A Big Data-Based Automated Management Method and System for Carbon Molecular Sieves Production

Abstract

This invention relates to the field of data management technology, specifically to a method and system for automated management of carbon molecular sieve production based on big data. The method includes the following steps: recording the adsorbed gas trajectory, dividing the path segments and marking directional changes, extracting particle configuration to identify interference fragments, extracting contours to track structural deformation, cross-locating process operations to identify overlapping areas, and sorting and reconstructing actions to generate an automatic control instruction structure table. In this invention, by combining particle boundary continuity and path overlap features to extract interference regions, the fine-grained expression of structural identification is enhanced. By utilizing structural jump trends and connecting lines to track and aggregate change trajectories, dynamic perception of the structural state is achieved. The structural strain region and operational behavior are cross-mapped in time and space to clarify process intervention targets. By combining operation sequence analysis and control chain reorganization, the execution logic and control response are optimized, constructing a multi-dimensional, interconnected automatic control strategy.

Description

Technical Field

[0001] This invention relates to the field of data management technology, and in particular to an automated management method and system for carbon molecular sieve production based on big data. Background Technology

[0002] The field of automatic control technology encompasses methods and technologies for the automated management and regulation of various industrial production processes, machinery operation, energy distribution and conversion, and environmental control systems. The core of this technology lies in using sensing, control, and feedback mechanisms to automatically execute preset commands onto actual objects, reducing human intervention and achieving efficient system operation. In industrial production, automatic control is widely used in manufacturing process control, automated machine operation, production line collaborative management, and status monitoring, often relying on the coordinated work of sensors, controllers, actuators, and control logic. Especially in complex processes, automatic control technology, by combining information acquisition and logical judgment, completes the construction of closed-loop control systems, providing fundamental support for intelligent management in modern factories.

[0003] Among them, the big data-based automated management method for carbon molecular sieve production refers to the collection, storage, and analysis of multi-dimensional data on key parameters such as raw material ratio control, carbonization temperature adjustment, and adsorption performance monitoring during the carbon molecular sieve production process. By constructing a historical data comparison model and an association rule analysis mechanism, production control parameters are automatically pushed to the execution system, and real-time parameter corrections and process scheduling are performed according to process requirements. This method constructs a multi-channel data stream through data acquisition nodes, deploys a decision logic engine at the production scheduling end, controls process conditions at each stage through rule matching and pattern recognition, and achieves information interaction and unified control between various devices through network transmission.

[0004] Current technologies for automated control of carbon molecular sieves lack a systematic classification of the adsorption behavior pathways and a clear labeling of directional change trends, making it difficult to accurately reveal the dynamic mapping relationship between microscopic disturbances and structures during adsorption. Path behavior is often presented as continuous parameters, failing to extract the inflection point types and intervention characteristics of directional abrupt changes. This makes it difficult to identify key disturbance regions within a short timeframe, affecting the real-time performance and targeted nature of the process response. In particle structure identification, relevant models focus on macroscopic trend analysis of the overall arrangement, lacking microscale processing of particle boundary deformation and connection extension characteristics, limiting the ability to express local changes in complex configurations. Furthermore, existing control strategies are generally based on the execution timing design of independent instructions, failing to interactively locate operational behavior with structural strain states. This leads to frequent overlapping operational segments in process control, making it difficult to effectively avoid execution conflicts caused by path interference. For example, executing multiple operations in areas with high structural disturbance incidence, without coupled analysis of the region and timing, can easily induce control lag and operational failures. The lack of ability to analyze operational conflict points and reconstruct the chain also makes the control output relatively rigid, making it difficult to adapt to the needs of multi-point coordination and dynamic adjustment of execution under the background of path interference, thus weakening the operational stability and control efficiency of the system in complex production scenarios. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and propose an automated management method and system for carbon molecular sieve production based on big data.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: an automated management method for carbon molecular sieve production based on big data, comprising the following steps:

[0007] S1: Record the continuous path trajectory of the gas adsorbed in the carbon molecular sieve, divide the path segments according to the change of airflow direction, identify abrupt changes in direction and screen key inflection points, number and classify the changing trends, and form a set of adsorption path markers.

[0008] S2: Call the adsorption path marker set, extract the particle configuration image of the corresponding region, analyze the continuity of the particle boundary, identify the interference area of ​​the trajectory and connection structure, and generate interference fragment group;

[0009] S3: Based on the spatial location data of the interference fragments, extract the surface arrangement contour, determine the horizontal and vertical jump trend of the boundary, track the connection line of continuous deformation between structures, analyze the consistency of the extension direction, and output the structural strain processing template sequence.

[0010] S4: Call the structural strain treatment template sequence to cross-locate the operation records in the carbon molecular sieve process flow, retrieve the thermal control, orientation and cooling operation segments, identify and classify the spatiotemporal overlapping parts, and form a process intervention window list.

[0011] S5: Based on the operation position and instruction content index in the process intervention window list, analyze and sort the action sequence of overlapping sections, identify operation conflict points and disassemble and reconstruct them to generate an automatic control instruction structure table for carbon molecular sieves.

[0012] As a further embodiment of the present invention, the adsorption path marker set includes path number information, direction change inflection points, path segment label classification, and offset overlap position characteristics; the interference segment group includes boundary continuity characteristics, particle connection configuration, path penetration region type, and structural interference differences; the structural strain treatment template sequence includes jump trend direction, structural contour morphology, connecting line change trajectory, and extension consistency characteristics; the process intervention window list includes execution area number, operation time index, process action type, and overlapping segment classification label; and the carbon molecular sieve automatic control instruction structure table includes operation sequence structure, action conflict decomposition group, execution chain combination form, and control instruction classification results.

[0013] As a further aspect of the present invention, the specific steps of S1 are as follows:

[0014] S101: Records the behavior trajectory of the gas adsorbed in the carbon molecular sieve. Based on the coordinate sequence and direction change data of the gas in the spatial node per unit time, it calls the coordinate difference and direction angle of adjacent time points for comparison, divides the direction change trend into multiple continuous path segments, and generates the path direction change interval value.

[0015] S102: Based on the path direction change interval value, obtain the spatial offset angle of continuous points within the path segment, call the abrupt change amplitude of the angle between adjacent points and compare it with the direction abrupt change benchmark value, filter the path segments that exceed the benchmark, and extract the node positions with offset angles higher than the average value to generate a path direction offset abrupt change point set.

[0016] S103: Based on the set of abrupt shifts in the path direction, call the adsorption structure position data corresponding to the node in the carbon molecular sieve, calculate the spatial overlap rate with the structure boundary, screen the abrupt shift segments with an overlap rate exceeding the overlap reference rate, number and classify the directional change trend, and obtain the adsorption path marker set.

[0017] As a further aspect of the present invention, the formula for calculating the spatial overlap rate with the structural boundary is specifically as follows:

[0018] ;

[0019] Where, η cr P represents the spatial overlap rate between the path direction of the mutation point and the structural boundary, N represents the total number of valid mutation points in the mutation point set, and P represents the spatial overlap rate between the path direction of the mutation point and the structural boundary. k b represents the three-dimensional coordinate vector of the k-th mutation point in the carbon molecular sieve.k P represents k The corresponding nearest structural boundary point coordinate vector, P k― B k ‖ represents the Euclidean distance between the two, V k n represents the path direction offset vector at the k-th mutation point. k Represents structural boundary point b k The unit normal vector at point α d The smoothing correction coefficient represents the magnitude of the path direction vector.

