A rotary distillation adaptive control method based on bubble state quantification

Abstract

The present application relates to a kind of based on bubble state quantitative rotary distillation adaptive control method, comprising the following steps: step one: bubble original image acquisition;Step two: bubble original image pre-processing obtains bubble to be analyzed image;Step three: bubble feature extraction is carried out to bubble to be analyzed image, step four: according to bubble coverage and boiling intensity adjustment distilling flask vacuum degree, bubble coverage is maintained in target interval, boiling intensity does not exceed maximum set value, in addition, the rate of change of bubble coverage with time does not exceed set value;Step five: when distilling flask vacuum degree adjustment cannot stabilize bubble coverage in target interval, while adjusting the temperature of water bath kettle immersed in the bottom of distilling flask.The bubble characteristics obtained in the distillation process are quantified in real time, so that the traditional artificial observation and judgment of fuzzy process are changed into programmable digital closed-loop control, which improves the process consistency, reduces the regulation delay time, and ensures the distillation safety.

Description

Technical Field

[0001] This invention relates to the field of rotary distillation control technology for energetic rubber compounds, specifically an adaptive control method for rotary distillation based on bubble state quantification. Background Technology

[0002] The rotary distillation process for energetic rubber compounds (such as nitrocellulose-based propellant slurries and composite modified dual-base propellant compounds) involves rotating and evaporating the material in a distillation flask under vacuum heating to remove excess solvents (such as acetone, ethyl acetate, and ethanol) from the rubber compound, bringing the compound concentration to the required level. During distillation, the characteristics of the bubbles generated by solvent boiling directly reflect the distillation state. Specifically: too few bubbles indicate a low vacuum or insufficient temperature, leading to incomplete solvent removal and potentially excessive solvent residue in the finished product; too many and violent bubbles indicate a high vacuum or excessive temperature, causing "bumping" in the distillation flask. In this case, the rubber compound can splash instantly into the condenser, resulting in material loss, potential equipment blockage, or even an explosion.

[0003] In existing technologies, rotary distillation apparatuses are generally controlled by manual visual inspection and manual adjustment of the vacuum valve. This means that the operator, through an explosion-proof observation window, judges the size, quantity, and boiling intensity of bubbles generated in the distillation flask based on experience, and then manually rotates the vacuum valve to adjust the vacuum level or the heating power of the water bath. This method has the following drawbacks:

[0004] First, it is highly subjective and lacks consistency. Different personnel have different judgment standards, and even the same person may make judgment errors due to factors such as fatigue and lack of focus on different shifts. Actual process testing shows that the coefficient of variation (CV) of bubble coverage between process batches is usually as high as 15% to 20%.

[0005] Second: Control response lag. From manual observation to manual adjustment, and then to the vacuum system response, there is usually a delay of 10 to 30 seconds, which makes it very easy to miss the optimal control window.

[0006] Third: Bubble state cannot be quantified. Due to the lack of quantifiable indicators for bubble state in existing technologies, closed-loop control cannot be formed, and data linkage with higher-level systems such as MES is impossible.

[0007] Fourth: Personnel spending long periods of time observing on-site violates the "unmanned" production safety principle required by many companies. Summary of the Invention

[0008] The purpose of this invention is to provide an adaptive control method for rotary distillation based on bubble state quantification. This method quantifies and outputs features such as bubble coverage, bubble quantity, average bubble area (i.e., average bubble size), and boiling intensity obtained during the distillation process in real time. This transforms the traditional fuzzy process of "manual visual inspection + experience judgment" into a programmable digital closed-loop control, which not only improves process consistency but also reduces the delay time of adjustment and control. At the same time, it can also minimize the occurrence of bumping accidents and ensure distillation safety.

