Dangerous chemical packaging quality prediction method and system based on deep learning

Abstract

The invention belongs to the technical field of chemical packaging supervision, and particularly relates to a dangerous chemical packaging quality prediction method and system based on deep learning. The system comprises a packaging material scanning detection module, a packaging material judgment analysis module, a sealing integrity analysis module, a man-machine collaborative decision-making module, a packaging management detection evaluation module and a packaging supervision terminal. The packaging material scanning and detecting module is used for scanning and detecting the surface of the packaging material of the dangerous chemical, and the packaging material judging and analyzing module and the sealing integrity analyzing module are used for analyzing the surface condition and the sealing integrity of the packaging material, so that the packaging quality of the dangerous chemical is effectively monitored and accurately judged; a powerful guarantee is provided for safe transportation and storage of the dangerous chemicals, reasonable improvement measures are taken in time according to the packaging management performance of the dangerous chemicals through accurate evaluation, and subsequent packaging quality and packaging stability and high efficiency can be guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of chemical packaging supervision, and specifically to a method and system for predicting the quality of hazardous chemical packaging based on deep learning. Background Art

[0002] Hazardous chemicals refer to highly toxic chemicals and other chemicals that are toxic, corrosive, explosive, flammable, or combustion-supporting, and are harmful to humans, facilities, and the environment. Since hazardous chemicals are flammable, explosive, toxic, and harmful, their packaging quality is directly related to their safe transportation and storage. Therefore, it is necessary to conduct a packaging quality assessment and analysis during the packaging process. Traditionally, hazardous chemical packaging quality testing relies primarily on manual inspection, which is inefficient and susceptible to human influence, leading to inaccurate test results. Currently, machine vision inspections struggle to adequately analyze hazardous chemical packaging quality, accurately assess hazardous chemical packaging management performance and on-site supervision, and provide timely warnings. This hinders ensuring packaging quality, stability, and efficiency, and hinders effective protection for the safe transportation and storage of hazardous chemicals. In view of the above technical defects, a solution is now proposed. Summary of the Invention

[0003] The purpose of the present invention is to provide a method and system for predicting the packaging quality of hazardous chemicals based on deep learning, which solves the problem that the existing technology is difficult to reasonably analyze the packaging quality of hazardous chemicals during machine vision inspection, and is unable to accurately evaluate the packaging management performance and on-site supervision status of hazardous chemicals and provide timely warnings, which is not conducive to ensuring packaging quality and packaging stability and efficiency, and is difficult to ensure the safe transportation and storage of hazardous chemicals.

[0004] To achieve the above object, the present invention provides the following technical solutions: A deep learning-based hazardous chemical packaging quality prediction system includes a packaging material scanning and detection module, a packaging material judgment and analysis module, a seal integrity analysis module, a human-machine collaborative decision-making module, a packaging management detection and evaluation module, and a packaging supervision terminal; The packaging material scanning and detection module scans and detects the surface of the packaging material of hazardous chemicals. The packaging material judgment and analysis module determines the surface condition of the packaging material through analysis and generates a material surface qualified signal or a material surface abnormality signal accordingly. The material surface abnormality signal is sent to the human-machine collaborative decision-making module, and the material surface qualified signal is sent to the sealing integrity analysis module. When the sealing integrity analysis module receives a qualified signal on the material surface, it identifies microcracks and sealing discontinuity areas based on the ultra-high-resolution local image of the sealing part of the packaging material. If microcracks or sealing discontinuity areas exist, a sealing abnormality signal is generated; otherwise, a sealing qualified signal is generated, and the sealing abnormality signal is sent to the human-machine collaborative decision-making module; the packaging management detection and evaluation module evaluates and analyzes the packaging management performance of hazardous chemicals in unit time, and generates a packaging management qualified signal or a packaging management alarm signal based on this, and sends the packaging management alarm signal to the packaging supervision end.