[0020] As a further aspect of the present invention, the specific steps of S2 are as follows:

[0021] S201: Call the position index of the identified path in the adsorption path marker set, locate the area covered by the path segment, obtain the two-dimensional grayscale data of the corresponding particle arrangement image, extract the particle edge intensity gradient in the image area, identify particles based on edge closure, and generate configuration boundary identification value.

[0022] S202: Based on the configuration boundary identification value, extract the connection segment between adjacent particle configuration boundaries, call the boundary shape change degree and boundary direction angle value at both ends of the connection segment, determine whether the boundary continuity is lower than the set connection continuity benchmark value, mark the path penetration area position of the corresponding connection segment, and obtain the interference connection area intensity value.

[0023] S203: Based on the intensity value of the interference connection region, call the path segment direction change value and the connection segment configuration arrangement density value, filter the intersection region where the direction change amplitude exceeds the trajectory offset change benchmark value and the arrangement density difference is higher than the configuration density difference threshold, divide the intersection region into structural difference segments, and obtain the interference segment group.

[0024] As a further aspect of the present invention, the specific steps of S3 are as follows:

[0025] S301: Call the spatial location data corresponding to the interference fragment group, extract the arrangement image of surface particles in the fragment region, obtain the grayscale contour edge pixel point set, fit the boundary according to the edge gradient intensity and closure, identify the outer contour shape of the particle arrangement region, and generate the surface contour boundary value.

[0026] S302: Based on the surface contour boundary value, obtain the coordinate sequence of continuous boundary points in the contour, calculate the extension jump amplitude of adjacent boundary segments in the horizontal and vertical directions respectively, determine whether the jump amplitude exceeds the horizontal jump reference value and the vertical jump reference value, filter the connecting lines corresponding to the continuous jump positions, and obtain the structural jump trend coefficient.

[0027] S303: Based on the structural jump trend coefficient, extract the position data of the connecting line segments that have undergone deformation, calculate the cosine value of the angle between the line segment and the direction, determine whether there is a directional consistency shift in the structural extension direction among multiple connecting segments, classify the path change sequence according to the jump range, and establish a structural strain treatment template sequence.

[0028] As a further aspect of the present invention, the formula for calculating the cosine value of the angle between the line segment and the direction is as follows:

[0029] ;

[0030] Where cos(β) represents the cosine of the angle between the line segment and the direction, m represents the estimated slope of the current connecting line segment on the two-dimensional projection plane, n represents the slope direction constant of the reference direction in the same projection plane, q represents the magnitude of the displacement component of the connecting line segment perpendicular to the projection plane, and |m·n| represents the absolute value of the slope product. This represents the combined directional difference between the direction of the connecting line segment and the reference direction. represents the sum of the magnitudes of the directional and perpendicular components of the connecting line segment, and |n| represents the absolute slope of the reference direction.

[0031] As a further aspect of the present invention, the specific steps of S4 are as follows:

[0032] S401: Call the regions listed in the structural strain treatment template sequence, obtain the process action index corresponding to the region in the carbon molecular sieve operation process, locate the execution time record of each process action, and cross-compare the time period with the corresponding spatial location data to generate the process action location interval value.

[0033] S402: Based on the process action positioning interval value, extract the execution segment information corresponding to the three types of actions: thermal control, orientation, and cooling. Compare the start and end times of the action segments with the spatial coordinates to determine whether there is a situation where the time and area coordinates of the three types of actions coincide. Filter out all overlapping segments and obtain the process action overlap ratio value.

[0034] S403: Based on the process action overlap ratio value, classify the segments whose overlap ratio exceeds the action overlap benchmark ratio, extract the operation position index covered, and combine the corresponding segment's change trend amplitude value to screen spatial positions with predictable operation adjustment, and establish a process intervention window list.

[0035] As a further aspect of the present invention, the specific steps of S5 are as follows:

[0036] S501: Call the operation position and instruction content index listed in the process intervention window list, extract the action type, start time, end time and path number corresponding to the overlapping execution segment, sort the operation actions under the same path in ascending order according to the action start time, obtain the order of process actions in the path segment, and generate the path operation sorting value.

[0037] S502: Based on the path operation sorting value, extract the overlapping segments of operation positions between differentiated paths, compare the action number and execution type within the same time period, determine whether there is a mutual exclusion logic conflict between actions, if there is no minimum interval time between adjacent action numbers, mark the segment as a conflict point, and classify and integrate the instruction positions corresponding to the conflict points to obtain the process action conflict position value.

[0038] S503: Based on the conflict position value of the process action, decompose all conflicting segments, rearrange the non-conflicting operation sequence, and connect the decomposed actions in time priority order into a continuous instruction sequence. Recombine and merge the operation chain of the path to establish an automatic control instruction structure table for carbon molecular sieve.

[0039] An automated management system for carbon molecular sieve production based on big data includes:

[0040] The adsorption trajectory recognition module acquires the gas flow path data of the adsorption region inside the carbon molecular sieve, calls the coordinate sequence and direction change value of the trajectory point, calculates the angle difference between adjacent points and divides the path segment, determines the direction change point of each segment and locates the start and end coordinates, compares the positional relationship between the coordinates and the adsorption structure space diagram, and generates an adsorption path marker set.

[0041] The interference fragment extraction module calls the path segment position in the adsorption path marker set, extracts the particle image of the corresponding region and detects the angle between adjacent edge lines, determines whether the angle change exceeds the continuous range, analyzes the angle between the path direction and the boundary extension, filters out the connected discontinuous particle combination regions, and generates interference fragment groups.

[0042] The strain trend analysis module calls the boundary coordinates in the interference segment group, extracts the contour line and calculates the horizontal and vertical change trends, identifies continuous offset positions and tracks the direction of the connecting line, collects the changing segments with the same direction, and generates a structural strain processing template sequence.

[0043] The process action positioning module calls the number in the structural strain treatment template sequence, extracts the time information of thermal control, orientation and cooling actions, determines whether there is time overlap under the same number, filters the adjustable time period and organizes it according to action type, and generates a process intervention window list.

[0044] The control chain generation module calls the action segments in the process intervention window list, extracts the number, action type and time data, breaks down the overlapping time periods under the same number, rearranges the action sequence and combines them into a complete execution chain, and generates the carbon molecular sieve automatic control instruction structure table.

[0045] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0046] In this invention, interference regions are extracted by combining particle boundary continuity and path overlap features, thereby enhancing the fine-grained expression of structure recognition. By utilizing structural jump trends and connecting lines to track and collect change trajectories, dynamic perception of structural state is achieved. The structural strain region and operational behavior are cross-mapped in time and space to clarify the process intervention target. By combining operation sequence analysis and control chain reorganization, the execution logic and control response are optimized, and a multi-dimensional linkage automatic control strategy is constructed. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a schematic diagram of the steps of the present invention;

[0049] Figure 2 This is a system module diagram of the present invention. Detailed Implementation

[0050] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0051] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0052] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0053] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0054] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0055] Please see Figure 1 A method for automated management of carbon molecular sieve production based on big data includes the following steps:

[0056] S1: Record the continuous behavior trajectory of the adsorbed gas in the carbon molecular sieve, divide the continuous path segments based on the difference in the direction of airflow change, judge the abrupt change behavior in the path segment and screen the inflection point of change, identify the intervention target segment based on the overlap relationship between the path offset position and the adsorption structure, number and classify the change trend in the determined path, establish a label record table for intervention treatment, and form an adsorption path label set.