[0009] The objective of this invention is achieved through the following technical solution:

[0010] An adaptive control method for rotary distillation based on bubble state quantization includes the following steps:

[0011] Step 1: Acquire raw images of the air bubbles inside the distillation flask;

[0012] Step 2: Extract the ROI region from the original bubble image, and then preprocess the ROI region to obtain the bubble image to be analyzed;

[0013] Step 3: Extract bubble features from the bubble image to be analyzed, specifically as follows:

[0014] Step 3.1: Process the bubble image to be analyzed based on dynamic threshold segmentation and obtain the final binary image;

[0015] Step 3.2: Mark connected components in the final binary graph and obtain bubble features through calculation, wherein the bubble features include:

[0016] 1. Number of bubbles Nb = Number of connected components;

[0017] 2: Bubble coverage C = (ΣAi) / A ROI Where ΣAi is the total area of ​​all bubbles in the final binary image, A ROI This refers to the area of ​​the ROI region in step two.

[0018] 3. Average bubble area Aavg = (ΣAi) / Nb;

[0019] 4. Boiling intensity F: Divide the final binary image into an N×N grid and calculate the variance of the LBP histogram for each grid. Then take the average value as the F value. The higher the F value, the more intense the boiling.

[0020] Step 4: Adjust the vacuum level of the distillation flask according to the bubble coverage C and boiling intensity F. The bubble coverage C is maintained within the target range [Clow, Chigh], while the boiling intensity F is limited to not exceeding the maximum set value Fmax. In addition, the rate of change of bubble coverage C with time t does not exceed the set value.

[0021] Step 5: When adjusting the vacuum level of the distillation flask cannot stabilize the bubble coverage C within the target range, simultaneously adjust the temperature of the water bath submerged at the bottom of the distillation flask.

[0022] In step one, the explosion-proof industrial camera in the image acquisition unit acquires real-time images of the liquid surface and the bubbles above the liquid surface in the distillation flask, and sends the original bubble images to the processing system of the image processing unit.

[0023] The image acquisition unit includes an LED light source and a polarizing filter. The angle between the axis of the explosion-proof industrial camera lens and the normal to the liquid surface of the distillation flask is controlled between 15° and 20°. The LED light source is coaxially mounted with the lens of the explosion-proof industrial camera. The polarizing filter is placed in front of the lens of the explosion-proof industrial camera, and a polarizing filter with an orthogonal direction is also installed in front of the LED light source.

[0024] In step two, the original bubble image undergoes preprocessing by the image processing unit, specifically as follows:

[0025] Step 2.1: ROI region extraction, determining the area within a specified range where the liquid surface in the distillation flask is located as the ROI region;

[0026] Step 2.2: Denoise and enhance the ROI region;

[0027] Step 2.3: Obtain the 0° and 90° polarization images of the ROI region after the above processing, and subtract the two to obtain the bubble image to be analyzed.

[0028] Step 3.1 specifically involves:

[0029] Step 3.1.1: Obtain the global pixel binarization threshold T0;

[0030] Step 3.1.2: For each pixel in the bubble image to be analyzed, calculate the mean μ and standard deviation σ of its designated neighborhood;

[0031] Step 3.1.3: Calculate the adaptive pixel threshold T1 according to the following formula (1):

[0032] T1=μ+M×σ (1;

[0033] In the above formula (1), the value of M ranges from -5 to -10;

[0034] Step 3.1.4: Obtain the final binary image based on T0 and T1, wherein the pixels in the bubble image to be analyzed that are greater than both T0 and T1 form the final binary image.

[0035] In step four:

[0036] When C < Clow and F < minimum set value Fmin, the frequency f of the vacuum pump inverter connected to the distillation flask increases, thereby increasing the pumping speed and internal vacuum of the distillation flask.

[0037] When C > Chigh or F > the maximum set value Fmax, the vent valve on the vacuum pump line connected to the distillation flask opens to the set opening degree, thereby reducing the vacuum degree of the distillation flask.

[0038] When the rate of change of C with time t is greater than the set value, that is, when dC / dt is greater than the set value, the vent valve on the vacuum pump pipeline will automatically open fully.

