Product quality intelligent management and control method and system based on information management system

Abstract

The application provides a product quality intelligent management and control method and system based on an information management system, relates to the technical field of product quality intelligent management and control, obtains a historical quality sampling record set of a plurality of quality index detection data units of different sampling time nodes of a to-be-managed-and-controlled product batch, performs quality fluctuation mode mining processing on the historical quality sampling record set, obtains quality index time sequence fluctuation characteristics and associated conduction characteristics of unqualified items in sampling, calls a preset quality risk diffusion deduction logic according to the quality index time sequence fluctuation characteristics and the associated conduction characteristics of the unqualified items in sampling to generate a quality risk diffusion path diagram, and determines risk conduction nodes and risk residence time lengths of each production link based on the path diagram. Finally, a directional quality intervention instruction containing a management and control intervention time sequence is generated according to the risk conduction nodes and the risk residence time lengths, and is distributed to a management and control terminal of a corresponding production link. The application realizes active deduction and directional intervention from historical data to a risk diffusion path, and improves the foresight and accuracy of quality management and control.

Description

Technical Field

[0001] This invention relates to the field of intelligent product quality control technology, and more specifically, to a method and system for intelligent product quality control based on an information management system. Background Technology

[0002] In the product manufacturing process, quality control is the core link to ensure product pass rate. Current technologies typically rely on statistical analysis of sampling data at a single point in time, such as using control charts to determine the fluctuation range of key quality indicators, or retrospectively tracing and rectifying specific production processes based on historical non-conforming product records. However, these methods only focus on whether the value of a single indicator at a certain moment exceeds the tolerance range, failing to systematically explore the continuous fluctuation trajectory of quality indicators over time, and also failing to reveal the potential correlation and transmission mechanism of non-conforming items between different quality indicators or different sampling batches. Furthermore, existing quality intervention measures are mostly adjustments to corresponding processes after non-conforming products are detected, representing passive response control. They lack the ability to proactively extrapolate the risk diffusion path between various production stages based on quality fluctuation patterns and the correlation characteristics of non-conforming items, and cannot identify risk transmission nodes and their residence time in each stage in advance. This results in delayed intervention timing and insufficient intervention precision, making it difficult to achieve proactive and targeted governance of product quality risks. Summary of the Invention

[0003] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide a product quality intelligent control method based on an information management system, the method comprising: Obtain the set of historical quality sampling inspection records corresponding to the batch of products to be controlled, wherein the set of historical quality sampling inspection records contains multiple quality indicator detection data units collected at different sampling time nodes; The historical quality sampling record set is subjected to quality fluctuation pattern mining processing to obtain the time-series fluctuation characteristics of the quality indicators of the batch of products to be controlled and the correlation transmission characteristics of the non-conforming items in the sampling inspection. Based on the time-series fluctuation characteristics of the quality indicators and the correlation transmission characteristics of the non-conforming items in the sampling inspection, a preset quality risk diffusion deduction logic is invoked to generate a quality risk diffusion path diagram for the batch of products to be controlled. Based on the quality risk diffusion path diagram, determine the risk transmission nodes and the risk retention time of each production link in the batch of products to be controlled; Based on the risk transmission node and the risk retention time of the risk transmission node, a targeted quality intervention instruction containing the timing of control intervention is generated, and the targeted quality intervention instruction is distributed to the control terminal of the corresponding production link.

[0004] Furthermore, embodiments of the present invention also provide a product quality intelligent control system based on an information management system, comprising: A processor; a machine-readable storage medium for storing machine-executable instructions of the processor; wherein the processor is configured to execute the above-described intelligent product quality control method based on an information management system by executing the machine-executable instructions.

[0005] In another aspect, embodiments of the present invention also provide a computer program product, the computer program product including machine-executable instructions, the machine-executable instructions being stored in a computer-readable storage medium, the processor of the product quality intelligent control system based on the information management system reading the machine-executable instructions from the computer-readable storage medium, the processor executing the machine-executable instructions, causing the product quality intelligent control system based on the information management system to execute the above-described product quality intelligent control method based on the information management system.

[0006] Based on the above, by acquiring historical quality sampling records of the batch of products to be controlled, containing multiple quality indicator detection data units at different sampling time points, and performing quality fluctuation pattern mining processing on them, the dynamic change patterns of quality indicators and the correlation transmission characteristics between sampling non-conforming items are fully captured at the time series level. On this basis, combining the time series fluctuation characteristics of quality indicators and the correlation transmission characteristics of sampling non-conforming items, a pre-set quality risk diffusion inference logic is invoked to generate a quality risk diffusion path diagram, realizing the mapping from discrete historical sampling data to a continuous risk diffusion topology, enabling the visualization and structured representation of the propagation mechanism of quality risk between various production links. Furthermore, based on this quality risk diffusion path diagram, the risk transmission nodes and risk retention duration of each production link are determined, transforming abstract quality fluctuations into risk node information with clear spatial location and time scale. Finally, based on the risk transmission nodes and their risk retention duration, targeted quality intervention instructions containing the timing of control intervention are generated and distributed to the control terminals of the corresponding production links, significantly improving the systematicness, foresight, and accuracy of product quality control. Attached Figure Description

[0007] Figure 1 This is a schematic diagram of the execution flow of the intelligent product quality control method based on an information management system provided in an embodiment of the present invention.

