Method for detecting suspension effect of particles in stirring reactor

Abstract

The invention belongs to the technical field of chemical equipment monitoring and control, and particularly relates to a method for detecting the suspension effect of particles in a stirring reactor, which comprises the following steps: preparing fluorescent beads with different sizes; liquid is injected into the stirring reactor, all the fluorescent beads are put into the liquid, and the liquid is stirred through a stirrer; the stirring reactor is irradiated by an ultraviolet lamp, so that the fluorescent beads in the stirring reactor emit fluorescence; a real-time monitoring module is adopted to collect images of the small fluorescent balls in the stirring reactor; analyzing the fluorescent ball image to obtain a motion track, an instantaneous rate and a direction of the fluorescent ball; obtaining the suspension effect of the particles in the liquid according to the motion trail, the instantaneous rate and the direction of the small fluorescent balls; through the high-speed camera and the intelligent image processing module, the particle suspension state can be captured in real time, and the detection efficiency is remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of chemical equipment monitoring and control, and particularly relates to a method for detecting the particle suspension effect in a stirred reactor. BACKGROUND

[0002] The stirred reactor is an important device in the fields of chemical industry, pharmaceutical industry, food industry and environmental protection, and its core function is to realize uniform mixing of materials, reaction promotion and heat and mass transfer. The mixing performance is one of the key indicators for measuring the effectiveness of the stirred reactor, and directly affects the reaction rate, conversion rate, product quality uniformity and production safety. The traditional monitoring method relies on indirect signals such as pH value, oxidation-reduction potential (ORP), pressure and torque, and these parameters are difficult to comprehensively and real-timely reflect the real disturbance state and mixing efficiency inside the fluid, and there is a certain lag and inaccuracy. In recent years, the solid-liquid mixing technology has been widely used in the fields of chemical industry, pharmaceutical industry, food processing and environmental protection. The evaluation of particle suspension effect is one of the key process parameters in the solid-liquid mixing process, and directly affects the mixing efficiency, reaction kinetics, product quality uniformity and energy efficiency of the system. How to quickly and accurately monitor the particle suspension state and particle motion characteristics, especially the instantaneous velocity of particle motion, has always been a hot issue of researchers. The traditional solid-liquid mixing monitoring method mainly relies on the collection of indirect signals, such as measuring the particle concentration gradient, viscosity change, conductivity change and other parameters to indirectly reflect the particle suspension state. However, these methods have limitations such as lack of real-time performance, complexity and the like. How to real-timely and accurately monitor the particle suspension state and the instantaneous velocity of particle motion is still a challenging research direction. SUMMARY

[0003] To solve the problems existing in the prior art, the present application provides a method for detecting the particle suspension effect in a stirred reactor, which comprises: preparing fluorescent small balls of different sizes; injecting a liquid into the stirred reactor, putting all the fluorescent small balls into the liquid, and stirring the liquid by a stirrer; irradiating the stirred reactor with an ultraviolet lamp to make the fluorescent small balls in the stirred reactor emit fluorescence; collecting images of the fluorescent small balls in the stirred reactor by using a real-time monitoring module; analyzing the images of the fluorescent small balls to obtain the motion trajectory, instantaneous velocity and direction of the fluorescent small balls; and obtaining the particle suspension effect in the liquid according to the motion trajectory, instantaneous velocity and direction of the fluorescent small balls.

[0004] The present application has the following advantages:

[0005] The application can capture the particle suspension state in real time through a high-speed camera and an intelligent image processing module, and significantly improves the detection efficiency; the application utilizes light detection technology, does not need to damage or intrude into the equipment, and reduces the damage to the equipment. The detection equipment has low cost and is suitable for large-scale industrial application. BRIEF DESCRIPTION OF DRAWINGS

[0006] Figure 1 It is a schematic diagram of the fluorescent ball of the application;

[0007] Figure 2 It is a schematic diagram of the particle suspension effect rapid monitoring device in the stirring reactor of the application;

[0008] Figure 3 It is an experimental diagram of the mixing module of the application;

[0009] Figure 4 It is a whole flow chart of the application. DETAILED DESCRIPTION

