An aluminum processing quality correlation analysis method based on human-machine-material method loop

Abstract

The application provides an aluminum processing quality correlation analysis method based on a man-machine-material-method loop, and belongs to the technical field of industrial data analysis.The application automatically generates a material node event through a device physical anchor point signal, and does not depend on a device absolute clock;cross-process data alignment is achieved through a material node identifier and a material flow turnover fingerprint; a material node clock label is used for data segment indexing; reliable data segments are screened through correlation readiness degree marking; and each process data segment is reversely correlated through quality detection results, so that automatic correlation analysis of man-machine-material-method loop factors and quality results is achieved, and aluminum processing quality traceability and process optimization efficiency are improved.

Description

Technical Field

[0001] This invention belongs to the field of industrial data analysis technology, specifically relating to a method for correlation analysis of aluminum processing quality based on man, machine, material, method, and environment. Background Technology

[0002] Aluminum processing encompasses multiple steps, including melting and casting, extrusion, aging, and surface treatment. Product quality is influenced by a combination of factors, including equipment, processes, materials, operations, and the environment. With the advancement of the Industrial Internet and intelligent manufacturing, quality control in aluminum processing is shifting from post-process sampling to full-process online management. Existing research has proposed multi-source data fusion based on real-time edge computing, multi-granularity data storage based on material family trees, and a four-dimensional traceability system encompassing time-point inspection, material feeding, and testing, enabling the synchronization of production data across all processes to a data platform within seconds. However, several technical challenges remain, including how to automatically align discrete process data into continuous data segments based on materials, how to achieve cross-process timeline alignment without relying on absolute equipment clocks, and how to correlate quality inspection results with process data from each process to support the correlation analysis of human, machine, material, method, and environmental factors.

[0003] Chinese patent CN118691167A discloses a quality traceability management system and method for the entire aluminum profile production process. It defines nodes throughout the production process using node definition units, resulting in a set of nodes including raw material procurement, smelting, extrusion, heat treatment, surface treatment, quality inspection, and transportation. It uses blockchain to store production data for each node and verifies data tampering through data comparison. While this solution has advantages in tracing the integrity of the traceability chain, its node set is manually defined at the workshop or process level, representing a management-level node division rather than material-level events automatically triggered by equipment signals. Data organization is based on production nodes and does not involve a mechanism for automatically dividing data segments based on adjacent material node events, nor does it provide specific details on achieving cross-process data alignment without relying on equipment absolute clocks. Its quality traceability primarily relies on blockchain data comparison and verification, and it does not yet involve reverse-linking quality inspection results to data segments of each process and conducting correlation analysis of human, machine, material, method, and environmental factors.

[0004] Furthermore, Chinese patent CN114862180A discloses a product quality data analysis and feedback method and system based on the entire product lifecycle. This method obtains the correlation mapping relationships between product objects in the design, manufacturing, and service stages, combines production stage quality data with manufacturing stage factors (manpower, machinery, materials, methods, and environment) into itemsets, sorts them in descending order of confidence based on association rule mining, and uses differential analysis to determine the source stage of quality problems. While this approach has reference value in using manpower, machinery, materials, methods, and environment factors for quality attribution, its data organization is based on product objects and BOM structure, and does not involve technical means to automatically generate material node events using equipment physical anchor point signals. Cross-stage data association is based on product object mapping relationships, and it does not disclose a mechanism for cross-process time axis alignment that does not rely on equipment absolute clocks, using material node clock tags. The analysis method uses association rule mining and differential statistics, but it does not involve reliability labeling and screening of data segments, nor does it involve technical paths for outputting correlation factors and effect sizes through difference testing.

[0005] In view of this, how to automatically generate material node events and unique identifiers from the physical anchor point signals of the equipment without relying on the absolute clock of the equipment, and realize the automatic segmentation, time axis alignment, reliability screening and quality correlation analysis of cross-process human, machine, material, method and environment data, is a technical problem that still needs to be solved in this field. Summary of the Invention

[0006] The purpose of this section is to outline some aspects of the embodiments of the present invention and to briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section, as well as in the abstract and title of the present application, to avoid obscuring the purpose of this section, the abstract and title of the invention. Such simplifications or omissions shall not be used to limit the scope of the present invention.

[0007] In view of the aforementioned existing problems, the present invention is proposed.

[0008] Therefore, the technical problem solved by this invention is: how to automatically generate material node events and unique identifiers based on the physical anchor point signals of the equipment without relying on the absolute clock of the equipment, so as to realize automatic segmentation, time axis alignment, reliability screening and quality correlation analysis of cross-process man, machine, material, method and environment data.

[0009] To address the aforementioned technical problems, the present invention provides the following technical solution: A method for quality correlation analysis of aluminum processing based on man, machine, material, method, and environment includes: generating material node events and corresponding material node identifiers based on the physical anchor point signals of aluminum processing equipment; Collect man-machine-material-method-environment (HMI) data containing material identifiers, associate the data records with the material node identifiers, and divide the data into segments based on adjacent material node events of the same material in the same process, generating material node clock tags for the data segments; construct the material node timeline for the data segments based on the material flow relationship, and attach an association readiness flag to the data segments; associate quality inspection results with the data segments according to the material identifiers, group the data segments based on the associated quality inspection results, select the data segments on the material node timeline whose association readiness flags meet preset conditions, and output the association relationship between HMI factors and quality inspection results.

