Picture identification method for PBX explosive water suspension granulation process and application
The MaskR-CNN model is used to identify and segment the particles and droplet images during the suspension granulation of explosive water, which solves the problem of difficulty in real-time monitoring and analysis in the prior art, and achieves safe and efficient product quality control.
Patent Information
- Application Number
- CN202410856676.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-28
- Publication Date
- 2024-11-15
AI Technical Summary
During the existing explosive water suspension preparation process, it is difficult to accurately monitor and analyze the morphology, size and distribution of particles, resulting in difficult to ensure product consistency and safety.
The MaskR-CNN deep learning network model is used to identify and segment the particles and droplet images acquired online. Multiple images are acquired through the online particle imager, the model is trained and predicted, and the target parameters of particles and droplets are obtained.
Real-time monitoring of the PBX explosive water suspension granulation process is achieved, which avoids operator danger, improves product quality and experimental efficiency, and makes up for the shortcomings of traditional methods that cannot analyze key quality parameters online.
Smart Images

Figure CN118968505A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of explosive water suspension preparation, and relates to a picture recognition method and application of a PBX explosive water suspension granulation process. Background Art
[0002] Polymer bounded explosives (PBX) are the most widely used type of mixed explosives with high energy density, excellent safety and mechanical properties. Water suspension granulation is one of the common methods for preparing PBX modeling powder. Monitoring the water suspension granulation process can provide real-time information on key parameters such as particle morphology, size, distribution and uniformity, so as to adjust operating parameters and process conditions in a timely manner, optimize the preparation process, and ensure product consistency and good performance. It can be seen that online process monitoring of water suspension granulation is crucial.
[0003] With the rapid development of high-speed camera technology and image processing algorithms, real-time acquisition of particle motion information in suspensions and analysis and interpretation of images is a promising method for online particle detection. However, traditional target detection algorithms still have difficulty in accurately identifying irregular targets that aggregate and overlap during water suspension granulation. Summary of the invention
[0004] In view of this, in order to solve the problem that important target parameters are difficult to obtain and analyze in the existing explosive water suspension preparation process, the present invention provides a PBX explosive water suspension granulation process image recognition method and application, and uses the MaskR-CNN deep learning network model to identify and segment the particle images and droplet images in the suspension obtained online.
[0005] To achieve the above-mentioned effects, the present invention provides, on one hand, an image recognition method for a PBX explosive water suspension granulation process, comprising: preparing PBX explosive by water suspension granulation of a main explosive, and acquiring multiple images of the granulation process at fixed time intervals using an online particle imager; randomly extracting some images from the multiple images acquired to form an image data set; training a Mask R-CNN model based on the image data set, and using the trained Mask R-CNN model to predict the multiple images acquired during the granulation process, thereby automatically identifying and segmenting particles and / or droplets in the image, and obtaining target parameters.
[0006] Optionally, the preparation of PBX explosive by water suspension granulation of the main explosive and the use of a linear particle imager to obtain multiple images of the granulation process at fixed time intervals specifically include: suspending the main explosive in ionized water to form a suspension solution; heating the suspension solution and adding a binder solution for a certain period of time, and then continuing to increase the temperature to evaporate and remove the solvent in the binder solution, so that the binder is coated on the surface of the explosive; during the entire granulation process, immersing the linear particle imager in the suspension solution to obtain images at fixed time intervals.
[0007] Optionally, the preparation of PBX explosive by water suspension granulation of the main explosive further comprises: the main explosive is 90-100g of cyclotrimethylene trinitramine explosive, and the volume of the ionized water is 300-400ml.
[0008] Optionally, the heating of the suspension solution and the addition of the binder solution for a certain period of time, and then continuing to increase the temperature to evaporate and remove the solvent in the binder solution further includes: the heating temperature is 50-55°C, the volume of the binder solution is 50-80ml, and the duration is 30-40min; the binder solution is prepared by dissolving 5g of cellulose acetate butyrate in 50-80ml of ethyl acetate; during the continuous process, heating is performed using a water bath, stirring is performed at a stirring rate of 300-500r / min, and the binder solution is slowly added dropwise to the suspension solution.