[0008] Figure 2 This is a schematic diagram of the visualization interface of the risk diffusion situation map of the production process in the product quality intelligent control method based on the information management system provided in the embodiment of the present invention.

[0009] Figure 3This is a schematic diagram of the interface for generating targeted quality intervention instructions in the product quality intelligent control method based on an information management system provided in this embodiment of the invention.

[0010] Figure 4 This is a schematic diagram of the traceability interception command interface in the product quality intelligent control method based on an information management system provided in this embodiment of the invention.

[0011] Figure 5 This is a schematic diagram of the multi-source risk fusion configuration interface in the product quality intelligent control method based on an information management system provided in this embodiment of the invention. Detailed Implementation

[0012] Figure 1 This is a flowchart illustrating a product quality intelligent control method based on an information management system, provided in one embodiment of the present invention. The method mines quality fluctuation patterns and correlation transmission characteristics from historical quality inspection records, invokes quality risk diffusion deduction logic to generate a quality risk diffusion path diagram, locates risk transmission nodes and their retention duration, and generates targeted quality intervention instructions containing control intervention timing, which are then distributed to the control terminals of the corresponding production stages. All quality inspection data, production equipment data, and environmental monitoring data involved in this embodiment originate from the enterprise's authorized production information management system, and the data collection and usage process strictly adheres to the enterprise's data security management system.

[0013] Step S110: Obtain the set of historical quality sampling inspection records corresponding to the batch of products to be controlled. The set of historical quality sampling inspection records contains multiple quality indicator detection data units collected at different sampling time nodes.

[0014] The historical quality inspection records are retrieved from the quality database module of the enterprise information management system in the form of a structured data table. Each row in this structured data table corresponds to one inspection record, and each column corresponds to a quality indicator detection data unit. Each inspection record row includes the inspection batch number, inspection timestamp, number of samples inspected, the detection value of each quality indicator, and a list of non-conforming items. The data type of the quality indicator detection data unit is either numerical data with units or enumerated data. The inspection timestamp records the precise moment each inspection operation occurred, used to construct the subsequent time series arrangement structure.

[0015] Step S120: Perform quality fluctuation pattern mining processing on the historical quality sampling record set to obtain the time-series fluctuation characteristics of the quality indicators of the batch of products to be controlled and the correlation transmission characteristics of the non-conforming items in the sampling inspection.

[0016] Step S121: Extract the timestamp of each quality indicator detection data unit in the historical quality sampling record set, construct the time series arrangement structure of the multiple quality indicator detection data units based on the timestamp, identify the numerical fluctuation trajectory of each quality indicator in the time series arrangement structure, and separate the periodically repeating quality indicator fluctuation segments from the numerical fluctuation trajectory.

[0017] Each sampling record in the historical quality inspection record set is arranged in ascending order by timestamp, forming a time series arrangement structure. For each quality indicator, the detection value at each sampling time node is extracted along this time series arrangement structure, constituting the numerical fluctuation trajectory of the quality indicator. A time series decomposition algorithm is used to decompose the numerical fluctuation trajectory of each quality indicator into a trend component, a periodic fluctuation component, and a random residual component. Autocorrelation analysis is performed on the periodic fluctuation component, and the lag number where the autocorrelation coefficient exceeds a preset autocorrelation threshold is extracted. Based on this lag number, periodically recurring quality indicator fluctuation segments are extracted from the periodic fluctuation component.

[0018] Step S122: Perform fluctuation amplitude span analysis and fluctuation direction conversion frequency analysis on the fluctuation segments of the quality indicators to obtain the fluctuation amplitude span and fluctuation direction conversion frequency of each quality indicator between adjacent sampling time nodes, and construct the time-series fluctuation characteristics of the quality indicators of the batch of products to be controlled based on the fluctuation amplitude span and the fluctuation direction conversion frequency.

[0019] For each quality indicator's fluctuation segment, the absolute value of the difference between the measured values ​​at adjacent sampling time points is calculated point by point along the time direction as the fluctuation range. The fluctuation direction switching frequency is the ratio of the number of times the measured value of the quality indicator changes from rising to falling or from falling to rising within the time length of the fluctuation segment to the time length. The mean and standard deviation of the fluctuation range of each quality indicator, along with the fluctuation direction switching frequency, are organized into a feature vector, which represents the time-series fluctuation characteristics of the quality indicator.

[0020] Step S123: Extract the non-conforming item markers from each sampling record in the historical quality sampling record set, count the co-occurrence frequency of different non-conforming item markers in the same sampling record, construct a non-conforming item co-occurrence frequency distribution table based on the co-occurrence frequency, and determine non-conforming item groups with stable co-occurrence relationships in the non-conforming item co-occurrence frequency distribution table.