[0010] In a preferred embodiment of the present invention, the generation of material node events and corresponding material node identifiers includes: configuring physical anchor point signals from the control system of the aluminum processing equipment, wherein the physical anchor point signals have signal identifiers, equipment identifiers, process identifiers, trigger conditions, timestamp accuracy, effective time windows, states before material state changes, and states after material state changes; determining the material state change generated when the physical anchor point signal is triggered as a material node event; and generating material node identifiers according to the process identifier, equipment identifier, signal identifier, trigger timestamp, material identifier, and sequence number.

[0011] As a preferred embodiment of the present invention, the method further includes: receiving manually entered information when the physical anchor point signal is missing, wherein the manually entered information includes event type, actual occurrence time, material identifier, process identifier, and equipment identifier, and the event type includes material node event and interval closure event; generating a manual material node identifier based on the process identifier, equipment identifier, actual occurrence time, material identifier, and sequence number in the manually entered information, and recording the actual occurrence timestamp and system entry timestamp; marking manually entered information that exceeds a set time limit as a severely missing state, and setting the data integrity status to severely missing and the overall readiness status to unavailable in the associated readiness flag.

[0012] As a preferred embodiment of the present invention, the data segmentation based on the material node identifier and adjacent material node events of the same material in the same process includes: collecting man, machine, material, method, and environment data, each data record including a collection timestamp, equipment identifier, process identifier, data category, data item identifier, value, quality code, and material identifier; when a material node event is triggered, the material node identifier is appended to the data record between the current material node event and the next material node event of the same material in the same process; if the next material node event has not yet been triggered, the data record is temporarily stored in a temporary cache, and after the next material node event is triggered, it is backfilled into the data segment corresponding to the current material node identifier according to the collection timestamp, the data segment version of the data segment is incremented, and only the data segment and subsequent data segments affected by the boundary are re-executed for data record sorting and material node clock tag generation; if the next material node event has not been triggered within the set waiting time limit and a man is received If the event type of the manually entered information is a range closure event or a process end signal, it is determined that the next material node event is missing. The range closure event or process end signal in the manually entered information is used as the range closure event. After generating the end material node identifier, the range is closed, and the data segment version of the data segment is incremented. If the collection timestamp is equal to the trigger time of the current material node event, it belongs to the current material node event. If the collection timestamp is equal to the trigger time of the next material node event, it belongs to the next material node event. If the collection timestamp falls between the trigger time of the current material node event and the trigger time of the next material node event, it belongs to the current material node event according to the left-closed and right-open interval. When there are multiple materials with the same collection timestamp, if the data record has its own material identifier, it is assigned according to the material identifier. If the data record has no material identifier, it is assigned according to the material identifier uniquely bound within the valid time window. If it cannot be uniquely bound, it is marked as pending confirmation and temporarily not involved in the data segment division.

[0013] In a preferred embodiment of the present invention, generating the material node clock tag for the data segment includes: sorting the data records within the data segment according to the acquisition timestamp, and sorting them according to the device identifier and data item identifier when the acquisition timestamps are the same; assigning a relative data point sequence number according to the sorting result; and generating the material node clock tag according to the material node identifier and the relative data point sequence number.

[0014] As a preferred embodiment of the present invention, the data segment is a single material data segment, and its data structure includes: segment identifier, data segment version, material identifier, starting material node identifier, ending material node identifier, process identifier, starting material node clock tag, ending material node clock tag, data record set, and quality tag field; the segment identifier is generated using the SHA-256 hash algorithm; the old version data segment is marked as invalid.

[0015] As a preferred embodiment of the present invention, associating the quality inspection result with the data segment based on the material identifier includes: the quality inspection result includes the material identifier, inspection time, inspection item, result, defect type, inspection equipment, and inspection standard; locating the data segment the material has experienced through the material identifier; writing the quality inspection result into the quality tag field of the data segment; and, in the event of a writing failure, setting the data integrity status of the data segment to severely missing, the overall readiness to unavailable, and outputting the reason for the writing failure.

[0016] In a preferred embodiment of the present invention, the associated readiness marker includes data integrity status, collection continuity status, and upstream / downstream material node matching status. The data integrity status is determined based on the missing rate of mandatory data items and the missing status of key items. A missing rate not greater than a first missing rate threshold and no key items are missing indicates complete data; a missing rate greater than the first missing rate threshold but not greater than a second missing rate threshold and no key items are missing indicates partial missing data; a missing rate greater than the second missing rate threshold or key items are missing indicates severe missing data. The collection continuity status is determined based on the time interval between adjacent data records of the same data item sorted by collection timestamp and the status of manual data entry. The overall collection continuity status of the data segment is determined by prioritizing key items and taking the worst-case scenario for non-key items. A time interval not greater than the product of the collection period and a first time interval multiple threshold indicates continuous data; a time interval greater than the product of the collection period and the first time interval multiple threshold but not greater than the product of the collection period and the second time interval multiple threshold indicates interrupted data; a time interval greater than the product of the collection period and the second time interval multiple threshold indicates severe data. If an interruption occurs, and manual data entry occurs within the set time limit, the system is considered to be in a manual data entry state. If manual data entry occurs after the set time limit, the data integrity status is "severely missing," the data collection continuity status is "manual data entry," and the overall readiness is "unavailable." The matching status of upstream and downstream material nodes is determined based on the material identifier and material flow fingerprint. A match is defined as the material flow fingerprint of a data segment matching the key items of the material flow fingerprints of the upstream and downstream data segments; a partial match is defined as the key items matching but the dynamic items not matching; and a non-match is defined as the key items not matching. The overall readiness is determined based on the data integrity status, the data collection continuity status, and the matching status of upstream and downstream material nodes: when the data integrity status is complete, the data collection continuity status is continuous, and the matching status of upstream and downstream material nodes is matched, the overall readiness is reliable. When the data integrity status is "severely missing," the data collection continuity status is "manual data entry," or the matching status of upstream and downstream material nodes is "non-match," the overall readiness is unavailable. In other cases, the overall readiness is used as a secondary reference.