[0009] Optionally, the step of heating the suspension solution and adding the binder solution for a certain period of time, and then continuing to increase the temperature to evaporate and remove the solvent in the binder solution further comprises: continuing to increase the temperature to 60-70°C.
[0010] Optionally, acquiring images at fixed time intervals further includes: taking a photo every 1-5 seconds.
[0011] Optionally, randomly extracting some images from the acquired multiple images to form an image data set further includes: labeling some images randomly extracted from the multiple images, and distinguishing the labels into particles and droplets; and expanding the total number of images in the image data set by changing the image resolution.
[0012] Optionally, the automatic identification and segmentation of particles and / or droplets in the image and obtaining target parameters further specifically include: automatically identifying and segmenting particles and / or droplets in the image, outputting an image that constructs a pixel-level mask and a rectangular box of the particles and / or droplets; performing statistical analysis on the pixel-level mask and the rectangular box of the particles and / or droplets in the image to obtain target parameters of the particles and / or droplets.
[0013] Optionally, the target parameters include one or more of quantity, area, perimeter, granularity, and aspect ratio.
[0014] Another aspect of the present invention provides an application of the image recognition method for the PBX explosive water suspension granulation process in the PBX explosive preparation process.
[0015] The beneficial effects of the present invention are:
[0016] (1) The PBX explosive water suspension granulation process is monitored and photographed in real time, and particles and droplets in the image are automatically identified and segmented, avoiding the potentially dangerous behavior of operators frequently removing particles for observation.
[0017] (2) It can count the number of particles, particle size distribution, and morphology in the suspension, which makes up for the deficiency of traditional PBX explosive water suspension granulation that cannot analyze key quality parameters online. It is of great significance to improve product quality and experimental efficiency. The acquisition of some target parameters will play a positive role in optimizing future process conditions and promote product design.
[0018] In addition, additional advantages, objects, and features of the present invention will be described in part in the following description, and will become apparent to those skilled in the art after studying the following, or may be learned from the practice of the present invention. The objects and other advantages of the present invention may be realized and obtained by the structures and method steps specifically pointed out in the written description and claims and the accompanying drawings.
[0019] Those skilled in the art will appreciate that the objectives and advantages that can be achieved with the present invention are not limited to the above specific description, and the above and other objectives that can be achieved by the present invention will be more clearly understood from the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] The drawings described herein are used to provide a further understanding of the present invention, constitute a part of this application, and do not constitute a limitation of the present invention.
[0021] Figure 1 This is a picture taken by an online particle imaging instrument in an embodiment of the present invention.
[0022] Figure 2 This is a graph of the predicted output of the trained Mask R-CNN model in an embodiment of the present invention.
[0023] Figure 3 Graph showing the changing trend of the number of particles and droplets in an embodiment of the present invention.
[0024] Figure 4 Graph showing particle size variation trends in an embodiment of the present invention.
[0025] Figure 5 Graph showing the changing trend of particle roundness and aspect ratio in an embodiment of the present invention. DETAILED DESCRIPTION
[0026] In order to make the purpose, technical scheme and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments and the accompanying drawings. The schematic embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention. It should also be noted that, in order to avoid blurring the present invention due to unnecessary details, only structures and / or processing steps closely related to the solutions according to the present invention are shown in the accompanying drawings, while other details that are not closely related to the present invention are omitted.
[0027] It should be emphasized that the term “include / comprises” when used herein refers to the presence of features, elements, steps or components, but does not exclude the presence or addition of one or more other features, elements, steps or components.