[0057] S2: Call the position index of the marked path in the adsorption path marker set, extract the configuration structure of the particle arrangement image in the area covered by the marked segment, analyze the continuity of the boundary morphology of adjacent particles, determine the interference area constructed by the connection between the path and the particles, extract the interference structure area by combining the change of path trajectory direction and the difference in configuration structure arrangement, and generate interference segment group.

[0058] S3: Call the spatial location data corresponding to the interference fragment group, extract the contour of the surface arrangement in the interference fragment, determine whether there is a lateral and longitudinal jump trend in the structural boundary area, track the position of the connecting line that continuously deforms, analyze the consistency of the extension direction between structures, collect the structural change trajectory according to the jump range, and output the structural strain processing template sequence.

[0059] S4: Call the regions listed in the structural strain treatment template sequence, cross-locate the execution time records of process actions in the carbon molecular sieve operation process, retrieve the execution segments of thermal control, orientation, and cooling operations, determine whether the action content overlaps in the same region and time, classify and mark overlapping segments, organize operation positions with adjustment estimates, and form a process intervention window list.

[0060] S5: Call the operation position and instruction content index in the process intervention window list, sort and analyze the action sequence in the overlapping execution segment, classify and integrate the positions where the operations interfere with each other between paths, determine the operation conflict points and disassemble them, recombine the operation execution chain according to the adjusted action sequence, organize and output the automatic control content, and obtain the carbon molecular sieve automatic control instruction structure table.

[0061] The adsorption path marker set includes path number information, directional change inflection points, path segment label classification, and offset overlap position characteristics. The interference segment group includes boundary continuity characteristics, particle connection configuration, path penetration area type, and structural interference differences. The structural strain treatment template sequence includes jump trend direction, structural contour morphology, connection line change trajectory, and extension consistency characteristics. The process intervention window list includes execution area number, operation time index, process action type, and overlapping segment classification label. The carbon molecular sieve automatic control instruction structure table includes operation sequence structure, action conflict decomposition group, execution chain combination form, and control instruction classification results.

[0062] The specific steps of S1 are as follows:

[0063] S101: Records the behavior trajectory of the gas adsorbed in the carbon molecular sieve. Based on the coordinate sequence and direction change data of the gas in the spatial node per unit time, it calls the coordinate difference and direction angle of adjacent time points for comparison, divides the direction change trend into multiple continuous path segments, and generates the path direction change interval value.

[0064] To record the behavior trajectory of adsorbed gas within a carbon molecular sieve, the three-dimensional movement path of the gas within the sieve must first be continuously tracked. Using a time step of 0.1 nanoseconds, the spatial coordinates of the gas molecules at each time point are collected. For example, at time t=0, the gas is recorded as being at coordinates (1.2, 0.8, 3.1), and at t=0.1 ns, it is at (1.3, 0.9, 3.3), and so on, recording all nodes along the entire adsorption path. Next, the directional change of the gas between two consecutive time points needs to be determined. Based on the direction of movement formed between each pair of consecutive coordinates, the degree of directional change in space is calculated sequentially. The angle difference between the previous and subsequent directions of movement is evaluated. If, between time points t=0.2 ns and 0.3 ns, the gas path direction shifts significantly to another spatial quadrant, this is considered a significant change. Based on the trend of directional change, the entire path is categorized, and directional change intervals are defined. The judgment criteria are based on 30°, 60°, and 90° as benchmarks to divide the angles into four levels: low change (0°~30°), medium change (30°~60°), high change (60°~90°), and abrupt change (above 90°). The angle change range here comes from the actual statistical distribution of the path segment angle changes in the simulation analysis. The selection criterion is the range that occupies the top 75% of the path segments, which is used to retain the change characteristics of most gas adsorption behaviors. In this process, the part of the continuous path segment where the angle change is consistently at the same level is classified as a single path segment. When the direction level changes, it is marked as the boundary of the path segment. For example, in an adsorption path with a total length of 50 nanoseconds, the analysis shows that it is divided into 16 path segments, of which there are 3 "abrupt change" segments, which are located near the corner structure of the molecular sieve. The final path segment information will record the start and end node numbers and direction change range values ​​of each segment, providing basic data for subsequent structural feature extraction.

[0065] S102: Based on the path direction change interval value, obtain the spatial offset angle of continuous points within the path segment, call the abrupt change amplitude of the angle between adjacent points and compare it with the direction abrupt change benchmark value, filter the path segments that exceed the benchmark, and extract the node positions with offset angles higher than the mean value to generate a path direction offset abrupt change point set.

[0066] Based on the aforementioned path direction change range, each node in each path segment is analyzed point by point. For the positional relationship between any two adjacent nodes, the change magnitude corresponding to their offset direction is calculated, and all offset angle changes within each segment are recorded. For example, path segment 4 contains 15 nodes, corresponding to 14 sets of adjacent point directional offset data, with offset angles of 18°, 22°, 19°, 40°, 85°, and 27°, respectively. In this set of data, the average value of all offset angles is first calculated, for example, the average is 34.7°. Simultaneously, a directional change baseline value of 45° is set. This baseline value is derived from the statistical results of multiple carbon molecular sieve adsorption simulation cases, and the top 25% quantile of offset angles in all path segments is selected as the baseline. The mutation baseline represents a typical characteristic node where structural boundary, turning, or obstruction behavior occurs. Then, the change value of each angle is compared. Any node with an angle greater than 45° is initially identified as a mutation node. In the example above, 85° constitutes a mutation point. In addition, nodes with an angle greater than the average value of 34.7° are further screened. For example, 40° and 85° both exceed the average value. These locations are further classified into the offset mutation point set. After summarizing this set, the position coordinates of the mutation point in the molecular sieve structure and the path segment number are recorded. For example, in the 4th segment, node 6 and node 8 are located at coordinates (2.5, 1.1, 5.9) and (2.7, 1.0, 6.3) respectively, thus forming a set of mutation points for structural boundary correlation analysis.

[0067] S103: Based on the set of abrupt shift points in the path direction offset, call the adsorption structure position data of the node in the carbon molecular sieve, calculate the spatial overlap rate with the structure boundary, screen the abrupt segments with an overlap rate exceeding the overlap reference rate, number and classify the directional change trend, and obtain the adsorption path marker set.

[0068] The formula for calculating the spatial overlap rate with the structural boundary is as follows:

[0069] ;

[0070] Where, η cr P represents the spatial overlap rate between the path direction of the mutation point and the structural boundary, N represents the total number of valid mutation points in the mutation point set, and P represents the spatial overlap rate between the path direction of the mutation point and the structural boundary. k b represents the three-dimensional coordinate vector of the k-th mutation point in the carbon molecular sieve. k P represents k The corresponding coordinate vector of the nearest structural boundary point, ‖P k ―B k ‖ represents the Euclidean distance between the two, V k n represents the path direction offset vector at the k-th mutation point. k Represents structural boundary point b k The unit normal vector at point α dA smoothing correction coefficient representing the magnitude of the path direction vector;

[0071] Parameter definition and data source

[0072] N=5: The number of effective mutation points in the mutation point set, extracted by the carbon molecular sieve adsorption path direction shift mutation detection algorithm. Effective points are screened based on the path mutation energy gradient threshold (threshold range 0.1-0.5eV / Å). The actual detection data is 5.

