Intelligent Management System for the Entire Process of Rubber Coating Products Based on Big Data

Abstract

This invention discloses an intelligent management system for the entire process of rubber coating products based on big data, belonging to the fields of intelligent manufacturing and industrial big data technology. The system includes a data scheduling module, which obtains the product batch code in response to the coating preparation start command and retrieves the full-process data identification matrix; a process acquisition module, which sends data reporting commands to the batching, coating, and vulcanization testing process equipment and receives real-time process parameters based on the full-process data identification matrix; a matrix mapping module, which fills the real-time process parameters into the corresponding dimension positions of the dynamic state matrix according to the parameter mapping relationship; a rank difference analysis module, which calculates the current matrix rank of the dynamic state matrix and compares it with the benchmark matrix rank to obtain the matrix rank deviation; a deviation positioning module, which locates the target dimension position of the parameter value deviation from the mapping relationship in the dynamic state matrix based on the matrix rank deviation; and an instruction generation module, which generates adjustment instructions based on the process identifier and parameter type identifier corresponding to the target dimension position.

Description

Technical Field

[0001] This invention relates to the fields of intelligent manufacturing and industrial big data technology, specifically to an intelligent management system for the entire process of rubber coating products based on big data. Background Technology

[0002] In the production of rubber-coated products, numerous process parameters are involved in stages such as coating formulation, application, and vulcanization testing. These parameters exhibit complex, implicit relationships. Existing end-to-end management solutions largely rely on independent threshold monitoring for each stage. A host computer sets upper and lower limits for raw material ratios in the batching stage, thickness and speed ranges in the coating stage, and temperature and pressure boundary values ​​in the vulcanization testing stage. An alarm is triggered when any parameter exceeds the preset range. However, this management technology based on single-parameter independent thresholds fails to reflect the coordination relationships between parameters across stages. For example, a minor adjustment to the proportion of a certain raw material in the batching stage, even if it does not exceed its own threshold, may, in combination with a specific coating speed, lead to uneven coating curing. This type of combined deviation is ignored in existing technologies. The quality of rubber-coated products is often the result of the synergistic effect of parameters across multiple stages. The lack of a holistic analysis of the interdependencies between parameters causes anomaly detection to lag until the final quality inspection stage, making it impossible to proactively identify deviations during the process. On the other hand, existing technologies lack a unified, structured way of representing data collected from each process segment. The batching, coating, and vulcanization testing segments use different data protocols and naming systems, with parameters stored independently according to their respective equipment's local identifiers, creating data silos. When attempting cross-segment correlation analysis, manual matching of parameter names against equipment documentation is required, which is labor-intensive, error-prone, and hinders automated full-process mapping and fusion. This situation makes it difficult to construct analytical models reflecting the intrinsic relationships between parameters across segments, even with a large amount of historical batch data, and to capture subtle anomalies through changes in the mathematical characteristics of multidimensional parameter relationships.

[0003] The problems to be solved by this invention focus on two aspects: how to organize heterogeneous real-time data from different work sections into a data structure with standard dimensional relationships through a parameter mapping mechanism, so that the parameters of the entire process can be expressed and calculated as a whole; and how to automatically detect anomalies in the implicit relationships between parameters based on the above data structure, and accurately locate the deviation position to trigger adjustment actions. Summary of the Invention

[0004] This invention provides an intelligent management system for the entire process of rubber coating products based on big data, aiming to solve the problems of lack of unified structured mapping of parameters across processes and difficulty in automatically locating deviations based on parameter relationships in existing technologies. By establishing a full-process data identification matrix and constructing a dynamic state matrix, deviation analysis using the matrix rank is used to identify abnormal parameter relationships, enabling deviation detection and location without human intervention. Adjustment instructions are generated for the deviation process and parameter type, thereby improving the intelligence level of the entire process management.

[0005] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides an intelligent management system for the entire process of rubber coating products based on big data. The system includes a data scheduling module, a work section acquisition module, a matrix mapping module, a rank difference analysis module, a deviation positioning module, and an instruction generation module. The data scheduling module responds to the coating preparation start command by obtaining the product batch code carried in the command and retrieving the full-process data identification matrix corresponding to the batch from the big data platform. The process acquisition module, based on the full-process data identification matrix, sends data reporting commands to the batching, coating, and vulcanization testing equipment to receive real-time process parameters returned by each equipment. The matrix mapping module fills the real-time process parameters into the corresponding dimension positions of the dynamic state matrix according to the parameter mapping relationship in the full-process data identification matrix. The rank difference analysis module calculates the current rank of the dynamic state matrix and compares it with the baseline rank to obtain the rank deviation. The deviation positioning module locates the target dimension position where the parameter value in the dynamic state matrix deviates from the mapping relationship based on the rank deviation. The instruction generation module generates an adjustment instruction based on the corresponding process identifier and parameter type identifier in the full-process data identification matrix for the target dimension position. This system maps real-time process parameters from multiple work sections and heterogeneous sources to a dynamic state matrix and uses the change in matrix rank to measure the synergistic relationship between parameters. It can automatically capture abnormal dimensions of parameter redundancy or missing parameters without relying on fixed thresholds, and realize integrated closed-loop control from data acquisition and deviation identification to instruction generation, which significantly improves the response speed and parameter adjustment accuracy of the entire rubber coating product production process.

[0006] As a technical solution of the present invention, when the data scheduling module obtains the product batch code, it receives the coating preparation start instruction issued by the host computer, parses the instruction type field to confirm that the current instruction is a preparation start type, extracts the timestamp subfield and the production line number subfield from the instruction data payload, and combines the two to generate the product batch code. When retrieving the full-process data identification matrix, the data scheduling module uses the production line number subfield in the product batch code as the index key, retrieves the corresponding production line equipment topology table from the production line configuration database of the big data platform, obtains the section equipment sequence of the production line corresponding to the batch and the data interface protocol of each section equipment according to the topology table, and then reads the data item name and data item unit declared in each data interface protocol in the order of the section equipment sequence, assembles the data item name and data item unit into the column vector of the full-process data identification matrix, and assembles the section equipment sequence into the row vector of the full-process data identification matrix. The process acquisition module extracts the first network address of the batching section equipment, the second network address of the coating section equipment, and the third network address of the vulcanization testing section equipment from the row vectors of the full-process data identification matrix. It then sends data reporting commands containing the data acquisition start time and duration to these network addresses. Upon arrival of the data acquisition start time, the module listens in parallel to the return data ports of each network address to receive the first real-time process parameters of the batching section equipment, the second real-time process parameters of the coating section equipment, and the third real-time process parameters of the vulcanization testing section equipment. If any real-time process parameter is not received within a preset timeout period, the matrix mapping module fills the corresponding position in the dynamic state matrix with a preset default flag value and records the timeout section information, ensuring the integrity of the state matrix construction and the traceability of process visibility issues.

