A lace fabric incoming quality evaluation method based on data analysis

Abstract

The application discloses a lace fabric incoming quality evaluation method based on data analysis and belongs to the technical field of quality management, and specifically comprises the following steps: receiving a process parameter change list, inquiring a pre-constructed lace fabric process knowledge graph, and identifying lace fabric incoming quality attributes affected by the changed process parameters; generating a predictive detection scheme based on the identified lace fabric quality attribute types; searching all batches of lace fabric inventory information in the current warehouse to determine the range of lace fabric batches that need to be detected; starting the predictive detection operation on the determined lace fabric batches, collecting quality data according to the detection parameters in the predictive detection scheme; and integrating the collected quality data with the lace fabric batch information to generate a quality evaluation record; and the application realizes a fundamental change from passive response to active prediction in lace fabric incoming quality evaluation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of quality management, and in particular to a lace fabric incoming quality evaluation method based on data analysis. BACKGROUND

[0002] In the textile and home decoration industry, the incoming quality of lace fabric directly determines the appearance, durability and market acceptance of downstream finished products. With the penetration of intelligent manufacturing technology in the textile field, the production process parameters of lace fabric need to be dynamically adjusted according to order requirements, and the warehouse inventory batches are increasing. The traditional quality evaluation mode relying on manual experience has been difficult to meet the real-time and accuracy requirements of batch production evaluation, and the industry urgently needs to build a data-driven systematic incoming quality evaluation system.

[0003] At present, the lace fabric incoming quality evaluation has formed a basic process framework: the downstream production system records process parameter change information, the warehouse management system stores batch inventory data, and the detection link collects surface images, physical properties and other data through standardized equipment. Some enterprises have also established a quality standard library to realize preliminary comparison of detection data and standard threshold. These methods can complete basic quality screening in the scene of stable process parameters and fewer inventory batches, to a certain extent, ensuring the smoothness of the production process.

[0004] However, the existing evaluation scheme is essentially a passive response, which is difficult to achieve active prevention, and has the following core limitations: first, in the face of downstream process adjustment, the system can only passively receive change information and cannot actively judge the impact of such changes on raw material quality. It must rely on manual decision to determine whether to detect, which not only easily misses key quality risk points due to insufficient manual experience, but also causes the prediction of quality problems to lag behind the process change rhythm; second, the detection scheme is mostly a fixed mode, which cannot flexibly adjust the detection detail level and sampling quantity according to the impact of process changes on quality. Either the detection is not sufficient for large impact changes, or the detection is redundant for small impact changes, which cannot adapt to the actual needs in different process change scenarios. SUMMARY

[0005] The purpose of the present application is to provide a lace fabric incoming quality evaluation method based on data analysis, which solves the following technical problems:

[0006] The existing lace fabric incoming quality evaluation scheme is essentially a passive response, and the system only passively receives change information when the downstream process is adjusted, and needs manual decision to determine whether to detect, which easily misses key quality risk points and lags behind the quality prediction. The detection scheme is fixed, and cannot flexibly adjust the detection detail level and sampling quantity according to the impact of process changes on quality.

[0007] The purpose of the present application can be achieved by the following technical solutions:

[0008] A lace fabric incoming quality evaluation method based on data analysis, comprising the following steps:

[0009] S1, receiving a process parameter change list, the process parameter change list containing changed process parameter items and their parameter values;

[0010] S2, according to the process parameter change list, querying the pre-constructed lace fabric process knowledge graph, identifying the lace fabric incoming quality attributes affected by the changed process parameters;

[0011] S3, based on the identified lace fabric quality attribute types, generating a predictive detection scheme, the predictive detection scheme containing a set of quality attributes that need to be detected and corresponding detection parameters;

[0012] S4, according to the predictive detection scheme, retrieving all batches of lace fabric inventory information in the current warehouse to determine the range of lace fabric batches that need to be detected;

[0013] S5, starting the predictive detection operation on the determined lace fabric batches, and performing quality data collection according to the detection parameters in the predictive detection scheme;

[0014] S6, integrating the collected quality data with the lace fabric batch information to generate a quality evaluation record.

[0015] As a further scheme of the present application: in S1, the specific process of receiving the process parameter change list is:

[0016] After the downstream production management system completes the process parameter configuration update, a process parameter change notification file is automatically generated, the process parameter change notification file adopts a structured text format, contains a change timestamp, a change operator identifier, a list of changed process parameter item names, records the parameter values before and after the change for each changed process parameter item, the system receives the process parameter change notification file through the file transmission service, parses the file content and verifies the data integrity, stores the changed process parameter items and their parameter values parsed as the process parameter change list, and the process parameter change list maintains a data table structure in the memory.

[0017] As a further scheme of the present application: in S2, the specific process of querying the pre-constructed lace fabric process knowledge graph is:

[0018] The lace fabric process knowledge graph adopts a graph structure of nodes and edges for data storage, the node types include process parameter nodes and lace fabric quality attribute nodes, the edge types represent the influence relationship of process parameter nodes on lace fabric quality attribute nodes, each edge contains an influence intensity coefficient and an influence direction identifier;

[0019] According to the changed process parameter item name in the process parameter change list, the corresponding process parameter node is located in the knowledge graph, all outgoing edges starting from the process parameter node are traversed, a set of directly connected lace fabric quality attribute nodes is obtained, according to the influence intensity coefficient in the edge attribute and the variation range of the process parameter value, the influence quantitative value of each lace fabric quality attribute node is calculated, the lace fabric quality attribute node with an influence quantitative value exceeding an activation threshold is marked as an affected attribute, and a lace fabric incoming quality attribute list is formed.