[0021] The algorithm iterates through the list of non-conforming item tags for each sampling record in the historical quality sampling record set. For each sampling record, it pairs all non-conforming item tags and counts the co-occurrence frequency of each tag pair. The co-occurrence frequency is divided by the total number of sampling records containing either tag in the pair to obtain the co-occurrence ratio. Non-conforming item tag pairs with a co-occurrence ratio exceeding a preset threshold are identified as having a stable association relationship. All non-conforming item tag pairs with stable association relationships are aggregated into a non-conforming item group using a transitive closure algorithm.

[0022] Step S124: Arrange the first occurrence time nodes of each non-conforming item in the non-conforming item group in a time sequence to obtain the triggering time sequence chain of the non-conforming item group. Calculate the inter-item propagation delay duration based on the time interval between the occurrence times of adjacent non-conforming items in the triggering time sequence chain, and extract the fluctuation range of the inter-item propagation delay duration.

[0023] For each group of non-conforming items, extract the timestamp of the first inspection that marked each non-conforming item as non-conforming in the historical quality inspection record set, and arrange them in chronological order to obtain the triggering time sequence chain. Calculate the time interval between the occurrence times of two adjacent non-conforming items in the triggering time sequence chain as the inter-item propagation delay duration. Calculate the maximum and minimum values ​​of the inter-item propagation delay duration at corresponding adjacent positions in all triggering time sequence chains; these maximum and minimum values ​​constitute the fluctuation range of the inter-item propagation delay duration.

[0024] Step S125: Construct the correlation transmission characteristics of the non-compliant items in the sampling inspection based on the fluctuation range of the triggering timing chain and the transmission delay duration between the items.

[0025] The order of non-conforming items in the triggering sequence chain and the fluctuation range of the inter-item transmission delay between adjacent items are organized into an associated transmission feature data object, which is the associated transmission feature of the non-conforming items in the sampling inspection.

[0026] Step S130: Based on the time-series fluctuation characteristics of the quality indicators and the correlation transmission characteristics of the non-conforming items in the sampling inspection, the preset quality risk diffusion deduction logic is invoked to generate the quality risk diffusion path diagram of the batch of products to be controlled.

[0027] Step S131: Obtain the production process topology structure corresponding to the batch of products to be controlled. The production process topology structure includes multiple process nodes and the material flow direction relationship between the multiple process nodes.

[0028] The production process topology corresponding to the batch of products to be controlled is obtained from the process management module of the enterprise information management system. This production process topology is a directed acyclic graph, where process nodes represent independent production links in the product processing, and directed edges represent the directional relationships of material flow from one process node to the next. Each process node contains a process identifier, a process name, and an identifier of the production workshop to which the process belongs.

[0029] Step S132: Map and associate each quality indicator in the time-series fluctuation characteristics of the quality indicators with the multiple process nodes to determine the quality monitoring process node corresponding to each quality indicator.

[0030] The identifier of the detection process node corresponding to each quality indicator is read from the quality standard configuration file of the enterprise information management system. Each quality indicator in the time-series fluctuation characteristics is mapped to the corresponding process node in the production process topology according to the correspondence. This process node is the quality monitoring process node for that quality indicator.

[0031] Step S133: Expand the trigger timing chain in the associated transmission characteristics of the non-conforming items in the sampling inspection into a project transmission network containing the initial triggering project node and subsequent transmission project nodes. According to the transmission direction relationship between each non-conforming item in the project transmission network, mark the diffusion direction of the non-conforming items in the production process topology.

[0032] Each non-conforming item in the trigger sequence chain is mapped to the corresponding quality monitoring process node in the production process topology. Along the trigger sequence chain from the initial trigger node to subsequent propagation nodes, diffusion flow direction markers for the non-conforming items are added between the corresponding process nodes in the production process topology. If there is no direct material flow direction relationship between two process nodes in the production process topology, diffusion flow direction markers are added along the shortest path in the production process topology.

[0033] Step S134: Convert the fluctuation amplitude span and fluctuation direction conversion frequency in the time-series fluctuation characteristics of the quality indicators into risk intensity parameters and risk change trend parameters of the quality monitoring process node, respectively. Calculate the risk diffusion probability of the quality monitoring process node spreading risk to its adjacent process nodes based on the risk intensity parameters and the risk change trend parameters.

[0034] The risk intensity parameter is proportional to the mean of the fluctuation range divided by the standard tolerance range of the quality indicator. The risk change trend parameter is proportional to the frequency of fluctuation direction reversal. For each quality monitoring process node, all its downstream adjacent process nodes are extracted. The risk diffusion probability pij = Si × Ti × wij, where Si is the normalized value of the risk intensity parameter, Ti is the normalized value of the risk change trend parameter, and wij is the material flow weight from process node i to process node j, which is equal to the proportion of material flowing from process node i to process node j to the total material quantity.

