KR101207754B1 - Measurements in transient mixing concentrations of two fuel oils using a quantitative flow visualization technique - Google Patents

Abstract

PURPOSE: A method for measuring the transient mixing concentration of two fuel oils using a quantitative flow visualization technique is provided to observe color variations of concentration generated in a process mixing different kinds of fuel oil, thereby providing a efficient quantitative measuring method with respect to a visual variation. CONSTITUTION: A method for measuring the transient mixing concentration of two fuel oils using a quantitative flow visualization technique is as follows. First oil and second oil(2) are uniformly mixed depending on concentration, thereby obtaining color information depending on coordinate positions of camera images caused by the concentration. Image information being observed by rapidly mixing the first and second oil is captured depending on time. The color information and image information are execution-processed depending on a neural network application method, thereby measuring the mix concentration of the first and second oil depending on the time and positions.

Description

Measurements in transient mixing concentrations of two fuel oils using a quantitative flow visualization technique

The present invention utilizes a neural network to quantify the relationship between color and concentration as a method for measuring the mixing concentration process by observing the change of color in the mixing process of heterogeneous fuel oils.

Stirring is a mixture of two materials having different physical and chemical properties, and the stirrer used for stirring is selected in consideration of the purpose and purpose of the stirring, the physical properties of the stirring liquid, the shape, dimensions, and rotation speed of the impeller. Such stirrers have been greatly used in the biochemical and environmental fields due to the recent industrial development. The mixed flow in the stirrer is very complicated and the mixed flow analysis and prediction is not easy.

However, if the concentration change with time of mixing flow can be known, the mixing time can be shortened and more perfect mixing is possible, so the invention has been made for the mixing concentration in the agitator.

On the other hand, as an experimental invention to understand the mixing phenomenon in the stirrer, the structure and trajectory of the vortex tip end vortex of the rush-ton turbine type using the photographic technique were analyzed but based on the qualitative evaluation.

Although there was an invention in which the turbulent flow characteristics were measured quantitatively with a laser Doppler flowmeter for the flow in the same rushton turbine stirrer, it was limited to the time average value due to the measurement method.

Although the flow characteristics obtained from these results have been applied to the stirrer design, it is necessary to invent the mechanism by which the passive scalar amount is mixed during turbulent mixing. Houcine et al. Applied PLIF (planar laser induced fluorescence) technique to measure the concentration field of a continuously stirred flow field, and showed that the mixing efficiency varies depending on the impeller type. Guillard et al. Obtained instantaneous concentration fields in batch stirred tanks using PLIF technique. The fluorescent concentrations were injected into bulk flow and impeller flow fields, and the instantaneous concentration fields obtained were statistically processed to present the macroscopic concentration field structure.

Jeong and Kim summarized the unsteady mixing characteristics of the entire concentration field in the rushton turbine stirrer using LIF technique in detail. Until now, most of the measurement analysis of the concentration field in the stirrer has been based on the LIF technique. The LIF technique forms a laser light source in a two-dimensional cross section and irradiates the concentration field so that the same concentration is achieved at the same measurement position. Nevertheless, the uncertainty in the measurement results is high due to the intensity variance of the brightness of the laser light source. In addition, most of the inventions related to concentration field measurement using the LIF technique were performed by injecting a Rhodamin B solution at an arbitrary position in a stirrer tank and measuring the diffusion process of the Rhodamin B solution. Is hard to see.

In the present invention, to build a visualization technique using an LCD monitor to ensure a uniform light source to compensate for the intensity variation (variance) caused by the use of a laser light source, the concentration field generated in the process of mixing different fuel oils It is applied to quantitative identification of temporal change.

In the present invention, to build a visualization technique using an LCD monitor to ensure a uniform light source to compensate for the intensity variation (variance) caused by the use of a laser light source, the concentration field generated in the process of mixing different fuel oils It is applied to quantitative identification of temporal change.