[0017] In a preferred embodiment of the present invention, the material node timeline of the data segment constructed based on the material flow relationship includes: in the material flow fingerprint, key items include alloy grade, aluminum rod batch, supplier, and casting furnace number; dynamic items include extrusion press number, aging furnace number, and surface treatment tank number; the key items are used to determine the consistency of material identity, and the dynamic items are used to record process flow differences; the material flow fingerprint is updated step by step according to the process, after casting including alloy grade, aluminum rod batch, supplier, and casting furnace number, after extrusion adding extrusion press number, after aging adding aging furnace number, and after surface treatment adding surface treatment tank number; the material node timeline is constructed according to the aluminum processing process flow sequence, where nodes are material node events and edges are material flow relationships; upstream and downstream data segments are associated through the material identifier and the material flow fingerprint.

[0018] As a preferred embodiment of the present invention, the correlation between the output of human, machine, material, method, and environment factors and the quality inspection results includes: selecting the data segment with overall reliable readiness on the material node time axis; constructing qualified and defective groups for the selected reliable data segments according to the quality inspection results for each inspection item; extracting the statistical characteristics of human, machine, material, method, and environment factors in the data segment; performing difference tests on the qualified group and the defective group, and outputting the correlation factors, statistics, significance values, and effect sizes.

[0019] The beneficial effects of this invention are as follows: Compared with the prior art, the technical effects of this invention are as follows: This invention automatically generates material node events through equipment physical anchor point signals, without relying on the absolute clock of the equipment; cross-process data alignment is achieved through material node identifiers and material flow fingerprints; material node clock tags are used for indexing within data segments; reliable data segments are filtered through correlation readiness markers; and data segments of each process are inversely correlated through quality inspection results, thereby realizing automatic correlation analysis between human, machine, material, method, and environmental factors and quality results, and improving the efficiency of aluminum processing quality traceability and process optimization. Attached Figure Description

[0020] Figure 1 This is a flowchart of an aluminum processing quality correlation analysis method based on man, machine, material, method, and environment, as described in this invention.

[0021] Figure 2 This is a flowchart of the material node event and material node identifier generation process in this invention.

[0022] Figure 3 This is a flowchart of the data segmentation and material node clock tag generation process in this invention. Detailed Implementation

[0023] The present application will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be construed as limiting the scope of protection of this application. Based on the spirit of the present invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of the present invention.

[0024] like Figures 1-3 As shown, the present invention provides a method for correlation analysis of aluminum processing quality based on man, machine, material, method, and environment, comprising: S1: Generate material node events and corresponding material node identifiers based on the physical anchor point signals of the aluminum processing equipment; S2: Collect human, machine, material, method, and environment data containing material identifiers, associate the data records with the material node identifiers, and divide the data segments by adjacent material node events of the same material in the same process, generating material node clock tags for the data segments; S3: Construct the material node time axis of the data segments based on the material flow relationship, and add association readiness markers to the data segments; S4: Associate the quality inspection results with the data segments based on the material identifiers, group the data segments by the associated quality inspection results according to the inspection items, select the data segments on the material node time axis that meet the preset conditions for association readiness markers, and output the association relationship between human, machine, material, method, and environment factors and quality inspection results.

[0025] In some embodiments, the aluminum processing quality correlation analysis method based on man, machine, material, method, and environment provided by the present invention includes the following steps in step S1: Step S11: Configure physical anchor point signals from the control system of the aluminum processing equipment. The physical anchor point signals are existing signals of the aluminum processing equipment, selected from signals readable by the PLC or SCADA system. Each physical anchor point signal has a signal identifier, equipment identifier, process identifier, trigger condition, timestamp accuracy, effective time window, state before material state change, and state after material state change. Trigger conditions include rising edge trigger, falling edge trigger, or threshold trigger. The timestamp accuracy is in milliseconds, and the timestamps are uniformly in UTC format, used only for local sorting and boundary determination, and not for cross-process absolute time alignment.

[0026] The effective time window (default 30 seconds) is used to limit the binding time of the material identifier after the physical anchor point signal is triggered; if the material identifier binding is not completed within the effective time window, the material node event is marked as pending confirmation. In some embodiments, physical anchor point signals are configured according to the process: the melting and casting process is configured with a furnace exit signal and a casting start signal; the extrusion process is configured with a material loading completion signal and an extrusion discharge signal; the aging process is configured with a furnace door closing signal and a furnace door opening signal; and the surface treatment process is configured with a fixture insertion signal and a fixture removal signal.

[0027] The process end signal is configured from the PLC or SCADA system and has a signal identifier, equipment identifier, process identifier, trigger condition, and timestamp accuracy. When the next material node event is not triggered within the set waiting time limit and the process end signal is triggered, the process end signal is used as the interval closure event, an end material node identifier is generated, and the interval is closed. The set waiting time limit is 10 minutes. If no manual supplementary information or process end signal is received after the set waiting time limit is exceeded, a manual confirmation prompt is output and the closure of the interval is temporarily suspended. If the preset maximum waiting time limit is exceeded (preferably 30 minutes), the interval is forcibly closed and the data segment is marked as severely missing.

[0028] Step S12: The material state change generated when the physical anchor point signal is triggered is determined as a material node event. Specifically, when the physical anchor point signal meets the triggering condition, the state before and after the material state change corresponding to the physical anchor point signal is read, and the state change is determined as a material node event.