[0073] P k With b k The coordinates of the mutation point and boundary point are output as three-dimensional coordinate data using carbon molecular sieve structure simulation software (such as MaterialsStudio), where p1=(1.2,3.4,0.8), b1=(1.0,3.0,0.7); p2=(2.5,4.1,1.3), b2=(2.3,4.0,1.2), and the remaining points are similar.

[0074] ‖Pk―Bk‖: Euclidean distance calculation, for example .

[0075] V k : Path direction offset vector, calculated by the path tracking algorithm, v1=(0.3-0.2,0.1), .

[0076] n k : Boundary normal vector, obtained through surface curvature analysis of the structure, n1=(0.5-1.1-1.5), has been normalized to a unit vector.

[0077] a d =0.3: Smoothing correction coefficient, set according to the standard deviation of the magnitude of the path direction vector (range 0.2-0.5), used to suppress magnitude fluctuations.

[0078] Calculation process of the example

[0079] Calculate a single mutation point term:

[0080] For k=1, calculate the numerator. .

[0081] denominator .

[0082] The contribution value at a single point is approximately 0.415 / 0.674 ≈ 0.616.

[0083] Summation and normalization:

[0084] Repeat the calculation from k=2 to k=5, assuming the contribution values ​​at each point are 0.616, 0.532, 0.481, 0.403, and 0.387 respectively, the sum is 0.616 + 0.532 + 0.481 + 0.403 + 0.387 = 2.419. After normalization, η cr =2.419 / 5=0.484. The result is related to the steps.

[0085] This result indicates that the spatial overlap ratio parameter η cr=0.484 If the overlap rate exceeds the preset baseline of 0.45, the mutation segment is selected as an effective adsorption path segment and classified into the directional change trend number T-12, and finally the adsorption path label set M12 is generated.

[0086] The specific steps of S2 are as follows:

[0087] S201: Call the position index of the identified path in the adsorption path marker set, locate the area covered by the path segment, obtain the two-dimensional grayscale data of the corresponding particle arrangement image, extract the particle edge intensity gradient in the image area, identify particles based on edge closure, and generate configuration boundary recognition value.

[0088] To call the location index of the identified path in the adsorption path marker set, the numbered path segment numbers must first be extracted, such as D_4_H_85, D_7_M_51, etc. The system obtains the spatial coordinate range corresponding to each number by looking up a table. For example, the path segment node number corresponding to number D_4_H_85 is from 120 to 145, and the corresponding coordinates are from (2.3, 1.0, 5.8) to (3.1, 1.4, 6.5). Then, this three-dimensional coordinate range is projected onto the two-dimensional image model. With the z-axis kept fixed, the x and y-axis positions are mapped to the pixel matrix coordinate system within the image area. For example, if the resolution is set to 512×512, the three-dimensional position (2.5, 1.1, z) is mapped to the image coordinate point (256, 285). Next, grayscale image data is extracted from this image area, and edge detection is performed by scanning the rate of change of grayscale values ​​in the two-dimensional pixel matrix. Each pixel is considered a particle edge point if the grayscale difference between its surrounding pixels and the pixels to its left, right, and above exceeds a set edge extraction threshold. The threshold is set to 25 grayscale levels, derived from the statistical average of particle edge contrast in typical adsorption images, and is set to the top 15 percentile of the total pixel grayscale change distribution. Furthermore, a closure judgment is performed on the identified edge point set to determine whether the edge points form a closed loop. If the distance between edge points in a certain region does not exceed 3 pixels and can be closed into an approximate ellipse or polygon, then that region is considered a single particle. A unique number is assigned to each closed boundary region, and the boundary closure of its contour is calculated. The closure range is divided into three levels: closure ≥ 0.9 for complete particles, 0.7~0.9 for partial particles, and < 0.7 for non-particle edges. Finally, the configuration boundary recognition value of each identified particle is output, and a configuration information index set for the corresponding image region of the path segment is established.

[0089] S202: Based on the configuration boundary identification value, extract the connection segment between the configuration boundaries of adjacent particles, call the boundary shape change degree and boundary direction angle value at both ends of the connection segment, determine whether the boundary continuity is lower than the set connection continuity benchmark value, mark the path penetration area position of the corresponding connection segment, and obtain the interference connection area intensity value.

[0090] Based on the configuration boundary recognition value, it is necessary to analyze the spatial border between adjacent particles. First, an edge buffer with a width of 3 pixels is set around the boundary contour of each particle, and the edge buffer is searched for overlapping areas with the boundaries of other particles. If there are no less than 5 consecutive overlapping pixels in the edge buffers of two particles, it is determined that a connection segment exists. Then, the boundary morphology parameters of the start and end points of each connection segment are extracted, including the local curvature value of the boundary and the edge direction. The boundary direction angle at both ends of each connection segment is calculated. The angle difference between the pixel connection direction and the horizontal axis is used as the judgment criterion. An angle less than 30° is considered consistent direction, 30°~60° is considered slight deflection, and more than 60° is considered discontinuous direction. At the same time, the curvature change rate of the boundary at both ends of the connection segment is calculated. If the change rate is greater than a set threshold of 0.25, the boundary morphology is considered inconsistent. The threshold is 1.5 times the statistical standard deviation of the curvature of the configuration boundary. Based on the above two judgment results, if the directional angle exceeds 60° and the curvature change exceeds 0.25, the continuity of the connection segment is judged to be insufficient. The system sets the connection continuity benchmark value as a standard combination of "angle < 45° and curvature change < 0.2". This value combination comes from the manual judgment samples of the grain connection points of multiple images, covering more than 85% of the boundary features of the identifiable effective connection segments. Any connection segment below this continuity standard is marked as the location of the path penetration area, and the average pixel gray value of the area is recorded as the intensity value of the interference connection area. For example, there are 4 connection segments in the image corresponding to the path segment D_4_H_85, of which 2 have angles of 78° and 92°, and boundary curvatures of 0.27 and 0.31, respectively. These 2 connection segments are identified as penetration areas, and the recorded intensity values ​​are 112 and 128, respectively.

[0091] S203: Based on the intensity value of the interference connection area, call the path segment direction change value and the connection segment configuration arrangement density value, filter the intersection area where the direction change amplitude exceeds the trajectory offset change benchmark value and the arrangement density difference is higher than the configuration density difference threshold, divide the intersection area into structural difference segments, and obtain the interference segment group.