In the method for measuring heterogeneous fuel oil transient mixing concentration distribution using quantitative flow visualization technology for measuring the change in concentration of two oils by mixing, the first oil 10 and the second oil 20 are Acquiring color information (R, G, B, H, S, I) by coordinate position (pixel unit) of the camera image according to each density by uniformly mixing each density (step 1 ); And capturing image information observed by rapidly mixing the first oil and the second oil with time ( step 2 ); And measuring the color and image information of steps 1 and 2 according to the neural network application method to measure the mixed concentration of the first and second oils by location and time ( step 3 ). do.

Detailed functional description of the present invention will be described in the specific content for carrying out the invention.

In the present invention, by observing the color change of the concentration field generated in the process of mixing different fuel oil, it proposes an effective quantitative measuring method for the time change.

1 is a cross-sectional view of an agitator and a cylindrical tank
2 is a perspective view of the entire experimental apparatus
3 shows images for color-concentration correction,
4 is a schematic diagram of neural network application
5 shows the sigmoid function used
6 is an image captured after the septum rupture
7 is a captured image one second after the septum rupture
8 is a graph of concentration change on the Y axis
9 is a captured image one second after the septum rupture
10 is a graph of concentration change along the X-axis horizontal line after 1 second of diaphragm rupture.
FIG. 11 is a graph of concentration distribution along a timeline at 84 mm points.

Hereinafter, preferred examples and experimental examples of the present invention will be described in detail.

However, the scope of the present invention is not limited thereto, and it is made clear in advance that any invention is equivalent to the invention described in the claims.

1. Experiment apparatus

1 shows an agitator for mixing heterogeneous fuel oils. The glass tank is filled with JetA1 oil and the tank top is filled with DMF oil. The size and pressure dimensions of the experimental apparatus can be variously modified, and it is natural that the measuring method of the present invention can be sufficiently applied even when mixing other than the oils.

In order to promote mixing, a stirring impeller is installed on the central axis of the tank as shown in FIG. 1. At the center of the tank bottom, a round rod is installed to activate stirring. The distance between this round rod and the impeller is 4mm, and the diameter of the impeller used in the experiment is 120mm, and it is HE-3 type.

The DMF oil storage tank installed at the top of the tank is designed to apply nitrogen gas at 1.9 bar pressure. A thin rupture disc (around 500 micrometers thick) is installed between the DMF oil and JetA1 oil.

This thin diaphragm is to burst instantly by the pressure of nitrogen gas applied from the top of the DMF oil storage tank.

The transient begins with the momentary opening of the solenoid valve installed at the outlet of the nitrogen gas cylinder.

2 shows the overall experimental apparatus. To visualize the overmixed state of DMF oil with JetA1 oil, DMF oil was stained with trace amounts (4 grams) of Rhodamin B solution. Colored DMF oils are dark pink in colour.

As a light source for visualizing the mixed state of deep red DMF oil, an LCD monitor is used at a distance of about 150 mm from the outermost surface of the cylinder, and is used to record an image of the mixed flow field visualized by the light source. Install on the other side of the camera.

In order to eliminate the influence of light other than the LCD monitor, the experiment is preferably performed in a dark room. The frame rate of the camcorder camera used to record the color change according to the DMF significance concentration was 60 fps (frame per second), and the color information was quantified on the host computer using the obtained camcorder image.

Table 1 shows the physical properties of the two fuel oils used in the mixing experiment. On the other hand, in the present invention, unlike the quantification of the diffusion process on the cross-section using a two-dimensional cross-sectional light source of the laser, as shown in Figure 2, the entire depth of field direction visible by the light source of the LCD monitor Focus on quantifying the visualization results. This is because the mixing flow field can be viewed as axisymmetric because the mixing process is a steady state flow field in which the stirrer is already rotating at 700 rpm in the cylindrical tank.

Fuel oil name

))Density (

)) Viscosity (cp)
Jet A1 808 1.313
DMF 944 0.92

2.Experimental Method

The experiment proceeds in two main ways.

First, a clinic to concentration calibration was performed to quantify the color change according to the mixed concentration of DMF oil, followed by an overmixing experiment by rupture of the diaphragm.

In the calibration experiment, the color change of the DMF oil was stored as the camcorder image described above, and the image was separated into three primary colors (R: red, G: green, and B: blue). The video value has 8 bits of color information per pixel, so it has a value between 0 and 256. Using these values, the values of hue (H: hue), saturation (S: saturation), and brightness (I: intensity) were obtained, respectively.