[0005] Furthermore, the specific analysis process of the packaging material judgment and analysis module is as follows: Based on the scanned image of the packaging material, its surface defects are obtained. If there are holes on the surface of the packaging material, a packaging material abnormality signal is generated. If there are no holes on the surface of the packaging material, scratches on the packaging material surface are obtained. The length, maximum width and maximum depth of the corresponding scratch are weighted and summed to calculate the scratch potential value. The scratch potential value is compared with a preset scratch potential threshold. If the scratch potential value exceeds the preset scratch potential threshold, the corresponding scratch is marked as a risk scratch. If there are risk scratches on the surface of the packaging material, a packaging material abnormality signal is generated. If there are no risk scratches on the surface of the packaging material, the concave and convex areas on the surface of the packaging material are obtained, and the area and concave and convex depth of the corresponding concave and convex areas are weightedly summed to calculate the concave and convex hidden danger value, and the concave and convex hidden danger value is numerically compared with the preset concave and convex hidden danger threshold. If the concave and convex hidden danger value exceeds the preset concave and convex hidden danger threshold, the corresponding concave and convex area is marked as a risky concave and convex area; if there is a risky concave and convex area on the surface of the packaging material, a packaging material abnormality signal is generated.

[0006] Furthermore, if there are no risky concave-convex areas on the surface of the packaging material, the surface of the packaging material is divided into a number of squares of equal area. If a corresponding square involves a surface defect that needs to be monitored, the corresponding square is marked as a non-safe square. The number of unsafe squares on the packaging material is obtained and its ratio with the total number of squares is calculated to obtain the square hidden danger value, and the square hidden danger value is numerically compared with the preset square hidden danger threshold. If the square hidden danger value exceeds the preset square hidden danger threshold, a packaging material abnormality signal is generated.

[0007] Furthermore, if the grid hidden danger value does not exceed the preset grid hidden danger threshold, the thickness of several positions on the packaging material is collected, the thickness of the corresponding position is marked as the position thickness detection value, the position thickness detection value of all positions is averaged to obtain the position thickness performance value, and the position where the position thickness detection value exceeds the preset position thickness detection threshold is marked as a trusted position, the ratio of the number of trusted positions on the packaging material to the total is obtained and marked as the trusted position measurement value, and the position thickness detection value with the smallest value is marked as the low position thickness detection value; The packaging material thickness risk coefficient is calculated by weighted summation of the bit thickness performance value, the trusted bit thickness measurement value and the bit thickness low detection value, and the packaging material thickness risk coefficient is numerically compared with the preset packaging material thickness risk coefficient threshold. If the packaging material thickness risk coefficient exceeds the preset packaging material thickness risk coefficient threshold, a packaging material qualified signal is generated; if the packaging material thickness risk coefficient does not exceed the preset packaging material thickness risk coefficient threshold, a packaging material abnormal signal is generated.

[0008] Furthermore, the specific analysis process of the packaging management detection and evaluation module is as follows: If the corresponding packaging material has abnormal surface signals or sealing signals, it will be marked as a packaging obstructor. The number of packaging obstructors per unit time is obtained and the ratio is calculated with the number of hazardous chemical packaging times per unit time to obtain a packaging obstruction detection value. The packaging obstruction detection value is then compared with a preset packaging obstruction detection threshold. If the packaging obstruction detection value exceeds the preset packaging obstruction detection threshold, a packaging management alarm signal is generated. If the packaging obstruction detection value does not exceed the preset packaging obstruction detection threshold, the packaging tube measurement value is obtained through analysis, and the packaging tube measurement value is compared with the preset packaging tube measurement threshold. If the packaging tube measurement value exceeds the preset packaging tube measurement threshold, a packaging management alarm signal is generated; if the packaging tube measurement value does not exceed the preset packaging tube measurement threshold, a packaging management qualified signal is generated.

[0009] Furthermore, the analysis and acquisition method of the packaging tube measurement value is as follows: When a material surface abnormality signal or a sealing abnormality signal of a corresponding packaging material is generated, the generation time of the corresponding signal is marked as the obstruction time, and the time when the next set of packaging materials is tested and analyzed is marked as the connection time. The time difference between the connection time and the obstruction time is calculated to obtain the obstruction response value; All the obstruction response values ​​within the unit time are obtained and their average is calculated to obtain the obstruction elimination value, and the obstruction response value is numerically compared with the preset obstruction response threshold. If the obstruction response value exceeds the preset obstruction response threshold, the corresponding obstruction response value is marked as an obstruction abnormal value, and the ratio of the number of obstruction abnormal values ​​and the number of obstruction response values ​​within the unit time is calculated to obtain the obstruction abnormality value; the packaging tube measurement value is obtained by weighted summation of the packaging obstruction detection value, the obstruction elimination value and the obstruction abnormality value.