[0029] Material node events characterize the objective state transitions of materials during a process. By using existing physical anchor point signals of the equipment as triggering criteria, the generation of material node events is automatically determined by equipment signals, without relying on manually defined management nodes, thereby improving the objectivity and real-time nature of event boundaries.

[0030] Step S13: Generate material node identifiers according to process identifier, equipment identifier, signal identifier, trigger timestamp, material identifier, and serial number.

[0031] Specifically, material identification is acquired upon physical anchor point signal triggering, via methods including barcode scanning, RFID reading, or reading from the PLC recipe. Serial numbers increment from 0 within the same equipment, signal, material, and millisecond. The format of the material node identification is: Process Identifier - Equipment Identifier - Signal Identifier - Trigger Timestamp - Material Identifier - Serial Number.

[0032] Through this generation rule, each material node event obtains a unique identifier and is associated with the material identifier, providing a foundation for subsequent data segmentation, material node clock tag generation, and quality association.

[0033] Step S14: When a physical anchor point signal is missing, manually entered information is received. This manually entered information includes the event type, actual occurrence time, material identifier, process identifier, and equipment identifier. The event type includes material node events and interval closure events. A manually entered material node identifier is generated based on the manually entered information, and the actual occurrence timestamp and system entry timestamp are recorded. The format of the manually entered material node identifier is: Process Identifier-Equipment Identifier-MANUAL-Actual Occurrence Timestamp-Material Identifier-Serial Number.

[0034] When dividing data segments, constructing material node timelines, and sorting, the actual occurrence timestamp is used as the time of the manually entered material node event. The system-entered timestamp is used to calculate the supplementary entry deadline and for traceability. If the actual occurrence timestamp falls into a closed data segment, the data segment management module automatically increments the data segment version, marks the old version data segment as invalid, and triggers the re-division of the data segment and the update of the clock tag. Simultaneously, the associated readiness flag, material node timeline, quality tag field, and statistical results already executed based on the data segment are updated for that data segment. Specifically, the data segment management module is used to detect whether the actual occurrence timestamp of manually entered information falls into a closed data segment and automatically increments the data segment version.

[0035] If the data segment has already been included in the qualified or defective group, the data segment screening, grouping, and difference testing will be re-executed; manual entries that exceed the set time limit will be marked as overdue manual entries, and the data integrity status will be set to severe missing and the overall readiness status will be set to unavailable.

[0036] In some embodiments, the time limit is set to 24 hours, starting from the actual timestamp of occurrence. By distinguishing between automatically generated and manually entered timestamps, and recording the actual timestamp of occurrence and the timestamp entered by the system, it is easier to determine the readiness level for subsequent association.

[0037] Therefore, this invention drives the generation of material node events and material node identifiers through equipment physical anchor point signals, without relying on the equipment's absolute clock; at the same time, it covers the situation where physical anchor point signals are missing through a manual supplementation mechanism, ensuring the integrity and traceability of material node event generation, and providing an objective and unique event benchmark for subsequent cross-process data alignment.

[0038] In some embodiments, the aluminum processing quality correlation analysis method based on man, machine, material, method, and environment provided by the present invention includes the following steps in step S2: Step S21: Configure the data collection items for personnel, machine, material, method, and environment data. Collection items include personnel data, machine data, material data, method data, and environment data. Each data item has a data item identifier, collection frequency, collection cycle (default 1s), data type, unit, whether it is a mandatory item, whether it is a critical item, its associated process, and the equipment it is bound to.

[0039] The data includes personnel identification, shift, and operating actions; the data includes process version and standard operating procedures; and the data includes ambient temperature, humidity, and cleanliness.

[0040] Furthermore, list the mandatory and critical data items by process. For example: Mandatory data items for the casting process include alloy grade, aluminum rod batch, supplier, casting furnace number, casting temperature, and casting speed; critical data items include alloy grade, aluminum rod batch, supplier, and casting furnace number. Mandatory data items for the extrusion process include extrusion press number, extrusion temperature, extrusion speed, and die number; critical data item includes extrusion temperature. Mandatory data items for the aging process include aging furnace number, aging temperature, and holding time; critical data item includes aging temperature. Mandatory data items for the surface treatment process include surface treatment tank number, bath concentration, and treatment time; critical data item includes surface treatment bath concentration.

[0041] Step S22: Real-time acquisition of data on personnel, machinery, materials, methods, and environment. Each data record includes a collection timestamp, equipment identifier, process identifier, data category, data item identifier, value, quality code, and material identifier. The material identifier is obtained by scanning, RFID reading, or reading from the PLC recipe when the physical anchor point signal is triggered.

[0042] If the material identifier cannot be obtained, the data record will be marked as pending confirmation. Unique binding within the valid time window (default 30s) means that only one material identifier is read within the valid time window, and the material identifier does not conflict with the current equipment identifier, process identifier, or the state before and after the material status change.

[0043] Among them, non-conflict is determined according to the following rules: (1) Equipment identifier matching: The equipment identifier bound to the read material identifier is consistent with the equipment identifier in the current physical anchor signal; (2) Process identifier matching: The process identifier bound to the read material identifier is consistent with the process identifier in the current physical anchor signal; (3) Material state before change matching: The material state before change corresponding to the read material identifier is consistent with the material state before change defined by the current physical anchor signal; (4) Material state after change matching: The material state after change corresponding to the read material identifier is consistent with the material state after change defined by the current physical anchor signal; (5) Uniqueness within time window: The read operation occurs within the effective time window after the physical anchor signal is triggered, and only one material identifier is read within the effective time window; (6) Not occupied: The material identifier is not bound to other material node events that have not been closed in the same process, or although it is bound, it has been released.

