles (HFSB) and visualizes the airflow. Synchronized with the CSIS, stereo colour cameras operate at a long exposure time to record the bubbles’ path and provide raw images for further processing. As shown Fig. 2, the streaks have different length due to different speed and each streak has a red start spot, a green middle spot and a blue end spot. The time sequence information of the image is extracted with an image processing algorithm and represented with two linking vectors on the raw image. Subsequently, a rectification-based 3D reconstruction algorithm with the calibration data calculates the 3D position of the bubbles’ path in a 3D space within a certain time period. Thus, the 3D velocity is captured.

The verification of CSPSV with a regular principal instrument, e.g., the HWA, and known moving objects have confirmed its accuracy. For moving objects verification, which is adopted by most visual based velocimetry, the relative error in most cases was less than 5% with the magnitude velocity ranging from 0.12~2.45 m/s. The test of the vortex flow has proven its capability at measuring complex flows and real-site large-room airflows.

Fig. 2 Flow chart of CSPSV

2.2 Quad-view CSPSV

The primary drawback of CSPSV with two cameras or a stereo CSPSV is the blind zone as shown Fig. 3. To illuminate the whole room with little effect, a light source is placed both on the top of the ceiling and under the transparent glass floor, projecting toward each other. This type of setup causes two bright zones on the ceiling and the floor marked in purple. When looking out of the chamber, these regions are over-exposed and can hardly record any bubbles’ trajectory, thus causing a blind zone for each camera set. The blind zone varies with the direction of the camera set. As in Fig. 3, the blind zone of upper camera set is marked blue; that of the lower camera set, orange; and the zone that can be captured by both camera sets, green. The view angles are presented with a dashed line that can facilitate the understanding of the effects of area overexposure caused by a light source.

As mentioned above, the measurement zone of a stereo CSPSV is always affected by over-exposed regions. While the flow could be crucial near light sources, one method to obtain a wider measurement space is to use more cameras to capture the movement from different angles and merge the measurement data together. In this way, a new quad-view CSPSV is proposed with four cameras that are set into two groups and operated synchronously. The upper cameras are primarily used to capture the air movements near the ceiling while the lower cameras primarily target the air movements near the floor. The airflow measurement zone sizes is 4m×2.5 m×0.8 m in the middle of the room.

Fig. 3 Comparison for blind zone captured by two pair of cameras

With these settings, both camera sets can capture the airflow in the middle of the room. If the results are simply added together, some vectors will become duplicated. To solve this problem, a data merge algorithm is designed as in Fig. 4. The streaks information from each camera set is reconstructed separately. Subsequently, the distance between each pair of 3D vectors in 3D spaces was calculated, which consists of two vectors from the same frame but in different sets, as in equation 1.

(1)

where X, Y, Z is the position of the vector in a 3D space; the subscripts s and e represent the start and end points of the vector; subscripts 1 and 2 are for the vectors from the upper camera set and lower camera set, respectively.

When the distance between two streaks are extremely close, the streak with a larger epipolar error is rejected. With this algorithm, the vectors captured by two camera sets can be merged together for a wider measurement space.

Fig. 4 Results algorithm from four cameras

The new quad-view CSPSV combined the measurement results of two normal stereo CSPSVs. Therefore, the measurement region is expanded, blind zone is minimized, and high accuracy of each stereo CSPSV is preserved. Compared with regular stereo PSV, CSPSV system modified the information gathering method by adding a CSIS. By this way, CSPSV can overcome the streak mismatching problem of regular stereo PSV, which is also the main reason for the large errors, and achieve a large measurement zone of several cubic meter with just two cameras. Although PTV with at least three cameras can provide higher accuracy compared with CSPSV, the accuracy of CSPSV is comparable and good enough for the study of indoor airflow pattern. Since there are a lot of buildings with even larger space which cannot be covered by a pair of cameras, this system is designed to accommodate more cameras to make up multiple-view CSPSV and may further conduct 3D3C measurements for flow pattern in large space like stadium.

3 Application of CSPSV

The CSPSV system has been used in many experiments and provided high quality 3D3C air flow field measurement results. Some tests are presented here and even more detailed data can be found in research papers listed in “5 Reference”.