[0035] Step S135: Input the risk diffusion probability and the diffusion direction of the non-conforming items into the preset quality risk diffusion deduction logic, perform risk state iterative update processing on the production process topology, and generate the risk state evolution sequence of each process node in the production process topology.

[0036] The pre-defined quality risk diffusion simulation logic employs a risk propagation algorithm based on discrete time steps. The initial risk state value is set as the risk intensity parameter of each quality monitoring process node at the initial time step, while the risk state value of other process nodes is 0. At each time step, for each process node, if its current risk state value is greater than a preset risk activation threshold, the risk is propagated to all downstream adjacent process nodes according to its risk diffusion probability. The propagation amount is the current risk state value multiplied by the risk diffusion probability. The risk state value of the downstream process node in the next time step is the sum of its own risk decay value from the previous time step and the received propagated risk value. This iteration is repeated until the risk state values ​​of all process nodes converge or the preset maximum iteration time step is reached. The sequence of risk state values ​​for each process node across all time steps constitutes a risk state evolution sequence.

[0037] Step S136: Extract the risk activation time and risk saturation time of each process node according to the risk state evolution sequence, and spatiotemporally splice the risk activation time and risk saturation time of each process node based on the material flow direction relationship of the production process topology to generate the quality risk diffusion path map.

[0038] The risk activation moment is the iteration time step when the risk state value in the risk state evolution sequence first exceeds the preset risk activation threshold. The risk saturation moment is the iteration time step after the risk state value in the risk state evolution sequence first reaches a certain proportion of its peak value. The risk activation moment and risk saturation moment of each process node are marked on the corresponding node in the production process topology. Adjacent process nodes are connected by directed edges along the material flow direction, and the risk diffusion probability and risk transmission direction are marked on the directed edges to form a quality risk diffusion path diagram.

[0039] Step S140: Based on the quality risk diffusion path diagram, determine the risk transmission nodes of each production link in the batch of products to be controlled and the risk retention time of the risk transmission nodes.

[0040] Step S141: Perform path branch point identification processing on the quality risk diffusion path diagram, extract the process nodes where the risk flow in the quality risk diffusion path diagram branches as candidate risk transmission nodes, and obtain the number of upstream risk inflow paths and the number of downstream risk outflow paths of the candidate risk transmission nodes in the quality risk diffusion path diagram.

[0041] Traverse each process node in the quality risk diffusion path graph and count the number of incoming and outgoing edges for that process node. The number of incoming edges represents the number of upstream risk inflow paths, and the number of outgoing edges represents the number of downstream risk outflow paths. If the number of outgoing edges is greater than 1, then the process node is a node where the risk flow has branched and is marked as a candidate risk propagation node.

[0042] Step S142: When the number of downstream risk outflow paths of the candidate risk transmission node is greater than the number of upstream risk inflow paths of the candidate risk transmission node, the candidate risk transmission node is determined as a risk transmission node.

[0043] Compare the number of downstream risk outflow paths (Nout) and the number of upstream risk inflow paths (Nin) for each candidate risk transmission node. If Nout is greater than Nin, the candidate risk transmission node has an amplifying and diffusion effect on the risk, and it is identified as a risk transmission node.

[0044] Step S143: Extract the risk activation time and risk saturation time corresponding to the risk transmission node in the quality risk diffusion path diagram, calculate the time difference between the risk saturation time and the risk activation time, use the time difference as the risk retention time of the risk transmission node, and associate and store the process identifier of the risk transmission node in the production process topology with the risk retention time.

[0045] Read the risk activation time (tact) and risk saturation time (tsat) of the risk transmission node from the quality risk diffusion path diagram. The risk residence time (tdur) is calculated as tsat - tact. Store the process identifier of the risk transmission node and the risk residence time (tdur) as key-value pairs in the risk transmission node information table.

[0046] Step S144: Based on the path connection relationship between each risk transmission node in the quality risk diffusion path diagram, determine the risk transmission level of the risk transmission node, extract the risk retention time of other risk transmission nodes at the same risk transmission level as the risk transmission node, and calculate the distribution central tendency parameter of the risk retention time within the same risk transmission level.

[0047] Using the risk source node in the quality risk diffusion path diagram as layer 0, the layer depth increases by 1 for each directed edge along the risk diffusion direction to determine the risk transmission layer of each risk transmission node. For all risk transmission nodes at the same risk transmission layer, the risk retention time (tdur) is collected, and the median of the risk retention time at that layer is calculated as the distribution central tendency parameter (Tmed).

[0048] Step S145: Determine the degree of abnormality of the risk transmission node based on the degree of deviation between the risk retention time of the risk transmission node and the distribution central tendency parameter.

[0049] The absolute deviation Ddev of the risk retention time tdur at the risk transmission node and the central tendency parameter Tmed at that level is calculated as: Ddev = absolute value of tdur - Tmed divided by Tmed. If Ddev exceeds a preset deviation threshold, the retention abnormality of the risk transmission node is marked as abnormal retention; otherwise, it is marked as normal retention.