[0044] If any of the above (1) to (6) is not met, or more than two material identifiers are read, it is determined to be a conflict, the data record is marked as pending confirmation, does not participate in the data segment division, and a manual confirmation prompt is output; if the material identifier is not completed within the valid time window, the data record is transferred to the unmatched cache, does not participate in the data segment division, and a manual confirmation prompt is output.

[0045] Step S23: Associate the data record with the material node identifier. Specifically, when the physical anchor point signal is triggered, the mapping table established in S1 is read to generate the material node identifier. The data's assigned interval is determined to be a left-closed, right-open interval. ,in This refers to the trigger time of the current material node event. This refers to the trigger time of the next material node event for the same material in the same process. If If a data record has a material identifier, it will be assigned to the next material node event. If multiple materials exist at the same data collection timestamp, the data record will be assigned according to the material identifier if it has its own material identifier. If the data record has no material identifier, it will be assigned according to the material identifier uniquely bound within the valid time window. If a unique identifier cannot be bound, it will be marked as pending confirmation.

[0046] Step S24: Divide the data into segments based on adjacent material node events. Two adjacent material node events include the first material node event and the second material node event. Using the first and second material node events as boundaries, divide the continuously collected man-machine-material-method-environment data into single-material data segments according to material identifiers. If segmentation by material identifier is not possible, mark it as a mismatch.

[0047] The single material data segment includes a segment identifier, data segment version, material identifier, starting material node identifier, ending material node identifier, process identifier, data record set, and quality label field; after the material node clock label is generated subsequently, the starting material node clock label and the ending material node clock label are written into the single material data segment.

[0048] After each update, the old version data segment is marked as invalid and will not participate in subsequent quality correlation analysis; the current version data segment is used for material node time axis, readiness mark, quality label and statistical test, and the initial version is 1.

[0049] The hash operation is SHA-256. The material identifier, process identifier, starting material node identifier, ending material node identifier, and data segment version are concatenated into a string in a set order. The SHA-256 value of the concatenated string is calculated, and the hexadecimal result is used as the segment identifier. If a hash collision occurs, the data segment version is incremented and the hash is recalculated.

[0050] Step S25: Generate material node clock tags for the data segment. Sort the data records within the data segment by acquisition timestamp; if the acquisition timestamps are the same, sort them by device identifier and data item identifier. Assign relative data point serial numbers according to the sorting results.

[0051] Define the material node clock tag as material node identifier#relative data point sequence number, for example, process identifier-equipment identifier-signal identifier-trigger timestamp-material identifier-serial number#0001, for indexing within the data segment.

[0052] Material node clock tags are used for indexing within data segments; cross-process data alignment uses material identifiers and material flow fingerprints as primary keys, independent of equipment absolute clocks; absolute timestamps are retained for sorting and boundary determination. Therefore, even if the absolute clocks of equipment in different processes are inconsistent, accurate indexing of data records within a data segment can still be achieved through material node identifiers and relative data point sequences.

[0053] In some embodiments, the aluminum processing quality correlation analysis method based on man, machine, material, method, and environment provided by the present invention includes the following steps in step S3: Step S31: Construct a material node timeline based on the material flow relationship. Specifically, the material flow fingerprint is generated by extracting key items and dynamic items from the process data records bound to the material identifier, and is updated after each process is completed and stored in conjunction with the material identifier; during comparison, the material flow fingerprint of the upstream and downstream data segments is read using the material identifier as an index, where key items include alloy grade, aluminum bar batch, supplier, and casting furnace number.

[0054] Dynamic items include extrusion press number, aging furnace number, and surface treatment tank number; key items are used to determine the consistency of material identity, and dynamic items are used to record differences in process flow; material flow fingerprints are updated step by step according to process, including alloy grade, aluminum rod batch, supplier, and melting furnace number after casting, extrusion press number is added after extrusion, aging furnace number is added after aging, and surface treatment tank number is added after surface treatment.

[0055] A material node timeline is constructed based on the flow sequence of aluminum processing operations. Nodes represent material node events, and edges represent material flow relationships. Upstream and downstream data segments are linked through material identifiers and material flow fingerprints. The timeline sorting is based on the causal relationship between the process flow sequence and material node events, and does not depend on the absolute time of the equipment. If parallel processes exist for the same material, they are sorted according to the process route sequence defined in the versioned process knowledge base. The versioned process knowledge base is a pre-configured database that stores process routes, standard operating procedures, process version numbers, and effective times. It records process nodes and allowed parallel processes in the process flow sequence using material identifiers or product models as indexes, and manages versions according to process version numbers. When the process route changes, a new version is generated while the old version is retained.

[0056] If the process route is not defined or there is a conflict, mark the upstream and downstream material nodes as mismatched and output a manual confirmation prompt.

[0057] The example path is: casting furnace exit node → extrusion feeding node → extrusion discharge node → aging furnace entry node → aging furnace exit node → surface treatment tank entry node → surface treatment tank exit node. If the flow relationship is unclear, mark the matching status of upstream and downstream material nodes as mismatched. Output material identifiers to an ordered data segment list.

[0058] By constructing a material node timeline that does not depend on the absolute time of the equipment, accurate correlation of data segments across processes can be achieved, avoiding time misalignment caused by inconsistent clocks of equipment in different processes.

[0059] Step S32: Attach an association readiness flag to the data segment. The association readiness flag includes data integrity status, acquisition continuity status, and upstream / downstream material node matching status. The data integrity status is determined based on the missing rate of mandatory data items and the missing status of key items, where the missing rate = number of missing mandatory data item records / number of mandatory data item records that should be collected.