[0010] Furthermore, the packaging management detection and evaluation module is communicatively connected to the packaging supervision and tracking module. The packaging management detection and evaluation module sends the packaging management alarm signal to the packaging supervision and tracking module. When the packaging supervision and tracking module receives the packaging management alarm signal, it analyzes the supervision performance of the hazardous chemical packaging and detection process within unit time, generates a packaging supervision qualified signal or a packaging supervision abnormal signal through analysis, and sends the packaging supervision abnormal signal to the packaging supervision end.

[0011] Furthermore, the specific analysis process of the packaging supervision and tracking module includes: The packaging and testing areas of hazardous chemicals are obtained and marked as target areas. The target areas are divided into several sub-areas. Surveillance images of each sub-area are obtained in real time through surveillance cameras. If there is no supervisor in the corresponding sub-area, the corresponding sub-area is judged to be in an unsupervised state. Obtain the total duration of the unsupervised state in the corresponding sub-area within the unit time and mark it as the unsupervised total time value, and obtain all single durations of the unsupervised state in the corresponding sub-area within the unit time, mark the number of occurrences of the single duration exceeding the preset single duration threshold as the unsupervised out-of-frequency value, and mark the single duration with the largest value as the unsupervised risk value; The unsupervised decision value is obtained by weighted summing up the unsupervised total time value, the unsupervised frequency difference value and the unsupervised risk value, and the unsupervised decision value is numerically compared with the preset unsupervised decision threshold. If the unsupervised decision value exceeds the preset unsupervised decision threshold, the corresponding sub-region is marked as a supervised neglect region; the number of supervised neglect regions is obtained and the ratio is calculated with the total number of sub-regions in the target region to obtain the supervised neglect detection value, and the supervised neglect detection value is numerically compared with the preset supervised neglect detection threshold. If the supervised neglect detection value exceeds the preset supervised neglect detection threshold, a packaging supervision abnormality signal is generated.

[0012] Furthermore, if the supervision neglect detection value does not exceed the preset supervision neglect detection threshold, the unsupervised decision values ​​of all sub-areas in unit time are averaged to obtain the packaging supervision evaluation value, and the packaging supervision evaluation value is numerically compared with the preset packaging supervision evaluation threshold. If the packaging supervision evaluation value exceeds the preset area supervision evaluation threshold, a packaging supervision abnormality signal is generated; if the packaging supervision evaluation value does not exceed the preset area supervision evaluation threshold, a packaging supervision qualified signal is generated.

[0013] Furthermore, the present invention also proposes a method for predicting the quality of hazardous chemical packaging based on deep learning, comprising the following steps: Step 1: Scan and inspect the surface of the packaging materials of hazardous chemicals; Step 2: Determine the surface condition of the packaging material; Step 3: Determine the sealing integrity of the sealing parts of the packaging material; Step 4: Evaluate packaging management performance for hazardous chemicals.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1. In the present invention, the packaging material scanning and detection module scans and detects the surface of the packaging materials of hazardous chemicals. The packaging material judgment and analysis module and the seal integrity analysis module analyze the surface condition and seal integrity of the packaging materials, providing strong guarantees for the safe transportation and storage of hazardous chemicals. By accurately evaluating the packaging management performance of hazardous chemicals and taking reasonable improvement measures in a timely manner, it is conducive to ensuring the subsequent packaging quality and packaging stability and efficiency. 2. In the present invention, the packaging management alarm signal is sent to the packaging supervision and tracking module through the packaging management detection and evaluation module. When the packaging supervision and tracking module receives the packaging management alarm signal, it analyzes the supervision performance of the hazardous chemical packaging and detection process within a unit time, and increases the supervision of the hazardous chemical packaging and detection process when a packaging supervision abnormality signal is generated, further ensuring the packaging quality of hazardous chemicals, with a high level of intelligence. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings; Figure 1 This is a system block diagram of Embodiment 1 of the present invention; Figure 2 This is a system block diagram of Embodiment 2 of the present invention; Figure 3 This is a flow chart of the method of embodiment 3 of the present invention. DETAILED DESCRIPTION