[0039] In step five, the water bath temperature is adjusted according to the following formula (2):

[0040] ΔTbath=Kt×(C-Ctarget) (2);

[0041] In the above formula (2), ΔTbath is the temperature change of the water bath, C is the bubble coverage rate, Kt is the change ratio factor, which is negative, and Ctarget is the target value of the bubble coverage rate, Ctarget=(Clow+Chigh) / 2.

[0042] 8. The adaptive control method for rotary distillation based on bubble state quantification according to claim 1, characterized in that: the mouth of the distillation flask is connected to the inlet of the rotary distillation apparatus, and the bottom of the distillation flask is immersed in a water bath, and the distillation flask is driven to perform rotary distillation by the rotary distillation apparatus.

[0043] The advantages and positive effects of this invention are as follows:

[0044] 1. This invention quantifies and outputs features such as bubble coverage, number of bubbles, average bubble area (i.e., average bubble size), and boiling intensity in real time during the distillation process. This transforms the traditional fuzzy process of "manual visual inspection + experience judgment" into programmable digital closed-loop control, which not only meets the requirements of unmanned production but also improves process consistency. In particular, actual process testing shows that the coefficient of variation of bubble coverage between batches can be reduced from about 18% in the manual method to about 5%, and the process consistency is significantly improved.

[0045] 2. The total delay time of this invention from image acquisition and feature extraction to vacuum adjustment and control is ≤2 seconds, which is far lower than the 10 to 30 seconds of manual response, and the response speed is greatly improved.

[0046] 3. This invention can automatically and quickly fully open the gas valve when signs of boiling occur by monitoring the rate of change of bubble coverage C (dC / dt). Actual process testing has shown that this can reduce boiling accidents by more than 80%.

[0047] 4. By maintaining the bubble coverage rate C within the target range [Clow, Chigh] and limiting the boiling intensity F to not exceed the maximum set value Fmax, and in conjunction with the aforementioned monitoring of the rate of change of bubble coverage rate C, this invention can more comprehensively ensure the safe operation of the distillation process.

[0048] 5. This invention integrates traditional image processing methods such as polarization difference dereflection, morphological reconstruction de-foaming, and CLAHE enhancement, which can ensure that the bubble recognition accuracy remains at a high level (e.g., above 92%) even under conditions such as foaming on the bottle wall and water mist. Compared with the recognition accuracy of pure threshold segmentation (e.g., 65%), this invention can effectively ensure the reliability of system control.

[0049] 6. All bubble characteristic values, distillation flask vacuum degree, water bath temperature and various control commands of this invention can be stored in the local database in real time and uploaded to the MES and safety situation platform through the interface. This can provide a complete visual evidence chain and data support for subsequent process optimization, accident analysis and quality management system. Attached Figure Description

[0050] Figure 1 This is a schematic diagram of the process of the method of the present invention.

[0051] Figure 2 This is a schematic diagram of the structure of the device of the present invention.

[0052] Figure 3 for Figure 2 A schematic diagram showing the assembly of a rotary distillation apparatus, a water bath, and a distillation flask.

[0053] Figure 4 This is a schematic diagram of the bubble image to be analyzed after preprocessing by the method of the present invention.

[0054] Figure 5 for Figure 4 A schematic diagram of the threshold segmentation process for the image to be analyzed, where bubbles are visible. Figure 1 ,

[0055] Figure 6 for Figure 4 A schematic diagram of the threshold segmentation process for the image to be analyzed, where bubbles are visible. Figure 2 ,

[0056] Figure 7 This is a schematic diagram illustrating the changes in bubble coverage, vacuum degree, and water bath temperature obtained according to one embodiment of the present invention.

[0057] Figure 8 This is a schematic diagram of a human-machine interface according to an embodiment of the present invention.

[0058] Among them, 1 is a rotary distillation apparatus, 2 is a water bath, 3 is a distillation flask, and 4 is an image acquisition unit. Detailed Implementation

[0059] The invention will now be described in further detail with reference to the accompanying drawings.