[0060] The number of mandatory data items to be collected = data segment duration / collection cycle.

[0061] Optionally, the first missing rate threshold is 5%, and the second missing rate threshold is 20%. A missing rate of no more than 5% and no missing key items is considered complete; a missing rate of more than 5% but no more than 20% and no missing key items is considered partially missing; a missing rate of more than 20% or missing key items is considered severely missing.

[0062] The continuity status of data collection is determined based on the time interval between adjacent data records of the same data item after sorting by collection timestamp and the status of manual data entry. Optionally, the first time interval multiple threshold is 3, and the second time interval multiple threshold is 10. A time interval not greater than the collection period × 3 is considered continuous; a time interval greater than the collection period × 3 but not greater than the collection period × 10 is considered interrupted; a time interval greater than the collection period × 10 is considered severely interrupted; manual data entry within the set time limit is considered a manual data entry status. The overall data collection continuity status of a data segment is determined by prioritizing critical items and taking the worst-case scenario for non-critical items. The worst-case scenario deteriorates in the order of continuous, interrupted, severely interrupted, and then manually entered. When there is manual data entry where the system entry timestamp exceeds the set time limit of the actual occurrence timestamp, the data collection continuity status is considered a manual data entry status (the worst-case scenario).

[0063] The matching status of upstream and downstream material nodes is determined based on the material identifier and material flow fingerprint of the data segment: Key item consistency means that the alloy grade, aluminum rod batch, supplier, and melting furnace number are completely identical in the material flow fingerprint of the upstream and downstream data segments; Dynamic item inconsistency means that at least one of the extrusion press number, aging furnace number, and surface treatment tank number is different; During comparison, the material identifier is used as the index, and the material flow fingerprint is compared level by level according to the process sequence; Key item consistency and dynamic item consistency constitute a match; Key item consistency but dynamic item inconsistency constitutes a partial match; Key item inconsistency constitutes a non-match. Overall readiness is determined based on data integrity status, acquisition continuity status, and upstream and downstream material node matching status: When the data integrity status is complete, the acquisition continuity status is continuous, and the upstream and downstream material node matching status is matched, the overall readiness is reliable; when the data integrity status is severely missing, the acquisition continuity status is manually supplemented, or the upstream and downstream material node matching status is non-matched, the overall readiness is unusable; in other cases, the overall readiness is for auxiliary reference.

[0064] The associated readiness flag is not weighted. On the material node timeline, each data segment carries the overall readiness. When upstream and downstream data segments are associated to form an associated combination through material identifier and material flow fingerprint, if the overall readiness of any data segment is unavailable, then the overall readiness of the associated combination is unavailable; if all are reliable, then the associated combination is reliable; otherwise, it is a supplementary reference.

[0065] By using this association readiness flag, the reliability of data segments can be effectively identified, unreliable data segments can be avoided from participating in quality association analysis, thereby improving the accuracy of the analysis results.

[0066] In some embodiments, the aluminum processing quality correlation analysis method based on man, machine, material, method, and environment provided by the present invention includes the following steps in step S4: Step S41: Associate the quality inspection results with the data segments. The quality inspection results include material identification, inspection time, inspection items, results, defect type, inspection equipment, and inspection standards. Locate the data segments the material has experienced using the material identification. If the quality inspection is a sampling inspection, map it to the data segment of the sampled material by material identification, or map it to the data segments of all materials within that batch by batch identification, and mark it as a sampling inspection result in the quality label field. When one data segment corresponds to multiple inspection results, write them separately to the quality label field according to the inspection items. When constructing qualified and defective groups, construct them separately according to the inspection items.

[0067] Write the quality inspection results into the quality label field of the data segment without overwriting the data record set of the data segment.

[0068] Write failures include not finding the data segment corresponding to the material identifier, the quality label field being locked, or storage anomalies. When a write fails, the data segment's data integrity status is set to critically missing, the overall readiness is set to unavailable, and the reason for the write failure is output. If the data segment has already participated in the construction of the qualified or defective group, the data segment screening, grouping, and difference verification are re-executed.

[0069] By using material identification as an index, the quality inspection results of the final inspection stage are written back to the data segments of each process. This ensures that although each data segment does not carry a quality label when it is generated, it automatically obtains the final quality label after the inspection is completed, thus realizing a complete association between process data and final quality results.

[0070] Step S42: On the material node time axis, filter data segments according to preset conditions, and construct qualified groups and defective groups based on quality inspection results.

[0071] On the material node timeline, data segments are filtered according to preset conditions: data segments with an overall readiness level of reliable are selected as usable data segments; data segments with an overall readiness level of auxiliary reference are used only as auxiliary references; and data segments with an overall readiness level of unusable are excluded. The preset condition is that the overall readiness level is reliable.

[0072] Based on the quality inspection results, qualified and defective groups are constructed from selected reliable data segments. The qualified group consists of data segments with qualified quality inspection results, and the defective group consists of data segments with defective quality inspection results. In some embodiments, when constructing the qualified and defective groups, the control groups are required to be consistent in material properties, i.e., the same alloy grade, the same aluminum rod batch or within the same batch, and the same supplier. If complete consistency is not possible, propensity score matching can be used. Propensity score matching uses alloy grade, batch, supplier, equipment model, and process version in the qualified and defective groups as covariates, uses logistic regression to estimate the propensity score, and performs 1:1 nearest neighbor caliper matching, with the caliper value set to 0.2 times the standard deviation of the propensity score. Matching variables include alloy grade, batch, supplier, equipment model, and process version. If there are missing values ​​in the covariates, the sample is deleted or multiple imputation is used. If matching fails, matching failure information is output, and stratified analysis or termination difference test is used.