[0092] Based on the obtained interference connection region intensity values, further filtering is needed by combining the path segment direction change values ​​and the connection segment configuration density. First, the direction change amplitude value of the path segment corresponding to each penetration region is extracted. For example, the direction change of path segment D_4_H_85 is 85°. Then, the configuration density within the penetration connection segment region is calculated. This is done by counting the number of particle centers within a 5×5 pixel area near the connection segment as the density benchmark. For example, if the density within a connection segment is 0.14 particles / pixel, and the average density of the corresponding path segment region is 0.08 particles / pixel, the density difference is calculated to be 0.06 particles / pixel. The trajectory offset change benchmark value is set to 70°. This value comes from the selection result of the 75th percentile in the path direction change angle distribution and is used to determine the severity of the direction offset and configuration. The density difference threshold is set to 0.05 per pixel, derived from a weighted average of the standard deviations of density changes in different particle regions. This indicates that the density change of the structure arrangement has reached a identifiable range. If the direction change of a path segment exceeds 70° and the density difference is higher than 0.05 per pixel, the connecting region is determined to be a structural difference intersection region. For example, in path segment D_4_H_85, the connecting segment that meets the conditions is position number A104, corresponding to an 85° direction change, a density difference of 0.06, and an intensity value of 128. Therefore, the connecting region number A104 is added to the structural difference segment set. Finally, a total of 3 intersection regions are selected in the entire path, corresponding to numbers A104, B209, and C302, forming an interference segment group for subsequent structural interference feature analysis.

[0093] The specific steps for S3 are as follows:

[0094] S301: Call the spatial location data corresponding to the interference fragment group, extract the arrangement image of surface particles in the fragment region, obtain the grayscale contour edge pixel point set, fit the boundary according to the edge gradient intensity and closure, identify the outer contour shape of the particle arrangement region, and generate the surface contour boundary value.

[0095] To retrieve the spatial location data corresponding to the interference fragment group, it is necessary to extract the spatial coordinate range corresponding to the interference fragment number. For example, the coordinate range corresponding to the number A104 is x=1.2-2.0, y=0.8-1.6, with the z-axis fixed at the 1.5 level. A two-dimensional image slice of this region on the z-value plane is extracted from the three-dimensional structure. Then, the surface particle arrangement information is obtained from this image. The grayscale values ​​in the pixel matrix within the region are extracted, and a grayscale threshold is set to separate the particle region from the background region. For example, a grayscale threshold of 80 is set; values ​​below this threshold are background, and values ​​above it are particle regions. The edge gradient value is calculated based on the abrupt changes in pixel grayscale in space. Points where the grayscale difference between adjacent pixels exceeds the set gradient baseline value are marked as edge points. This gradient baseline value is set to 20, derived from the statistical average of grayscale changes at the edge, adjusted upwards by 10. This process yields a preliminary set of grayscale contour edge pixels. Then, contour fitting is performed on this set. During fitting, the length and closure of the connecting lines between each set of edge points are statistically analyzed to determine if a continuous circular boundary is formed. If there is a region where the average distance between adjacent points is no more than 2 pixels and the overall boundary closure rate is higher than 0.85, it is determined to be a complete particle outer contour. This closure threshold is derived from the 75th percentile of the complete boundary closure rate in the particle sample image. Further morphological extraction is performed on each closed contour region, extracting parameters such as the curvature, area, and number of boundary points, and recording the outer contour boundary values. For example, in region A104, 5 closed boundary particles are identified, with a curvature range of 0.12~0.21 and an average of 32 boundary points. Finally, a structured set of surface contour boundary values ​​is formed for trend analysis.

[0096] S302: Based on the surface contour boundary value, obtain the coordinate sequence of continuous boundary points in the contour, calculate the extension jump amplitude of adjacent boundary segments in the horizontal and vertical directions respectively, determine whether the jump amplitude exceeds the horizontal jump reference value and the vertical jump reference value, filter the connecting lines corresponding to the continuous jump positions, and obtain the structural jump trend coefficient.

[0097] Based on the surface contour boundary values, the spatial coordinate sequence of the boundary points is extracted from the outer contour of each particle. After being sorted by contour direction, it is recorded as {P1, P2, P3...Pn}. A boundary segment is formed between every two adjacent points. The jump amplitude of these boundary segments in the horizontal (x-direction) and vertical (y-direction) directions is calculated. For example, if boundary points P1=(1.25, 1.12) and P2=(1.29, 1.18), the horizontal jump amplitude is 0.04 and the vertical jump amplitude is 0.06. After calculating the horizontal and vertical jump amplitudes of all adjacent boundary segments, the jump judgment benchmark value is called for judgment. The horizontal jump benchmark value is set to 0.08, and the vertical jump benchmark value is 0.1. This setting is based on the mean and variance of the jump amplitudes of all boundary segments. The standard deviation of the variance in the entire image data is taken with a 10% upward adjustment as the jump judgment benchmark. A threshold is set, and when the horizontal or vertical jump amplitude of a certain boundary segment exceeds the above-mentioned benchmark value, the segment is marked as a jump segment. For example, in the particle profile numbered A104, the jump amplitudes of boundary segments P7 to P8 are 0.13 (horizontal) and 0.15 (vertical), both exceeding the benchmark value. Therefore, this segment is marked as a jump boundary segment. Connecting lines are established in multiple consecutive jump segments to form a structural jump path by connecting the boundary segments between jump points. The jump trend is quantitatively characterized by statistically analyzing the total offset value of the jump direction and the rate of change of direction in consecutive jump segments, and calculating the structural jump trend coefficient. For example, in numbered A104, there are 5 groups of consecutive jump segments with a direction change angle range of 45°~78° and an average change amplitude of 61.2°. The trend coefficient is set as the average value of the direction change angle, and the final structural jump trend coefficient is 61.2.

[0098] S303: Based on the structural jump trend coefficient, extract the position data of the connecting line segments that have undergone deformation, calculate the cosine value of the angle between the line segment and the direction, determine whether there is a directional consistency shift in the structural extension direction among multiple connecting segments, classify the path change sequence according to the jump range, and establish a structural strain treatment template sequence.

[0099] The formula for calculating the cosine of the angle between a line segment and a direction is as follows:

[0100] ;

[0101] Where cos(β) represents the cosine of the angle between the line segment and the direction, m represents the estimated slope of the current connecting line segment on the two-dimensional projection plane, n represents the slope direction constant of the reference direction in the same projection plane, q represents the magnitude of the displacement component of the connecting line segment perpendicular to the projection plane, and |m·n| represents the absolute value of the slope product. This represents the combined directional difference between the direction of the connecting line segment and the reference direction. This represents the sum of the magnitudes of the directional and perpendicular components of the connecting line segment, where |n| represents the absolute slope of the reference direction.

[0102] Parameter acquisition and calculation methods:

[0103] m: Represents the estimated slope of the current connecting line segment on the two-dimensional projection plane. The slope on the projection plane is calculated by measuring the coordinates of the two endpoints of the connecting line segment. Let the coordinates of the two endpoints be (x1, y1) and (x2, y2), then the formula for calculating the slope m is:

[0104] ;

[0105] n: Represents the slope direction constant of the reference direction within the same projection plane. The slope is calculated by measuring the coordinates of two points along the reference direction. Let the coordinates of the two points along the reference direction be (x3, y3) and (x4, y4), then the formula for calculating the slope n is:

[0106] ;

[0107] q: Represents the magnitude of the displacement component of the connecting line segment perpendicular to the projection plane. The displacement component is calculated by measuring the coordinate difference between the two endpoints of the connecting line segment perpendicular to the projection plane. Let the coordinates of the two endpoints in the vertical direction be z1 and z2, then the formula for calculating the displacement component q is:

[0108] ;

[0109] Specific numerical calculation examples:

[0110] Let the coordinates of the two endpoints of the connecting line segment be (2,3,5) and (5,7,9), and the coordinates of the two points in the reference direction be (0,0) and (4,8).