[0007] When filling real-time process parameters into the dynamic state matrix, the matrix mapping module extracts the combination key of the section identifier and parameter type identifier stored in each matrix element of the full-process data identifier matrix. Using the section source identifier and data item name carried by the first real-time process parameter as the first query key, it matches the first combination key in the full-process data identifier matrix to obtain the first row index and first column index of the first real-time process parameter in the dynamic state matrix. Similarly, using the section source identifier and data item name carried by the second real-time process parameter as the second query key, it matches the second row index and second column index; using the section source identifier and data item name carried by the third real-time process parameter as the third query key, it matches the third row index and third column index. The values ​​of each real-time process parameter are then stored in the positions determined by the corresponding row and column indices in the dynamic state matrix. The resulting dynamic state matrix maintains a structurally consistent mapping with the full-process data identifier matrix, providing a standardized data foundation for subsequent deviation analysis based on matrix operations.

[0008] When calculating the matrix rank deviation, the rank difference analysis module performs row echelon simplification on the dynamic state matrix, counts the number of non-zero row vectors in the dynamic state matrix, and determines this number as the current matrix rank. It then retrieves the standard state matrix corresponding to the product batch code from the big data platform and performs the same row echelon simplification on the standard state matrix to obtain the baseline matrix rank. The baseline matrix rank is the inherent parameter dimension of the qualified production state under the corresponding formula for that product batch; in the qualified state, there are no effective linearly independent parameter combinations higher than this dimension. The difference between the current matrix rank and the baseline matrix rank is calculated to obtain the matrix rank deviation. When the matrix rank deviation is positive, it indicates the presence of redundant parameter dimensions in the dynamic state matrix, suggesting unexpected parameter changes or interference. When the matrix rank deviation is negative, it indicates the presence of missing parameter dimensions in the dynamic state matrix, suggesting that some key parameters were not effectively collected or were abnormally lost. Preferably, the standard state matrix adopts the mean matrix of the full-process data identification matrix of historical compliant batches corresponding to the product batch code stored in the big data platform, so that the rank of the benchmark matrix truly reflects the ideal state of multi-stage parameter collaboration under the production line and formula conditions, and improves the reliability and pertinence of deviation judgment.

[0009] The deviation localization module performs differentiated localization operations based on the positive or negative attribute of the matrix rank deviation. When the matrix rank deviation is positive, a linear correlation test is performed on each row vector of the dynamic state matrix to identify redundant row vectors that are linearly correlated with other row vectors. The row position of this redundant row vector in the dynamic state matrix is ​​determined as the target dimension position. When the matrix rank deviation is negative, the number of non-zero elements in each column vector of the dynamic state matrix is ​​counted. Column vectors with fewer than a preset column density threshold are marked as sparse column vectors. The column position of this sparse column vector in the dynamic state matrix is ​​determined as the target dimension position. The target dimension position is recorded in an abnormal dimension localization table, which includes a section field, a parameter field, and a deviation direction field. This localization method directly correlates with structural anomalies within the matrix, accurately separating the specific section parameters causing the rank deviation and avoiding erroneous disturbances to normal parameters.

[0010] The instruction generation module reads the segment field value and parameter field value corresponding to the target dimension position from the anomaly dimension location table. Using the segment field value as the key, it queries the row vector of the full-process data identification matrix to find the matching target row and extract the target segment identifier. Using the parameter field value as the key, it queries the column vector of the full-process data identification matrix to find the matching target column and extract the target parameter type identifier. Based on the deviation direction field, the target segment identifier, and the target parameter type identifier, it assembles an instruction frame for adjustment instructions. This instruction frame includes an opcode field, a segment address field, a parameter code field, and an adjustment direction field. The resulting adjustment instructions directly point to the segment and parameter type that need correction and clearly define the direction of adjustment. This allows for seamless delivery to the corresponding segment equipment for adaptive correction, ensuring that the batching accuracy, coating uniformity, and vulcanization process stability of rubber coating products remain dynamically optimal across the entire process, continuously guaranteeing the consistency of product batch quality.

[0011] The technical effects and advantages provided by the present invention in the above technical solution are as follows: In constructing and mapping the full-process data identification matrix, a batch-coded production line equipment topology table is used. The data item names and units declared in the data interface protocol of each work segment are read sequentially according to the equipment sequence. The data item names and units are assembled into column vectors of the matrix, and the work segment equipment sequence is assembled into row vectors of the matrix, forming the full-process data identification matrix. This matrix assigns a combination key consisting of the work segment identifier and the parameter type identifier to each process parameter. When receiving real-time process parameters returned from the batching, coating, and vulcanization testing sections, the work segment source identifier and data item name carried by each parameter are used as query keys. The corresponding combination key is matched in the full-process data identification matrix to obtain the row and column indices of each parameter in the dynamic state matrix, and then the parameter values ​​are accurately filled into the corresponding dimension positions. This mapping method unifies the semantically heterogeneous and inconsistent parameters of different work sections into matrix elements under the same mathematical framework. The work section affiliation and parameter type relationships between parameters are preserved through the row and column positions of the matrix. As a result, the parameter structures of different batches of products are comparable, providing a structured data foundation for subsequent algebraic analysis of cross-work section associations.

[0012] In terms of rank difference analysis and deviation location, a row echelon simplification operation is performed on the dynamic state matrix after real-time process parameters are filled in. The number of non-zero row vectors is counted as the current matrix rank. The standard state matrix corresponding to the batch code of the product is retrieved from the big data platform and the rank of the benchmark matrix is ​​calculated. The two are compared to obtain the matrix rank deviation. When the matrix rank deviation is positive, it indicates that there are redundant parameter dimensions in the dynamic state matrix. Further linear correlation tests are performed on each row to identify redundant row vectors that are linearly correlated with other row vectors, and their row positions are used as the target dimension positions. When the matrix rank deviation is negative, it indicates that there are missing parameter dimensions. The number of non-zero elements in each column vector is counted, and sparse column vectors with the number of non-zero elements below the column density threshold are marked, and their column positions are used as the target dimension positions.