[0010] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0011] A particle suspension effect detection method in a stirring reactor, as shown in Figure 4 , the method comprises the following steps: preparing fluorescent balls of different sizes; injecting a liquid into the stirring reactor, putting all the fluorescent balls into the liquid, and stirring the liquid by a stirrer; irradiating the stirring reactor by a UV lamp to make the fluorescent balls in the stirring reactor emit fluorescence; collecting fluorescent ball images in the stirring reactor by a real-time monitoring module; analyzing the fluorescent ball images to obtain the motion trajectory, instantaneous speed and direction of the fluorescent balls; and obtaining the particle suspension effect in the liquid according to the motion trajectory, instantaneous speed and direction of the fluorescent balls.

[0012] Embodiment 1

[0013] Fluorescent ball: a good quality luminescent powder on the market is selected for preparation, to ensure stable and reliable luminescent performance, and to be able to stably emit fluorescence under UV irradiation. The schematic diagram of the fluorescent ball is as shown in Figure 1 .

[0014] Mixing module: a stirring shaft and a rigid paddle are installed inside the container. The stirring shaft is connected to a prime mover through a coupling, and the rigid paddle is installed on the node of the stirring shaft.

[0015] Real-time monitoring module: Choose high-resolution, high-frame-rate high-speed camera to ensure that it can clearly capture the dynamic behavior of fluorescent beads under ultraviolet light. At the same time, prepare a high-performance computer for receiving and storing image data collected by high-speed camera.

[0016] Intelligent image analysis module: Install MATLAB processing software on the computer and write corresponding image processing and analysis algorithms for real-time analysis of images collected by high-speed camera to extract disturbance pattern features of fluorescent beads.

[0017] In this embodiment, the schematic diagram of the device for quickly monitoring the particle suspension effect in the stirred reactor is as shown in Figure 2 The device includes a UV light 2, a rigid turbine blade 3, a stirring shaft 4, a container 5, a prime mover 6, a high-speed camera 7, a computer device 8, and an image analysis calculation module 9. The UV light 2 is distributed around the container 5 to provide UV light for the fluorescent beads 1 in the container. The rigid turbine blade 3 is connected to one end of the stirring shaft 4, and the other end of the stirring shaft 4 is connected to the prime mover 6. The prime mover 6 rotates the stirring shaft 4, thereby rotating the rigid turbine blade 3, so that the fluorescent beads 1 are uniformly mixed in the solution. The high-speed camera 7 is used to collect the position of the fluorescent beads 1 in the container 5 and generate images input to the computer device 8. The image analysis calculation module 9 is arranged inside the computer device 8. The image analysis calculation module 9 analyzes and calculates the multiple fluorescent bead images to obtain the particle suspension effect in the stirred reactor.

[0018] Embodiment 2

[0019] Preparation of fluorescent beads of different sizes includes: obtaining an appropriate amount of ZnS∶Cu 2+ @Al2O3 luminescent powder, mixing the luminescent powder with polydimethylsiloxane PDMS according to a mass ratio of 1:10 to obtain a mixed colloid; pouring the mixed colloid into a spherical mold of different sizes and drying; and correcting the beads after the colloid is solidified to obtain fluorescent beads of different sizes.

[0020] Specifically, luminescent powder with excellent luminescent performance is selected. According to design requirements, the luminescent powder is mixed with the colloid and molded using a mold. The colloid in the mold is treated to mainly remove impurities such as bubbles and dust. The flexible blade shaped by the mold is placed in a well-ventilated and dust-free environment for drying or natural air drying. The drying temperature and time are controlled to avoid incomplete solidification or cracking. After solidification, the fluorescent effect of the fluorescent beads is checked. If necessary, local repair or re-preparation is performed using the colloid to ensure that the fluorescent effect is durable and the roundness is high. At this time, the prepared fluorescent beads are the fluorescent beads 1.

[0021] The fluorescent beads are placed in the stirring container, and the high-speed camera is used for image capture.