[0111] Calculate the slope of the connecting line segment. ;

[0112] Calculate the slope in the reference direction ;

[0113] Calculate vertical displacement components ;

[0114] Substituting the above values ​​into the formula, calculate cos(β):

[0115] ;

[0116] Results Explanation:

[0117] The results indicate that the cosine of the angle between the connecting line segment and the reference direction is approximately 0.931, suggesting a relatively consistent direction between them. This value can be used to determine whether there is a shift in the consistency of the structural extension direction among multiple connecting segments, and to classify the path variation sequences according to the jump range, thereby establishing a structural strain treatment template sequence.

[0118] The specific steps of S4 are as follows:

[0119] S401: Call the regions listed in the structural strain treatment template sequence, obtain the corresponding process action index of the region in the carbon molecular sieve operation process, locate the execution time record of each process action, and cross-compare the time period with the corresponding spatial location data to generate the process action location interval value.

[0120] To access the regions listed in the structural strain treatment template sequence, the spatial coordinate segments corresponding to each template number must be retrieved one by one. For example, the coordinate range corresponding to template T_A104_1 is x=1.2-1.8, y=0.9-1.5. This coordinate range is mapped to the equipment number and spatial operation segment index information in the carbon molecular sieve operation process. The operation node identifier number corresponding to each coordinate point is retrieved. For example, the operation number can be mapped to HT-012 (thermal control), DR-003 (orientation), etc. Then, the time period of each process action number is retrieved in the operation record table to obtain its precise start and end time. For example, DR-003 is executed from 03:15 to 03:28, and HT-012 is executed from 03:00 to 03:12. Then, the time periods of these actions are cross-referenced with the spatial location data listed in the template to construct the sequence for each action. The system establishes a correspondence between the time of an action and its corresponding spatial region. If the time period of a certain process action coincides with the time label of the coordinate region listed in the template, it is determined that the corresponding action has occurred. The system's time matching standard is that the start and end time range of the action overlaps with any consecutive 5 seconds within the time period of the coordinate segment being collected. For example, if the spatial location segment jumps between 03:10 and 03:18, and HT-012 is executed between 03:00 and 03:12, the last 2 minutes of which coincide with the jump time, then it is recorded as a successful process action location matching. Finally, the system outputs the action index number corresponding to each template and its interval on the spatial and time axes, generating the process action location interval value. For example, T_A104_1 corresponds to process HT-012, time period 03:10 to 03:12, location coordinates x=1.2+1.5, y=0.9+1.2.

[0121] S402: Based on the process action positioning interval value, extract the execution segment information corresponding to the three types of actions: thermal control, orientation, and cooling. Compare the start and end time of the action segment with the spatial coordinates to determine whether there is a situation where the time and area coordinates of the three types of actions coincide at the same time. Filter out all overlapping segments and obtain the process action overlap ratio value.

[0122] Based on the aforementioned process action positioning intervals, three key operation types are extracted: thermal control, orientation, and cooling actions. The corresponding action numbers begin with HT, DR, and CL, respectively. Each action type segment is selected, and its start and end times and the spatial coordinate range it operates on are extracted. For example, thermal control HT-012 has an execution time of 03:00-03:12, corresponding to coordinates x=1.2-1.5, y=0.9-1.2; orientation DR-003 has an execution time of 03:10-03:28, corresponding to coordinates x=1.3-1.8, y=1.0-1.4; and cooling CL-005 has an execution time of 03:25-03:40, corresponding to coordinates x=1.5-2.0, y=1.1-1.5. The spatial and temporal overlap of the three types of actions is then checked. First, the time axes are compared; if the start and end times of any two actions intersect, then... For example, if the time intervals of HT and DR overlap by 2 minutes between 03:10 and 03:12, then the spatial coordinate range is compared to determine if there is an intersection between the x-axis and y-axis. If the x-axis intersection is greater than 0.1 and the y-axis intersection is greater than 0.1, it is considered spatially overlapping. This value is set based on the minimum spacing of the particle distribution unit cell boundary. Further, all combinations of segments that overlap in both time and space in the three types of operations are statistically analyzed, and their overlap ratio is calculated. The overlap ratio is defined as the ratio of the area of ​​the overlapping spatial region to the area of ​​the total operation region. For example, DR and HT overlap in the region of x=1.3+1.5, y=1.0+1.2, with an area of ​​0.04. The total area of ​​the two is 0.25, resulting in an overlap ratio of 0.16. Finally, all segments that meet the dual conditions of time and space overlap are summarized, and the corresponding overlap ratio values ​​are output and recorded as a list of process action overlap ratio values.

[0123] S403: Based on the process action overlap ratio, classify the segments whose overlap ratio exceeds the action overlap benchmark ratio, extract the index of the covered operation position, and combine the change trend amplitude value of the corresponding segment to screen the spatial positions with predictable operation adjustment and establish a process intervention window list.

[0124] Based on the process action overlap ratio, all segments with overlap ratios exceeding the baseline overlap ratio were extracted for analysis. This baseline ratio was set at 0.12, derived from the median value plus one standard deviation of the process action interval overlap data from multiple carbon molecular sieve operations. Specifically, the average overlap ratio was calculated from 20 data sets to be 0.08, with a standard deviation of 0.02. Therefore, the baseline was set as 0.08 + 0.04 = 0.12. All segments with overlap ratios greater than 0.12 were recorded and categorized. For example, the overlap ratio of T_A104_1 with HT-012 and DR-003 was 0.16, greater than the baseline, and thus listed as a valid intervention segment. The operational space location covered by this segment, such as x = 1.3 + 1.5, y = 1.0 + 1.2, and time 03:10 - 03:12, was recorded as the operational location. The index is used to simultaneously call the structural change trend magnitude value corresponding to that location. For example, the jump trend coefficient of T_A104_1 is 61.2. The trend coefficient is compared with whether it exceeds the set change trend benchmark value, which is set to 50 and comes from the 75th percentile value of the statistical distribution of structural jump angle. If the trend coefficient is higher than 50 and the corresponding segment overlap ratio is greater than 0.12, then the location is determined to be a high-interference point with predictable adjustment. Finally, all such operation locations are combined with relevant trend values ​​to establish a process intervention window list. For example, the record items in the list are: number T_A104_1, time 03:10 03:12, coordinates x=1.3 1.5, y=1.0 1.2, trend 61.2, overlap ratio 0.16, which are used for subsequent intervention strategy deployment and operation plan generation.

[0125] The specific steps of S5 are as follows:

[0126] S501: Call the operation location and instruction content index listed in the process intervention window list, extract the action type, start time, end time and path number corresponding to the overlapping execution segment, sort the operation actions under the same path in ascending order according to the action start time, obtain the sequence of process actions in the path segment, and generate the path operation sorting value.