[0013] By using changes in matrix rank to characterize alterations in the linear relationships between parameters, this approach can detect anomalies where individual parameter values, while not exceeding independent thresholds, have deviated from their combined relationships. For example, the linear synergy that some parameters should maintain may be broken, leading to redundant relationships, or the absence of expected parameter responses may result in insufficient information dimensions. Based on the identified row or column positions, the corresponding work segment and parameter type identifiers are traced within the full-process data identifier matrix, and adjustment command frames are assembled by combining these with the deviation direction field. This method does not rely on manually set fixed thresholds but automatically determines the adjustment targets through implicit changes in parameter dependency structures. This makes deviation detection more sensitive and targeted, and the generated adjustment commands can directly point to the work segments and parameter items requiring intervention, reducing reliance on manual experience analysis. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0015] Figure 1 This is a schematic diagram of the intelligent management system for the entire process of rubber coating products based on big data. Figure 2 This is a flowchart of the data acquisition and parallel monitoring process of the work section's data acquisition module; Figure 3 This is a flowchart illustrating the rank difference analysis and deviation location process; Figure 4 This is a schematic diagram of real-time process parameter data acquisition based on the equipment in the work section; Figure 5 It is a curve comparing the singular values ​​of the standard state matrix and the dynamic state matrix; Figure 6 This is a distribution diagram of parameter redundancy and missing deviations in the rubber coating product manufacturing process. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] See Figure 1 This invention provides an intelligent management system for the entire process of rubber coating products based on big data, including a data scheduling module, a process acquisition module, a matrix mapping module, a rank difference analysis module, a deviation location module, and an instruction generation module. The data scheduling module, in response to a coating preparation start command, obtains the product batch code carried in the coating preparation start command and retrieves the corresponding full-process data identification matrix from the big data platform based on the product batch code. The process acquisition module, based on the full-process data identification matrix, sends data reporting commands to the batching, coating, and vulcanization testing equipment respectively, and receives real-time process parameters returned by each process equipment. The matrix mapping module fills the real-time process parameters into the corresponding dimension positions of the dynamic state matrix according to the parameter mapping relationship in the full-process data identification matrix. The rank difference analysis module calculates the current matrix rank of the dynamic state matrix and compares the current matrix rank with the baseline matrix rank of the dynamic state matrix to obtain the matrix rank deviation. The deviation location module, based on the matrix rank deviation, locates the target dimension position where the parameter value in the dynamic state matrix deviates from the mapping relationship. The instruction generation module generates adjustment instructions based on the section identifier and parameter type identifier corresponding to the target dimension position in the full-process data identifier matrix.

[0018] Example 1: In implementation, the data scheduling module receives the paint preparation start command from the host computer. The paint preparation start command is a fixed-length binary message containing a command type field and a command data payload. The command type field occupies 2 bytes; when the command type field value is 0x01, it indicates a configuration start type command. The data scheduling module parses the command type field of the paint preparation start command, reads its value, and confirms that the received paint preparation start command is a preparation start type command when the read value equals 0x01. From the command data payload of the paint preparation start command, the product batch code field is extracted according to byte offset. The first 8 bytes of the product batch code field are a timestamp subfield, storing a Unix timestamp, and the last 4 bytes are a production line number subfield, storing an unsigned integer. The extracted timestamp subfield value is converted to a decimal string, and the extracted production line number subfield value is converted to a decimal string. Underscores are inserted as separators between the number strings for string concatenation, and the concatenated string is used as the product batch code.

[0019] In some embodiments, after obtaining the product batch code, the data scheduling module performs the operation of retrieving the full-process data identifier matrix from the big data platform. It extracts the production line number subfield from the product batch code, uses the value of the production line number subfield as the index key, and initiates a key-value retrieval request to the production line configuration database of the big data platform. The production line configuration database stores data using the production line number as the key and the production line equipment topology table as the value. The production line equipment topology table records the sequence of equipment sections included in the production line and the data interface protocol access path for each section. The sequence of equipment sections is an ordered list of equipment identifiers, which sequentially includes the equipment identifiers for the batching section, coating section, and vulcanization testing section. The data scheduling module reads the data interface protocol files for the batching section equipment, coating section equipment, and vulcanization testing section equipment sequentially through the data interface protocol access paths recorded in the production line equipment topology table. The data interface protocol files are in JSON format, where an array structure declares a list of data item names, and another array structure declares a list of data item units at the corresponding positions. The data scheduling module parses the data interface protocol of each piece of equipment in the work section according to the order of the equipment sequence, extracts all data item names and data item units declared by the equipment in that work section, and binds the data item names and data item units one by one into data item tuples.

[0020] In practice, all data item tuples from all workstation equipment are merged into a single set. During the merge, duplicate data item names are removed, retaining only the first occurrence of the data item name and its associated unit, forming a column vector set. The total number of data item tuples in the column vector set is... , The number of deduplicated data item names declared for all section equipment. Each section equipment identifier in the section equipment sequence is treated as a row vector element, and the number of row vector elements is the total number of section equipment in the section equipment sequence. Construct a full-process data identification matrix based on row vectors and column vectors. Full-process data identification matrix The row dimension corresponds to the equipment in the work section, and the column dimension corresponds to the data item. Full-process data identification matrix. The construction method is represented by the following formula: ; in, This indicates the total number of equipment in the section's equipment sequence. This represents the total number of duplicate data item pairs in the column vector set after deduplication. For row index variables, The value range is from 1 to integers, For column index variables, The value range is from 1 to integers, Represents the full-process data identification matrix The Middle line, number Column matrix elements, matrix elements The value is determined by the first The equipment identifier string for each work section, the colon separator, and the... The key is a combination of data item name strings concatenated sequentially. This concatenated key is then entered into the end-to-end data identifier matrix. At the corresponding position, complete the retrieval and construction of the full-process data identification matrix.

[0021] See Figure 4 In the graph, the horizontal axis represents the data acquisition time in seconds, ranging from 0 to 3600 seconds, and the vertical axis represents the real-time process parameter values, with units determined according to the data interface protocol. The graph shows the changing trends of two key parameters over time for the batching, coating, and vulcanization testing equipment. The curves are distinguished by different colors and markings, and the legend indicates the corresponding equipment section and parameter number for each curve.

[0022] In the batching section, parameter 1 (blue solid dot) gradually increased from an initial value of approximately 50 to approximately 85 at the end of the data collection, showing a generally stable upward trend. Parameter 2 (orange dashed square) remained around 30 with minimal fluctuations, indicating that this parameter was relatively stable during the data collection period. In the coating section, parameter 1 (green dashed triangle) slowly increased from approximately 60 to approximately 80, with a significant abrupt change occurring at approximately 1800 seconds of data collection, where the parameter value rapidly jumped to above approximately 80 and remained stable. Parameter 2 (red dotted inverted triangle) generally fluctuated around approximately 45 without significant trend changes. In the vulcanization testing section, parameter 1 (purple solid rhombus) continuously increased from approximately 100 to a final value of approximately 170, showing a clear increasing trend, reflecting significant dynamic changes in this section's parameters. Parameter 2 (brown dashed pentagon) remained stable around approximately 80 with minimal fluctuations.