[0060] like Figures 2-3 As shown, the apparatus of the present invention includes a rotary distillation apparatus 1, a water bath 2, a distillation flask 3, an image acquisition unit 4, an image processing unit, an adjustment component, and a controller, etc. Wherein:

[0061] 1. The mouth of the distillation flask 3 containing the material is connected to the port of the rotary distillation apparatus 1. At the same time, the distillation flask 3 is immersed in the water bath 2. After the water bath 2 is heated to the set temperature, the rotary distillation apparatus 1 is started and drives the distillation flask 3 to perform rotary distillation under internal vacuum conditions.

[0062] Second: The image acquisition unit 4 includes an explosion-proof industrial camera (1" CMOS, 8 megapixels), a ring-shaped LED light source (10W power, 5000K color temperature) and a polarizing filter, which are used to acquire clear, low-reflection bubble images inside the distillation flask 3 during the distillation process.

[0063] Among them, such as Figures 2-3 As shown in this embodiment, a bracket is provided on one side of the rotary distillation apparatus 1, and the explosion-proof industrial camera is mounted on the bracket. The angle between the lens axis of the explosion-proof industrial camera and the normal to the liquid surface of the distillation flask is controlled at 15° to 20° to avoid main reflection from the flask wall.

[0064] The LED light source is coaxially mounted with the lens of the explosion-proof industrial camera, and its brightness is adjustable from 0 to 100%. A polarizing filter is placed in front of the lens of the explosion-proof industrial camera, and a polarizing filter with orthogonal direction is also installed in front of the LED light source to eliminate specular reflection.

[0065] In this embodiment, the explosion-proof industrial camera is used for continuous acquisition at a frame rate of 30fps, with automatic control of exposure time (range 0.1 to 10ms).

[0066] Third: The image processing unit adopts an industrial control computer, which is electrically connected to the explosion-proof industrial camera in the image acquisition unit 4. The bubble image acquired by the explosion-proof industrial camera is sent to the industrial control computer, and then the industrial control computer runs image preprocessing, bubble segmentation, feature extraction and other programs for processing. In addition, the internal program of the industrial control computer can also include a deep learning model as needed to improve control accuracy.

[0067] Fourthly, the regulating components include a vacuum pump for evacuating the interior of the distillation flask 3 and a heater for controlling the heating temperature of the water bath 2. The vacuum pump is equipped with a frequency converter, and a venting valve (proportional control valve) is installed on the vacuum pump pipeline connected to the distillation flask 3. When the vacuum inside the distillation flask 3 is too low, the pumping speed can be increased by adjusting the frequency converter to improve the vacuum level. When the vacuum inside the distillation flask 3 is too high, the venting valve can be opened to quickly reduce the vacuum level. The heater can adjust the heating power, thereby regulating the heating temperature of the water bath 2.

[0068] 5. The controller is electrically connected to the image processing unit. It is used to receive the parameter signal after the final processing by the image processing unit, and then execute the control algorithm according to the corresponding parameter signal and output the drive signal to control the corresponding device in the above-mentioned adjustment component to make timely adjustments.

[0069] Furthermore, this invention can be used with a human-machine interface to display parameters such as bubble images, characteristic curves, and vacuum settings in real time, depending on actual needs, while also allowing for manual adjustment. The human-machine interface can be a touchscreen panel for convenient adjustment.

[0070] All of the above-mentioned devices are technologies known in the art and are commercially available products.

[0071] The method of the present invention includes the following steps:

[0072] Step 1: Acquire the original image of the air bubbles inside the distillation flask.

[0073] like Figure 2 As shown, during the distillation process, the explosion-proof industrial camera in the image acquisition unit 4 acquires real-time images of the liquid surface and the bubbles above the liquid surface in the distillation flask 3, and sends the original bubble images to the image processing unit.

[0074] Step 2: Preprocess the original bubble image to obtain the bubble image to be analyzed.

[0075] The image processing unit preprocesses the original bubble image, specifically as follows:

[0076] Step 2.1: ROI (Region of Interest) extraction.