[0028] On the one hand, an embodiment of the present invention provides an image recognition method for a PBX explosive water suspension granulation process, comprising: preparing the PBX explosive by water suspension granulation of a main explosive, and acquiring multiple images of the granulation process at fixed time intervals using a linear particle imager; randomly extracting some images from the acquired multiple images to form an image data set; training a Mask R-CNN model based on the image data set, and using the trained Mask R-CNN model to predict the acquired multiple images of the granulation process, thereby automatically identifying and segmenting particles and / or droplets in the image, and obtaining target parameters.
[0029] Optionally, the main explosive is subjected to water suspension granulation to prepare PBX explosive, and a linear particle imager is used to obtain multiple images of the granulation process at fixed time intervals. Specifically, the main explosive is floated in ionized water to form a suspension solution; the suspension solution is heated and a binder solution is added, and after a certain period of time, the temperature is continued to be increased to evaporate and remove the solvent in the binder solution, so that the binder is coated on the surface of the explosive; during the entire granulation process, the linear particle imager is immersed in the suspension solution to obtain images at fixed time intervals.
[0030] Optionally, the preparation of PBX explosive by water suspension granulation of the main explosive further includes: the main explosive is 90-100g of cyclotrimethylene trinitramine explosive, and the volume of ionized water is 300-400ml; as an alternative for the same use, the main explosive can also be selected from other explosives, including but not limited to 3-nitro-1,2,4-triazole-5-one (NTO), hexanitrohexaazaisowurtzitane (HNIW), etc.
[0031] Optionally, the suspension solution is heated and a binder solution is added thereto for a certain period of time, and then the temperature is continued to be raised to evaporate and remove the solvent in the binder solution, further comprising: a heating temperature of 50-55°C, a volume of the binder solution of 50-80ml, and a duration of 30-40min; in fact, the duration can also be selected within a certain range, such as 20-40min, and the binder solution is prepared by dissolving 5g of cellulose acetate butyrate in 50-80ml of ethyl acetate; during the continuous process, heating is performed in a water bath, stirring is performed at a stirring rate of 300-500r / min, and the binder solution is slowly added dropwise to the suspension solution.
[0032] Optionally, the step of heating the suspension solution and adding a binder solution for a certain period of time, and then continuing to increase the temperature to evaporate and remove the solvent in the binder solution further comprises: continuing to increase the temperature to 60-70°C.
[0033] Optionally, acquiring images at fixed time intervals further includes: taking a photo every 1-5 seconds.
[0034] Optionally, randomly extracting some images from the acquired multiple images to form an image data set further includes: labeling some images randomly extracted from the multiple images, and distinguishing the labels into particles and droplets; and expanding the total number of images in the image data set by changing the image resolution.
[0035] Optionally, automatically identifying and segmenting particles and / or droplets in the image and obtaining target parameters further specifically include: automatically identifying and segmenting particles and / or droplets in the image, outputting an image that constructs a pixel-level mask and a rectangular box of the particles and / or droplets; performing statistical analysis on the pixel-level mask and the rectangular box of the particles and / or droplets in the image to obtain target parameters of the particles and / or droplets.
[0036] Optionally, the target parameters include one or more of number, area, perimeter, granularity, and aspect ratio.
[0037] On the one hand, an embodiment of the present invention provides an application of the above-mentioned PBX explosive water suspension granulation process image recognition method in the PBX explosive preparation process.
[0038] The specific embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings. It should be noted that the following embodiments are only illustrative and not restrictive, and the protection scope of the present invention cannot be limited thereto.
[0039] 95g of the main explosive was suspended in 400ml of ionized water, and an online particle imager was inserted into the suspension to take one picture per second. The online particle imager was connected to a computer to collect images of the PBX explosive water suspension preparation process in real time. The solution was heated to 50°C, and 50ml of the binder solution was added. The process lasted for about 40 minutes, and the temperature was raised to 65°C to evaporate and remove the solvent.