[0050] Step S150: Generate a targeted quality intervention instruction containing the timing of control intervention based on the risk transmission node and the risk retention time of the risk transmission node, and distribute the targeted quality intervention instruction to the control terminal of the corresponding production link.

[0051] Step S151: Obtain the process identifier of the risk transmission node in the production process topology, and query the control terminal address and control terminal communication protocol type corresponding to the risk transmission node based on the process identifier.

[0052] Query the control terminal address and control terminal communication protocol type corresponding to the process identifier from the equipment management module of the enterprise information management system. The control terminal address is a combination of network IP address and port number, and the control terminal communication protocol type is either Message Queuing Telemetry Transport Protocol or Hypertext Transfer Protocol.

[0053] Step S152: Determine the control intervention sequence of the risk transmission node based on the risk retention time of the risk transmission node. The control intervention sequence includes the control initiation time and the control duration.

[0054] The control activation time is the current time plus a preset lead time. The control duration is equal to the risk retention time (tdur) multiplied by the control duration coefficient. If the abnormal retention level is marked as abnormal retention, the control duration coefficient is taken as a larger value, and the lead time for control activation is taken as a smaller value to intervene as early as possible.

[0055] Step S153: Obtain the number of risk inflow paths and the number of risk outflow paths corresponding to the risk transmission node, and determine the risk amplification index of the risk transmission node based on the number of risk inflow paths and the number of risk outflow paths.

[0056] The risk amplification index Aamp = Nout / (Nin + 1), where Nout is the number of downstream risk outflow paths and Nin is the number of upstream risk inflow paths. The larger the Aamp, the stronger the risk amplification effect of the risk transmission node.

[0057] Step S154: Generate the control measure intensity level parameter of the risk transmission node based on the risk retention time and the risk amplification index.

[0058] The control measure intensity level parameter Lctrl = ceil(α × tdur_norm + β × Aamp_norm), where tdur_norm is the normalized value of the risk dwell time, Aamp_norm is the normalized value of the risk amplification index, α and β are preset weighting coefficients, and ceil is the round-up function. Lctrl is an integer level value.

[0059] Step S155: Call the preset control instruction template library, match the corresponding control instruction template according to the control measure intensity level parameter, fill the control intervention sequence and the process identifier of the risk transmission node into the control instruction template, and generate a targeted quality intervention instruction.

[0060] The control instruction template library stores control instruction templates corresponding to different control measure intensity levels. Each control instruction template includes a control measure description field, control parameter configuration fields, and control execution requirement fields. Based on the Lctrl search, a matching control instruction template is retrieved, and the control initiation time, control duration, and process identifier are filled into the corresponding placeholders in the control instruction template to generate a targeted quality intervention instruction.

[0061] Step S156: Perform protocol encapsulation processing on the targeted quality intervention instruction according to the communication protocol type of the control terminal, generate an encapsulated instruction data packet that matches the communication protocol type of the control terminal, and send the encapsulated instruction data packet to the control terminal corresponding to the control terminal address through the message distribution channel.

[0062] If the communication protocol type of the control terminal is Message Queuing Telemetry Transport Protocol (MQTP), the targeted quality intervention instruction is serialized into a JSON string and then encapsulated into a MQTP publish message packet. If the communication protocol type of the control terminal is Hypertext Transfer Protocol (HTTP), the targeted quality intervention instruction is encapsulated into an HTTP request body. The encapsulated instruction data packet is sent to the control terminal address through the message distribution channel, triggering the control terminal to execute the quality intervention operation corresponding to the control measure intensity level parameter.

[0063] Step S157: Receive the control execution feedback information returned by the control terminal, and update the risk status marker of the risk transmission node in the quality risk diffusion path diagram according to the control execution feedback information.

[0064] The control execution feedback information includes the control execution completion status and the quality sampling inspection data of the process node after control. Based on the quality sampling inspection data after control, the risk intensity parameter of the risk transmission node is recalculated. If the risk intensity parameter drops below the preset safety threshold, the risk status of the risk transmission node is updated to the controlled status, and the quality risk diffusion inference logic is re-executed to update the risk status evolution sequence of the downstream process nodes.

[0065] Step S210: Obtain the historical supply quality data set of the raw material suppliers corresponding to the batch of products to be controlled. The historical supply quality data set includes the supply batch records of each raw material supplier and the quality inspection pass rate data of each batch of supply.

[0066] Historical supply quality data sets of raw material suppliers are extracted from the supplier management module of the enterprise information management system. This historical supply quality data set is organized using supplier ID as the primary key, with each supplier ID linked to a list of supply batch records. Each supply batch record includes the batch number, supply timestamp, supply quantity, number of samples inspected, and inspection pass rate. The inspection pass rate is calculated by dividing the number of qualified samples by the total number of samples inspected.

[0067] Step S220: Based on the quality inspection pass rate data of each batch of supply in the historical supply quality data set, construct the time series change curve of the quality pass rate of each raw material supplier, perform trend inflection point detection processing on the time series change curve of the quality pass rate, and identify the trend inflection point moment when the quality pass rate changes from a stable state to a declining state in the time series change curve of the quality pass rate.