[0016] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0017] Example 1: Figure 1 As shown, the present invention proposes a dangerous chemical packaging quality prediction system based on deep learning, which includes a packaging material scanning and detection module, a packaging material judgment and analysis module, a sealing integrity analysis module, a human-machine collaborative decision module, a packaging management detection and evaluation module, and a packaging supervision terminal; The packaging material scanning and detection module scans and detects the surface of hazardous chemical packaging materials (mainly used to collect scanned images and thickness data). The packaging material judgment and analysis module determines the surface condition of the packaging materials through analysis, and generates a qualified material surface signal or a material surface abnormality signal based on the analysis. The material surface abnormality signal is then sent to the human-machine collaborative decision-making module. The human-machine collaborative decision-making module displays the sealing abnormality signal and issues a prompt to remind on-site personnel to handle the corresponding packaging materials to ensure the packaging quality of hazardous chemicals. The specific analysis process of the packaging material judgment and analysis module is as follows: Surface defects of the packaging material are obtained based on a scanned image of the packaging material. If holes are present on the surface of the packaging material, indicating that the surface condition of the packaging material is poor, a packaging material abnormality signal is generated. If no holes are present on the surface of the packaging material, scratches on the packaging material surface are obtained, and a scratch hidden danger value is calculated by weighted summing the length, maximum width, and maximum depth of the corresponding scratch. Specifically, a corresponding preset weight coefficient is assigned to the length, maximum width, and maximum depth of the scratch, and the length, maximum width, and maximum depth of the scratch are multiplied by the corresponding preset weight coefficients, and the sum of the three product results is marked as the scratch hidden danger value. It should be noted that the larger the scratch risk value, the higher the overall packaging risk brought by the corresponding scratch. The scratch risk value is compared with the preset scratch risk threshold. If the scratch risk value exceeds the preset scratch risk threshold, it indicates that the overall packaging risk brought by the corresponding scratch is high, and the corresponding scratch is marked as a risk scratch. If there is a risk scratch on the surface of the packaging material, it indicates that the surface condition of the corresponding packaging material is poor, and a packaging material abnormality signal is generated. If there are no risk scratches on the surface of the packaging material, the concave and convex areas on the packaging material surface are obtained, and the area and depth of the corresponding concave and convex areas are weighted and summed to calculate the concave and convex risk value. That is, the area and depth of the concave and convex areas are respectively assigned corresponding preset weight coefficients, and the area and depth of the concave and convex areas are respectively multiplied by the corresponding preset weight coefficients. The sum of the two sets of product results is marked as the concave and convex risk value. It should be noted that the larger the concave and convex risk value, the higher the overall packaging risk brought by the corresponding concave and convex area. The concave-convex hidden danger value is numerically compared with the preset concave-convex hidden danger threshold. If the concave-convex hidden danger value exceeds the preset concave-convex hidden danger threshold, it indicates that the packaging risk brought by the corresponding concave-convex area is generally high, and the corresponding concave-convex area is marked as a risky concave-convex area; if there is a risky concave-convex area on the surface of the packaging material, it indicates that the surface condition of the corresponding packaging material is poor, and a packaging material abnormality signal is generated.

[0018] And, if the surface of the packaging material does not have the risk uneven area, the surface of the packaging material is divided into a plurality of equal-area squares, if the corresponding square is related to the surface defect required to be monitored, the corresponding square is marked as a non-safe square; the number of non-safe squares on the packaging material is obtained and a ratio calculation is performed between the number of non-safe squares and the total number of squares to obtain a square hidden danger value, the square hidden danger value is compared with a preset square hidden danger threshold value, if the square hidden danger value exceeds the preset square hidden danger threshold value, it indicates that the surface condition of the corresponding packaging material is poor, and a packaging material abnormal signal is generated.