[0040] 203 images were randomly selected and manually annotated using Labme software to generate json files. The Mask R-CNN model was trained according to the ratio of 8:2 between the training set and the validation set. In addition, in order to improve the accuracy and robustness of the Mask R-CNN model, the image data was expanded to 609 by changing the resolution of the acquired images. Specifically, ResNet-101 can be used as the backbone network as a feature extractor to extract features. After the input image passes through ResNet-101, a five-layer feature map will be obtained. The structure of the FPN feature pyramid network is adopted. The high-level features with smaller resolution are first reduced in dimension by 1×1 convolution (reducing the amount of calculation), and then upsampled to the same size of the previous feature map, and then element-by-element addition is performed to perform feature fusion; the obtained feature map is input into the RPN to generate a proposal consisting of a category label (foreground or background) and a coordinate position (based on the input image). Considering the shape of particles and droplets and the possibility of overlap, the aspect ratios of the anchors are set to 0.5, 1, and 2, respectively. The diameters of particles and droplets in the image are mainly between 100 and 3000 μm, so the anchor scale is set to five levels: 32, 64, 128, 256, and 512 pixels to accommodate particles and droplets of different sizes. Finally, the ROIAlign operation is performed on the remaining ROI for classification and Mask generation.
[0041] Use the trained model to predict 1003 images in the process. Figure 1 This is a picture taken by an online particle imaging instrument. Figure 2 This is the output of the trained Mask R-CNN model prediction. The predicted image successfully recognizes and segments particles and droplets, and displays the mask and rectangular box of each particle and droplet.
[0042] The number of particles and droplet masks in the image is counted to obtain the change in the number of particles and droplets, such as Figure 3 The trend of the number of particles and droplets shows that during the addition of the binder, the particles agglomerate and the number decreases. During the temperature-raising and dissolving process, some particles break and the number increases. The number of droplets indirectly reflects the concentration of the binder solution. With the addition of the binder solution, the number of droplets gradually increases. During the temperature-raising and dissolving process, the solvent evaporates and the number of droplets gradually decreases.
[0043] Write a program to calculate the number of pixels in the mask of the particles in each image and the number of pixels on the mask boundary to estimate the area S and circumference L of the particles. According to the conversion rule of 1 pixel = 4.6 microns, find out the true size of the particles. Consider the particles as circles and find the equivalent radius of the particles according to the area formula of the circle. Take every 50 images as a group and calculate the average.
[0044] Figure 4 This is a trend chart of particle size change. It can be seen that the particle size first increases, then slightly decreases, and finally remains stable. This reflects that the particles first agglomerate under the action of the binder, then some particles break during the temperature increase and finally remain stable.
[0045] According to the area S and perimeter L of the particles obtained above, the roundness C and aspect ratio Ar of the particles are calculated using the formulas C=4*π*S / L2 and Ar=H / W. Figure 5 It is a trend chart of the change of particle roundness and aspect ratio. It can be seen that in the whole process, the particle roundness is always distributed around 0.8, and the particle aspect ratio is distributed around 1, and there is no obvious change, indicating that the particles in the whole process are approximately round.
[0046] The present invention is mainly aimed at the fact that the water suspension granulation process cannot be monitored in real time by existing technical means, and can only be manually viewed by manually scooping up some samples. This poses a serious safety risk. Use an online particle imager to obtain pictures of the water suspension granulation process, input them into the Mask R-CNN model, and segment and identify the particle and droplet images. The pictures taken during the water suspension granulation process have the characteristics of high particle concentration and serious overlap. Manual identification requires measuring the shape and size of the particles one by one, which is difficult to operate. Traditional image recognition algorithms are difficult to accurately segment particles and droplets in this complex environment. The present invention innovatively proposes an image recognition method for the PBX explosive water suspension granulation process, introduces deep learning methods into the acquisition of parameters in the explosive water suspension granulation process, and achieves a safe and efficient technical effect that cannot be achieved by existing technologies.