[0068] For each raw material supplier, the supplier's supply batch records are sorted in ascending order by supply timestamp. A time-series curve of the quality pass rate is plotted with the supply timestamp on the horizontal axis and the sampling pass rate on the vertical axis. A sliding window piecewise linear fitting algorithm is used to detect trend inflection points on the time-series curve of the quality pass rate. The time-series curve of the quality pass rate is divided into continuous time window segments, and the slope value of each segment is obtained by least-squares linear fitting. The slope value is compared segment by segment from the starting segment. When the slope value of several consecutive segments changes from close to 0 to a continuously negative value, and the absolute value of the negative value exceeds a preset slope threshold, the starting time point of that turning point segment is identified as the trend inflection point.

[0069] Step S230: Extract the quality inspection pass rate data of multiple consecutive batches of raw material suppliers after the trend inflection point, and calculate the quality degradation rate parameter of the raw material supplier based on the quality inspection pass rate data of the multiple consecutive batches of raw material suppliers.

[0070] Starting from the trend inflection point, the quality inspection pass rate values ​​of the subsequent Q consecutive batches supplied by the raw material supplier are extracted, forming a quality pass rate decline sequence. An exponential decay fitting is applied to the quality pass rate decline sequence, with the fitting function being y = y0 × exp(-λ × t), where y0 is the quality pass rate value at the trend inflection point, λ is the quality decay rate parameter, and t is the batch number starting from the trend inflection point. The least squares method is used to solve for the value of λ as the quality decay rate parameter for the raw material supplier.

[0071] Step S240: Obtain the raw material supplier identifier corresponding to the raw material batch used in the batch of products to be controlled, match the corresponding quality degradation rate parameter according to the raw material supplier identifier, and deduce the contribution of the raw material quality in the batch of products to be controlled to the diffusion process of finished product quality risk based on the quality degradation rate parameter and the production time node of the batch of products to be controlled.

[0072] Extract the raw material supplier identifier corresponding to the raw material batch used in the bill of materials of the product batch to be controlled, and match the corresponding quality degradation rate parameter λ from the calculation result of step S230. Based on the time difference Δt between the production time node of the product batch to be controlled and the trend inflection point of the corresponding raw material supplier, calculate the remaining quality level of the raw material Qrem = Q0 × exp(-λ × Δt), where Q0 is the initial quality level of the raw material. The contribution Craw = 1 - Qrem represents the degree of contribution of raw material quality degradation to the finished product quality risk. Superimpose the contribution Craw onto the risk intensity parameter of the initial risk node in the quality risk diffusion path diagram, i.e., the first process node using the raw material batch, and update the risk intensity parameter Snew = Sori × (1 + Craw), where Sori is the original risk intensity parameter. Based on the updated risk intensity parameter, re-execute steps S135 to S136 to update the risk state evolution sequence of each process node in the quality risk diffusion path diagram.

[0073] Step S310: Obtain the set of production equipment operation status records corresponding to the batch of products to be controlled. The set of production equipment operation status records contains time series data of operation parameters of each production equipment during the production period of the batch of products to be controlled. Perform abnormal pattern segment extraction processing on the time series data of operation parameters to obtain a set of abnormal segments of operation parameters of the production equipment during the production period of the batch of products to be controlled.

[0074] Time-series data of operating parameters for each production device are extracted from the equipment management module of the enterprise information management system. These parameters include spindle speed, feed rate, operating temperature, and vibration amplitude. For each operating parameter's time-series data, its moving mean and moving standard deviation are calculated, with a preset window width. Continuous data points deviating from the moving mean by more than K times the moving standard deviation are marked as abnormal segments of the operating parameter. Each abnormal segment records the start time, end time, duration, and peak deviation of the abnormality. All abnormal segments of the operating parameters constitute a set of abnormal operating parameter segments.

[0075] Step S320: Based on the duration and frequency of occurrence of abnormal segments in the set of abnormal operating parameters, construct the operational stability deviation feature of the production equipment, associate the operational stability deviation feature of the production equipment with the process nodes in the production process topology, and determine the affected process nodes affected by the operating status of the production equipment.

[0076] For each production equipment, the average duration and frequency of all abnormal segments in its set of abnormal operating parameters are calculated. The operational stability deviation feature Ddev = τ1 × Tavg + τ2 × F, where Tavg is the average duration of abnormal segments, F is the frequency of abnormal segments, and τ1 and τ2 are preset weighting coefficients. The process node corresponding to each production equipment is queried from the process route configuration in the enterprise information management system. The operational stability deviation feature Ddev is associated with this process node, marking it as an affected process node.

[0077] Step S330: Obtain the risk activation time corresponding to the affected process node in the quality risk diffusion path diagram, compare the overlap between the risk activation time of the affected process node and the time range of the abnormal operating parameters of the production equipment, adjust the risk diffusion probability of the affected process node according to the overlap comparison result, and increase the risk diffusion probability of the affected process node when the overlap exceeds the preset overlap threshold.