[0019] Further, if the square hidden danger value does not exceed the preset square hidden danger threshold value, the thickness of a plurality of positions on the packaging material is collected, the thickness of the corresponding position is marked as a position thickness detection value, the position thickness detection values of all positions are averaged to obtain a position thickness performance value, and the positions whose position thickness detection values exceed a preset position thickness detection threshold value are marked as trusted positions, the number of trusted positions on the packaging material is obtained and is marked as a trusted position measurement value, and the minimum position thickness detection value on the corresponding packaging material is marked as a position thickness low detection value; The packaging material thickness risk coefficient is calculated by weighted summation of the position thickness performance value, the trusted position measurement value and the position thickness low detection value; that is, the position thickness performance value, the trusted position measurement value and the position thickness low detection value are respectively assigned with corresponding preset weight coefficients, and the position thickness performance value, the trusted position measurement value and the position thickness low detection value are respectively multiplied by the corresponding preset weight coefficients, and the sum of the three groups of product results is marked as the packaging material thickness risk coefficient; it should be noted that the smaller the value of the packaging material thickness risk coefficient, the worse the thickness condition of the corresponding packaging material, and the more unfavorable it is to ensure the packaging quality of the corresponding packaging material; The packaging material thickness risk coefficient is compared with a preset packaging material thickness risk coefficient threshold value, if the packaging material thickness risk coefficient exceeds the preset packaging material thickness risk coefficient threshold value, it indicates that the surface condition of the corresponding packaging material is generally good, and a packaging material qualified signal is generated; if the packaging material thickness risk coefficient does not exceed the preset packaging material thickness risk coefficient threshold value, it indicates that the thickness condition of the corresponding packaging material is not good and the surface condition is poor, and a packaging material abnormal signal is generated.

[0020] The packaging material scanning detection module sends the material surface qualified signal to the sealing integrity analysis module, the sealing integrity analysis module receives the material surface qualified signal, recognizes micro-cracks and sealing discontinuous areas based on the ultra-high resolution local image of the sealing part of the packaging material, if there are micro-cracks or sealing discontinuous areas, a sealing abnormal signal is generated, if there are no micro-cracks and sealing discontinuous areas, a sealing qualified signal is generated, and the sealing abnormal signal is sent to the man-machine collaborative decision-making module, the man-machine collaborative decision-making module displays the sealing abnormal signal and issues a prompt to remind the on-site personnel to process the corresponding packaging material, and further ensure the packaging quality of dangerous chemicals.

[0021] The packaging management detection and evaluation module evaluates and analyzes the packaging management performance of hazardous chemicals within a unit of time, generates a packaging management qualified signal or a packaging management alarm signal based on the evaluation, and sends the packaging management alarm signal to the packaging supervision end. Upon receiving the packaging management alarm signal, the packaging supervision end issues a corresponding warning to remind back-end personnel to conduct timely investigation and analysis of the cause and take reasonable improvement measures to ensure subsequent packaging quality and packaging stability and efficiency. The specific analysis process of the packaging management detection and evaluation module is as follows: If the corresponding packaging material has abnormal surface signals or sealing signals, it will be marked as a packaging obstructor. The number of packaging obstructors per unit time is obtained and the ratio is calculated with the number of hazardous chemical packaging times per unit time to obtain the packaging obstruction detection value. The packaging obstruction detection value is then compared with the preset packaging obstruction detection threshold. If the packaging obstruction detection value exceeds the preset packaging obstruction detection threshold, it indicates that the packaging management performance of hazardous chemicals per unit time is poor and corresponding optimization and improvement measures need to be taken in a timely manner, and a packaging management alarm signal is generated. If the package obstruction detection value does not exceed the preset package obstruction detection threshold, then when a material surface abnormality signal or a seal abnormality signal is generated for the corresponding packaging material, the time at which the corresponding signal is generated is marked as the obstruction time, and the time at which the next set of packaging material is tested and analyzed is marked as the connection time. The time difference between the connection time and the obstruction time is calculated to obtain an obstruction response value. The larger the value of the obstruction response value, the slower the processing efficiency for the corresponding package obstructor. Obtain all the obstruction response values ​​within a unit time and calculate their average to obtain an obstruction elimination value, and compare the obstruction response value with a preset obstruction response threshold. If the obstruction response value exceeds the preset obstruction response threshold, mark the corresponding obstruction response value as an obstruction abnormal value, and calculate the ratio of the number of obstruction abnormal values ​​to the number of obstruction response values ​​within a unit time to obtain an obstruction abnormal value; The packaging tube measurement value is calculated by weighted summing the packaging obstruction detection value, obstruction elimination value, and obstruction abnormality value. Specifically, the packaging obstruction detection value, obstruction elimination value, and obstruction abnormality value are each assigned a corresponding preset weight coefficient, and the packaging obstruction detection value, obstruction elimination value, and obstruction abnormality value are each multiplied by the corresponding preset weight coefficient. The sum of the three product results is then marked as the packaging tube measurement value. It should be noted that the larger the value of the packaging tube measurement value, the worse the overall performance of hazardous chemical packaging management per unit time. The packaging tube measurement value is compared with the preset packaging tube measurement threshold. If the packaging tube measurement value exceeds the preset packaging tube measurement threshold, it indicates that the packaging management performance of hazardous chemicals in unit time is generally poor, and a packaging management alarm signal is generated; if the packaging tube measurement value does not exceed the preset packaging tube measurement threshold, it indicates that the packaging management performance of hazardous chemicals in unit time is generally good, and corresponding optimization and improvement measures need to be taken in a timely manner, and a packaging management qualified signal is generated.