[0023] Example 2: In specific implementation, please refer to Figure 2 The process acquisition module extracts network addresses from the row vectors of the full-process data identifier matrix. Each row of the full-process data identifier matrix corresponds to a process device. In addition to the process device identifier, the row vector also contains the addressable communication address information of that process device in the network. The process acquisition module traverses the row vectors of the full-process data identifier matrix and reads the row offset address field of the row containing the process device corresponding to the batching process device identifier. The row offset address field stores the first network address of the batching process device, which is an IPv4 address string in dotted decimal format. The process acquisition module continues to read the row offset address field of the row containing the process device corresponding to the coating process device identifier to obtain the second network address of the coating process device. The second network address is also an IPv4 address string in dotted decimal format. The section acquisition module reads the row offset address field of the row where the section equipment corresponding to the equipment identifier of the sulfurization testing section is located in the row vector, and obtains the third network address of the sulfurization testing section equipment. The third network address is an IPv4 address string in dotted decimal format.

[0024] In some embodiments, the section acquisition module constructs a data reporting instruction. The data reporting instruction is a structured data message, with the instruction body consisting of an instruction header and a data acquisition configuration field. The instruction header contains a command code field, with a value of 0xA1, indicating that the instruction type is a data reporting instruction. The data acquisition configuration field contains the data acquisition start time and the data acquisition duration. The data acquisition start time is calculated by reading the system's current clock and adding a preset start delay interval, which is set to 3 seconds. The data acquisition duration is determined based on the expected process duration of the coating preparation operation and is set to 3600 seconds. The section acquisition module encapsulates the constructed data reporting instruction into three data packets using the UDP protocol, with the first network address, second network address, and third network address as destination addresses. The destination IP field of the first data packet is filled with the first network address, the destination IP field of the second data packet is filled with the second network address, and the destination IP field of the third data packet is filled with the third network address. The destination port of all three data packets is filled with the instruction receiving port agreed upon by each section's equipment, and the instruction receiving port number is set to 5100. The process acquisition module sends the first data packet to the batching process equipment corresponding to the first network address, sends the second data packet to the coating process equipment corresponding to the second network address, and sends the third data packet to the vulcanization detection process equipment corresponding to the third network address.

[0025] In practice, the section acquisition module continuously polls the system clock, comparing the current time of the system clock with the data acquisition start time. When the current time of the system clock equals the data acquisition start time, the section acquisition module triggers a parallel listening operation. The section acquisition module creates three network socket resources: a first socket, a second socket, and a third socket. The first socket is bound to the local network interface and corresponds to the return data port of the batching section equipment, with the return data port number of the batching section equipment set to 5201. The second socket is bound to the local network interface and corresponds to the return data port of the coating section equipment, with the return data port number of the coating section equipment set to 5202. The third socket is bound to the local network interface and corresponds to the return data port of the vulcanization testing section equipment, with the return data port number of the vulcanization testing section equipment set to 5203. The section acquisition module sets the first, second, and third sockets to non-blocking mode and registers them to the same I / O multiplexing monitor for parallel listening. The I / O multiplexing monitor calls the epoll system call provided by the operating system, and the epoll instance listens for input events of the file descriptors of the above three sockets.

[0026] In some embodiments, when the I / O multiplexing monitor detects an input event from the file descriptor corresponding to the first socket, the process acquisition module reads the first real-time process parameter data stream sent from the batching section equipment through the first socket. The first real-time process parameter data stream is a binary-encoded byte sequence, which is parsed according to the data interface protocol of the batching section equipment to obtain a first real-time process parameter object. When the I / O multiplexing monitor detects an input event from the file descriptor corresponding to the second socket, the process acquisition module reads the second real-time process parameter data stream sent from the coating section equipment through the second socket. The second real-time process parameter data stream is a binary-encoded byte sequence, which is parsed according to the data interface protocol of the coating section equipment to obtain a second real-time process parameter object. When the I / O multiplexing monitor detects an input event from the file descriptor corresponding to the third socket, the process acquisition module reads the third real-time process parameter data stream sent from the vulcanization testing section equipment through the third socket. The third real-time process parameter data stream is a binary-encoded byte sequence, which is parsed according to the data interface protocol of the vulcanization testing section equipment to obtain a third real-time process parameter object. The first real-time process parameter object contains a process source identifier field and multiple key-value pairs consisting of data item names and parameter values. The value of the process source identifier field is the batching section equipment identifier. The second real-time process parameter object contains a process source identifier field and multiple key-value pairs consisting of data item names and parameter values. The value of the process source identifier field is the coating section equipment identifier. The third real-time process parameter object contains a process source identifier field and multiple key-value pairs consisting of data item names and parameter values. The value of the process source identifier field is the vulcanization testing section equipment identifier.

[0027] In practice, the section acquisition module continuously receives process parameters until the data acquisition duration expires. The acquisition completion time for each parameter in the first, second, and third real-time process parameter objects is recorded. This completion time is obtained by the section acquisition module from the system clock after all parameter objects have been received. If the listening socket corresponding to a parameter does not generate an input event within a preset timeout period (set to 5% of the data acquisition duration, i.e., 180 seconds), the section acquisition module marks this parameter as a timeout parameter, stops waiting for it, and sends a timeout notification to the matrix mapping module. The timeout notification includes the section device identifier and data item name corresponding to the timeout parameter. At the moment the data acquisition duration expires, the section acquisition module closes the first, second, and third sockets, releases the epoll instance handle of the I / O multiplexing monitor, and completes the parameter reception process. The section acquisition module then transmits the received first, second, and third real-time process parameter objects, along with the timeout notifications corresponding to the timeout parameters, to the matrix mapping module.

[0028] Optionally, when caching the received real-time process parameter objects locally, the section acquisition module allocates a circular buffer in memory. The capacity of the circular buffer is determined based on the total number of bytes of the received real-time process parameter objects, and the calculation method for the capacity of the circular buffer is expressed by the following formula: ; in, This represents the total capacity of the circular buffer, in bytes. This indicates the total number of section devices being monitored in parallel. The value is 3. This is a variable representing the equipment serial number of the work section, with a value range from 1 to... integers, Indicates the first The total number of data packets expected to be reported by each piece of equipment during the data acquisition period is calculated by dividing the data acquisition period by the reporting cycle of that piece of equipment. The reporting cycle is read from the data interface protocol of that piece of equipment. Indicates the first The length of a single data packet reported by a piece of equipment in a work section is measured in bytes. The length of a single data packet reported is read from the data interface protocol of the equipment in that work section.