[0078] Extract the risk activation time `tact` from the quality risk diffusion path diagram for the affected process nodes. Iterate through the set of abnormal operating parameter segments of the production equipment corresponding to the affected process node, checking whether the time range [tstart, tend] of each abnormal segment overlaps with `tact`, i.e., satisfying `tstart ≤ tact ≤ tend`. If overlapping abnormal segments exist, calculate the degree to which the overlapping time length covers the risk activation time. If the overlap exceeds a preset overlap threshold, increase the risk diffusion probability of the affected process node by multiplying it by (1 + ω × Ddev), where ω is an adjustment coefficient.

[0079] Step S410: Obtain the production environment monitoring data set corresponding to the batch of products to be controlled. The production environment monitoring data set includes the temperature and humidity monitoring value sequence and the air cleanliness monitoring value sequence of the production environment during the production period of the batch of products to be controlled. Perform environmental parameter fluctuation pattern analysis on the temperature and humidity monitoring value sequence and the air cleanliness monitoring value sequence to extract the transient events and gradual drift trends of environmental parameters in the production environment during the production period of the batch of products to be controlled.

[0080] Temperature and humidity monitoring data sequences and air cleanliness monitoring data sequences were extracted from the environmental monitoring module of the enterprise information management system. A mutation point detection algorithm was used to identify transient events in the temperature, humidity, and air cleanliness monitoring data sequences where the values ​​changed by more than a preset mutation threshold within a short period. The occurrence time, duration, and magnitude of each transient event were recorded. A linear trend analysis algorithm was used to fit the overall trend of the temperature, humidity, and air cleanliness monitoring data sequences, extracting the direction and rate of the gradual drift trend of the environmental parameters.

[0081] Step S420: Determine the environmental disturbance impact window of the production environment based on the occurrence time and duration of the transient event of the environmental parameters.

[0082] For each transient event of an environmental parameter, the start time of the environmental disturbance impact window is the occurrence time of the transient event, and the end time of the environmental disturbance impact window is the occurrence time of the transient event plus the duration multiplied by the preset window expansion coefficient.

[0083] Step S430: Determine the cumulative direction of environmental parameter offset in the production environment based on the direction and rate of the gradual drift trend of the environmental parameters.

[0084] The direction of the gradual drift trend of environmental parameters is the direction of deviation from the center value of the standard environmental parameter interval, and it is divided into positive deviation and negative deviation. The rate of cumulative deviation of environmental parameters is equal to the slope of the linear fit of the gradual drift trend of environmental parameters. The greater the rate of deviation from the center value of the standard interval, the more significant the impact of the cumulative deviation of environmental parameters on environmentally sensitive process nodes.

[0085] Step S440: Obtain a list of process nodes in production state within the environmental disturbance impact window, and align the risk status of each process node in the process node list with the environmental disturbance impact window in the time dimension.

[0086] Query the production planning and scheduling data to find a list of process nodes that are in production within the environmental disturbance impact window. Align the risk state evolution sequence of each process node in the quality risk diffusion path diagram with the environmental disturbance impact window in terms of time dimension, and extract the risk state data segments that fall within the environmental disturbance impact window from the risk state evolution sequence.

[0087] Step S450: Obtain the set of process nodes corresponding to the sensitive quality indicators in the cumulative direction of the environmental parameter offset, and proportionally amplify the risk diffusion probability of each process node in the set of process nodes according to the rate of the gradual drift trend of the environmental parameter.

[0088] From the quality characteristic configuration file of the enterprise information management system, query the set of process nodes corresponding to the quality indicators that are sensitive to temperature and humidity and air cleanliness. For each process node in the set, multiply its risk diffusion probability by a proportional amplification factor, where proportional amplification factor = 1 + ζ × vdrift, where vdrift is the rate normalized value of the gradual drift trend of the environmental parameter, and ζ is the amplification ratio coefficient.

[0089] Step S510: Obtain the set of quality complaint records for historical similar product batches corresponding to the product batch to be controlled. The set of quality complaint records for historical similar product batches includes the batch number of the complained product, the type of quality problem complained about, and the time when the complaint occurred.

[0090] Extract a collection of historical quality complaint records for similar product batches from the after-sales management module of the enterprise information management system. Historical similar product batches are identified by screening market-released product batches that share the same product model, manufacturing process, or intended use as the batch to be controlled. Each quality complaint record includes the batch number of the complained product, a description of the quality issue, and a timestamp indicating when the complaint occurred.

[0091] Step S520: Classify and aggregate the historical quality complaint records of the same product batch according to the complaint quality problem type to obtain the complaint event sequence corresponding to different complaint quality problem types. Perform complaint time interval distribution analysis on the complaint event sequence corresponding to each complaint quality problem type to obtain the central tendency parameter and dispersion parameter of the complaint time interval.