[0022] Example 2: Figure 2 As shown, the difference between this embodiment and the first embodiment is that the packaging management detection and evaluation module is communicatively connected to the packaging supervision and tracking module. The packaging management detection and evaluation module sends a packaging management alarm signal to the packaging supervision and tracking module. When the packaging supervision and tracking module receives the packaging management alarm signal, it analyzes the supervision performance of the hazardous chemical packaging and detection process within a unit time. The module generates a qualified or abnormal packaging supervision signal through analysis and sends it to the packaging supervision terminal. Upon receiving the abnormal packaging supervision signal, the terminal issues a corresponding warning to remind back-end personnel to optimize subsequent on-site supervision and strengthen supervision of the hazardous chemical packaging and testing process, thereby further ensuring the packaging quality of hazardous chemicals. The specific analysis process of the packaging supervision tracking module is as follows: The packaging and testing areas of hazardous chemicals are obtained and marked as target areas. The target areas are divided into several sub-areas. Surveillance images of each sub-area are obtained in real time through surveillance cameras. If there is no supervisor in the corresponding sub-area, the corresponding sub-area is judged to be in an unsupervised state. Obtain the total duration of the unsupervised state in the corresponding sub-area within the unit time and mark it as the unsupervised total time value, and obtain all single durations of the unsupervised state in the corresponding sub-area within the unit time, mark the number of occurrences of the single duration exceeding the preset single duration threshold as the unsupervised out-of-frequency value, and mark the single duration with the largest value as the unsupervised risk value; The unsupervised decision value is calculated by weighted summing the unsupervised total time value, the unsupervised frequency difference value, and the unsupervised risk value; that is, the unsupervised total time value, the unsupervised frequency difference value, and the unsupervised risk value are respectively assigned corresponding preset weight coefficients, and the unsupervised total time value, the unsupervised frequency difference value, and the unsupervised risk value are respectively multiplied by the corresponding preset weight coefficients, and the sum of the three sets of product results is marked as the unsupervised decision value; it should be noted that the larger the value of the unsupervised decision value, the worse the overall supervision status of the corresponding sub-area in unit time; The unsupervised decision value is numerically compared with the preset unsupervised decision threshold. If the unsupervised decision value exceeds the preset unsupervised decision threshold, it indicates that the overall supervision status of the corresponding sub-area in unit time is poor, and the corresponding sub-area is marked as a supervision neglect area; the number of supervision neglect areas is obtained and the ratio is calculated with the total number of sub-areas in the target area to obtain the supervision neglect detection value, and the supervision neglect detection value is numerically compared with the preset supervision neglect detection threshold. If the supervision neglect detection value exceeds the preset supervision neglect detection threshold, it indicates that the supervision status of the packaging and testing process of hazardous chemicals in unit time is poor, and a packaging supervision abnormality signal is generated.

[0023] Furthermore, if the supervision neglect detection value does not exceed the preset supervision neglect detection threshold, the unsupervised decision values ​​of all sub-areas in unit time are averaged to obtain the packaging supervision evaluation value, and the packaging supervision evaluation value is numerically compared with the preset packaging supervision evaluation threshold. If the packaging supervision evaluation value exceeds the preset regional supervision evaluation threshold, it indicates that the supervision status of the packaging and testing process of hazardous chemicals in unit time is not good, and a packaging supervision abnormality signal is generated; if the packaging supervision evaluation value does not exceed the preset regional supervision evaluation threshold, it indicates that the supervision status of the packaging and testing process of hazardous chemicals in unit time is good, and a packaging supervision qualified signal is generated.

