An automatic demulsification control algorithm based on emulsification state identification

Abstract

The application discloses an automatic demulsification control algorithm based on emulsification state recognition, comprising the following steps: collecting video stream data after liquid chemical reaction; extracting an ROI region through a deep learning segmentation network; classifying the ROI region through a deep learning classification network to obtain a liquid emulsification judgment result; segmenting the liquid emulsification judgment result through a deep learning segmentation algorithm to obtain a specific emulsification layer region; calculating the proportion of the emulsification layer region image in the ROI region; calculating demulsifier parameters according to the proportion; calculating stirring speed according to the proportion; and performing demulsification operation according to the demulsifier parameters and the stirring speed. The application controls the content of the demulsifier and the stirring speed in real time through the deep learning segmentation network and the deep learning classification network, so as to adapt to different experimental scenes. Through self-feedback setting, the efficiency and reliability of the system are ensured.

Description

Technical Field

[0001] This invention relates to the field of image recognition technology, and in particular to an automatic demulsification control algorithm based on emulsification state recognition. Background Technology

[0002] In chemical experiments, when two immiscible liquids coexist, the presence of surfactants promotes the formation of an emulsion layer, blurring the originally clear liquid interface. This phenomenon is the emulsification that demulsification addresses. To restore the liquids to their separate states, demulsification is performed by adding demulsifiers or extending the settling time.

[0003] Demulsification is crucial in chemical experiments, directly impacting the smooth execution of subsequent key operations such as liquid separation and transfer. Failure to effectively break down the emulsion will severely disrupt the experimental process, hindering the accurate separation and extraction of target components, and consequently affecting the reliability and validity of the experimental results.

[0004] However, in the field of chemical demulsification, traditional demulsification control methods currently rely primarily on subjective human judgment. This approach is highly subjective, making it difficult to precisely control the demulsification process. The varying levels of experience among operators can easily lead to significant differences in demulsification results. Furthermore, human judgment cannot track the dynamic changes in the emulsion state in real time, lacking precise data support. It is difficult to flexibly adjust key parameters such as the amount of demulsifier and the stirring speed according to actual conditions, resulting in low demulsification efficiency and difficulty in guaranteeing the accuracy of the final experimental results. This significantly restricts the high-quality conduct of chemical experiments and the reliability of research findings. Summary of the Invention

[0005] This invention provides an automatic demulsification control algorithm based on emulsion state recognition, addressing the problem that traditional demulsification control methods in the prior art mainly rely on subjective human judgment. This approach is highly subjective, making it difficult to accurately control the demulsification process. Different operators with varying levels of experience can easily lead to significant differences in demulsification results. Human judgment cannot track the dynamic changes in the emulsion state in real time, lacks accurate data support, and makes it difficult to flexibly adjust key parameters such as demulsifier dosage and stirring speed according to actual conditions. This results in low demulsification efficiency and makes it difficult to guarantee the accuracy of the final experimental results.

[0006] On one hand, embodiments of the present invention provide an automatic demulsification control algorithm based on emulsification state recognition, including:

[0007] Acquire video stream data following a liquid chemical reaction;

[0008] Regions of Interest (ROI) are extracted using a deep learning segmentation network.

[0009] The ROI region is classified by a deep learning classification network to obtain a liquid emulsification determination result;

[0010] The liquid emulsification determination result is segmented by a deep learning segmentation algorithm to obtain a specific emulsion layer region;

[0011] The proportion of the emulsion layer region image in the ROI region is calculated;

[0012] The demulsifier parameter is calculated according to the proportion;

[0013] The stirring speed is calculated according to the proportion;

[0014] The demulsification operation is performed according to the demulsifier parameter and the stirring speed.

[0015] In a possible implementation, the extracting the ROI region of interest by the deep learning segmentation network comprises:

[0016] The video frame data of the video stream data is segmented by a YOLOv8-Seg deep learning segmentation network to remove the background and obtain the ROI region of interest.

[0017] In a possible implementation, the classifying the ROI region by the deep learning classification network to obtain the liquid emulsification determination result comprises:

[0018] The ROI region is classified by a YOLOv8-cls deep learning classification network to obtain an emulsification state result and a non-emulsification state result;

[0019] The demulsification completion result is obtained according to the non-emulsification state result.

[0020] In a possible implementation, the segmenting the liquid emulsification determination result by the deep learning segmentation algorithm to obtain the specific emulsion layer region comprises:

[0021] The ROI region obtaining the emulsification state result is regionally segmented by a SegFormer semantic segmentation network to obtain an emulsion layer region image.

[0022] In a possible implementation, the calculating the proportion of the emulsion layer region image in the ROI region comprises:

[0023] The proportion is calculated according to the pixel area of the emulsion layer region image in the ROI region.