[0077] In this embodiment, the image processing unit automatically determines the rectangular area (1200×600 pixels) within the specified range where the liquid surface is located as the observation and processing area, that is, the ROI area. In addition, since the liquid level changes slightly during the distillation process, the image processing unit can use dynamic edge detection (such as Canny edge detection) to update the liquid surface boundary of the ROI area in real time as needed. The above are technologies known in the art.

[0078] Step 2.2: Denoise and enhance the ROI region.

[0079] The image processing unit first performs median filtering (kernel 5×5) on the ROI region to remove noise, then enhances local contrast using a contrast-limited adaptive histogram equalization method (CLAHE, clipLimit=2.0, gridSize=8×8), and finally performs Gaussian filtering (σ=1.2) to smooth the texture. All of the above processing is implemented through software and is well-known in the art.

[0080] Step 2.3: After performing the above processing on both the 0° and 90° polarized images of the ROI region, the image processing unit subtracts the two images to obtain an image that suppresses specular reflection components. This is the bubble image to be analyzed. (See also...) Figure 4 As shown. The above processing is implemented through a software program and is a technique known in the art.

[0081] In addition, in this step, polarized images at different angles are obtained through a polarizing filter in front of the lens of an explosion-proof industrial camera. This is a well-known technique in the field. For example, polarized images at different angles can be obtained sequentially by rotating the polarizing filter in front of the lens of an explosion-proof industrial camera.

[0082] In this step, such as Figure 4 As shown, for the dark areas caused by material buildup on the walls of the distillation flask, the present invention can recover bubble information by using morphological reconstruction (mask = original image, marker = image after material buildup) as needed. This processing is implemented by software program and is a well-known technology in the field.

[0083] Step 3: Extract bubble features from the bubble image to be analyzed. Specifically:

[0084] Step 3.1: As Figures 5-6 As shown, the bubble image to be analyzed is processed based on dynamic threshold segmentation, and the final binary image is obtained. Wherein:

[0085] Step 3.1.1: The image processing unit calculates and obtains the global pixel binarization threshold T0 of the bubble image to be analyzed using the Otsu algorithm. The Otsu algorithm is a well-known technique in the art.

[0086] Step 3.1.2: The image processing unit calculates the mean μ and standard deviation σ of a defined neighborhood (31×31 pixels) for each pixel in the bubble image to be analyzed. This is a technique known in the art.

[0087] Step 3.1.3: Calculate the adaptive pixel threshold T1 according to the following formula (1):

[0088] T1=μ+M×σ (1;

[0089] In the above formula (1), the value of M ranges from -5 to -10, preferably -8.

[0090] Step 3.1.4: Obtain the final binary image based on T0 and T1, wherein the pixels in the bubble image to be analyzed that are greater than both T0 and T1 form the final binary image.

[0091] All of the above processes are implemented through software programs.

[0092] In addition, in this step, the final binary image can be further filtered to remove weak noise as needed. This is achieved through software programs and is a well-known technique in the field.

[0093] In this step, morphological processing can be performed on adjacent bubbles in the final binary image as needed to facilitate subsequent overall analysis of bubble clusters. Specifically, in this embodiment: an opening operation (5×5 rectangular kernel) removes isolated noise points; a closing operation (11×11 rectangular kernel) aggregates neighboring bubble regions with a spacing ≤ 11 pixels. This is implemented through software and is a technique known in the art.

[0094] The final binary image obtained after the above processing in this embodiment can be found in [reference]. Figure 5 As shown.

[0095] Step 3.2: Mark connected components in the final binary graph and calculate the bubble features. Specifically:

[0096] Step 3.2.1: The image processing unit marks the connected components of the final binary graph. This is a well-known technique in the art, wherein, for example... Figure 6 As shown, the number of connected components is equal to the number of bubbles.

[0097] Step 3.2.2: The image processing unit calculates the area Ai, perimeter Pi, roundness (4πAi / Pi^2), and centroid of each bubble, i=1, 2... and the maximum number of connected components.