[0127] The process intervention window list is used to retrieve the index of operation locations and instructions. All action types associated with a specific operation location, such as thermal control, orientation, and cooling, are extracted, along with their corresponding start and end times. These actions are then categorized according to their path numbers. For example, operations under path number T_A104_1 might include HT-012 and DR-003, where HT-012 runs from 03:00 to 03:12 and DR-003 runs from 03:10 to 03:28. These actions are then sorted in ascending order by their start time, ensuring that operations within the same path can be performed sequentially. Executing HT-012, followed immediately by DR-003, is based on the temporal correlation of actions, ensuring the continuity and logic of operations. In this way, the system can generate the sequence of process actions within each path segment. This sequence reflects the implementation time of the operations, further facilitating the monitoring and adjustment of the process flow. The final output is the sequence of process actions in each path segment, i.e., the path operation sorting value. For example, the sorting value of T_A104_1 lists HT-012 (03:00-03:12) and DR-003 (03:10-03:28) arranged in chronological order.

[0128] S502: Based on the path operation sorting value, extract the overlapping segments of operation positions between differentiated paths, compare the action number and execution type within the same time period, determine whether there is a mutual exclusion logic conflict between actions, if there is no minimum interval time between adjacent action numbers, mark the segment as a conflict point, and classify and integrate the instruction positions corresponding to the conflict points to obtain the process action conflict position value.

[0129] Based on the path operation sorting value, possible overlapping segments of operation positions between each path are further extracted, and the action number and execution type are compared within the same time period to determine whether there is a mutually exclusive logical conflict between actions. For example, in the time period from 03:10 to 03:12, the HT-012 thermal control action and the DR-003 orientation action of the T_A104_1 path overlap in time. At this time, the minimum interval time is checked. If the interval time between the two actions is less than the set minimum interval time standard, which is usually set to 2 minutes to prevent operation conflicts and equipment overload, the operation in this time period is marked as a conflict point. This judgment process is calculated by comparing the start and end times of the actions. For example, HT-012 and DR-003 are only 0 minutes apart, which is obviously lower than the 2-minute standard, so they are marked as conflict points. The instruction positions of the conflict points are classified and integrated, and the process action conflict position value is output. For example, the conflict occurs in the T_A104_1 path, time 03:10 to 03:12, operation type HT-012 and DR-003.

[0130] S503: Based on the conflict position value of the process action, decompose all conflicting segments, rearrange the non-conflicting operation sequence, and connect the decomposed actions into a continuous instruction sequence in time priority order. Recombine and merge the operation chain of the path to establish an automatic control instruction structure table for carbon molecular sieve.

[0131] Based on the conflict location values ​​of process actions, all operation segments marked as conflicting are analyzed and decomposed in detail. The order of non-conflicting operations is then rearranged to ensure that each operation can be performed according to time priority, without waiting for the previous operation to finish completely. In this way, the operation sequence can be adjusted into a continuous, conflict-free instruction sequence. For example, the operation time of HT-012 in T_A104_1 is adjusted to 03:00 to 03:08, and the operation time of DR-003 is adjusted to 03:09 to 03:28, avoiding direct time overlap between the two. During the process, the specific execution time or sequence of operations may need to be adjusted according to the actual situation to ensure the smoothness and efficiency of the entire process. Finally, the operation chain of the path is recombined and merged to establish an optimized automatic control instruction structure table for carbon molecular sieves that meets the actual needs of operation. This table not only guides daily operations but also serves as a basis for fault response and process adjustment. For example, the structure table may include operation T_A104_1, time 03:00 to 03:08 (HT-012), 03:09 to 03:28 (DR-003), providing clear operation guidelines and time arrangements.

[0132] Please see Figure 2 An automated management system for carbon molecular sieve production based on big data includes:

[0133] The adsorption trajectory recognition module acquires the gas flow path data of the adsorption region inside the carbon molecular sieve, calls the coordinate sequence and direction change value of the trajectory point, calculates the angle difference between adjacent points and divides the path segment, determines the direction change point of each segment and locates the start and end coordinates, compares the positional relationship between the coordinates and the adsorption structure space diagram, and generates an adsorption path marker set.

[0134] The interference fragment extraction module calls the path segment position in the adsorption path marker set, extracts the particle image of the corresponding region and detects the angle between adjacent edge lines, determines whether the angle change exceeds the continuous range, analyzes the angle between the path direction and the boundary extension, filters out the connected discontinuous particle combination regions, and generates interference fragment groups.

[0135] The strain trend analysis module calls the boundary coordinates in the interference segment group, extracts the contour line and calculates the horizontal and vertical change trends, identifies the continuous offset position and tracks the direction of the connecting line, collects the changing segments with the same direction, and generates a structural strain processing template sequence.

[0136] The process action positioning module calls the number in the structural strain treatment template sequence, extracts the time information of thermal control, orientation and cooling actions, determines whether there is time overlap under the same number, filters the adjustable time period and organizes it according to action type, and generates a process intervention window list.

[0137] The control chain generation module calls the action segments in the process intervention window list, extracts the number, action type and time data, breaks down the overlapping time periods under the same number, rearranges the action sequence and combines them into a complete execution chain, and generates the carbon molecular sieve automatic control instruction structure table.

[0138] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A big data-based carbon molecular sieve production automation management method, characterized in that, The method comprises the following steps: S1: record the continuous path trajectory of the adsorbed gas in the carbon molecular sieve, divide the path segments according to the change of the gas flow direction, identify the direction mutation and screen the key inflection points, number and classify the change trend, and form an adsorption path marker set; S2: call the adsorption path marker set, extract the particle configuration image of the corresponding area, analyze the continuity of the particle boundary, identify the interference area of the trajectory and the connecting structure, and generate an interference segment group; S3: according to the spatial position data of the interference segment, extract the surface arrangement contour, judge the horizontal and vertical jump trend of the boundary, track the connecting line of the continuous deformation between structures, analyze the consistency of the extension direction, and output a structure strain processing template sequence; S4: call the structure strain processing template sequence, cross-position the operation records in the carbon molecular sieve process, search for the heat control, orientation, and cooling operation paragraphs, identify the space-time overlapping part and classify the marker, and form a process intervention window list; S5: according to the operation position and instruction content index in the process intervention window list, analyze and sort the action sequence of the overlapping segment, identify the operation conflict point and perform disassembly and reconstruction, and generate a carbon molecular sieve automatic control instruction structure table.

2. The big data based automated management method of carbon molecular sieve production according to claim 1, wherein, The adsorption path marker set includes path number information, direction change inflection point, path segment label classification, and offset overlap position feature. The interference segment group includes boundary continuity feature, particle connection configuration, path penetration area type, and structure interference difference. The structure strain processing template sequence includes jump trend direction, structure contour shape, connecting line variation track, and extension consistency feature. The process intervention window list includes execution area number, operation time index, process action type, and overlapping segment classification label. The carbon molecular sieve automatic control instruction structure table includes operation sequence structure, action conflict disassembly group, execution chain combination form, and control instruction classification result.

3. The big data based automated management method of carbon molecular sieve production according to claim 1, wherein, The specific steps of S1 are: S101: record the behavior trajectory of the adsorbed gas in the carbon molecular sieve, according to the coordinate sequence and direction change data of the gas in the space node within a unit time, call the coordinate difference value and direction angle of adjacent time points for comparison, divide into multiple continuous path segments according to the direction change trend, and generate path direction change interval value; S102: based on the path direction change interval value, obtain the spatial offset angle of the continuous points in the path segment, call the mutation amplitude of the adjacent point angle and compare it with the direction mutation reference value, screen the path segments exceeding the reference, and extract the node position with the offset angle higher than the average value, generate a path direction offset mutation point set; S103: according to the path direction offset mutation point set, call the adsorption structure position data corresponding to the node in the carbon molecular sieve, calculate the spatial coincidence rate with the structure boundary, screen the mutation segments with the coincidence rate exceeding the coincidence reference rate, number and classify the direction change trend, and obtain the adsorption path marker set.