The LCD monitor's light used for visualization is the color information of each tank location (R, G, B, H, S, I) even though the concentration is the same because the reflected light is not uniform in the upper and lower parts of the stirrer tank and from side to side. ), The color-to-concentration calibration is nonlinear. Therefore, in the present invention, a neural network having excellent performance in nonlinear mapping is used for color-concentration quantification.

(1) calibration experiment

를 가득 채우고 DMF를 0 vol%(부피농도)부터 2vol%씩 증가시키며 50 vol%까지 변화시켜 가면서 캠코드 영상을 기록하였으며, 54 vol%부터는 4vol%씩 증가시켜가면서 100vol%까지의 캠코드 영상을 기록하였다. 이러한 방식으로 총 30개의 색-농도 관계를 나타내는 영상을 얻었다. In the calibration experiment, the relationship between the change of color according to the concentration of DMF is obtained. JetA1 oil 24.9 inside the glass cylinder

The camcode image was recorded by changing the DMF from 0 vol% (volume concentration) to 2 vol% and increasing to 50 vol%, and the camcode image was recorded up to 100 vol% from 54 vol% to 4 vol%. . In this way, a total of 30 color-density images were obtained.

3 shows a calibration experiment image for obtaining a color-concentration relationship. The impeller was stirred at 700 rpm for a long time (30 seconds or more) so that DMF oil was uniformly mixed with JetA1 oil in all regions in the tank. The images for each concentration were used as calibration images using an average of 200 images for about 5 seconds from the time when the mixed state in the tank was judged to be uniform. The criterion of uniformity was considered to be uniform when no difference was found by comparing the color information of each position in the tank. Color information (R, G, B, H, S, I) of 200 average images in total can be used as input data of color-concentration relationship by neural network method as follows.

(2) color-concentration quantification using neural networks

As described above, even though the same DMF oil concentration may be different in color information (R, G, B, H, S, I) for each position of the tank, the relationship between the color and the concentration has nonlinearity. In order to quantitatively evaluate the color-concentration relationship with nonlinearity, it is preferable to use a neural network which is widely used for nonlinear mapping.

4 shows a neural network in the present invention. The neural network is composed of input layer, middle layer, and output layer. A total of eight pieces of information (R, G, B, H, S, I and positions X and Y in the image) are input, and the middle layer is 20, and the output layer is It consisted of one. That is, the input layer is composed of 8 neurons, the middle layer is 20 neurons, and the output layer is composed of 1 neuron. The following describes the calculation process of the neural network applied to obtain the relationship between color and density. Each neuron can be represented as a model that accepts signals from other neurons. The relation can be expressed as

는 결합계수, bias는 각 뉴런에 주어지는 오차값을 나타낸다. 각 뉴런(neuron)의 출력값에 대하여 연결망값(net value)에 신경망의 임계 함수적인 동작을 잘 표현할 뿐만 아니라 미분가능하고 수학적으로 편리한 특성을 가지고 있는 것으로 알려져 있는 수학식(2)와 같은 시그모이드함수(Sigmoid function)⁹⁾를 사용하여 구하였다.
here,

Where is the coupling coefficient and bias is the error value given to each neuron. A sigmoid such as Equation (2), which is known to not only express the critical functional operation of the neural network at the net value for the output value of each neuron, but also has differentiable and mathematically convenient characteristics. Sigmoid function ⁹ was used.

Here, a and b are arbitrary constants. In the present invention, a function having a bias (0.166666) in the output value is used as shown in FIG. 5 to improve the computational power in the saturation period of learning.

The neural network acquires the desired knowledge by repetitive learning. In the present invention, the error back-propagation algorithm is a method of adjusting the connection weight and bias until the learning error is reduced to a desired level. algorithm: The following BP method) was used.

, i=1~8)와 결합하중(v) 및 바이어스(

)의 합을 나타낸 것이며, 이를 시그모이드 함수로 전이하여 수학식(4)와 같이 중간층에서의 출력데이터가 구해진다. Equation (3) represents the input data in the input layer (

, i = 1 to 8) and combined load (v) and bias (

) Is converted into a sigmoid function, and the output data in the intermediate layer is obtained as shown in Equation (4).