[0098] Step 3.2.3: The image processing unit removes extremely small noise points with an area smaller than a set value (<10 pixels in this embodiment).

[0099] Step 3.2.4: Extract bubble features, including:

[0100] 1. Number of bubbles Nb = Number of connected components;

[0101] 2: Bubble coverage C = (ΣAi) / A ROI Where ΣAi is the total area of ​​all bubbles, A ROI This refers to the area of ​​the ROI region in step two.

[0102] 3: Average bubble area (i.e., average bubble size) Aavg=(ΣAi) / Nb;

[0103] 4. Boiling Intensity F: Using Local Binary Pattern (LBP) texture analysis, the final binary image is divided into an N×N (16×16 in this embodiment) grid, and the variance of the LBP histogram for each grid is calculated. The average value is then taken as the F value. The above calculation is a well-known technique in the art. For the present invention, the higher the F value, the more intense the bubble formation and bursting.

[0104] All of the above processing is implemented through software programs.

[0105] Step 4: Adjust the vacuum level inside the distillation flask 3 according to the bubble coverage rate C and the boiling intensity F. The bubble coverage rate C is maintained within the target range [Clow, Chigh], while the boiling intensity F is limited to not exceeding the maximum set value Fmax. In addition, the rate of change of bubble coverage rate C with time t (i.e., dC / dt) does not exceed the set value.

[0106] In this embodiment:

[0107] Controlled variable: Vacuum degree inside distillation flask 3, with an adjustment range of 3–50 kPa. In this embodiment, the vacuum degree closed-loop control cycle is 200 ms. The system performs adaptive fine-tuning based on a reference vacuum degree process curve, where the reference curve is: initially 10 kPa held for 10 min, then gradually reduced to 3 kPa held for 30 min. All of the above are techniques known in the art.

[0108] Controlled variables: frequency f (0-50Hz) of the vacuum pump inverter and opening degree V (0-100%) of the vent valve.

[0109] When C < Clow and F < 0.5 (i.e., the minimum set value Fmin), the system determines that the vacuum inside the distillation flask 3 is too low, and then controls the frequency f of the frequency converter to increase the pumping speed of the vacuum pump and improve the vacuum inside the distillation flask.

[0110] When C > Chigh or F > 0.7 (i.e., the maximum set value Fmax), the system determines that there is a risk of bumping in distillation flask 3, and then controls the vent valve to open to the set opening degree to reduce the vacuum degree inside the distillation flask. In addition, when the rate of change of C with time t is too fast, that is, when dC / dt is greater than the set value, this is usually a precursor to bumping. The vent valve can automatically and quickly open fully to minimize the occurrence of bumping accidents.

[0111] This invention takes into account the nonlinear factors of bumping risk and employs gain-scheduled PID control. It uses the deviation between the target bubble coverage value Ctarget=(Clow+Chigh) / 2 and the real-time value as input. Based on the bubble coverage range and boiling intensity, it automatically switches the PID parameter group and actuator, thereby achieving adaptive vacuum adjustment. This PID control is a well-known technique in the art. Specific details of this embodiment are shown in the table below:

[0112] C < 5% and F < 0.5 0.8 0.05 0.2 Vacuum pump frequency increases 5% ≤ C ≤ 30% and F < 0.5 0.5 0.02 0.1 Keep C > 30% or F > 0.7 1.5 0.01 0.5 The vent valve opening increases dC / dt > 15% / s (signs of impending boiling) 2 0 1 The valve opens rapidly and fully in 1 second.

[0113] A graph obtained from one embodiment of the present invention can be found here. Figure 7 As shown.

[0114] Step 5: When the vacuum adjustment of distillation flask 3 fails to stabilize C within the target range, the controller simultaneously adjusts the heater of water bath 2 according to the following formula (2), thereby adjusting the temperature of water bath 2:

[0115] ΔTbath=Kt×(C-Ctarget) (2);

[0116] In the above formula (2), ΔTbath is the temperature change of water bath 2, C is the real-time value of bubble coverage, and Kt is the change ratio factor, which is negative. That is, when the bubble coverage C is too high, the temperature of water bath 2 is reduced, and vice versa. In this embodiment, Kt=-0.5℃ / %, Ctarget is the target value of bubble coverage, Ctarget=(Clow+Chigh) / 2.