[0020] As a further scheme of the application: in S3, the specific process of generating the predictive detection scheme is:

[0021] Read each lace fabric quality attribute identifier in the lace fabric incoming quality attribute list, query the detection rule library, obtain the standard detection method number and the reference detection parameter corresponding to each lace fabric quality attribute identifier, and adaptively adjust the reference detection parameter according to the influence quantitative value calculated in the lace fabric process knowledge graph, the detection parameter adjustment includes increasing the sampling point density, improving the detection precision level, and prolonging the detection observation time, the adjusted detection parameter is bound with the lace fabric quality attribute identifier, a detection task item is formed, all detection task items are combined into a predictive detection scheme, and the predictive detection scheme includes a scheme number, a generation timestamp and a detection task item list.

[0022] As a further scheme of the application: the specific process of adaptively adjusting the reference detection parameter is:

[0023] According to the size of the influence quantitative value, the detection parameter adjustment level is determined, the influence quantitative value is divided into a plurality of continuous intervals, each interval corresponds to an adjustment level, the detection parameter adjustment level includes a first level adjustment, a second level adjustment and a third level adjustment, each adjustment level corresponds to a set proportion of detection sampling point quantity and a set level of detection precision.

[0024] As a further scheme of the application: in S4, the specific process of determining the lace fabric batch range that needs to be detected is:

[0025] Query the lace fabric inventory database in the warehouse management system, obtain the inventory records of all in-stock lace fabric batches, the inventory records include lace fabric batch number, storage time, storage position and fabric component specification, according to the lace fabric quality attributes involved in the predictive detection scheme, combined with the lace fabric component specification, the correlation score of each lace fabric batch with the detection scheme is calculated, the correlation score threshold is set, the lace fabric batches with a correlation score greater than or equal to the threshold are screened, the lace fabric batch range that needs to be detected is formed, and the lace fabric batch range is sorted in descending order of correlation score.

[0026] As a further scheme of the present application: the specific process of the correlation score calculation is:

[0027] A correlation weight matrix of lace fabric component specifications and quality attributes is constructed, the rows of the matrix correspond to different lace fabric component specification types, the columns of the matrix correspond to different lace fabric quality attributes, and the weight values in the matrix represent the sensitivity of the lace fabric component specifications to the quality attributes;

[0028] For each lace fabric batch, the lace fabric component specification type thereof is extracted, the corresponding row vector is located in the weight matrix, the lace fabric quality attribute list involved in the predictive detection scheme is extracted from the row vector, the corresponding weight values are extracted from the row vector, all the extracted weight values are weighted and summed, the weight is the influence quantization value of each quality attribute in the predictive detection scheme, and the weighted sum result is normalized to obtain the correlation score of the lace fabric batch.

[0029] As a further scheme of the present application: in S5, the specific process of starting the predictive detection operation is:

[0030] A detection task dispatch order is generated according to the range of lace fabric batches, the detection task dispatch order contains a list of lace fabric batch numbers and a copy of the predictive detection scheme corresponding to each batch, the detection task dispatch order is distributed to an automatic detection equipment control center, the automatic detection equipment control center analyzes the detection task dispatch order, allocates detection equipment resources, and the detection equipment performs quality data collection according to the detection parameters in the predictive detection scheme, the collection process includes lace fabric surface image acquisition, physical performance testing, and chemical component analysis, and the collected raw data is standardized in format and converted into quality data files with a unified structure.

[0031] As a further scheme of the present application: the specific process of generating the quality evaluation record is:

[0032] The quality data file corresponding to each lace fabric batch is read, the quality index measurement value is extracted, the quality index measurement value is compared with the allowable range in the lace fabric quality standard library, the quality evaluation conclusion of a single lace fabric batch is generated, the quality evaluation conclusion includes a qualified judgment result and an abnormal index note, the quality evaluation conclusions of all lace fabric batches are summarized to form a total quality evaluation record, and the total quality evaluation record includes an evaluation date, a number of lace fabric batches involved, a total qualified rate statistic, and a detailed evaluation data table.

[0033] The present application has the following advantages:

[0034] The application solves the queuing delay problem caused by the need to re-initiate the detection task due to process adjustment in the traditional quality evaluation process by automatically analyzing the downstream process parameter change list and querying the lace fabric process knowledge graph, accurately identifying the affected incoming material quality attributes, and dynamically generating a predicted detection scheme based on the influence quantization value. By constructing the correlation weight matrix of the lace fabric component specification and quality attributes, the correlation score of each batch and the detection scheme is calculated, and intelligent filtering of the detection range is realized, avoiding resource waste of full batch detection. The system adaptively optimizes the sampling point density and detection accuracy requirements according to the detection parameter adjustment level, and performs standardized data collection through automatic equipment, ensuring the consistency and reliability of the detection results. Finally, the evaluation record is generated by automatically integrating the quality data and batch information, forming a closed-loop management from process change to quality feedback, significantly improving the timeliness and accuracy of the lace fabric incoming material quality evaluation, and providing timely and reliable data support for downstream process adjustment. BRIEF DESCRIPTION OF DRAWINGS

[0035] The application will be further described below with reference to the accompanying drawings.

[0036] Figure 1 is a flowchart of the application. DETAILED DESCRIPTION

[0037] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the application.

[0038] Please refer to Figure 1 The application is a lace fabric incoming material quality evaluation method based on data analysis, which comprises the following steps:

[0039] S1, receiving a process parameter change list, wherein the process parameter change list comprises changed process parameter items and their parameter values;

[0040] S2, querying a pre-constructed lace fabric process knowledge graph according to the process parameter change list, and identifying the lace fabric incoming material quality attributes affected by the changed process parameters;

[0041] S3, generating a predicted detection scheme based on the identified lace fabric quality attribute types, wherein the predicted detection scheme comprises a set of quality attributes that need to be detected and corresponding detection parameters;

[0042] S4, according to the prediction type detection scheme, retrieve lace fabric inventory information of all batches in the current warehouse, determine the range of lace fabric batches that need to be detected;

[0043] S5, start the prediction type detection operation on the determined lace fabric batches, and perform quality data collection according to the detection parameters in the prediction type detection scheme;

[0044] S6, integrate the collected quality data with the lace fabric batch information to generate a quality evaluation record.