)의 합으로 되는 수학식 (5)가 재차 시그모이드 함수로 전이되어 수학식(6)으로 표현된 출력층에서의 출력데이터(z)가 구해지게 된다.
Next, the output (y) and the combined load (w) and bias (

Equation (5), which is the sum of N, is transferred to the sigmoid function again to obtain the output data z in the output layer represented by Equation (6).

) 및 바이어스계수(

)를 수학식(8)과 같이 갱신하여 반복계산을 수행하면서 오차가 가장 작은 최적화된 결합하중과 바이어스가 구해지면 학습을 종료시키게 된다. Finally, the combined load between the input layer and the middle layer, the middle layer and the output layer so that the error between the output data that can be expressed as shown in Equation (7) and the desired output data (t: teaching) is reduced (

) And bias coefficient (

) Is updated as shown in Equation (8), and the learning is terminated when the optimized combined load and bias having the smallest error are found.

(3) transient mixing experiment

를 가득 담은 상태에서, 도 1에 나와 있는 바와 같이 교반용 임펠러를 700rpm으로 회전시켰다. For the transient mixing experiments, the DMF oil from the moment of rupture of the thin rupture disc was imaged with a Camcode camera on the JetA1 oil. JetA1 oil 22.2 inside the glass cylinder

In the state filled with, the stirring impeller was rotated at 700 rpm as shown in FIG.

(14.5vol%의 농도)가 들어 있으며 격막이 파열되는 순간부터 JetA1유와 혼합되기 시작한다. During the rotation of the impeller, the diaphragm was ruptured by pressurizing the upper tank containing DMF oil with 1.9 bar of nitrogen gas. The upper tank has DMF oil 3.75

(Concentration of 14.5 vol%) and begin to mix with JetA1 oil from the moment the diaphragm ruptures.

In the experiment of the present invention, the rupture due to the pressurization was performed twice due to the characteristics of the diaphragm, and the time difference between the initial partial rupture (primary rupture) and the later all rupture (secondary rupture) was 3.34 s.

FIG. 6 shows a time-continuous image of a transient mixing state in which DMF oil is mixed with JetA1 oil from the moment when the diaphragm ruptures. For this image, the concentration change in real time was obtained using the color-density calibration data obtained in the calibration experiment.

6. Experimental Results

(14.5vol%의 농도에 상당)의 DMF유가 교반탱크로 유입되기 된 시점으로부터 1초 후의 영상이며, 이 영상에 대하여 앞에서 기술한 신경망을 이용한 색-농도 교정데이터를 이용하여 얻어낸 농도 분석 결과를 도 8에 나타내었다. 가로축은 DMF유의 농도를 나타내고 임펠러의 하단부 위치를 Y=0mm로 하여 교반기 탱크바닥까지의 깊이 (mm)를 나타낸다.
Fig. 7 shows the 3.75 rupture of the rupture disc, which was in the upper tank.

This image is 1 second after the DMF oil (corresponding to the concentration of 14.5 vol%) flows into the stirring tank. The results of the concentration analysis obtained using the color-concentration correction data using the neural network described above are shown. 8 is shown. The horizontal axis represents the concentration of the DMF oil and the depth (mm) to the bottom of the stirrer tank with the lower position of the impeller as Y = 0 mm.

, 76

, 121

위치에서의 수직선상의 농도변화를 나타내는데 격막파열 시점에서 1초 후의 농도 분포를 나타낸다. 그림으로부터 교반탱크의 상단부, 즉 DMF가 들어 있었던 상부탱크의 바로 아래에는 X위치에 상관없이 약 5vol%이상의 DMF유가 존재하고 있음을 알 수 있다. 교반기탱크 아래쪽에는 DMF유의 농도가 2vol% 미만으로 낮음을 알 수 있다.
As shown in FIG. 8, the graphs of FIG. 8 are respectively 32 in radial positions from the center of the agitator tank.