[0117] Furthermore, Kt can be set according to actual needs, as long as the temperature change rate can be controlled within the set range to prevent thermal shock. In this embodiment, the temperature change rate is limited to ±1℃ / min.

[0118] This invention quantifies and outputs in real time the characteristics of bubble coverage, bubble quantity, average bubble area (i.e., average bubble size), and boiling intensity during the distillation process. This transforms the traditional fuzzy process of "manual visual inspection + experience-based judgment" into programmable digital closed-loop control. Actual process testing shows that using this method reduces the batch-to-batch bubble coverage variation coefficient from approximately 18% using manual methods to approximately 5%, significantly improving process consistency. Furthermore, the total delay time from image acquisition and feature extraction to vacuum regulation control is ≤2 seconds, far lower than the 10-30 seconds of manual response, resulting in a substantial improvement in response speed. Simultaneously, by monitoring the rate of change of bubble coverage C (dC / dt), this invention can automatically and rapidly fully open the gas valve when signs of impending boiling appear (in this embodiment, dC / dt > 15% / s). This can reduce boiling accidents by more than 80%, thereby maximizing distillation safety. Finally, this invention integrates traditional image processing methods such as polarization difference dereflection, morphological reconstruction de-fogging, and CLAHE enhancement, which can ensure that the bubble recognition accuracy remains at a high level (over 92% in this embodiment) even under harsh conditions such as bottle wall fogging and condensation mist. Compared with the recognition accuracy of pure threshold segmentation (65% in actual process testing), this invention can effectively ensure the reliability of system control.

[0119] A graph obtained in one embodiment of the present invention is shown below. Figure 7 As shown, the solid line represents the bubble coverage rate C, the dashed line represents the vacuum degree P, and the dotted line represents the water bath temperature T. Figure 7 As shown at point A in the diagram, at the 18th minute, due to the large-scale evaporation of solvent, the bubble coverage rate C surged from 18% to 35% within 3 seconds. The system detected dC / dt = 5.7% / s, which did not reach the boiling threshold of 15% / s, but C > 30%. The venting valve immediately opened by 5%, and the vacuum level increased from 8 kPa to 20 kPa. Simultaneously, the water bath 2 decreased in temperature at a rate of -0.5℃ / min. After 5 seconds, the bubble coverage rate C dropped back to 22%, and no actual boiling occurred. Therefore, this invention, by maintaining the bubble coverage rate C within the target range [Clow, Chigh] while limiting the boiling intensity F to not exceed the maximum set value Fmax, combined with monitoring the rate of change of the bubble coverage rate C, can more comprehensively ensure the safe operation of the distillation process.

[0120] The human-machine interface of this embodiment can be found in [reference]. Figure 8 As shown.

Claims

1. An adaptive control method for rotary distillation based on bubble state quantization, characterized in that: Includes the following steps: Step 1: Acquire raw images of the air bubbles inside the distillation flask; Step 2: Extract the ROI region from the original bubble image, and then preprocess the ROI region to obtain the bubble image to be analyzed; Step 3: Extract bubble features from the bubble image to be analyzed, specifically as follows: Step 3.1: Process the bubble image to be analyzed based on dynamic threshold segmentation and obtain the final binary image; Step 3.2: Mark connected components in the final binary graph and obtain bubble features through calculation, wherein the bubble features include:

1. Number of bubbles Nb = Number of connected components; 2: Bubble coverage C = (ΣAi) / A ROI Where ΣAi is the total area of ​​all bubbles in the final binary image, A ROI This refers to the area of ​​the ROI region in step two.