[0045] In the S1, the specific process of receiving the process parameter change list is:

[0046] The downstream production management system has a parameter configuration update listening module, which monitors the modification state of the system internal process parameter configuration file in real time. When the operator completes the process parameter adjustment through the parameter configuration interface of the production management system and submits and saves, the configuration update listening module captures the write completion event of the configuration file, and then triggers the process parameter change notification file generation process. The system calls the preset file generation interface, and the interface constructs the file content according to the structured text format specification, ensuring the standardization and parsability of data organization.

[0047] In the structured text format, the change timestamp field strictly follows the ISO8601 standard format, accurately records the completion time of the parameter change operation, and provides a time reference for subsequent tracing; The change operator identification field fills in the unique user identification code in the system, which is bound to the operator's permission account, and can be associated with the specific operator information; The list of changed process parameter item names exists in the form of a string array, and each element in the array corresponds to the standard name of a changed process parameter item. The name is consistent with the parameter item naming in the system internal process parameter database. For each changed process parameter item in the list, the file will separately open a data segment to record the parameter value before the change and the parameter value after the change. The value type strictly matches the definition type of the parameter item, such as floating point or integer storage for numerical parameters, and preset enumeration value string storage for enumeration parameters, to ensure the accuracy of the parameter value and the consistency of the data type.

[0048] After the process parameter change notification file is generated, the downstream production management system sends it to the quality evaluation system corresponding to the application through a file transfer service. The file transfer service is implemented based on an industrial-level data transmission protocol, common protocols include SFTP, FTP or a RESTful file transfer interface based on HTTP, a data verification mechanism is enabled during transmission, the CRC32 check value or MD5 hash value of the file is calculated, and the check value is transmitted with the file to verify the integrity of the file at the receiving end. The file receiving module of the quality evaluation system continuously monitors the specified transmission port or file receiving directory. When a new file is detected, the receiving process is automatically triggered, the data integrity verification is first completed by comparing the file check values of the transmission end and the receiving end, if the verification fails, a retransmission request is sent to the downstream production management system, if the verification passes, the file parsing process is started.

[0049] The file parsing process is executed by a structured text parser. The parser calls the corresponding parsing algorithm according to the format type of the file, and extracts the change timestamp, change operator identifier, change process parameter item name list and pre-change and post-change values of each parameter item in the file field by field. During the parsing process, data legality verification is performed synchronously, including field missing check (to ensure that the core field is not missing), data type verification (to verify that the value type of each field matches the pre-set definition), parameter value rationality verification (to check whether the pre-change and post-change parameter values are within the allowed value range of the process parameter item), if a legality problem is found, a parsing exception log is generated and the process is suspended, and manual troubleshooting is required; if the parsing and verification are passed, the extracted change process parameter item and its corresponding pre-change and post-change parameter values, change timestamp, operator identifier and other data are stored as a process parameter change list.

[0050] The process parameter change list is maintained in memory as a two-dimensional data table structure. The data table is implemented using an in-memory database or a dynamic array. The table structure includes fixed column fields: parameter item ID (a unique identifier associated with the system process parameter database), parameter item name, pre-change parameter value, post-change parameter value, change timestamp, operator ID. The data table uses the parameter item ID as the primary key index to build a hash index structure, supporting fast query and call of specific parameter item change information by subsequent modules. At the same time, the data table sets a data expiration mechanism to only retain the parameter change data required by the current evaluation process, avoiding excessive memory resource occupation.

[0051] In the S2, the specific process of querying the pre-constructed lace fabric process knowledge graph is:

[0052] The lace fabric process knowledge graph is constructed based on a graph database, adopts a node-edge graph structure model to realize data storage, and the storage engine of the graph database manages the association relationship between nodes and edges through an adjacency table or an adjacency matrix to ensure efficient association query performance. The nodes in the knowledge graph are divided into two categories, namely process parameter nodes and lace fabric quality attribute nodes, and both types of nodes include basic attributes and extended attributes: the basic attributes of the process parameter nodes include parameter item ID, parameter item name, parameter type (such as numerical type, enumeration type, and Boolean type), and parameter value range, and the extended attributes include process action description of the parameter and common adjustment scenarios; the basic attributes of the lace fabric quality attribute nodes include quality attribute ID, quality attribute name, and attribute description (such as “fabric tensile strength” and “pattern clarity”), and the extended attributes include detection method identifier of the quality attribute and associated downstream finished product quality requirements.

[0053] The edges in the knowledge graph are used to represent the influence relationship between the process parameter nodes and the lace fabric quality attribute nodes, and each edge includes fixed attributes: influence intensity coefficient (the value range is 0-1, and the larger the value is, the higher the influence degree of the process parameter on the corresponding quality attribute is, and the coefficient value is determined based on historical experimental data, industry standard documents, and expert experience value), and influence direction identifier (enumeration type, including “positive influence”, “negative influence”, and “bidirectional influence”, which respectively represent that the quality attribute value increases, decreases, or presents different influence directions according to the parameter value interval when the process parameter value increases), in addition, the edge also includes source identifier of the influence relationship (such as “experimental data derivation”, “expert experience setting”, and “industry standard regulation”) and update timestamp, which are used to trace the validity and timeliness of the influence relationship.

[0054] When the quality evaluation system obtains the process parameter change list, the knowledge graph query module is called to start the association query process. The query module first extracts the changed process parameter item name in the process parameter change list as a query keyword, calls the node query interface of the graph database, and the interface performs an exact query by matching the “parameter item name” attribute of the process parameter node. If there are multiple process parameter nodes with the same name but different parameter types or value ranges, the query module will further compare the value ranges of the parameter item before and after the change in the process parameter change list with the “parameter value range” attribute of the node to filter out the unique matching process parameter node and complete node positioning; if no matching node is found, a node mismatch log is generated to prompt that the knowledge graph data needs to be supplemented and improved.