[0029] Example 3: In practical implementation, the matrix mapping module extracts the combination key of the work segment identifier and parameter type identifier stored in each matrix element of the full-process data identifier matrix. line, number Each matrix element stores a composite key, which is constructed by concatenating a work segment identifier string, a separator, and a parameter type identifier string, with a colon character as the separator. The matrix mapping module traverses all rows and columns of the entire process data identifier matrix, reading the composite key of each matrix element into memory to build a composite key index table. The composite key index table records the composite key string and its corresponding row number. and column number The correspondence, line number The value range is from 1 to the total number of rows in the entire process data identifier matrix. Integer, column number The value range is from 1 to the total number of columns in the entire process data identifier matrix. Integers.

[0030] The matrix mapping module obtains the first real-time process parameter object transmitted by the section acquisition module. This object contains a section source identifier field and multiple data item names and their corresponding parameter values. The module iterates through each data item name in the first real-time process parameter object, extracting the string corresponding to the data item name and the string from the section source identifier field. It then concatenates these strings in the following order: section source identifier string first, colon separator in the middle, and data item name string last, to obtain the first query key. The module then searches the composite key index table for a composite key that exactly matches the first query key. When a matching composite key is found, it reads the row number from the composite key index table as the first row index and the column number as the first column index. Finally, it stores the parameter value corresponding to the data item name in the first real-time process parameter object at the position determined by the first row and first column indices in the dynamic state matrix.

[0031] The matrix mapping module processes the second real-time process parameter object in the same way. The section source identifier field of the second real-time process parameter object is the coating section equipment identifier. The matrix mapping module concatenates the coating section equipment identifier string with the name string of each data item in the second real-time process parameter object using a colon separator to generate the second query key. It retrieves the matching second query key from the composite key index table, obtaining the corresponding row number as the second row index and the column number as the second column index. The corresponding parameter values ​​from the second real-time process parameter object are then stored in the positions corresponding to the second row and second column indices in the dynamic state matrix.

[0032] The matrix mapping module processes the third real-time process parameter object in the same way. The section source identifier field of the third real-time process parameter object is the equipment identifier for the sulfurization testing section. The matrix mapping module concatenates the string of the sulfurization testing section equipment identifier with the name string of each data item in the third real-time process parameter object using a colon separator to generate the third query key. It then retrieves the matching third query key from the composite key index table, obtaining the corresponding row number as the third row index and the column number as the third column index. Finally, it stores the corresponding parameter values ​​from the third real-time process parameter object in the positions corresponding to the third row and third column indices in the dynamic state matrix.

[0033] Example 4: In specific implementation, please refer to Figure 3 The rank difference analysis module obtains the dynamic state matrix that has been constructed and filled by the matrix mapping module. The number of rows in the dynamic state matrix is ​​denoted as... , The value is equal to the total number of rows in the entire process data identifier matrix, that is, the total number of equipment in the section equipment sequence. The number of columns in the dynamic state matrix is ​​denoted as... , The value is equal to the total number of columns in the entire process data identifier matrix, that is, the total number of data items after deduplication. The rank difference analysis module performs row step reduction on the dynamic state matrix. The row step reduction operation is achieved by sequentially performing column pivoting Gaussian elimination: the rank difference analysis module starts from the first column of the dynamic state matrix and searches for the row number from the current processing row to the current processing row in the current processing column. The row containing the element with the largest absolute value in the range is swapped with the currently processed row; if the current processed column contains elements from the current processed row to the... The absolute value of all elements within the row range is less than the preset non-zero threshold. If the current column is not found, skip the current column and move the column pointer one column to the right to continue processing; if the element with the largest absolute value is greater than or equal to the non-zero threshold, then... If the maximum element is selected as the pivot, the non-zero elements in the current processing column of all rows below it are eliminated using the pivot row. This elimination is performed by multiplying the pivot row by a coefficient and subtracting it from the target row. The coefficient is obtained by dividing the current column element value of the target row by the pivot value. After elimination, the processing row pointer is moved down one row and the processing column pointer is moved up one column. Non-zero threshold. Set as The setting is based on the fact that the real-time process parameter values ​​stored in the dynamic state matrix are all floating-point numbers, and the cumulative range of absolute error in floating-point operations is within a certain range. Scale, using As a threshold, it can effectively distinguish between zero elements and minimal quantities caused by numerical errors.

[0034] After the row echelon simplification operation is completed, the rank difference analysis module traverses all row vectors of the final echelon matrix and counts the elements in the row vectors whose maximum absolute value is greater than the non-zero threshold. The number of rows. When the maximum absolute value of all elements in a row vector exceeds a non-zero threshold. If the row is zero, it is considered a non-zero row vector; otherwise, it is considered a zero row vector. The number of non-zero row vectors is determined as the rank of the current matrix, denoted as . Current matrix rank The calculation method is expressed by the following formula: ; in, This represents the total number of rows in the dynamic state matrix. This represents the total number of columns in the dynamic state matrix. This is a row index variable, with a value range from 1 to... integers, This is a column index variable, with values ​​ranging from 1 to... integers, Represents the first state in the dynamic state matrix. line, number The numerical values ​​of the column elements. Indicates the first All lines Take the absolute value of each element and then take the maximum value. The threshold for determining non-zero values. Values , This is an indicator function. The indicator function takes the value 1 when the condition inside the parentheses is true, and takes the value 0 when the condition is false.

[0035] In some embodiments, the rank difference analysis module retrieves the standard state matrix corresponding to the product batch code from the big data platform. The historical batch database of the big data platform stores production data for multiple historically compliant batches archived by product category. Historically compliant batches refer to batches where, after the entire process of coating preparation, application, and vulcanization testing is completed, quality inspection determines that the three core indicators—coating thickness, coating adhesion, and vulcanization degree—all fall within a preset acceptable range. The construction process of the standard state matrix is ​​as follows: The big data platform queries all historically compliant batches belonging to the same product category as the current product batch code, extracts the process parameter matrix corresponding to the full-process data identifier matrix for each historically compliant batch, and fills in the matrix after mapping the real-time process parameters reported by each section's equipment during actual production according to the full-process data identifier matrix; the arithmetic mean of the values ​​of elements corresponding to the same row and column positions in all process parameter matrices is calculated, and the resulting mean matrix is ​​the standard state matrix. The rank difference analysis module downloads the standard state matrix from the big data platform, performs the same row step reduction operation as described above on the standard state matrix, and obtains the rank of the baseline matrix, denoted as . .