3. Average bubble area Aavg = (ΣAi) / Nb; 4. Boiling intensity F: Divide the final binary image into an N×N grid and calculate the variance of the LBP histogram for each grid. Then take the average value as the F value. The higher the F value, the more intense the boiling. Step 4: Adjust the vacuum level of the distillation flask according to the bubble coverage C and boiling intensity F. The bubble coverage C is maintained within the target range [Clow, Chigh], while the boiling intensity F is limited to not exceeding the maximum set value Fmax. In addition, the rate of change of bubble coverage C with time t does not exceed the set value. Step 5: When adjusting the vacuum level of the distillation flask cannot stabilize the bubble coverage C within the target range, simultaneously adjust the temperature of the water bath submerged at the bottom of the distillation flask.

2. The adaptive control method for rotary distillation based on bubble state quantization according to claim 1, characterized in that: In step one, the explosion-proof industrial camera in the image acquisition unit acquires real-time images of the liquid surface and the bubbles above the liquid surface in the distillation flask, and sends the original bubble images to the processing system of the image processing unit.

3. The adaptive control method for rotary distillation based on bubble state quantization according to claim 2, characterized in that: The image acquisition unit includes an LED light source and a polarizing filter. The angle between the axis of the explosion-proof industrial camera lens and the normal to the liquid surface of the distillation flask is controlled between 15° and 20°. The LED light source is coaxially mounted with the lens of the explosion-proof industrial camera. The polarizing filter is placed in front of the lens of the explosion-proof industrial camera, and a polarizing filter with an orthogonal direction is also installed in front of the LED light source.

4. The adaptive control method for rotary distillation based on bubble state quantization according to claim 3, characterized in that: In step two, the original bubble image undergoes preprocessing by the image processing unit, specifically as follows: Step 2.1: ROI region extraction, determining the area within a specified range where the liquid surface in the distillation flask is located as the ROI region; Step 2.2: Denoise and enhance the ROI region; Step 2.3: Obtain the 0° and 90° polarization images of the ROI region after the above processing, and subtract the two to obtain the bubble image to be analyzed.

5. The adaptive control method for rotary distillation based on bubble state quantization according to claim 1, characterized in that: Step 3.1 specifically involves: Step 3.1.1: Obtain the global pixel binarization threshold T0; Step 3.1.2: For each pixel in the bubble image to be analyzed, calculate the mean μ and standard deviation σ of its designated neighborhood; Step 3.1.3: Calculate the adaptive pixel threshold T1 according to the following formula (1): T1=μ+M×σ (1; In the above formula (1), the value of M ranges from -5 to -10; Step 3.1.4: Obtain the final binary image based on T0 and T1, wherein the pixels in the bubble image to be analyzed that are greater than both T0 and T1 form the final binary image.

6. The adaptive control method for rotary distillation based on bubble state quantization according to claim 1, characterized in that: In step four: When C < Clow and F < minimum set value Fmin, the frequency f of the vacuum pump inverter connected to the distillation flask increases, thereby increasing the pumping speed and internal vacuum of the distillation flask. When C > Chigh or F > the maximum set value Fmax, the vent valve on the vacuum pump line connected to the distillation flask opens to the set opening degree, thereby reducing the vacuum degree of the distillation flask. When the rate of change of C with time t is greater than the set value, that is, when dC / dt is greater than the set value, the vent valve on the vacuum pump pipeline will automatically open fully.

7. The adaptive control method for rotary distillation based on bubble state quantization according to claim 1, characterized in that: In step five, the water bath temperature is adjusted according to the following formula (2): ΔTbath=Kt×(C-Ctarget) (2); In the above formula (2), ΔTbath is the temperature change of the water bath, C is the bubble coverage rate, Kt is the change ratio factor, which is negative, and Ctarget is the target value of the bubble coverage rate, Ctarget=(Clow+Chigh) / 2.

8. The adaptive control method for rotary distillation based on bubble state quantization according to claim 1, characterized in that: The mouth of the distillation flask is connected to the inlet of the rotary distillation apparatus, and the bottom of the distillation flask is immersed in a water bath. The distillation flask is driven by the rotary distillation apparatus to perform rotary distillation.

Images