[0055] After the process parameter node positioning is completed, the query module calls the edge traversal interface of the graph database to start the out-edge traversal process from the target process parameter node. The traversal algorithm adopts a breadth-first traversal strategy to preferentially obtain all out-edges directly connected to the process parameter node, thereby avoiding irrelevant data interference caused by excessive traversal depth. During the traversal process, the interface returns the attribute information (influence intensity coefficient, influence direction identifier) of each out-edge and the information of the terminal node (i.e., the lace fabric quality attribute node) connected by the out-edge in real time. The query module temporarily stores these information in the memory cache area to form an associated data set of "process parameter node-edge-quality attribute node".

[0056] Subsequently, the query module calculates the influence quantitative value of each lace fabric quality attribute node based on the associated data set in the cache area. The calculation process is based on the variation range of the process parameter value, which is determined by the relative change relationship between the "changed parameter value" and the "original parameter value" in the process parameter change list (such as the absolute difference value, the ratio value, etc., the specific calculation logic is preset according to the parameter type), and then combined with the influence intensity coefficient of the edge to generate the influence quantitative value through a preset logical rule (such as the variation range and the influence intensity coefficient are positively correlated, and both determine the size of the influence quantitative value). The influence quantitative value ensures that the quantitative result can objectively reflect the actual influence degree of the process parameter change on the quality attribute.

[0057] The query module is built-in with an activation threshold configuration unit, which is set based on the enterprise process quality requirements and historical quality risk data. Different types of quality attribute nodes can correspond to different activation thresholds (such as the activation threshold of a key quality attribute is lower, and the activation threshold of a non-key quality attribute is higher). After calculating the influence quantitative value of each quality attribute node, the query module compares it with the activation threshold corresponding to the node. If the influence quantitative value exceeds the activation threshold, the quality attribute node is marked as "affected attribute", and the core information such as the quality attribute ID, quality attribute name, and standard threshold range of the node is extracted. If the activation threshold is not exceeded, it is determined that the quality attribute is not affected by the current process parameter change and is not included in the subsequent process.

[0058] Finally, the query module sorts all the quality attribute nodes marked as "affected attribute" according to the quality attribute ID and arranges them into a lace fabric incoming quality attribute list. The list is stored in JSON array format, and each element in the array contains the quality attribute ID, quality attribute name, influence quantitative value, and influence direction identifier. This facilitates the direct calling of the S3 generation of predictive detection scheme module, and the list is also stored in the temporary data directory of the system and retained until the current batch quality evaluation process is completed, so as to facilitate subsequent traceability queries.

[0059] In the S3, the specific process of generating a predictive detection scheme is as follows:

[0060] First, the lace fabric incoming quality attribute identification reading process is started, and the system structured data reading interface is called to access the quality attribute list output in S2. The list is stored in the system temporary data area, and the data area is configured with role-based access control strategy, which only allows the current module to read data through the encrypted authorized service link to prevent unauthorized tampering or access. The list is organized in a standardized structured format, and each entry contains lace fabric quality attribute identification, impact quantization value, impact direction identification and other core information. The reading interface uses streaming parsing technology to extract quality attribute identification line by line to avoid memory resource overload caused by full data loading, and improves reading efficiency through data sharding processing.

[0061] During the reading process, dual validity verification is performed simultaneously: first, the coding compliance of the quality attribute identification is verified. The identification needs to follow the fixed length coding rules specified by the enterprise process standard. The code is composed of letters and numbers to ensure uniqueness in the whole system. If the code format is abnormal, the module generates a verification exception log immediately, clearly marking the location of the abnormal identification and the type of violation. Second, the effectiveness of the impact quantization value is verified. It needs to confirm that the quantization value is not empty and within the reasonable value interval preset by the system. If there is a null value or an out-of-limit value, the module triggers the alarm mechanism to send an exception notification to the administrator terminal, and suspends the reading process until the exception data is corrected and restarted.

[0062] After completing the quality attribute identification extraction, the module calls the detection rule library query interface to initiate associated retrieval. The detection rule library is built based on enterprise-level relational databases, using a storage architecture that partitions by quality attribute category. The core data table contains fields such as quality attribute identification, standard detection method number, baseline sampling point density, baseline detection precision level, baseline detection observation time, detection method application scope, and rule update timestamp. The quality attribute identification is used as the primary key to build an index to ensure query efficiency. The query interface uses an index-driven retrieval mechanism, using the quality attribute identification as the search condition to quickly locate the target data through the database primary key index, avoiding delays caused by full table scanning.

[0063] After the query result is returned, integrity verification is performed: if a quality attribute identification does not match the corresponding detection rule, the interface automatically generates a rule missing report containing the missing rule's quality attribute identification and associated process parameter information, which is sent to the administrator through the system message push component. The administrator needs to supplement the rule data according to the enterprise detection standard before requerying. If the result is complete, the interface extracts the standard detection method number and baseline detection parameters to the module's dedicated memory cache area, which uses a key-value pair mapping structure to establish an associated index with "quality attribute identification-detection rule data" to provide fast data support for subsequent parameter adjustment.

[0064] Subsequently, the module adaptively adjusts the benchmark detection parameters based on the influence quantification value in the process knowledge graph. The adjustment dimensions include increasing the sampling point density, improving the detection precision level, and prolonging the detection observation time. After adjustment, the module generates a unique detection task ID for each quality attribute identifier. The ID is generated using a globally unique identification mechanism to ensure traceability. The adjusted detection parameters are then bound with the quality attribute identifier and the standard detection method number to construct a detection task item containing the task ID, quality attribute identifier, standard detection method number, adjusted parameters, and task creation timestamp.