[0073] By screening reliable data segments, the results of quality correlation analysis can be avoided from being contaminated by unreliable data segments.

[0074] Step S43: Extract statistical features of the human, machine, material, method, and environment factors in the data segment. Extract statistical features for each data segment by sorting the data records according to the material node clock labels within the data segment; extract cross-process factors using material identifiers and material flow fingerprints as primary keys, aligned with the material node time axis.

[0075] Statistical features include: for continuous variables, extracting the mean, standard deviation, maximum, minimum, slope, or fluctuation range; for operation time sequence features, extracting the mean, variance, or edit distance between the operation sequence and the standard operating procedure; for categorical variables, extracting the frequency distribution. Standard operating procedures are read from the versioned process knowledge base by process node.

[0076] Step S44: Perform a difference test on the qualified and defective groups. For multiple data segments of the same material, cluster them according to material identifier or take only the final process data segment as an independent sample. Output the correlation factors, statistics, significance values, and effect sizes. In some embodiments, continuous variables are first tested for normality, and the Mann-Whitney U test or t-test is used based on the test results; categorical variables are tested using the chi-square test. Calculate the effect size, including Cohen's d or Cramér's V. Set the significance level (default 0.05), and perform Benjamini-Hochberg correction on all p-values ​​of the same data category for the same process. A q-value less than 0.05 after correction is considered significant.

[0077] When the corrected significance value is less than the significance level and the effect size is greater than the preset effect size threshold, the corresponding human, machine, material, method, and environment factors are marked as related factors. The preset effect size threshold is Cohen's d ≥ 0.5 or Cramér's V ≥ 0.3.

[0078] Output the correlation factors, statistics, significance values, and effect sizes. If the sample size of the qualified or defective group is less than the preset sample size threshold, output an under-sample information message. The preset sample size threshold is 30 by default. If the sample size of any group is less than 30, output the under-sample information message and stop the difference test, without outputting the correlation factors, statistics, significance values, and effect sizes; or output only descriptive statistics without outputting the correlation factors.

[0079] Through the above discrepancy test, factors related to quality inspection results can be accurately identified from the data on manpower, machinery, materials, methods, and environment, providing a basis for aluminum processing quality traceability and process optimization.

[0080] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0081] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for correlation analysis of aluminum processing quality based on man, machine, material, method, and environment, characterized in that, include: Based on the physical anchor point signals of the aluminum processing equipment, generate material node events and corresponding material node identifiers; Collect human, machine, material, method, and environment data containing material identifiers, associate the data records with the material node identifiers, and divide the data into segments based on adjacent material node events of the same material in the same process, and generate material node clock tags for the data segments; Construct a material node timeline for the data segment based on the material flow relationship, and attach an associated readiness flag to the data segment; The quality inspection results are associated with the data segment based on the material identifier. The data segment is grouped by the associated quality inspection results. The data segment whose association readiness mark meets the preset conditions is selected on the material node time axis. The association relationship between human, machine, material, method, and environment factors and quality inspection results is output.

2. The aluminum processing quality correlation analysis method based on man, machine, material, method, and environment according to claim 1, characterized in that, The generated material node event and the corresponding material node identifier include: Configure physical anchor point signals from the control system of aluminum processing equipment. The physical anchor point signals have signal identifiers, equipment identifiers, process identifiers, trigger conditions, timestamp accuracy, effective time windows, states before material state changes, and states after material state changes. The material state change generated when the physical anchor point signal is triggered is determined as a material node event; Material node identifiers are generated based on process identifiers, equipment identifiers, signal identifiers, trigger timestamps, material identifiers, and serial numbers.

3. The aluminum processing quality correlation analysis method based on man, machine, material, method, and environment as described in claim 1, is characterized in that, Also includes: When the physical anchor point signal is missing, manually recorded information is received. The manually recorded information includes event type, actual occurrence time, material identifier, process identifier, and equipment identifier. The event type includes material node event and interval closure event. Based on the process identifier, equipment identifier, actual occurrence time, material identifier, and serial number in the manually entered information, a manual material node identifier is generated, and the actual occurrence timestamp and system entry timestamp are recorded; Manually entered data that exceeds the set time limit will be marked as severely missing, and the data integrity status will be set to severely missing and the overall readiness status will be set to unavailable in the associated readiness flag.

4. The aluminum processing quality correlation analysis method based on man, machine, material, method, and environment according to claim 1, characterized in that, The data segments are divided according to the material node identifier and adjacent material node events of the same material in the same process, including: Collect data on personnel, machinery, materials, methods, and environment. Each data record includes a collection timestamp, equipment identifier, process identifier, data category, data item identifier, value, quality code, and material identifier. When a material node event is triggered, the material node identifier is appended to the data record between the current material node event and the next material node event for the same material in the same process; If the next material node event has not yet been triggered, the data record will be temporarily stored in a temporary cache. After the next material node event is triggered, it will be backfilled into the data segment corresponding to the current material node identifier according to the collection timestamp, the data segment version will be incremented, and the data record sorting and material node clock tag generation will be re-executed only for the data segment and subsequent data segments affected by the boundary. The subsequent data segment affected by the boundary refers to the adjacent data segment on the material node timeline that is after the data segment and has the same material identifier. If the next material node event is not triggered within the set waiting time limit and the event type of the manually recorded information is a section closure event or a process end signal, it is determined that the next material node event is missing. The section closure event or process end signal in the manually recorded information is used as the section closure event. After generating the end material node identifier, the section is closed, and the data segment version of the data segment is incremented. If the collected timestamp is equal to the trigger time of the current material node event, it belongs to the current material node event; if the collected timestamp is equal to the trigger time of the next material node event, it belongs to the next material node event; if the collected timestamp falls between the trigger time of the current material node event and the trigger time of the next material node event, it belongs to the current material node event according to the left-closed and right-open interval. When multiple materials exist at the same data collection timestamp, if the data record has its own material identifier, it will be assigned according to the material identifier; if the data record does not have a material identifier, it will be assigned according to the material identifier uniquely bound within the valid time window. If it cannot be uniquely bound, it will be marked as pending confirmation and will not participate in the data segment division for the time being.