[0047] The specific description above further illustrates the purpose, technical solutions and beneficial effects of the invention in detail. It should be understood that the above is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. A method for image recognition of a PBX explosive water suspension granulation process, characterized in that: include: The main explosive is subjected to water suspension granulation to prepare PBX explosive, and a line particle imager is used to obtain multiple images of the granulation process at fixed time intervals; Randomly extracting some images from the acquired multiple images to form an image data set; The Mask R-CNN model is trained based on the image data set, and the trained Mask R-CNN model is used to predict the multiple images obtained in the granulation process, thereby automatically identifying and segmenting the particles and / or droplets in the image, and obtaining the target parameters.
2. The image recognition method for PBX explosive water suspension granulation process according to claim 1 is characterized in that: The method of preparing PBX explosive by water suspension granulation of the main explosive and obtaining multiple images of the granulation process at fixed time intervals using a linear particle imager specifically includes: suspending the main explosive in ionized water to form a suspension solution; The suspension solution is heated and a binder solution is added thereto. After a certain period of time, the temperature is further increased to evaporate and remove the solvent in the binder solution, so that the binder is coated on the surface of the explosive; During the whole granulation process, the line particle imager was immersed in the suspension solution to acquire images at fixed time intervals.
3. The image recognition method for PBX explosive water suspension granulation process according to claim 2 is characterized in that: The method of suspending the main explosive in ionized water to form a suspension solution further comprises: The main explosive is 90-100g of cyclotrimethylene trinitramine explosive, and the volume of the ionized water is 300-400ml.
4. The image recognition method for PBX explosive water suspension granulation process according to claim 3 is characterized in that: The step of heating the suspension solution and adding the binder solution for a certain period of time, and then continuing to increase the temperature to evaporate and remove the solvent in the binder solution further comprises: The heating temperature is 50-55°C, the volume of the binder solution is 50-80ml, and the duration is 30-40min; the binder solution is prepared by dissolving 5g of cellulose acetate butyrate in 50-80ml of ethyl acetate; during the duration, heating is performed using a water bath, while stirring at a stirring rate of 300-500r / min, and the binder solution is slowly added dropwise to the suspension solution.
5. The image recognition method for PBX explosive water suspension granulation process according to claim 4 is characterized in that: The step of heating the suspension solution and adding the binder solution for a certain period of time, and then continuing to increase the temperature to evaporate and remove the solvent in the binder solution further comprises: Continue to raise the temperature to 60-70°C.
6. The image recognition method for PBX explosive water suspension granulation process according to claim 5 is characterized in that: The step of acquiring images at fixed time intervals further comprises: Take a picture every 1-5 seconds.
7. The image recognition method for PBX explosive water suspension granulation process according to claim 1 is characterized in that: The step of randomly extracting some images from the acquired multiple images to form an image data set further comprises: Randomly extracting some images from the plurality of images for labeling, and distinguishing the labels into particles and droplets; By changing the image resolution, the total number of images in the image dataset is expanded.
8. The image recognition method for PBX explosive water suspension granulation process according to claim 1 is characterized in that: The automatic identification and segmentation of particles and / or droplets in the image and obtaining target parameters further includes: By automatically identifying and segmenting particles and / or droplets in the image, an image with pixel-level masks and rectangular boxes of particles and / or droplets is output; Statistical analysis is performed on the pixel-level masks and rectangular boxes of particles and / or droplets in the image to obtain target parameters of the particles and / or droplets.
9. The image recognition method for PBX explosive water suspension granulation process according to claim 8, characterized in that: The target parameters include one or more of number, area, perimeter, granularity, and aspect ratio.
10. Application of the image recognition method for the PBX explosive water suspension granulation process according to claims 1 to 9 in the preparation process of PBX explosive.
Citation Information
Patent Citations
A nanoparticle size measurement method based on an improved Mask R-CNN
CN109948712A
Starch particle identification method based on Mask R-CNN
CN113887309A
Mask R-CNN mineral particle identification and particle size detection method based on improved mask
CN114897816A
Chemical fertilizer particle instance segmentation method based on improved Mask R-CNN
CN116777937A
new process for coating explosives
FR1469198A