[0024] In a possible implementation, the calculating the demulsifier parameter according to the proportion comprises:

[0025] According to the area proportion in the emulsion layer region and the ROI region, the emulsifier addition amount of the demulsifier is calculated.

[0026] In a possible implementation, the calculating the stirring speed according to the proportion comprises:

[0027] According to the area proportion in the emulsion layer region and the ROI region, the stirring speed required for demulsification is calculated.

[0028] In a possible implementation, the demulsification operation according to the demulsifier parameter and the stirring speed comprises:

[0029] The liquid is left to stand;

[0030] Returning to the step of collecting the video stream data, the detection is performed again.

[0031] An automatic demulsification control algorithm based on emulsion state recognition in the application has the following advantages:

[0032] (1) The content of the demulsifier and the stirring speed are controlled in real time through a deep learning segmentation network and a deep learning classification network to adapt to different experimental scenarios.

[0033] (2) Through self-feedback setting, the efficiency and reliability of the system are ensured. BRIEF DESCRIPTION OF DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description only some embodiments of the present application, and for those skilled in the art, on the premise of not paying creative labor, other drawings can also be obtained according to these drawings.

[0035] Figure 1 A flow chart of an automatic demulsification control algorithm based on emulsion state recognition provided by the embodiments of the present application. DETAILED DESCRIPTION

[0036] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0037] Figure 1A flowchart of an automatic demulsification control algorithm based on emulsification state recognition provided by the embodiment of the present application; the embodiment of the present application provides an automatic demulsification control algorithm based on emulsification state recognition, comprising:

[0038] Collecting video stream data after liquid chemical reaction;

[0039] Extracting an ROI region through a deep learning segmentation network;

[0040] Classifying the ROI region through a deep learning classification network to obtain a liquid emulsification determination result;

[0041] Segmenting the liquid emulsification determination result through a deep learning segmentation algorithm to obtain a specific emulsion layer region;

[0042] Calculating the proportion of the emulsion layer region image in the ROI region;

[0043] Calculating a demulsifier parameter according to the proportion;

[0044] Calculating a stirring speed according to the proportion;

[0045] Performing a demulsification operation according to the demulsifier parameter and the stirring speed.

[0046] The extraction of the ROI region of interest through the deep learning segmentation network comprises:

[0047] Segmenting the video frame data of the video stream data through a YOLOv8-Seg deep learning segmentation network to remove the background and obtain the ROI region of interest.

[0048] The classification of the ROI region through the deep learning classification network to obtain the liquid emulsification determination result comprises:

[0049] Classifying the ROI region using a YOLOv8-cls deep learning classification network to obtain an emulsification state result and a non-emulsification state result;

[0050] Obtaining a demulsification completion result according to the non-emulsification state result.

[0051] The segmentation of the liquid emulsification determination result through the deep learning segmentation algorithm to obtain the specific emulsion layer region comprises:

[0052] Segmenting the ROI region obtaining the emulsification state result through a SegFormer semantic segmentation network to obtain an emulsion layer region image.

[0053] The calculation of the proportion of the emulsion layer region image in the ROI region comprises:

[0054] According to the pixel area of the emulsion layer region image in the ROI region, the proportion is calculated.

[0055] The emulsion breaking agent parameter is calculated according to the proportion.

[0056] The emulsifier addition amount of the emulsion breaking agent is calculated according to the area proportion in the emulsion layer region and the ROI region.

[0057] The stirring speed is calculated according to the proportion.

[0058] The stirring speed required for emulsion breaking is calculated according to the area proportion in the emulsion layer region and the ROI region.

[0059] After the emulsion breaking operation according to the emulsion breaking agent parameter and the stirring speed, the following is included:

[0060] The liquid is left to stand;

[0061] The step of collecting the video stream data is returned to, and the detection is performed again.

[0062] Exemplarily, the video stream data after the liquid chemical reaction is collected in real time by a fixed industrial camera.

[0063] The region of interest (ROI) is extracted from the video frame using the YOLOv8-Seg semantic segmentation network, effectively removing background irrelevant information, and improving the accuracy and efficiency of detection.

[0064] The YOLOv8-Seg is a model in the YOLOv8 series that integrates target detection and instance segmentation functions, and the structure includes Backbone (C2f module and SPPF module to extract multi-scale features), Neck (PAN-FPN structure to fuse multi-scale features), and Head (detection head + segmentation head).

[0065] The extracted ROI region is classified using the YOLOv8-cls deep learning classification network to determine whether it belongs to an emulsified state or a non-emulsified state.