[0065] After all task items are generated, consistency verification is performed: check if there are duplicate task items for the same quality attribute identifier and remove redundant entries by ID deduplication; check the matching of the adjusted parameters and the influence quantification value to ensure that the parameter adjustment range is logical; check the validity of the standard detection method number and confirm its availability by comparing it with the detection method database code. Task items that pass the verification are arranged in ascending order of ID to form a task item list, which is then combined into a predictive detection scheme. The scheme contains core elements such as scheme number (globally unique), generation timestamp (ISO8601 standard format), and detection task item list. After generation, it is written into the scheme database and synchronized to the high-frequency access cache area, and at the same time, the "scheme ready" event is triggered to notify the S4 link to start the batch range determination process.

[0066] The specific process of adaptively adjusting the benchmark detection parameters is as follows:

[0067] The detection parameter adjustment level is determined according to the size of the influence quantification value. The influence quantification value is divided into multiple continuous intervals. The interval division comprehensively considers the production history detection data of lace fabrics, quality risk assessment reports, and industry expert technical experience. Multiple continuous intervals are set, and administrators can dynamically adjust them through the system background. The first level adjustment corresponds to the first interval, the second level adjustment corresponds to the second interval, and the third level adjustment corresponds to the third interval.

[0068] Each adjustment level corresponds to a preset parameter adjustment rule. The rule is stored in a configuration file in a standardized text format, which clearly specifies the specific adjustment method.

[0069] First-level adjustment: suitable for scenarios with small influence quantification values. The benchmark sampling point density is increased by a system preset proportion, with the proportion set to balance detection precision and cost. The benchmark detection precision level is improved by one notch. The precision level is divided into basic, standard, precise, and super-precise levels according to enterprise standards. An increase of one notch switches to the adjacent higher level. The benchmark detection observation time is extended by a preset duration, with the duration set to capture the basic change characteristics of the quality attribute.

[0070] Second level adjustment: suitable for the medium scenario of affecting the quantization value, the adjustment amplitude is higher than the first level. The reference sampling point density is increased by a higher preset ratio, the sampling data amount is increased to enhance the reliability of the result; the reference detection accuracy level is improved by two grades, further reducing the detection error; the reference detection observation time is extended by a longer preset time, ensuring the complete recording of the dynamic change process of the quality attribute.

[0071] Third level adjustment: suitable for the larger scenario of affecting the quantization value, the adjustment amplitude is the highest. The reference sampling point density is increased by the maximum preset ratio allowed by the system, and strictly controlled within the maximum sampling density range supported by the detection equipment hardware; the reference detection accuracy level is improved to the highest grade (ultra-high precision level) to obtain the highest precision data; the reference detection observation time is extended to the maximum preset time, covering the possible change interval of the quality attribute.

[0072] In the S4, the specific process of determining the lace fabric batch range that needs to be detected is:

[0073] First, start the docking process with the warehouse management system, establish a communication link with the lace fabric inventory database through an industrial-level data interaction protocol. This communication link uses an encryption transmission mechanism to ensure the security and integrity of the data during transmission, and also needs to pass through the identity authentication and permission verification of the warehouse management system to obtain the reading permission of the inventory data, allowing access only to the inventory records related to lace fabrics, to avoid unauthorized access to other categories of data.

[0074] The inventory database uses a relational database architecture, and the core data table is stored in "year-season" partitions to optimize the query performance of large amounts of data. The table structure includes fields such as lace fabric batch number (batch number is a unique identifier automatically generated by the system, composed of production year, production line number and serial number), storage time (accurate to the minute level, following the ISO8601 standard format), storage location (using a "warehouse area-shelf-shelf location" three-level coding rule), fabric composition specifications (including fiber type proportion, yarn count, etc. Key information). When executing the query operation, call the preset structured query statement to extract all in-stock lace fabric batch inventory records with "in-stock status" as the filtering condition. The query statement associates the batch number and storage state fields through the index optimization mechanism to avoid query delays caused by full table scanning.

[0075] After the data extraction is completed, the inventory record preprocessing process is started: first, data integrity checking is performed to check whether each record is missing core fields (if the batch number or fabric composition specification is missing, it is marked as an abnormal record), data format errors (such as warehouse entry time not conforming to the standard format), etc. Problems, generate logs for abnormal records and temporarily store them in the abnormal data buffer area, and after manual investigation and completion, they are re-integrated into the processing flow; second, data standardization processing is performed to convert unstructured descriptions in fabric composition specifications (such as "high cotton content") into standardized numerical formats (such as "cotton 85%"), ensuring that the composition specification data of different batches is comparable, and the standardization rules are based on enterprise fabric composition coding specifications and support dynamic updates through the system management background.

[0076] After preprocessing is completed, the correlation score calculation interface is called to associate the fabric composition specification data of each in-stock batch with the lace fabric quality attributes involved in the predictive detection scheme, triggering the correlation score calculation process. After the correlation score calculation of all batches is completed, the system's preset correlation score threshold is read, which is set by the quality control department based on historical detection data and quality risk assessment results, and supports dynamic adjustment according to fabric categories or detection scenarios. Compare the correlation score of each batch with the threshold, and select batches with a score greater than or equal to the threshold to form an initial detection batch set.

[0077] Subsequently, the initial detection batch set is sorted in descending order of correlation score, and the sorting algorithm uses a stable sorting strategy to ensure that batches with the same score are sorted in descending order of warehouse entry time, and batches with more recent warehouse entry times are processed first. After sorting is completed, the final detection batch range is output in a structured data format, including batch number, correlation score, warehouse entry time, storage location, etc. Fields, and is stored in the detection task scheduling database at the same time, and a batch range confirmation log is generated, recording key information such as selection criteria, sorting rules, threshold values, etc. to provide a basis for subsequent process tracing.