[0024] Example 3: Figure 3 As shown, the difference between this embodiment and the first and second embodiments is that the present invention proposes a method for predicting the quality of hazardous chemical packaging based on deep learning, comprising the following steps: Step 1: Scan and inspect the surface of the packaging materials of hazardous chemicals; Step 2: Determine the surface condition of the packaging material; Step 3: Determine the sealing integrity of the sealing part of the packaging material; Step 4: Evaluate packaging management performance for hazardous chemicals.

[0025] The working principle of the present invention is as follows: when in use, the packaging material surface of the hazardous chemicals is scanned and detected by the packaging material scanning and detection module, the packaging material judgment and analysis module judges the surface condition of the packaging material through analysis, and when a qualified material surface signal is generated, the sealing integrity analysis module judges the sealing integrity of the sealing part of the packaging material, thereby realizing effective monitoring and accurate judgment of the packaging quality of hazardous chemicals, providing strong guarantees for the safe transportation and storage of hazardous chemicals, and evaluating and analyzing the packaging management performance of hazardous chemicals per unit time through the packaging management detection and evaluation module, and reminding the background personnel to conduct cause investigation and analysis in time and make reasonable improvement measures when a packaging management alarm signal is generated, which is conducive to ensuring the subsequent packaging quality and packaging stability and efficiency, and has a high level of intelligence.

[0026] The threshold or the preset value, the preset range and the like in the technical solution of the present application are set for result comparison and analysis, so as to determine whether it is good or not. The size of the threshold or the preset value, the preset range and the like is set by combining the big model analysis of sample data and artificial experience, and is also adjusted appropriately according to the seasonal or common influence conditions. The preset weight coefficient and the influence factor are set according to the influence of each parameter on the result, and the specific value is allocated to finally reflect the influence of the result. The preset weight coefficient and the influence factor are also set by combining the big model analysis of sample data and artificial experience, and are also adjusted appropriately according to the seasonal or common influence conditions.

[0027] The preferred embodiments of the present application disclosed above are only used to help explain the present application, and the preferred embodiments do not describe all the details and do not limit the present application to the specific embodiments. Obviously, many modifications and changes can be made according to the content of the present application. The present application selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present application, so that the skilled in the art can well understand and use the present application. The present application is limited by the claims and the entire scope and equivalents thereof.

Claims

1. A dangerous chemical packaging quality prediction system based on deep learning, characterized by: It includes a packaging material scanning and detection module, a packaging material judgment and analysis module, a sealing integrity analysis module, a human-machine collaborative decision-making module, a packaging management detection and evaluation module, and a packaging supervision terminal; the packaging material scanning and detection module scans and detects the surface of the packaging material of hazardous chemicals, the packaging material judgment and analysis module determines the surface condition of the packaging material through analysis, and sends abnormal surface material signals to the human-machine collaborative decision-making module, and sends qualified surface material signals to the sealing integrity analysis module; When the sealing integrity analysis module receives a qualified signal on the material surface, it identifies microcracks and sealing discontinuity areas based on the ultra-high-resolution local image of the sealing part of the packaging material. If microcracks or sealing discontinuity areas exist, a sealing abnormality signal is generated; otherwise, a sealing qualified signal is generated, and the sealing abnormality signal is sent to the human-machine collaborative decision-making module; the packaging management detection and evaluation module evaluates and analyzes the packaging management performance of hazardous chemicals in unit time, and generates a packaging management qualified signal or a packaging management alarm signal based on this, and sends the packaging management alarm signal to the packaging supervision end.

2. The system for predicting the quality of hazardous chemical packaging based on deep learning according to claim 1 is characterized in that: The specific analysis process of the packaging material judgment and analysis module is as follows: Surface defects of the packaging material are detected based on the scanned image. If holes are present on the surface of the packaging material, a packaging material abnormality signal is generated. If holes are not present on the surface of the packaging material, the corresponding scratch is marked as a risk scratch if the scratch potential value exceeds a preset scratch potential threshold. If there are risk scratches on the surface of the packaging material, a packaging material abnormality signal is generated; if there are no risk scratches on the surface of the packaging material, the corresponding concave-convex area is marked as a risk concave-convex area when the concave-convex risk value exceeds the preset concave-convex risk threshold; If there are risky concave-convex areas on the surface of the packaging material, a packaging material abnormality signal is generated.