[0036] The rank difference analysis module calculates the rank of the current matrix. With the rank of the benchmark matrix The difference is used to obtain the matrix rank deviation. , Matrix rank deviation The sign and magnitude of represent the deviation of the dynamic state matrix from the standard state matrix. When At that time, the rank difference analysis module determines that there are redundant parameter dimensions in the dynamic state matrix; when At that time, the rank difference analysis module determined that there were missing parameter dimensions in the dynamic state matrix. Matrix rank deviation. The numerical value and sign are transmitted to the deviation positioning module.

[0037] In practical implementation, the deviation positioning module receives the matrix rank deviation. and the dynamic state matrix. When When the value is positive, the deviation positioning module performs a linear correlation consistency check on each row vector of the dynamic state matrix: First, it performs singular value decomposition on the row vectors of the reference state matrix and records the relative singular values ​​(the ratio of the relative singular value to the maximum singular value) that are less than the redundancy judgment threshold. The row vector combinations are approximately linearly correlated under qualified production conditions. Then, the same singular value decomposition is performed on the row vectors of the current dynamic state matrix to identify the row vector combinations that were originally approximately linearly correlated, where the relative singular value is greater than or equal to the redundancy threshold. The combination of these parameters determines the corresponding row vector as the target dimension position for the redundant parameter, and this position is recorded in the target dimension position set. For each identified redundant row vector, the deviation location module adds a new record to the abnormal dimension location table. The section field in the abnormal dimension location table is filled with the section equipment identifier corresponding to the redundant row vector. The section equipment identifier is obtained by looking up the row vector of the full process data identifier matrix through the row index of the dynamic state matrix. The parameter field is filled with the parameter type identifier corresponding to the column containing the non-zero element in the redundant row vector. The parameter type identifier is obtained by looking up the column vector of the full process data identifier matrix through the column index. The deviation direction field is filled with the string "redundant".

[0038] When the matrix rank deviation When the value is negative, the deviation positioning module counts the number of non-zero elements in each column vector of the dynamic state matrix. Each column of the dynamic state matrix is ​​a column vector, and the deviation positioning module counts the absolute values ​​of each column vector that are greater than the non-zero threshold. The number of elements is the number of non-zero elements. The deviation positioning module compares the number of non-zero elements with a preset column density threshold. Comparison, preset column density threshold Set as the total number of rows in the dynamic state matrix Twenty percent, that is The setting is based on the premise that a certain proportion of valid parameter values ​​are required in the vulcanization detection section and the batching section to ensure sufficient observation of parameter dimensions. When the number of non-zero elements is less than 20% of the total number of rows, the parameter dimension corresponding to that column is judged as insufficiently observed. If the number of non-zero elements is less than the preset column density threshold... The deviation positioning module marks the column vector as a sparse column vector and records the column position of the sparse column vector in the dynamic state matrix into the target dimension position set. For each sparse column vector, the deviation positioning module adds a new record to the abnormal dimension positioning table. The section field in the abnormal dimension positioning table is filled with the section equipment identifier to which the parameter type of the sparse column vector belongs. The section equipment identifier is obtained by looking up the row vector of the full process data identifier matrix by reversing the row index of the non-zero element of the sparse column vector. The parameter field is filled with the parameter type identifier corresponding to the sparse column vector. The parameter type identifier is obtained by looking up the column vector of the full process data identifier matrix by reversing the column index of the sparse column vector. The deviation direction field is filled with the string "missing".

[0039] The deviation location module passes the complete anomaly dimension location table to the instruction generation module. Each record in the anomaly dimension location table contains three parts: the section field value, the parameter field value, and the deviation direction field value. The deviation direction field value is fixed to one of two discrete strings: "redundant" or "missing".

[0040] See Figure 5In the figure, the horizontal axis represents the singular value index, and the vertical axis uses a logarithmic scale to represent the magnitude of the corresponding singular value. The vertical axis is labeled "Singular Values ​​(Logarithmic)" and the scale covers the range from 10^1 to 10^-9. The figure contains two curves: the blue solid line represents the singular values ​​of the standard state matrix, and the red dashed line represents the singular values ​​of the dynamic state matrix. The legends are labeled "Standard State Matrix Singular Values" and "Dynamic State Matrix Singular Values," respectively.

[0041] From the trend of the curves, the singular values ​​of both matrices show a characteristic of first decreasing sharply and then approaching zero. Furthermore, within the index range of approximately 0 to 200, the magnitudes of the singular values ​​of the two matrices largely overlap, indicating that the numerical characteristics of the dynamic state matrix in this dimension are highly consistent with those of the standard state matrix. As the index increases to the range of 200 to 400, the singular values ​​gradually decrease to the order of 10^-6 to 10^-8. The singular values ​​of the standard state matrix (blue solid line) show a relatively stable trend and gradually approach zero, while the singular values ​​of the dynamic state matrix (red dashed line) fluctuate significantly, indicating that the dynamic state matrix exhibits many subtle changes and fluctuations in this dimension.

[0042] In the high-order interval of 400 to 500, the singular values ​​of the standard state matrix remain at an extremely low level of approximately 10^-9, approaching zero, indicating that the higher-order dimensions of the matrix have virtually no significant variance information, consistent with the theoretical expectation of matrix rank. In contrast, the singular values ​​of the dynamic state matrix exhibit dramatic fluctuations and occasional rebounds in this interval, suggesting abnormal fluctuations in the higher-order dimensions of the dynamic state matrix, possibly corresponding to redundant or anomalous parameter dimensions detected by the rank difference analysis module.

[0043] Example 5: In practice, the instruction generation module obtains the anomaly dimension location table from the deviation location module. The anomaly dimension location table is stored using a linear list data structure. Each record in the linear list is a triplet, containing a section field value, a parameter field value, and a deviation direction field value. The instruction generation module reads the first record in the anomaly dimension location table, extracts the section field value (a text string representing the section equipment identifier), extracts the parameter field value (a text string representing the parameter type identifier), and extracts the deviation direction field value (a fixed value of either "redundant" or "missing").

[0044] In practical implementation, the instruction generation module uses the extracted work segment field value as the query key to perform matching and retrieval in the row vector set of the full-process data identifier matrix. Each row of the full-process data identifier matrix corresponds to a work segment device, and the row vector structure contains a work segment device identifier field and a work segment device network address field. The instruction generation module traverses all row vectors of the full-process data identifier matrix, comparing the string of the work segment device identifier field in the row vector with the work segment field value string. When the two strings are completely identical, the instruction generation module determines the current row vector as the target row. The instruction generation module extracts the target work segment identifier from the work segment device identifier field of the target row. The target work segment identifier is a full-length work segment device identifier string. It also extracts the target work segment network address from the work segment device network address field of the target row. The target work segment network address is the addressable address of the work segment device in the current network.