4. The big data based automated management method of carbon molecular sieve production according to claim 3, wherein, The calculation formula of the spatial coincidence rate with the structure boundary is specifically: ； where η cr represents the spatial coincidence rate of the mutation point path direction and the structure boundary, N represents the total number of effective mutation points in the mutation point set, P k represents the three-dimensional coordinate vector of the kth mutation point in the carbon molecular sieve, b k represents the P k corresponding nearest structure boundary point coordinate vector, ‖P k― b k ‖ represents the Euclidean distance between the two, V k represents the path direction offset vector at the kth mutation point, n k represents the unit normal vector at the structure boundary point b k , ɑ d represents the smoothing correction coefficient of the path direction vector module length.

5. The big data based automated management method of carbon molecular sieve production according to claim 1, wherein, The specific steps of S2 are: S201: Call the position index of the identified path in the adsorption path marker set, locate the area covered by the path segment, obtain the two-dimensional gray scale data of the corresponding particle arrangement image, extract the particle edge intensity gradient in the image area, and identify the particles according to the edge closure degree to generate the configuration boundary recognition value; S202: Based on the configuration boundary recognition value, extract the connection segment between adjacent particle configuration boundaries, call the boundary shape change degree and boundary direction angle value at both ends of the connection segment, judge whether the boundary continuity is lower than the set connection continuity reference value, and mark the path penetration area position of the corresponding connection segment, and obtain the interference connection area intensity value; S203: According to the interference connection area intensity value, call the path segment direction change value and the connection segment configuration arrangement density value, filter the intersection area whose direction change amplitude exceeds the trajectory offset change reference value and whose arrangement density difference is higher than the configuration density difference threshold, divide the intersection area into structure difference segments, and obtain the interference segment group.

6. The big data based carbon molecular sieve production automation management method of claim 1, wherein, The specific steps of S3 are: S301: Call the spatial position data corresponding to the interference segment group, extract the arrangement image of the surface particles in the segment area, obtain the gray scale contour edge pixel set, fit the boundary according to the edge gradient intensity and closure degree, and identify the outer contour shape of the particle arrangement area to generate the surface contour boundary value; S302: Based on the surface contour boundary value, obtain the coordinate sequence of the continuous boundary points in the contour, calculate the extension jump amplitude of adjacent boundary segments in the transverse and longitudinal directions respectively, judge whether the jump amplitude exceeds the transverse jump reference value and the longitudinal jump reference value, filter the connection line corresponding to the continuous jump position, and obtain the structure jump trend coefficient; S303: According to the structure jump trend coefficient, extract the position data of the connection line segment that occurs deformation, calculate the included angle cosine value between the line segment and the direction, judge whether there is a direction consistency deviation phenomenon among multiple connection segments in the structure extension direction, and classify the path change sequence according to the jump range, and establish a structure strain processing template sequence.

7. The big data based automated management method of carbon molecular sieve production according to claim 6, wherein, The included angle cosine value calculation formula between the line segment and the direction is: ； wherein cos(β) represents a cosine value of an included angle between the line segment and the direction, m represents a slope estimation value of the current connection line segment in a two-dimensional projection plane, n represents a slope direction constant of the reference direction in the same projection plane, q represents a displacement component amplitude of the connection line segment in a direction perpendicular to the projection plane, and |m·n| represents an absolute value item of a slope product, represents a comprehensive direction difference between the connection line segment direction and the reference direction, represents a module length sum of the connection line segment direction component and the perpendicular component, and |n| represents an absolute slope of the reference direction.

8. The big data based carbon molecular sieve production automation management method of claim 1, wherein, The specific steps of S4 are: S401: Call the region listed in the structure strain processing template sequence, obtain the process action index corresponding to the region in the carbon molecular sieve operation process, locate the execution time record of each process action, and cross compare the time period with the corresponding spatial position data to generate the process action positioning interval value; S402: Based on the process action positioning interval value, extract the execution paragraph information corresponding to the three types of actions of heat control, orientation and cooling, compare the start and end time and spatial coordinates of the action paragraph, judge whether there is a case of time and region coordinates coinciding at the same time in the three types of actions, filter all overlapping segments, and obtain the process action overlap ratio value; S403: According to the process action overlap ratio value, classify the position of the segment whose overlap ratio exceeds the action overlap reference ratio, extract the operation position index covered, and combine the change trend amplitude value of the corresponding segment to filter the spatial position with operation adjustment predictability, and establish a process intervention window list. 9.The big data based automated management method of carbon molecular sieve production according to claim 1, wherein, The specific steps of S5 are: S501: Call the operation position and instruction content index listed in the process intervention window list, extract the action type, start time, end time and path number corresponding to the coincident execution segment, arrange the operation actions in the same path in ascending order according to the action start time, obtain the order of process actions in the path paragraph, and generate a path operation sorting value; S502: Based on the path operation sorting value, extract the operation position coincidence segment between the differential paths, compare the action number and execution type in the same time period, judge whether there is a mutual exclusion logic conflict, if there is no minimum interval time between adjacent action numbers, mark the segment as a conflict point, and classify and integrate the instruction positions corresponding to the conflict points to obtain a process action conflict position value; S503: According to the process action conflict position value, the action is disassembled and processed in all conflict paragraphs, the operation order without conflict is rearranged, and the disassembled action is connected in time priority order to form a continuous instruction sequence, the operation chain of the path is recombined and merged, and a carbon molecular sieve automatic control instruction structure table is established.

10. A big data based carbon molecular sieve production automation management system, characterized in that, The carbon molecular sieve production automatic management method based on big data according to any one of claims 1-9, the system comprises: An adsorption trajectory recognition module obtains gas flow path data of an adsorption area inside the carbon molecular sieve, calls a coordinate sequence and a direction change value of a trajectory point, calculates an angle difference between adjacent points and divides a path segment, judges each direction mutation point and locates start and end coordinates, compares the coordinates with a position relationship in an adsorption structure space diagram, and generates an adsorption path marker set; An interference segment extraction module calls a path segment position in the adsorption path marker set, extracts a corresponding area particle image and detects an adjacent edge line included angle, judges whether the included angle change exceeds a continuity range, analyzes a path direction and a boundary extension included angle, screens a discontinuous particle combination area, and generates an interference segment group; A strain trend analysis module calls boundary coordinates in the interference segment group, extracts a contour line and calculates a horizontal and vertical change trend, identifies a continuous offset position and tracks a connection line direction, collects strain paragraphs with consistent directions, and generates a structure strain processing template sequence; A process action positioning module calls a number in the structure strain processing template sequence, extracts time information of a thermal control, orientation and cooling action, judges whether there is time coincidence under the same number, screens an adjustable time period and organizes according to an action type, and generates a process intervention window list; A control chain generation module calls an action segment in the process intervention window list, extracts a number, an action type and time data, disassembles an overlapping time period under the same number, rearranges an action order and combines into a complete execution chain, and generates a carbon molecular sieve automatic control instruction structure table.

Images