3.1 Ceiling Fan test

Experimental setup and typical image of fan test.

a) Setup of four cameras

b) Location of cameras and chamber

Camera setup and relation with the chamber

Vectors induced by ceiling fan

Accumulated colour streaks of 200 images in the measurement zone

Turbulence intensity distribution in the middle section of fan-induced airflow

C:\Users\CRH-CV\AppData\Local\Temp\Image.png

Vorticity w _zin the middle section of the fan-induced airflow

Vorticity w _z in the middle section of the fan-induced airflow

Velocity contour through fan’s center

Vectors on different horizontal planes

(a) 10 cm above the fan blades

(b) 5 cm below the fan blades

(d) 20 cm below the fan blades

(e) 30 cm below the fan blades

(f) 40 cm below the fan blades

(g) 50 cm below the fan blades

(h) 60 cm below the fan blades

(i) 70 cm below the fan blades

(j) 80 cm below the fan blades

(k) 90 cm below the fan blades

(l) 100 cm below the fan blades

(m) 120 cm below the fan blades

(n) 140 cm below the fan blades

(o) 160 cm below the fan blades

(p) 180 cm below the fan blades

Vector field at different heights

3.2 Isothermal jet test

Averaged results on the middle section

3.3 Exhalation track under displacement ventilation

Averaged results on the middle section:

3.4 Mix ventilation with side wall inlet and opposite outlet

Averaged results on the middle section:

3.5 Vortex test in wind tunnel

During the experiment, the main tunnel fan supplies 260 m³/h of air into the tunnel, while the circulation fan circulates 523 m³/h of air within the tunnel. Hence, there will be a large vortex within the test zone.

Averaged results on the middle section:

3.6 Portable CSPSV system and application in Ice cubic

System setup in Ice cubic for ice curling of winter Olympics 2022

Accumulated colour streaks of the air inlet

Velocity vectors of the air inlet

3.7 Split air conditioner

Accumulated airflow trajectories from the room air conditioner

Downward jet

Horizontal jet

3.8 Tracing of pollutants exhaled by the human body

Speed vectors

No separation Partition beside the patient Partition beside the doctor

4 Commercial application

4.1 TICA (The largest enterprise in the field of industrial air conditioning in China)

Based on TICA’s requirements for airflow laboratory, we have discussed and finalized the laboratory functions and airflow measurement scheme. The construction and implementation have been successfully completed. The CSPSV system has been deployed and utilized as the core system for measuring the air flow field over the long term.

4.2 Wuxi Freshair AQ Technology Co.,Ltd (The leading enterprise in the field of hospital air conditioning in China)

C:\Users\GPU-CV\AppData\Local\Temp\WeChat Files\a6830ccb61972d34f443ff56229549d.jpg

Based on the testing requirements for airflow organization in hospital settings according to Freshair AQ Technology and considering their existing laboratory conditions, we have discussed and determined the main scenarios in the laboratory along with corresponding airflow measurement schemes. By employing a side-mounted light source configuration, we were able to measure the airflow organization in various medical scenarios. The CSPSV system has been deployed and utilized as the core system for measuring the air flow field over the long term.

4.3 Daikin Industries (The global No.1 air conditioning manufacturer)

In our collaborative project with Daikin, We utilized the CSPSV system to conduct a series of actual airflow field measurements under various operating conditions for the new indoor terminals developed by Daikin. This provided direct data support for the use and further optimization of the products.

5 Published paper

Wang H, Li X, Shao X, Wang B, Lin Y. A colour-sequence enhanced particle streak velocimetry method for air flow measurement in a ventilated space. Build Environ 2017; 112: 77–87.

Wang H, Wang G, Li X. High-performance color sequence particle streak velocimetry for 3D airflow measurement. Appl Opt 2018; 57: 1518–1523.

Wang H, Zhang H, Hu X, Luo M, Wang G, Li X, Zhu Y. Measurement of airflow pattern induced by ceiling fan with quad-view colour sequence particle streak velocimetry. Build Environ 2019; 152: 122–134.

Wang H , Wang G , Li X .Implementation of demand-oriented ventilation with adjustable fan network[J].Indoor and Built Environment, 2020, 29(4):1420326X1989711.DOI:10.1177/1420326X19897114.

6 AWARD

7 Contact

Should you have any questions please contact with us:

Xianting Li

xtingli@tsinghua.edu.cn

Huan Wang

wanghuan4610@126.com


(a) 70RPM	(b) 140RPM

(c) 210RPM	(d) 280RPM

(e) 350RPM
Result vector field for downward blowing fan


(a) 140RPM	(b) 210RPM

(c) 280RPM	(d) 350RPM
Result vector field for upward blowing fan


(a) 70RPM	(b) 140RPM

(c) 210RPM	(d) 280RPM

(e) 350RPM
Velocity magnitude on the middle section for downward blowing fan


(a) 140RPM	(b) 210RPM

(c) 280RPM	(d) 350RPM
Velocity magnitude on the middle section for upward blowing fan

00	00
(a) 70RPM	(b) 140RPM
00	00
(c) 210RPM	(d) 280RPM
00
(e) 350RPM
Result vector field on the middle section for downward blowing fan