[0066] The YOLOv8-cls is a lightweight model designed for image classification in the YOLOv8 series, and its structure consists of Backbone (feature extraction), Neck (multi-scale fusion, which can be simplified or omitted in classification tasks), and Head (classification output). Backbone uses C2f module instead of traditional C3 module to reduce redundant parameters by optimizing gradient flow and improve efficiency by combining depth separable convolution. The Head part uses global average pooling (GAP) and fully connected layer to output class probability, supporting Anchor-Free mechanism without preset anchor box.

[0067] If the YOLOv8-cls deep learning classification network identifies as a non-emulsification state, it is determined that the demulsification is completed, and there is no emulsion layer in the solution; if it is identified as an emulsification state, it enters the next step of processing.

[0068] For the ROI region identified as an emulsification state, the specific area of the emulsion layer is further segmented by the SegFormer semantic segmentation network, providing a basis for subsequent quantitative analysis.

[0069] The SegFormer is a semantic segmentation model based on the Transformer architecture, composed of an encoder (MixTransformer, hierarchical extraction of multi-scale features) and a decoder (All-MLP structure, only through linear layer fusion features). Its core design includes Overlap Patch Embedding (retaining local context information) and Efficient Self-Attention. The parameter amount is extremely low, suitable for high-precision segmentation of complex scenes.

[0070] Among them, the model configuration of the SegFormer model is as shown in the following code:

[0071]

[0072] The area ratio of the emulsion layer to the overall ROI region liquid pixel area is calculated. The formula is as follows:

[0073]

[0074] Among them, is the segmented emulsion layer area. W and H are the width and height of the ROI region respectively. R is the area ratio.

[0075] The dosage of the demulsifier is dynamically adjusted by the area ratio R. The formula is as follows:

[0076]

[0077] Among them, is the emulsifier addition amount. t is the frame position of the current video frame. represents the preset initial demulsifier content. represents the area ratio calculated in the t frame.

[0078] The stirring speed is dynamically adjusted by the area ratio R. The formula is as follows:

[0079]

[0080] Among them, is the stirring speed. t is the frame position of the current video frame. represents the preset initial stirring speed. represents the area ratio calculated in the t frame.

[0081] Based on the demulsifier amount and agitation speed obtained above, the agitation equipment is controlled. After standing for N seconds, return to the initial step and continue to monitor the liquid state until demulsification is complete.

[0082] While the preferred embodiments of the application have been described, additional variations and modifications can be made to these embodiments by those skilled in the art once they learn of the basic inventive concepts. Such additional variations and modifications should be considered as within the scope of the application as defined by the claims appended hereto.

[0083] It will be apparent to those skilled in the art that various modifications and variations can be made to the present application without departing from the spirit or scope of the application. Thus, it is intended that the present application cover modifications and variations of this application provided they come within the scope of the appended claims and their equivalents.

Claims

1. An automatic demulsification control algorithm based on emulsion state recognition, characterized in that, The method comprises: collecting video stream data after liquid chemical reaction; extracting ROI region through deep learning segmentation network; classifying the ROI region through deep learning classification network to obtain liquid emulsification determination result; segmenting the liquid emulsification determination result through deep learning segmentation algorithm to obtain specific emulsion layer region; calculating the proportion of the emulsion layer region image in the ROI region; calculating demulsifier parameters according to the proportion; calculating stirring speed according to the proportion; performing demulsification operation according to the demulsifier parameters and the stirring speed; the method further comprises: resting the liquid; returning to the step of collecting the video stream data to re-detect; the method further comprises: using YOLOv8-cls deep learning classification network to classify the ROI region to obtain emulsion state result and non-emulsion state result; obtaining demulsification completion result according to the non-emulsion state result; the method further comprises: using SegFormer semantic segmentation network to segment the ROI region to obtain emulsion layer region image.

2. The automatic demulsification control algorithm based on emulsion state recognition according to claim 1, characterized in that, the method further comprises: using YOLOv8-Seg deep learning segmentation network to segment the video frame data of the video stream data to remove the background and obtain the ROI region of interest.

3. The automatic demulsification control algorithm based on emulsion state recognition according to claim 1, characterized in that, the method further comprises: calculating the proportion according to the pixel area of the emulsion layer region image in the ROI region.

4. The automatic break-up control algorithm based on emulsion state recognition according to claim 1, characterized in that, the method further comprises: calculating the emulsifier addition amount of the demulsifier according to the area proportion in the emulsion layer region and the ROI region.

5. The automatic break-up control algorithm based on emulsion state recognition according to claim 1, characterized in that, the method further comprises: calculating the stirring speed required for demulsification according to the area proportion in the emulsion layer region and the ROI region.

Images