[0078] The specific process of the correlation score calculation is as follows:

[0079] A correlation weight matrix of lace fabric composition specifications and quality attributes is constructed, which is based on lace fabric production process data, historical quality detection results, and industry expert experience. The row dimension of the matrix corresponds to all defined lace fabric composition types, including fiber type combinations (such as cotton polyester blend, cotton ammonia blend), yarn count intervals (such as 20S-30S, 30S-40S), weaving density levels, etc. The column dimension corresponds to all lace fabric quality attributes, including tensile strength, pattern clarity, color fastness, elastic recovery rate, etc.

[0080] The weight values in the matrix are determined by multi-dimensional data fusion: first, the fluctuation amplitude of the quality attribute under different component specifications in the historical detection data is counted, the greater the fluctuation amplitude, the higher the sensitivity of the component specification to the quality attribute, and the higher the corresponding weight value; second, industry experts and enterprise process engineers are invited to calibrate the weight value, and the weight deviation caused by extreme data is adjusted according to the actual production experience; third, the weight value is iteratively updated based on the newly added production and detection data, and the update cycle is synchronized with the enterprise process optimization cycle. The weight matrix is stored in a matrix database, which supports efficient row vector retrieval and numerical updating. At the same time, a mapping table of component specification type and row index, quality attribute and column index is established to improve the data positioning efficiency.

[0081] When it is necessary to calculate the correlation score of a batch, first, the fabric component specification type of the batch is extracted from the inventory preprocessing data, and the corresponding row vector is located in the weight matrix through the component specification-row index mapping table. Subsequently, the lace fabric quality attribute list in the predictive detection scheme is read, and the column index of each attribute in the list in the matrix is determined through the quality attribute-column index mapping table. The weight value corresponding to the column index is extracted from the located row vector to form the weight value set of the batch.

[0082] The influence quantization value corresponding to each quality attribute is obtained from the predictive detection scheme, which is matched with the corresponding element in the above weight value set as a weight coefficient, and the weighted summation process is started. In the summation process, the product of each weight value and the corresponding influence quantization value is accumulated to obtain the original summation result. Subsequently, the original summation result is normalized, the normalization range is set to 0-1, and the linear normalization strategy is adopted to map the original result to the interval to ensure that the correlation scores of different batches have a unified comparison benchmark.

[0083] After the normalization processing is completed, the correlation score obtained is checked for effectiveness, and whether the score is within the interval of 0-1 is checked. If it is out of the interval, it is judged as a calculation abnormality, and the weight value and influence quantization value are re-called for secondary calculation. The correlation score that passes the verification is bound with the corresponding batch number, stored in the correlation score result database, and a calculation log is generated to record the matrix version, weight value set, influence quantization value set, original result, normalization parameter and other information, ensuring that the calculation process is traceable.

[0084] In the S5, the specific process of starting the predictive detection operation is:

[0085] First, based on the lace fabric batch range determined, a detection task dispatch order is generated. The dispatch order is constructed in a structured data format, and the core content includes a lace fabric batch number list and a predicted detection scheme copy corresponding to each batch. The batch number list is arranged in descending order of correlation score to ensure that high-priority batches are given priority in the detection process. The detection scheme copy retains all fields of the original scheme, including scheme number, adjusted detection parameters (sampling point density, detection accuracy level, observation time), and standard detection method number, to avoid detection operation deviations caused by missing scheme information. The dispatch order is generated with the addition of task identification, generation timestamp, and task priority fields. The task identification uses a globally unique coding rule composed of detection date and task sequence code, which facilitates subsequent process tracking. The task priority is set according to the batch correlation score, with higher scores indicating higher priority, supporting the preferential scheduling of high-priority tasks by detection equipment.

[0086] After the dispatch order is generated, it is distributed to the automatic detection equipment control center through an industrial-level data transmission protocol. Before transmission, the dispatch order is compressed to reduce bandwidth usage, and a data check code is added. The control center verifies data integrity using the check code upon receipt. If the verification fails, it will send a retransmission request to the sender until the data is completely received. In addition, transmission logs are recorded during the dispatch process, including sending time, receiving time, and transmission status, providing a basis for subsequent transmission exception troubleshooting.

[0087] After the automatic detection equipment control center receives the dispatch order, it starts the parsing process. First, syntax checking is performed to check if the structured format of the dispatch order meets the preset specifications, such as missing fields or incorrect data types. If there are syntax issues, the order is marked as invalid and an exception parsing log is generated. Second, integrity checking is performed to confirm whether the detection scheme copy for each batch contains all necessary parameters. If the sampling point density is not specified, the order is considered incomplete and the parsing is suspended while notifying relevant personnel to supplement the information. After successful parsing, the control center extracts the detection requirements (such as surface image acquisition, physical performance testing, and chemical composition analysis) and detection parameters for each batch from the dispatch order, establishing an association mapping table of "batch number-detection requirement-detection parameter" to provide data support for equipment resource allocation.

[0088] Subsequently, the detection equipment resource allocation is carried out. The control center is built-in with an equipment management module, which collects the running state (idle / busy / fault) of all automated detection equipment, supported detection items and current load rate data in real time. Two core principles are followed in the allocation: one is the equipment capacity matching principle, that is, according to the detection requirements, the equipment with corresponding functions is allocated, for example, the surface image acquisition task is allocated to the visual detection equipment equipped with a high-resolution industrial camera, the physical performance test task is allocated to the special equipment such as the tensile testing machine and the wear testing machine, and the chemical composition analysis task is allocated to the near-infrared spectroscopy detection equipment; the other is the load balancing principle, which avoids the detection delay caused by the single equipment bearing too many tasks, and the new task is allocated to the idle equipment with lower load rate by calculating the ratio of the current task amount to the maximum processing capacity of the equipment. After the equipment allocation is completed, the control center sends the detection instruction to the corresponding equipment, the instruction contains the target batch number, the associated detection parameters and the detection sequence requirements, and the equipment state is updated to "busy" to prevent repeated allocation.