[0022] Example 3

[0023] The mixing module is assembled according to the design requirements, and a transparent fluid is added to the container 5. Start the motor 6 and set the rotation speed of the stirring shaft. After starting the stirring, turn on the ultraviolet light source to irradiate the fluorescent beads. Start the high-speed camera to collect image data in the stirring reactor in real time and transmit it to the computer for storage. Use the intelligent image analysis module to process and analyze the collected image data, extract the disturbance mode characteristics of the fluorescent beads, such as movement speed, trajectory, etc. According to the analysis results, evaluate the suspension effect of the particles, and adjust the rotation speed of the stirring shaft or optimize the design of the stirring reactor as needed.

[0024] Example 4

[0025] The present application is applied to a stirring reactor in a certain large chemical enterprise. The reactor is mainly used for mixing and reacting two kinds of chemical raw materials, one being solid particles and the other being liquid.

[0026] The system is constructed and operated according to the methods of Examples 1, 2 and 3, and the movement patterns of the beads of different sizes, quantities and densities at different rotation speeds of the reactor are compared and analyzed. Based on the collected data and analysis results, the rotation speed, particle size and quantity that can improve the mixing performance of the stirring reactor and make the material mixing more uniform and efficient are selected.

[0027] The present application adopts a non-invasive monitoring method, which avoids the interference and damage of traditional measurement methods on the stirring process, reduces production cost and improves production efficiency.

[0028] Example 4

[0029] In this embodiment, the analysis of the fluorescent bead image includes:

[0030] Step 1, non-local mean denoising and wavelet threshold filtering are used to denoise and filter the fluorescent bead image, and an enhanced fluorescent bead image is obtained;

[0031] Step 2, the enhanced fluorescent bead image is processed using a recognition algorithm based on gray threshold and shape factor, and the spatial coordinates of the particles are identified;

[0032] Step 3, images at different times are obtained and used as time series images; for the time series images, the same position and pixel at different times are extracted, the pixel-level displacement is obtained using the cross-correlation method; a polynomial model is established for the known displacement pixels in the local window, the coefficients are solved by the least squares method, and the interpolation is completed by substituting the coordinates of the unknown displacement pixels.

[0033] The local window is demarcated, including: taking the pixel of the current displacement to be completed as the center, selecting a local window of 7x7 size (the number of valid displacement pixels in the window is not less than the number of polynomial coefficients). Then, the pixels with reliable displacement (u, y direction displacement) obtained by the cross-correlation method are screened in the window, and their coordinates (x i ,y i ) and corresponding displacement values (u i ,v i ) are recorded. For the mapping relationship between the coordinates of the x-y plane and the displacement, a polynomial equation is established. Taking the x direction displacement u as an example, the model is:

[0034]

[0035] where a0-a5 are the coefficients to be solved (the same for v direction).

[0036] Substitute (x_i, y_i, u_i) of the valid pixels in the window into the model, construct an overdetermined equation system, minimize the sum of squares of errors by the least square method, and solve all the polynomial coefficients. Finally, substitute the pixel coordinates (x0, y0) of the displacement to be completed in the window into the solved polynomial, and the fitted displacement value (u0, v0) of the pixel can be directly calculated to obtain the sub-pixel level displacement. The velocity vector is calculated according to the pixel level displacement, the sub-pixel level displacement and the time interval;

[0037] x direction velocity component:

[0038]

[0039] y direction velocity component:

[0040]

[0041] wherein, is the x direction pixel level displacement (integer), is the y direction pixel level displacement (integer) is the x direction sub-pixel level displacement (decimal, range [0, 1)), is the y direction sub-pixel level displacement (decimal, range [0, 1)), Δt is the time interval (time difference between adjacent two frames of images, unit: s), and k is the physical pixel equivalent (actual physical length corresponding to 1 pixel, such as μm / px).

[0042] Step 4, set the speed value threshold, calculate the area in the pixel flow field that is less than the speed value threshold according to the speed vector, and make the area white, while the rest of the area is processed using grayscale; use the image measurement function to calculate the area of the white area: first count the number of white pixels in the image, then convert the pixel equivalent to the actual physical area, and the mainstream image processing library (OpenCV, MATLAB) has a ready-to-use function that can be directly called. Take the rgb2gray function in MATLAB as an example: first calculate the pixel area and actual physical area of the white area in the image; input: img-single channel grayscale image (white=255, grayscale=0-254); pixel_scale-physical pixel equivalent; output: pixel_area-pixel area (px2); physical_area-actual physical area; verify single channel, if color, convert to grayscale size (img, 3) = 3, then img = rgb2gray (img); generate a binary image (white = 1, others = 0), and count the number of white pixels; binary_img = img = 255; pixel_area = sum (binary_img (:)); physical_area = pixel_area * (pixel_scale ^ 2). Then compare the area ratio of gray and white areas under different conditions to get the suspension effect of the particles. The suspension effect is shown in Figure 3