3. The system for predicting the quality of hazardous chemical packaging based on deep learning according to claim 2 is characterized in that: If there are no risky concave-convex areas on the surface of the packaging material, the number of unsafe squares on the packaging material is obtained and the ratio is calculated with the total number of squares to obtain the square hidden danger value. If the square hidden danger value exceeds the preset square hidden danger threshold, a packaging material abnormality signal is generated.

4. The system for predicting the quality of hazardous chemical packaging based on deep learning according to claim 3 is characterized in that: If the grid hidden danger value does not exceed the preset grid hidden danger threshold, the packaging material thickness risk coefficient is calculated by weighted summing the bit thickness performance value, the trusted bit measurement value, and the bit thickness low detection value. If the packaging material thickness risk coefficient exceeds the preset packaging material thickness risk coefficient threshold, a packaging material qualified signal is generated; Otherwise, a packaging material abnormality signal is generated.

5. The system for predicting the quality of hazardous chemical packaging based on deep learning according to claim 1 is characterized in that: The specific analysis process of the packaging management detection and evaluation module is as follows: The number of package obstructors per unit time is obtained and the ratio thereof with the number of packages of hazardous chemicals per unit time is calculated to obtain a package obstruction detection value. If the package obstruction detection value exceeds a preset package obstruction detection threshold, a package management alarm signal is generated. If the package obstruction detection value does not exceed the preset package obstruction detection threshold, the package tube measurement value is obtained through analysis. If the package tube measurement value exceeds the preset package tube measurement threshold, a package management alarm signal is generated; otherwise, a package management qualified signal is generated.

6. The system for predicting the quality of hazardous chemical packaging based on deep learning according to claim 5 is characterized in that: The specific method for analyzing and obtaining the packaging tube measurement value is as follows: all the obstruction response values ​​within a unit time are obtained and their average is calculated to obtain the obstruction elimination table value, and the number of obstruction abnormality values ​​per unit time is calculated by ratio calculation to the number of obstruction response values ​​to obtain the obstruction abnormality value; the packaging tube measurement value is obtained by weighted summation of the packaging obstruction detection value, the obstruction elimination table value and the obstruction abnormality value.

7. The system for predicting the quality of hazardous chemical packaging based on deep learning according to claim 5 is characterized in that: The packaging management detection and evaluation module is communicatively connected to the packaging supervision and tracking module. The packaging management detection and evaluation module sends the packaging management alarm signal to the packaging supervision and tracking module. When the packaging supervision and tracking module receives the packaging management alarm signal, it generates a packaging supervision qualified signal or a packaging supervision abnormal signal through analysis, and sends the packaging supervision abnormal signal to the packaging supervision end.

8. The system for predicting the quality of hazardous chemical packaging based on deep learning according to claim 7 is characterized in that: The specific analysis process of the packaging supervision and tracking module includes: The unsupervised decision value is calculated by weighted summing up the unsupervised total time value, the unsupervised frequency difference value and the unsupervised risk value. If the unsupervised decision value exceeds the preset unsupervised decision threshold, the corresponding sub-region is marked as a supervised neglect region. The number of supervised neglect regions is obtained and the ratio is calculated with the total number of sub-regions in the target region to obtain the supervised neglect detection value. If the supervised neglect detection value exceeds the preset supervised neglect detection threshold, a packaging supervision abnormality signal is generated.

9. The system for predicting the quality of hazardous chemical packaging based on deep learning according to claim 8, characterized in that: If the supervision neglect detection value does not exceed the preset supervision neglect detection threshold, the unsupervised decision values ​​of all sub-areas in unit time are averaged to obtain the packaging supervision evaluation value. If the packaging supervision evaluation value exceeds the preset regional supervision evaluation threshold, a packaging supervision abnormality signal is generated; otherwise, a packaging supervision qualified signal is generated.

10. A method for predicting the quality of hazardous chemical packaging based on deep learning, characterized in that: The following steps are involved: Step 1: Scan and inspect the surface of hazardous chemical packaging materials; Step 2: Determine the surface condition of the packaging materials; Step 3: Determine the sealing integrity of the sealing parts of the packaging materials; Step 4: Evaluate the packaging management performance for hazardous chemicals.

Images