[0045] In some embodiments, the instruction generation module uses the extracted parameter field values ​​as query keys to perform matching and retrieval in the column vector set of the full-process data identifier matrix. Each column of the full-process data identifier matrix corresponds to a parameter type, and the column vector structure contains a parameter type identifier field and a parameter code field. The instruction generation module traverses all column vectors of the full-process data identifier matrix, comparing the string of the parameter type identifier field in the column vector with the parameter field value string. When the two strings are completely identical, the instruction generation module determines the current column vector as the target column. The instruction generation module extracts the target parameter type identifier from the parameter type identifier field of the target column. The target parameter type identifier is a full-length parameter type identifier string. It also extracts the target parameter code from the parameter code field of the target column. The target parameter code is an integer code value. The value range of the parameter code field is an integer from 0 to 65535. The parameter codes are assigned in ascending order according to the registration order of the parameter types in the big data platform during the initial system configuration.

[0046] In practice, the instruction generation module assembles instruction frames for adjustment instructions based on the deviation direction field value, the target section identifier, and the target parameter type identifier. The instruction frame is a fixed-length binary byte sequence, and its format is defined as follows: the first byte is the opcode field; the second and third bytes are combined to form the section address field; the fourth and fifth bytes are combined to form the parameter encoding field; the sixth byte is the adjustment direction field; and the seventh and eighth bytes are combined to form the frame check field.

[0047] The value of the opcode field is determined by the value of the deviation direction field. When the deviation direction field value is "redundant", the opcode field is assigned the hexadecimal value 0xA2, which represents a reduction adjustment command for the section parameter. When the deviation direction field value is "missing", the opcode field is assigned the hexadecimal value 0xA1, which represents an incremental adjustment command for the section parameter. The section address field is obtained by converting the communication address of the section equipment corresponding to the target section identifier. The conversion method is as follows: the instruction generation module retrieves the corresponding section address integer value by using the target section identifier as the key in the section identifier-to-section address mapping table maintained during the system initialization phase. The integer value of the section address is an integer ranging from 1 to 65534. This integer value is encoded into two bytes in big-endian byte order, with the high byte stored in the second byte position of the instruction frame and the low byte stored in the third byte position of the instruction frame. The parameter encoding field is directly filled with the integer value of the target parameter encoding. The encoding method also adopts big-endian byte order, with the high byte stored in the fourth byte position of the instruction frame and the low byte stored in the fifth byte position of the instruction frame.

[0048] The value of the adjustment direction field is calculated using the following formula: ; in, The value of the adjustment direction field is a single-byte unsigned integer. The direction field value is stored at the sixth byte of the instruction frame. When the value is 0x00, it instructs the corresponding execution section to lower the corresponding parameter value by a step amount, adjusting the value of the direction field. When the value is 0x01, it instructs the corresponding execution section to increase the corresponding parameter value by a step increment. This indicates the value of the deviation direction field read from the current record in the anomaly dimension location table. The data type is string. It can only take the values ​​of two specific string constants: "redundant" or "missing".

[0049] In practice, the frame check field covers the first six bytes of binary data in the instruction frame. It is calculated using a cyclic redundancy check algorithm, with the calculation polynomial being the CRC-16-CCITT standard polynomial. The instruction generation module fills the calculated check value into the seventh and eighth bytes of the instruction frame in big-endian byte order. The instruction generation module then sends the assembled adjustment instruction through the communication channel of the section acquisition module to the section equipment control interface pointed to by the target section network address, completing the generation and issuance of the adjustment instruction.

[0050] Optionally, after assembling an adjustment instruction, the instruction generation module continues to read the next record in the anomaly dimension location table, repeating the operations of matching section field values, matching parameter field values, and assembling instruction frames until all records in the anomaly dimension location table have been processed into corresponding adjustment instructions and sent.

[0051] See Figure 6 In the graph, the horizontal axis represents the parameter number, ranging from 0 to 500, and the vertical axis represents the percentage of deviation (%). The legend shows two types of deviation data points: "redundant parameter deviation" (red dots) and "missing parameter deviation" (blue squares). The red dots are distributed in the positive range of the vertical axis, with the deviation percentage concentrated between approximately 6% and 30%, and the overall fluctuation is relatively uniform, without obvious clustering or trending increases or decreases. This indicates that redundant parameter deviation is distributed across all parameter numbers and its amplitude is constant. The blue squares are distributed in the negative range of the vertical axis, with the deviation percentage concentrated between approximately -30% and -5%, also showing a discrete and uniform distribution without obvious staged clustering. This indicates that missing parameter deviation appears at multiple parameter number points and its amplitude remains stable. The two types of deviation data points are independently distributed on both sides of the vertical axis, with the zero point marked by a gray dashed line, further highlighting the directional distinction between redundant and missing deviations.

[0052] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A big data-based intelligent management system for the entire process of rubber coating products, characterized in that: The system includes: The data scheduling module, in response to the coating preparation start command, obtains the product batch code carried in the coating preparation start command, and retrieves the full-process data identification matrix corresponding to the batch product from the big data platform according to the product batch code; The process acquisition module sends data reporting instructions to the batching process equipment, coating process equipment and vulcanization testing process equipment respectively, based on the full-process data identification matrix, in order to receive real-time process parameters returned by each process equipment. The matrix mapping module fills the real-time process parameters into the corresponding dimension positions of the dynamic state matrix according to the parameter mapping relationship in the full-process data identifier matrix. The rank difference analysis module calculates the current matrix rank of the dynamic state matrix and compares the current matrix rank with the reference matrix rank of the dynamic state matrix to obtain the matrix rank deviation. The deviation positioning module locates the target dimension position of the parameter value deviation from the mapping relationship in the dynamic state matrix based on the matrix rank deviation. The instruction generation module generates adjustment instructions based on the section identifier and parameter type identifier corresponding to the target dimension position in the full-process data identifier matrix.

2. The intelligent management system for the entire process of rubber coating products based on big data as described in claim 1, characterized in that, The step of responding to the coating preparation start command by obtaining the product batch code carried in the coating preparation start command specifically includes: Receive the paint preparation start command issued by the host computer, the paint preparation start command includes an instruction type field and an instruction data payload; Parse the instruction type field to confirm that the current instruction is a configuration startup type; Extract the product batch code field from the instruction data payload. The product batch code field includes a timestamp subfield and a production line number subfield. The timestamp subfield and the production line number subfield are combined to generate the product batch code.