[0043] Step 5, denoising processing of time series images, calculation of two-dimensional box fractal dimension of denoised time series images, and binarization of images; determine the box size according to the image pixel size; calculate the number of objects in each box in the image; directly reference the imboxft function in MATLAB to filter the image with boxes, use the sum function to calculate the number of pixels with a value of 1 in each box, and plot the relationship between box size and box number on a log-log graph; linearly fit the data in the log-log graph and calculate the slope; determine the complexity of the image according to the slope.

[0044] Fractal dimension F d It can be estimated by least squares fitting of log (Nr) and log (1 / r):

[0045]

[0046] Where y = log (Nr), x = log (1 / r), r is the partition ratio, and Nr is the total number of boxes in all grids.

[0047] Step 6, establish the correlation function between the suspension state and the stirring speed, and determine the critical suspension speed N c .​

[0048] In the embodiment, the non-local mean denoising and wavelet threshold filtering are adopted to perform denoising filtering processing on the fluorescent microsphere image, including similarity measurement (Euclidean distance): for a pixel (i, j) to be denoised in the image I (denoted as a center block P(i, j)), searching for an image block P(k, l) corresponding to any pixel (k, l) in the window, and the similarity between the two is measured by a normalized Euclidean distance:

[0049]

[0050] wherein m is a similarity block size (in the fluorescent image, m=3~5, too small is easy to misjudge the similarity, and too large increases the calculation amount).

[0051] Weight calculation:

[0052]

[0053] wherein h is a smoothing parameter (in the fluorescent image, h=5~15, the larger h is, the stronger the noise reduction is, but the more blurred the details are, and h needs to be adjusted according to the noise intensity); and Ω is a search window (in the fluorescent image, Ω=7~11, which covers the local similar area of the microsphere, so as to avoid introducing background noise due to too large search range).

[0054] Denoised pixel value:

[0055]

[0056] The final pixel value is the weighted sum of all pixels in the search window, and the similar block has a larger weight.

[0057] In the embodiment, the enhanced fluorescent microsphere image is processed by using an identification algorithm based on a gray threshold and a shape factor, including:

[0058] Input: denoised / enhanced fluorescent microsphere single-channel gray image I (microsphere gray value > background gray value): based on the image gray histogram, a threshold T is manually selected (usually taking the valley value between the "background peak value" and the "microsphere peak value"). The adaptive / manual threshold is set, the image is binarized, and a candidate region containing microspheres and a small amount of impurities is obtained.

[0059] Binarization rule:

[0060]

[0061] Post-segmentation post-processing: remove small noise points by morphological operation (dilation + erosion), fill small ball internal holes; label all independent candidate connected regions, calculate the geometric characteristics (area, perimeter, circumscribed rectangle, etc.) of each region; find all independent white connected regions in the binary image (each region corresponds to 1 candidate target), and extract the geometric characteristics of the region. Erosion (Erosion) and then dilation (Dilation) (opening operation), formula based on pixel traversal of structure element:

[0062] Erosion formula:

[0063]

[0064] Dilation formula:

[0065]

[0066] Shape factor screening: set the shape factor threshold, eliminate non-circular regions, and retain the true fluorescent beads; output the bead target region mask, the center coordinates of each bead, the radius, etc. Core parameters (for subsequent displacement calculation).