, 76

, 121

It shows the concentration change on the vertical line at the position, which shows the concentration distribution 1 second after the time of diaphragm rupture. From the figure, it can be seen that there is more than about 5 vol% of DMF oil at the upper end of the stirring tank, that is, directly below the upper tank containing the DMF, regardless of the X position. Below the stirrer tank it can be seen that the concentration of DMF oil is low, less than 2vol%.

In particular, near the vertical axis of rotation (X = -32mm) and Y = -80mm, the concentration of DMF oil is suddenly high and then decreases, which means that the DMF oil introduced by the impeller is 1 second after the diaphragm rupture. Y = -80mm). Looking at the concentration change on three positions (X-axis), the closer the tank center axis, the greater the change in concentration. This means that the change in concentration is large around the tank center axis due to the influence of the impeller, and the change in concentration is slower as it moves away from the center of rotation.

, 84

, 153

의 세 위치에서의 반경방향으로의 농도 변화를 나타낸다. 탱크 상부에 해당되는 (X=-12

)에서의 농도는 약 5vol%로 수평방향으로 균일하게 분포되었으며 하부보다 농도가 높은 분포를 보였다.
The concentration change on the horizontal line in FIG. 7 was examined. FIG. 9 shows an image at the same time (1 second after rupture) as shown in FIG. 7, and the result of analyzing the concentration distribution using the image is shown in FIG. 10. 10 is based on the lower position Y = 0mm of the impeller 12

, 84

, 153

The concentration change in the radial direction at three positions is shown. (X = -12 corresponding to the top of the tank

) Concentration was about 5 vol% uniformly distributed in the horizontal direction, the concentration was higher than the lower portion.

)의 위치에서의 농도는 1vol% 전후로 매우 낮은 농도분포를 보였다.
As it goes down, the concentration decreases and the left and right are asymmetrically in the Y = -84mm position, which can be seen from the horizontal point where the color change of FIG. 9 is large. It can be interpreted that the lowered DMF oil passes through this point. Lower part of the tank (Y = -153

At the position of), the concentration distribution was very low around 1 vol%.

에서의 수평선상의 농도변화는 반경방향으로 약 5vol%정도로 균일하게 나타남을 알 수 있는데, 이는 원래 교반기 탱크내부에 들어 있었던 JetA1유에 유입되는 DMF유(3.75

)가 14.5vol%에 해당된다는 점을 고려하면 약 DMF유의 1/3(=5/14.5)의 즉 30%가 이 지역에 균일하게 혼합되어 있음을 의미한다.
Y = 12 close to the impeller

It can be seen that the concentration change on the horizontal line in the surface is uniformly about 5 vol% in the radial direction, which is DMF oil (3.75)

), Which is equivalent to 14.5 vol%, means that 1/3 of the DMF oil (= 5 / 14.5), or 30%, is uniformly mixed in this region.

의 위치에서의 시간 변화(1, 4, 7, 10, 13초)에 따른 농도본포 변화를 나타낸다. 11 is the middle point of the tank, Y = 84

It shows the change of concentration main with the change of time (1, 4, 7, 10, 13 seconds) at the position of.

As time passes from the time when the diaphragm ruptures, the concentration value increases and after 10 seconds, the concentration of about 13 vol% is uniformly distributed in the horizontal direction, which is close to 14.5 vol% of the DMF oil originally introduced. 90% of DMF oil is mixed in JetA1. The concentration distributions at 10 and 13 seconds show that they are uniform in the radial direction, which means that uniform mixing occurs over time. In particular, the 13 second case shows a concentration distribution around 14.5 vol% corresponding to the amount of the original DMF oil, indicating that the minimum time for the complete mixing of the DMF and JetA1 oils is 13 seconds.

The concentration distribution of DMF oil at 4 seconds shows a uniform distribution in the horizontal direction. Considering that the distribution is at the middle point of the tank, the DMF oil is evenly mixed with JetA1 oil from about 4 seconds. We can see that it takes about 9 seconds to start and then 13 seconds, the point of complete mixing. This suggests that compulsory mixing is necessary with an impeller-like mechanism in order to mix DMF oil and JetA1 oil, which means that the mixing time can be shortened according to the mixing performance of the impeller.