3. The intelligent management system for the entire process of rubber coating products based on big data as described in claim 2, characterized in that, Based on the product batch code, the full-process data identification matrix corresponding to the batch of products is retrieved from the big data platform, specifically including: Using the production line number subfield in the product batch code as the index key, the corresponding production line equipment topology table is retrieved from the production line configuration database of the big data platform; Based on the production line equipment topology table, obtain the sequence of equipment sections of the production line corresponding to the product batch code and the data interface protocol of each equipment section; According to the equipment sequence of the section, read the data item name and data item unit declared in the data interface protocol of each section equipment in sequence; The data item name and the data item unit are assembled into a column vector of the full-process data identifier matrix, and the section equipment sequence is assembled into a row vector of the full-process data identifier matrix.

4. The intelligent management system for the entire process of rubber coating products based on big data as described in claim 3, characterized in that, Based on the aforementioned full-process data identification matrix, data reporting instructions are sent to the batching section equipment, coating section equipment, and vulcanization testing section equipment, respectively, to receive real-time process parameters returned by each section equipment, specifically including: Extract the first network address of the batching section equipment, the second network address of the coating section equipment, and the third network address of the vulcanization testing section equipment from the row vectors of the full-process data identification matrix; The data reporting instructions are sent to the first network address, the second network address, and the third network address, respectively. The data reporting instructions include the data collection start time and the data collection duration. When the data acquisition start time arrives, the return data ports of the first network address, the second network address, and the third network address are monitored in parallel. The return data port receives the first real-time process parameters of the batching section equipment, the second real-time process parameters of the coating section equipment, and the third real-time process parameters of the vulcanization detection section equipment.

5. The intelligent management system for the entire process of rubber coating products based on big data as described in claim 4, characterized in that, The real-time process parameters are filled into the corresponding dimension positions of the dynamic state matrix according to the parameter mapping relationship in the full-process data identification matrix, specifically including: Extract the combination key of the work segment identifier and parameter type identifier stored in each matrix element of the full-process data identifier matrix; Using the section source identifier and data item name carried by the first real-time process parameter as the first query key, a first combination key consistent with the first query key is matched in the full-process data identifier matrix to obtain the first row index and first column index of the first real-time process parameter in the dynamic state matrix; Using the section source identifier and data item name carried by the second real-time process parameter as the second query key, a second combination key consistent with the second query key is matched in the full-process data identifier matrix to obtain the second row index and the second column index of the second real-time process parameter in the dynamic state matrix; Using the section source identifier and data item name carried by the third real-time process parameter as the third query key, a third combination key consistent with the third query key is matched in the full-process data identifier matrix to obtain the third row index and third column index of the third real-time process parameter in the dynamic state matrix; The values ​​of the first real-time process parameters are stored in the positions corresponding to the first row index and the first column index in the dynamic state matrix. The values ​​of the second real-time process parameters are stored in the positions corresponding to the second row index and the second column index in the dynamic state matrix. The values ​​of the third real-time process parameters are stored in the positions corresponding to the third row index and the third column index in the dynamic state matrix.

6. The intelligent management system for the entire process of rubber coating products based on big data as described in claim 5, characterized in that, Calculate the current rank of the dynamic state matrix and compare it with the baseline rank of the dynamic state matrix to obtain the rank deviation, specifically including: Perform row echelon simplification on the dynamic state matrix, count the number of non-zero row vectors in the dynamic state matrix, and determine the number of non-zero row vectors as the rank of the current matrix; The standard state matrix corresponding to the product batch code is retrieved from the big data platform, and the row step reduction operation is performed on the standard state matrix to obtain the rank of the baseline matrix; Calculate the difference between the current matrix rank and the reference matrix rank to obtain the matrix rank deviation. When the matrix rank deviation is positive, it is determined that there are redundant parameter dimensions in the dynamic state matrix; When the matrix rank deviation is negative, it is determined that there is a missing parameter dimension in the dynamic state matrix.

7. The intelligent management system for the entire process of rubber coating products based on big data as described in claim 6, characterized in that, Based on the matrix rank deviation, locate the target dimension position where the parameter values ​​in the dynamic state matrix deviate from the mapping relationship, specifically including: When the matrix rank deviation is positive, a linear correlation test is performed on each row vector of the dynamic state matrix to identify row vectors that are linearly correlated in the baseline state matrix but become linearly independent in the current dynamic state matrix. The row position of the redundant row vector in the dynamic state matrix is ​​determined as the target dimension position. When the matrix rank deviation is negative, the number of non-zero elements in each column vector of the dynamic state matrix is ​​counted, and the column vectors with fewer than the preset column density threshold are marked as sparse column vectors. The column position of the sparse column vector in the dynamic state matrix is ​​determined as the target dimension position. The target dimension position is recorded in the abnormal dimension positioning table, which includes a section field, a parameter field, and a deviation direction field.

8. The intelligent management system for the entire process of rubber coating products based on big data as described in claim 7, characterized in that, Based on the segment identifier and parameter type identifier corresponding to the target dimension position in the full-process data identifier matrix, an adjustment instruction is generated, specifically including: Read the section field value and parameter field value corresponding to the target dimension position from the anomaly dimension location table; Using the section field value as the key, query the matching target row in the row vector of the full-process data identifier matrix, and extract the target section identifier from the target row; Using the parameter field value as the key, query the matching target column in the column vector of the full-process data identifier matrix, and extract the target parameter type identifier from the target column; Based on the deviation direction field, the target section identifier, and the target parameter type identifier, an instruction frame for the adjustment instruction is assembled. The instruction frame includes an opcode field, a section address field, a parameter encoding field, and an adjustment direction field.

9. The intelligent management system for the entire process of rubber coating products based on big data as described in claim 5, characterized in that, When there is a timeout parameter among the first real-time process parameter, the second real-time process parameter, and the third real-time process parameter that has not been received within a preset timeout period, a preset default flag value is filled into the corresponding position in the dynamic state matrix, and the timeout section information is recorded.

10. The intelligent management system for the entire process of rubber coating products based on big data as described in claim 6, characterized in that, The standard status matrix is ​​the mean matrix of the process parameter matrices of historical compliant batches corresponding to the product batch codes stored in the big data platform. The process parameter matrix is ​​a numerical matrix of the same dimension obtained by filling in the real-time process parameters of the historical compliant batches according to the mapping relationship of the full-process data identification matrix.

Images