[0067] Circularity C: measure the proximity of the region to a perfect circle: the smaller the perimeter, the larger the area, and the closer the circularity to 1; Non-circular regions (such as rectangles, irregular shapes) have a larger perimeter and a smaller circularity. Its ideal value is C=1. Its expression is:

[0068]

[0069] Aspect ratio AR: measure the aspect ratio of the circumscribed rectangle of the region: the circumscribed rectangle of a perfect circle is a square, and the aspect ratio = 1; The aspect ratio of elongated impurities is much less than 1. Its ideal value is AR=1. Its expression is:

[0070]

[0071] Compactness Comp: measure the proportion of the region to the circumscribed rectangle: the compactness of a perfect circle ≈ 0.785 (π / 4); Irregular regions have a lower compactness. Its ideal value is Comp=0.785. Its expression is:

[0072]

[0073] Solidity S: measure the solidity of the region: fluorescent beads are solid, and the solidity ≈ 1; Hollow impurities or broken regions have a lower solidity. Its ideal value is S=1. Its expression is:

[0074]

[0075] where, is the area of the region convex hull.

[0076] In the embodiment, the pixel-level displacement is obtained by using the cross-correlation method, which includes: taking the fluorescent ball in the reference frame as a template, searching for the best matching position in the target frame search area by normalized cross-correlation (NCC), calculating the difference between the center coordinates of the two frames of balls, and obtaining the pixel-level displacement:

[0077] (1) Template and search area definition:

[0078] Template ROI (reference frame I t , ball center , equivalent radius )

[0079] side length , coordinate range

[0080] (2) Normalized cross-correlation (NCC)

[0081]

[0082] wherein is the average gray value of the template, is the average gray value of the search area sub-region.

[0083] In the embodiment, the correlation function between the suspension state and the stirring speed is established, which includes: quantifying the experimental data of the suspension state (suspension rate) and the stirring speed, and fitting to obtain a nonlinear correlation function, which is commonly described by a logarithmic or S-shaped curve:

[0084] (1) Logarithmic correlation function (applicable to low and medium speed interval, suspension rate gradually rises)

[0085]

[0086] wherein η is the suspension rate (white suspension area / total area, 0~1), n is the stirring speed (r / min), n0 is the critical speed (the lowest speed that just starts to suspend), a and b are experimental fitting coefficients (positive numbers, reflecting the rate of suspension rate with speed growth).

[0087] (2) S-shaped correlation function (adapt to the whole speed interval, including the saturation stage)

[0088]

[0089] wherein η max is the saturated suspension rate (the highest suspension degree, ≈1), k is the growth rate coefficient, n c is the characteristic speed (the speed when the suspension rate reaches η max / 2).

[0090] The above examples further illustrate the objects, technical solutions and advantages of the present application. It should be understood that the above examples are only preferred embodiments of the present application and are not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made to the present application within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for detecting a particle suspension effect in a stirred reactor, characterized by, include: Prepare fluorescent microspheres of different sizes; Liquid is injected into the stirred reactor, all the fluorescent beads are placed in the liquid, and the liquid is stirred by a stirrer; An ultraviolet lamp is used to irradiate a stirred reactor, causing fluorescent spheres inside the reactor to fluoresce. A real-time monitoring module is used to acquire images of the fluorescent spheres inside the stirred reactor. The images of the fluorescent spheres are analyzed to obtain their trajectory, instantaneous velocity, and direction. Based on the trajectory, instantaneous velocity, and direction of the fluorescent spheres, the effect of particle suspension in the liquid is obtained.

2. The method of claim 1, wherein the method is characterized by: The preparation of fluorescent small balls with different sizes comprises: obtaining ZnS:Cu 2+ @Al2O3 photoluminescence powder, mixing the photoluminescence powder and polydimethylsiloxane (PDMS) according to a mass ratio of 1:10 to obtain a mixed colloid; pouring the mixed colloid into spherical molds with different sizes and drying; and correcting the small balls after the colloid is solidified to obtain fluorescent small balls with different sizes.

3. The method of claim 1, wherein the method is characterized by: The agitator includes a prime mover, an agitator shaft, and a rigid turbine blade; the rigid turbine blade is connected to one end of the agitator shaft, and the other end of the agitator shaft is connected to the prime mover. When the prime mover rotates, it drives the agitator shaft and the rigid turbine blade to rotate.

4. The method of claim 1, wherein the method is characterized by: The real-time monitoring module uses a high-resolution, high-frame-rate high-speed camera.