5. The aluminum processing quality correlation analysis method based on man, machine, material, method, and environment according to claim 1, characterized in that, The material node clock tag that generates the data segment includes: The data records within the data segment are sorted by the collection timestamp, and if the collection timestamps are the same, they are sorted by the device identifier and the data item identifier. Assign relative sequence numbers to the data points according to the sorting results; A material node clock tag is generated based on the material node identifier and the relative data point sequence number.

6. The aluminum processing quality correlation analysis method based on man, machine, material, method, and environment according to claim 1, characterized in that, The data segment is a single material data segment, and its data structure includes: segment identifier, data segment version, material identifier, starting material node identifier, ending material node identifier, process identifier, starting material node clock label, ending material node clock label, data record set, and quality label field; The segment identifier is generated using the SHA-256 hash algorithm; The old version data segment is marked as invalid.

7. The aluminum processing quality correlation analysis method based on man, machine, material, method, and environment according to claim 6, characterized in that, Linking the quality inspection results to the data segment based on the material identifier includes: The quality inspection results include the material identification, inspection time, inspection items, results, defect type, inspection equipment, and inspection standards; The material identifier is used to locate the data segment the material has experienced; Write the quality inspection results into the quality label field of the data segment; When a write operation fails, the data integrity status of the data segment is set to critically missing, the overall readiness is set to unavailable, and the reason for the write failure is output.

8. The aluminum processing quality correlation analysis method based on man, machine, material, method, and environment according to claim 1, characterized in that, The associated readiness markers include data integrity status, acquisition continuity status, and upstream and downstream material node matching status; The data integrity status is determined based on the missing rate of mandatory data items and the missing status of key items. The data is considered complete if the missing rate is not greater than the first missing rate threshold and there are no missing key items. The data is partially missing if the missing rate is greater than the first missing rate threshold but not greater than the second missing rate threshold and there are no missing key items. The data is severely missing if the missing rate is greater than the second missing rate threshold or there are missing key items. The continuity of data collection is determined based on the time interval between adjacent data records of the same data item after sorting by collection timestamp and the status of manual data entry. The overall continuity of data collection is determined by prioritizing key items and taking the worst state for non-key items. Among them, a time interval not greater than the product of the collection period and the first time interval multiple threshold is considered continuous; a time interval greater than the product of the collection period and the first time interval multiple threshold but not greater than the product of the collection period and the second time interval multiple threshold is considered interrupted; a time interval greater than the product of the collection period and the second time interval multiple threshold is considered severely interrupted; and manual data entry within the set time limit is considered a manual data entry status. When there is manual data entry exceeding the set time limit, the data integrity status is severely lacking, the data collection continuity status is manual data entry, and the overall readiness is unavailable. The matching status of upstream and downstream material nodes is determined based on the material identifier and material flow fingerprint. The material flow fingerprint of the data segment is matched if the key items of the material flow fingerprint of the upstream and downstream data segments are the same. If the key items are the same but the dynamic items are different, it is a partial match. If the key items are different, it is a non-match. The overall readiness is determined based on the data integrity status, the data acquisition continuity status, and the upstream and downstream material node matching status: when the data integrity status is complete, the data acquisition continuity status is continuous, and the upstream and downstream material node matching status is matched, the overall readiness is reliable; when the data integrity status is severely missing, the data acquisition continuity status is manually supplemented, or the upstream and downstream material node matching status is mismatched, the overall readiness is unusable; in other cases, the overall readiness is used as an auxiliary reference.

9. The aluminum processing quality correlation analysis method based on man, machine, material, method, and environment as described in claim 8, is characterized in that, The material node time axis for constructing the data segment based on material flow relationships includes: In the material flow fingerprint, key items include alloy grade, aluminum bar batch, supplier, and casting furnace number; dynamic items include extrusion press number, aging furnace number, and surface treatment tank number; the key items are used to determine the consistency of material identity, and the dynamic items are used to record process flow differences. The material flow fingerprint is updated step by step according to the process. After melting and casting, it includes alloy grade, aluminum bar batch, supplier, melting and casting furnace number. After extrusion, it adds extrusion machine number. After aging, it adds aging furnace number. After surface treatment, it adds surface treatment tank number. A material node timeline is constructed according to the flow sequence of aluminum processing procedures. Nodes are material node events, and edges are material flow relationships. Upstream and downstream data segments are associated through the material identifier and the material flow fingerprint.

10. The aluminum processing quality correlation analysis method based on man, machine, material, method, and environment according to claim 8, characterized in that, The correlation between the output factors (human, machine, material, method, environment) and the quality inspection results includes: Select the data segment with overall reliable readiness on the material node time axis; construct qualified and defective groups for the selected reliable data segments according to the quality inspection results for each inspection item; extract the statistical characteristics of human, machine, material, method, and environmental factors in the data segment; Perform a difference test on the qualified group and the defective group, and output the correlation factors, statistics, significance values ​​and effect sizes.

Images