[0089] After receiving the instruction, the detection equipment executes the quality data collection according to the parameters in the predictive detection scheme. The collection process covers three core links: one is the lace fabric surface image acquisition, an industrial camera with a resolution not less than 20 million pixels is used, combined with a ring-shaped light source to eliminate shadow interference, the camera selects the collection area uniformly on the fabric surface according to the set sampling point density, 3-5 images are taken for each area to ensure the image clarity, the image data is stored in a lossless format, and the collection time and camera parameters (focal length, exposure) are recorded; two is the physical performance test, according to the standard detection method number, the corresponding test program is called, such as the electronic tensile testing machine is used for tensile strength test, the tensile speed and force value measurement accuracy are controlled according to the set detection accuracy level, the force value change data is recorded in real time, and after the test is completed, the tensile strength, elongation at break and other indexes are automatically calculated; three is the chemical composition analysis, a near-infrared spectrometer is used for non-destructive detection of the fabric, the spectral scanning range covers the characteristic wavelength of the target component (such as fiber type, dye content), the scanning data is sampled at a predetermined interval to ensure that the data can accurately reflect the component proportion, and the original data collected is processed by format standardization to convert into a unified structure of quality data file.

[0090] In the S6, the specific process of generating the quality evaluation record is:

[0091] Read the quality data file corresponding to each batch of lace fabric. Access the storage server through the file service interface. The interface uses a batch number-based retrieval mechanism. Input the batch number to locate the corresponding quality data file, avoiding reading delays caused by full file traversal. Perform file validity check during reading process. Check if the file is damaged (such as missing file header, data truncation), and if the file format meets the standardized requirements. If the file is damaged, trigger the data recovery process to retrieve backup data from the local cache of the detection device and regenerate the file. If the format is abnormal, return the format error log and notify the data standardization link to reprocess. After file reading is completed, establish data cache according to batch number to ensure the efficiency of data call in subsequent operations.

[0092] Subsequently, the quality index measurement values in the quality data file are extracted. The extraction process is based on the pre-set field mapping rules, which associate the structured data in the file with the standard quality index fields. For example, extract the "tensile strength" and "elongation at break" fields from the physical performance data, extract the "pattern clarity score" and "surface flaw number" fields from the surface image analysis results, and extract the "fiber composition ratio" and "dye content" fields from the chemical composition data. Perform data type check simultaneously during extraction to ensure that the data type (such as numerical type, enumeration type) of the measurement value is consistent with the requirements of the standard index field. For example, "tensile strength" should be a floating-point data. If character type data appears, it is marked as an abnormal value and temporarily stored in the abnormal data list for manual review.

[0093] After extraction, compare the quality index measurement values with the allowed range in the lace fabric quality standard library. The quality standard library uses a relational database architecture. The core data table contains quality attribute ID, quality attribute name, allowed upper limit value, allowed lower limit value, and judgment rule (such as "must meet upper limit ≥ measurement value ≥ lower limit" and "single indicator unqualified, whole unqualified") fields. The allowed range in the standard library is based on industry standards (such as GB / T series fabric standards) and enterprise internal control standards, and supports dynamic adjustment according to fabric categories and application scenarios. During comparison, associate the measurement value with the standard allowed range according to "quality attribute ID". Judge whether the measurement value is within the allowed range for each indicator: if all indicators meet the requirements, the batch is preliminarily determined to be qualified; if there is at least one indicator exceeding the allowed range, it is preliminarily determined to be unqualified, and the indicator name, measurement value, and allowed range that exceed the range are recorded.

[0094] Based on the comparison result, the quality evaluation conclusion of a single lace fabric batch is generated. The evaluation conclusion includes the pass / fail determination result ("pass" or "fail") and abnormal index note: the pass conclusion only needs to determine the result; the fail conclusion needs to note the abnormal index information in detail, including the abnormal index name, the measured value and the deviation of the allowed range (such as "tensile strength measured value 2.8MPa, lower than the allowed lower limit 3.0MPa"), the process related prompt that may cause abnormality (such as "suspected to be related to yarn count adjustment"). The evaluation conclusion is generated by adding the conclusion identifier, the generation timestamp, the evaluation personnel identifier (if involving manual review), the conclusion identifier is bound to the corresponding batch number to ensure uniqueness and traceability.

[0095] After the quality evaluation conclusions of all single batches are generated, they are summarized to form the total quality evaluation record. The core content of the total record includes the evaluation date, the number of lace fabric batches involved, the overall pass rate statistics, and the detailed evaluation data table: the evaluation date follows the ISO8601 standard format, accurate to the hour level; the batch number statistics is the total batch number of this detection, distinguishing between "evaluation completed" and "to be reviewed" states; the overall pass rate is calculated according to "the number of qualified batches / the total number of evaluated batches", and the result is rounded to two decimal places; the detailed evaluation data table adopts a tabular structure, including the batch number, the pass determination result, the abnormal index note, and the evaluation time field of each batch, arranged in ascending order of batch number.

[0096] The above has carried out the detailed description to one embodiment of the present application, but the content described is only the preferred embodiment of the present application, cannot be considered for limiting the implementation scope of the present application. Any equivalent changes and improvements made according to the scope of the present application should still belong to the patent coverage range of the present application.