5. The method of claim 1, wherein the method is characterized by: Analysis of the fluorescent sphere images includes: Step 1: The fluorescent sphere image is denoised using nonlocal mean denoising and wavelet threshold filtering to obtain the enhanced fluorescent sphere image. Step 2: The enhanced fluorescent sphere image is processed using a recognition algorithm based on grayscale threshold and shape factor to identify the spatial coordinates of the particles; Step 3: Acquire images at different times and use them as time series images; extract images of the same position and pixels at different times from the time series images, and use the cross-correlation method to obtain pixel-level displacement; establish a polynomial model for known displacement pixels within a local window, and solve the polynomial model using the least squares method to obtain the coefficients; substitute the coordinates of unknown displacement pixels into the polynomial model to complete the interpolation; construct an overdetermined system of equations based on the interpolated data, and solve for all polynomial coefficients by minimizing the sum of squared errors using the least squares method; substitute the pixel coordinates (x0, y0) of the displacement to be completed within the window into the solved polynomial to obtain the fitted displacement value (u0, v0) of that pixel, thus obtaining the sub-pixel-level displacement; calculate the velocity vector based on the pixel-level displacement, the sub-pixel-level displacement, and the time interval. Step 4: Set a velocity threshold. Calculate the region in the pixel flow field that is less than the velocity threshold based on the velocity vector, and turn this region white, while the remaining regions are processed in grayscale. Use an image measurement function to calculate the area of ​​the white region. By comparing the proportion of gray and white regions under different conditions, the suspension effect of the particles is obtained. Step 5: Denoise the time series image, calculate the two-dimensional box fractal dimension of the denoised time series image, and binarize the image; determine the box size based on the image pixel size; calculate the number of objects in each box in the image; use the imboxft function to perform box filtering on the image, use the sum function to calculate the number of pixels with a value of 1 in each box, and plot the relationship between box size and box number on a log-log plot; linearly fit the data in the log-log plot and calculate the slope; determine the image complexity based on the slope. Step 6, the correlation function between the suspension state and the stirring speed is established according to the complexity of the image, and the critical suspension speed N is determined by fitting the curve c .

6. The method of claim 5, wherein the method is characterized by: The non-local mean denoising and wavelet threshold filtering are adopted to perform denoising filtering processing on the fluorescent ball image, including: searching for an image block P(k, l) corresponding to any pixel (k, l) in a search window of a pixel (i, j) to be denoised in the image I, and calculating the similarity of the two; constructing a weight according to the similarity; filtering the pixel based on the weight to obtain a denoised pixel value; and obtaining a final denoised image by taking the weighted sum of all pixels in the search window as the pixel value.

7. The method of claim 5, wherein the method is characterized by: The enhanced fluorescent ball image is processed by using a recognition algorithm based on a gray threshold and a shape factor, including: performing binaryzation processing on the denoised fluorescent ball single-channel gray image I to obtain a candidate region containing the fluorescent balls and a small amount of impurities; removing small noise points and filling holes in the fluorescent balls by morphological operation; marking all independent candidate connected regions, and calculating the geometric features of each region; finding all independent white connected regions in the binary image, and extracting the geometric features of the regions; setting a shape factor threshold, removing non-circular regions, and retaining the real fluorescent balls; and outputting a ball target region mask, the center coordinates of each ball, and a radius parameter.

8. The method of claim 5, wherein the method is characterized by: The pixel-level displacement is obtained by using a cross-correlation method, including: taking the reference frame fluorescent ball as a template, searching for the best matching position in the target frame search region by normalized cross-correlation, calculating the difference between the center coordinates of the two frames of fluorescent balls, and obtaining the pixel-level displacement.

9. The method of claim 5, wherein the method is characterized by: The correlation function between the suspension state and the stirring speed is established, including: quantifying experimental data of the suspension state and the stirring speed, and fitting to obtain a nonlinear correlation function; wherein the nonlinear correlation function is a logarithmic correlation function or an S-shaped correlation function.

10. The method of claim 9, wherein the method is characterized by: The logarithmic correlation function is: ； wherein η is the suspension rate, n is the stirring speed, n0 is the critical speed, and a and b are experimental fitting coefficients; The S-shaped correlation function is: ； where η max is the saturation suspension rate, k is the growth rate coefficient, n c is the characteristic rotational speed.

Images