A concentration field measurement technique was established using visualization techniques to quantitatively evaluate the transient mixing state of heterogeneous liquids. The following results were obtained in the process of applying it to the transient mixing concentration field measurement between DMF oil and JetA1 oil. Neural network-based color-concentration correction method using color information (R, G, B, H, S, I) of DMF oil and position (X, Y) in tank, and visible light source according to experimental environment even at the same concentration field A density field measurement technique was constructed to consider the color effects of

As a result of measuring the change in concentration distribution of DMF oil in the tank with the change of time from the time of rupture of the diaphragm, it was found that the time until the DMF oil was completely mixed with JetA1 oil was about 13 seconds.

In addition, the time when DMF oil started to be uniformly mixed with JetA1 oil in the tank was about 4 seconds from the time when the diaphragm was ruptured, and after that, about 9 seconds were required for complete mixing. This means that the mixing characteristics of the impeller used to improve the mixing performance should be considered.

Claims (2)

In the method of measuring the concentration change of the two oils are mixed by color, but using a neural network, to measure the heterogeneous fuel oil excessive mixed concentration distribution, the first oil (10) and the second oil (20) Acquiring color information (R, G, B, H, S, I) by coordinate position (pixel unit) of the camera image according to each density by uniformly mixing each density (step 1 ); Capturing the image information observed by rapidly mixing the first oil and the second oil according to time ( step 2 ); and calculating and processing the color information and the image information of the first and second steps according to a neural network application method. Including the step 2 ( step 3 ) of measuring the mixing concentration of the oil by location and time,
In the first step , in order to obtain a change in color according to the mixing concentration of the first oil and the second oil, a plurality of images showing the color and the concentration relationship are captured, but uniform in all regions of the tank containing the mixed oil. as to ensure that the oil is mixed, in the respective concentrations image uses an image the average for a predetermined time a plurality of images from the time it is determined that the mixing conditions tank to the uniform state in the calibration image for the third step, the observed image A neural network is used for the quantitative evaluation of the color-concentration relationship, wherein the neural network includes an input layer, a plurality of intermediate layers, and an output layer composed of neurons. , B, H, S, I and positions X, Y) in the image are input, and neurons in the neural network receive signals from other neurons and are represented by Equation 1,

(One) (

Is the coupling coefficient and bias is the error value given to each neuron)
Sigmoid function (Equation 2) is used for the output value of each neuron, but a function that biases the output value is used to improve the computational power in the saturation interval of learning.

(2)
Input data in the input layer (

, i = 1 to 8) and combined load (v) and bias (

After the summation of Eq. (Eq. 3), it is transferred to the sigmoid function to obtain the output data in the intermediate layer (Eq. 4),

Output (y) and coupling load (w) and bias (

The resulting value of Equation 5, which is the sum of N, is transferred to the sigmoid function again, and the output data z is obtained from the output layer represented by Equation 6, and the output data and the desired output represented by Equation (7) Combined load between input and middle layers, middle and output layers to reduce error from data (

) And bias coefficient (

) Is updated by the method of Equation (8), and the repetition calculation is performed, but the repetition is terminated when the optimized coupling load and bias having the smallest error are obtained. The heterogeneous fuel oil transient using quantitative flow visualization technology Mixed concentration distribution measurement method.
The method according to claim 1,
Experimental apparatus for mixing the 1,2 oil,
The first oil is formed to fall to the upper portion of the second oil, is separated by the diaphragm so as to be rapidly mixed, and the diaphragm is instantaneously ruptured immediately after the first oil receives a predetermined pressure, 2 color the oil to visualize the transient mixing of the oil,
An LCD monitor is used as a light source for visualizing a transient mixed state, and the LCD monitor is installed at a predetermined distance from the rear surface of the second oil tank, and an image of the mixed flow field visualized by the LCD monitor as a light source is recorded. LCD monitor is installed on the opposite side of the camera.
Characterized by quantifying the visualization result over the entire depth of field direction by using the LCD monitor as a light source.
A method for measuring the heterogeneous concentration distribution of heterogeneous fuel oils using quantitative flow visualization.

Images