Claims

1. A lace fabric incoming quality evaluation method based on data analysis, characterized in that, The method comprises the following steps: S1, receiving a process parameter change list, the process parameter change list containing changed process parameter items and their parameter values; S2, according to the process parameter change list, querying a pre-constructed lace fabric process knowledge graph to identify the incoming quality attributes of the lace fabric affected by the changed process parameters; S3, based on the identified lace fabric incoming quality attribute types, generating a predictive detection scheme, the predictive detection scheme containing a set of quality attributes that need to be detected and corresponding detection parameters; S4, according to the predictive detection scheme, retrieving the inventory information of all batches of lace fabric in the current warehouse to determine the range of batches of lace fabric that need to be detected; S5, starting the predictive detection operation on the determined batches of lace fabric, and performing quality data collection according to the detection parameters in the predictive detection scheme; S6, integrating the collected quality data with the lace fabric batch information to generate a quality evaluation record; In S3, the specific process of generating the predictive detection scheme is as follows: Read each lace fabric incoming quality attribute identifier in the lace fabric incoming quality attribute list, query the detection rule library to obtain the standard detection method number and reference detection parameters corresponding to each lace fabric incoming quality attribute identifier, adaptively adjust the reference detection parameters according to the calculated influence quantitative value in the lace fabric process knowledge graph, the detection parameter adjustment includes increasing the sampling point density, improving the detection precision level and prolonging the detection observation time, binding the adjusted detection parameters with the lace fabric incoming quality attribute identifier to form a detection task item, and combining all detection task items into a predictive detection scheme; In S4, the specific process of determining the range of batches of lace fabric that need to be detected is as follows: Query the lace fabric inventory database in the warehouse management system to obtain the inventory records of all in-stock batches of lace fabric, the inventory records containing the batch number, storage time, storage position and fabric composition specification of the lace fabric, calculate the correlation score of each batch of lace fabric with the detection scheme according to the lace fabric incoming quality attributes involved in the predictive detection scheme and in combination with the lace fabric composition specification, and screen the batches of lace fabric with a correlation score greater than or equal to the correlation score threshold to form the range of batches of lace fabric that need to be detected; The specific calculation process of the correlation score is as follows: Construct a correlation weight matrix of lace fabric composition specifications and quality attributes, the rows of the matrix corresponding to different lace fabric composition specification types, the columns of the matrix corresponding to different lace fabric incoming quality attributes, and the weight values in the matrix representing the sensitivity of the lace fabric composition specification to the quality attribute; For each batch of lace fabric, extract its lace fabric composition specification type, locate the corresponding row vector in the correlation weight matrix, extract the corresponding weight values from the row vector according to the list of lace fabric incoming quality attributes in the predictive detection scheme, and normalize the weighted sum of all extracted weight values to obtain the correlation score of the batch of lace fabric.

2. The lace fabric quality evaluation method based on data analysis according to claim 1, wherein, In S1, the specific process of receiving the process parameter change list is as follows: After the downstream production management system completes the process parameter configuration update, a process parameter change notification file is automatically generated, the process parameter change notification file adopts a structured text format, and contains a change timestamp, a change operator identifier, and a list of changed process parameter item names, for each changed process parameter item, records the parameter value before the change and the parameter value after the change, the system receives the process parameter change notification file through a file transmission service, parses the file content and verifies the data integrity, stores the changed process parameter items and their parameter values parsed into a process parameter change list, and the process parameter change list maintains a data table structure in the memory.

3. The lace fabric quality evaluation method based on data analysis according to claim 1, wherein, In S2, the specific process of querying the pre-constructed lace fabric process knowledge graph is as follows: The lace fabric process knowledge graph adopts a graph structure of nodes and edges for data storage, the node types include process parameter nodes and lace fabric incoming material quality attribute nodes, the edge types represent the influence relationship of the process parameter nodes on the lace fabric incoming material quality attribute nodes, each edge contains an influence intensity coefficient and an influence direction identifier; According to the changed process parameter item name in the process parameter change list, the corresponding process parameter node is located in the knowledge graph, all outgoing edges starting from the process parameter node are traversed, the set of directly connected lace fabric incoming material quality attribute nodes is obtained, according to the influence intensity coefficient in the edge attribute and the change range of the process parameter value, the influence quantization value of each lace fabric incoming material quality attribute node is calculated, the lace fabric incoming material quality attribute nodes with influence quantization values exceeding the activation threshold are marked as affected attributes, to form a lace fabric incoming material quality attribute list.

4. The lace fabric quality evaluation method based on data analysis according to claim 1, wherein, The specific process of adaptively adjusting the reference detection parameters is as follows: According to the size of the influence quantization value, the detection parameter adjustment level is determined, the influence quantization value is divided into multiple continuous intervals, each interval corresponds to an adjustment level, the detection parameter adjustment level includes a first level adjustment, a second level adjustment and a third level adjustment, each adjustment level corresponds to a set proportion of detection sampling point quantity and a set level of detection accuracy.

5. The lace fabric quality evaluation method based on data analysis as claimed in claim 1, wherein, In S5, the specific process of starting the predictive detection operation is as follows: According to the lace fabric batch range, a detection task dispatch order is generated, the detection task dispatch order contains a list of lace fabric batch numbers and a copy of the predictive detection scheme corresponding to each batch, the detection task dispatch order is distributed to the automatic detection equipment control center, the automatic detection equipment control center analyzes the detection task dispatch order, allocates detection equipment resources, and the detection equipment executes quality data collection according to the detection parameters in the predictive detection scheme, the collection process includes lace fabric surface image acquisition, physical performance testing and chemical composition analysis, the collected raw data is standardized and converted into a unified structure quality data file.

6. The lace fabric quality evaluation method based on data analysis according to claim 1, wherein, In S6, the specific process of generating a quality evaluation record is as follows: The quality data file corresponding to each lace fabric batch is read, the quality index measurement value is extracted, the quality index measurement value is compared with the allowable range in the lace fabric quality standard library, the quality evaluation conclusion of the single lace fabric batch is generated, the quality evaluation conclusion includes the qualified judgment result and the abnormal index note, the quality evaluation conclusions of all lace fabric batches are summarized to form the total quality evaluation record, and the total quality evaluation record includes the evaluation date, the number of lace fabric batches involved, the total qualified rate statistics and the detailed evaluation data table.

Images