[0092] Quality complaint records are categorized and aggregated according to the type of quality issue. Complaint records of the same quality issue type are arranged in ascending order by the timestamp of the complaint occurrence to form a complaint event sequence. The time interval between adjacent complaint events in the complaint event sequence is calculated, and the median of all complaint time intervals is used as a parameter of central tendency. The coefficient of variation of all complaint time intervals is used as a parameter of dispersion.

[0093] Step S530: Construct a distribution model of the latent period of quality hazards corresponding to each type of complaint quality problem based on the central tendency parameter and dispersion parameter of the complaint time interval.

[0094] For each type of quality complaint, a normal distribution model is constructed. The mean of the normal distribution is the central tendency parameter, and the standard deviation is the dispersion parameter multiplied by the central tendency parameter. This normal distribution model is the distribution model of the latent period of quality hazards corresponding to this type of quality complaint, describing the probability distribution of the time interval from product leaving the factory to the occurrence of the complaint.

[0095] Step S540: Match the non-conforming item types corresponding to each risk transmission node in the quality risk diffusion path diagram of the batch of products to be controlled with the complaint quality problem types, obtain the risk activation time of the successfully matched risk transmission nodes, and deduce the complaint outbreak prediction time window of the quality hidden dangers corresponding to the risk transmission nodes after the batch of products to be controlled is put into the market based on the quality hidden danger incubation period distribution model.

[0096] Each non-conforming item type corresponding to a risk transmission node in the quality risk diffusion path diagram of the product batch to be controlled is matched against the complaint quality issue type. For successfully matched risk transmission nodes, their risk activation time is obtained. Starting from the planned shipment time of the product batch to be controlled, the predicted delay time of the complaint outbreak corresponding to that risk transmission node is sampled from the quality hazard incubation period distribution model. The start time of the complaint outbreak prediction time window is the time corresponding to the lower quantile of the planned shipment time plus the predicted delay time of the complaint outbreak, and the end time is the time corresponding to the upper quantile of the planned shipment time plus the predicted delay time of the complaint outbreak.

[0097] Step S550: Generate an after-sales quality proactive intervention instruction for the batch of products to be controlled based on the complaint outbreak prediction time window. The after-sales quality proactive intervention instruction includes the proactive recall trigger time and the proactive recall coverage.

[0098] The earliest start time within the predicted complaint outbreak time window corresponding to all risk transmission nodes is taken as the proactive recall trigger time. The recall coverage is the range of product batches corresponding to risk transmission nodes whose predicted complaint outbreak time window falls within the preset market circulation cycle. The proactive after-sales quality intervention instruction includes two fields: proactive recall trigger time and recall coverage, and is sent to the enterprise's after-sales service management module.

[0099] For example, step S610: Obtain the production plan scheduling data and warehousing and logistics flow data corresponding to the batch of products to be controlled. The production plan scheduling data includes the planned start time and planned completion time of each production link in the batch of products to be controlled. The warehousing and logistics flow data includes the entry time, exit time and flow destination of semi-finished products and finished products in the batch of products to be controlled.

[0100] Production planning and scheduling data and warehouse logistics data are obtained from the production planning and warehouse management modules of the enterprise information management system, respectively. The production planning and scheduling data is organized by process node number, and the warehouse logistics data is organized by material batch number.

[0101] Step S620: Construct the spatiotemporal trajectory of material flow for the batch of products to be controlled based on the production plan scheduling data and the warehousing and logistics flow data.

[0102] Using the planned start and finish times of each process node in the production scheduling data as the production time axis, and the inbound and outbound times in the warehousing logistics data as the warehousing time axis, the process nodes are connected in series according to the material flow direction to construct a complete spatiotemporal trajectory of material flow from raw material input through each production stage to finished product warehousing. Each spatiotemporal point in the material flow trajectory includes a timestamp, process node identifier, material batch number, and destination information.

[0103] Step S630: Map the risk activation time of each risk transmission node in the quality risk diffusion path diagram to the material flow spatiotemporal trajectory, and determine the current position and current state of the semi-finished or finished product corresponding to the risk transmission node in the material flow spatiotemporal trajectory.

[0104] The activation time of each risk transmission node is located on the timeline of the material flow spatiotemporal trajectory. Based on the time location, the current position and status of the semi-finished or finished products that have flowed through the corresponding process node of the risk transmission node in the material flow spatiotemporal trajectory at that risk activation time are determined. If the current position is warehouse inventory, the current status is marked as pending shipment; if the current position is a downstream process node, the current status is marked as work in progress; if the current position is a customer delivery address, the current status is marked as delivered.

[0105] Step S640: Based on the risk retention time of the risk transmission node and the material flow speed in the material flow time-space trajectory, calculate the quantity of material that has flowed out of the production link corresponding to the risk transmission node within the risk retention time interval.

[0106] Material flow rate is calculated from the time interval and spatial distance between adjacent points in the material flow spatiotemporal trajectory. Within the risk retention time interval, the quantity of material that has flowed out of the production link corresponding to the risk transmission node = material flow rate × risk retention time × output per unit time.

[0107] Step S650: Obtain the destination information of the materials that have been transferred in the material flow time and space trajectory, and generate a list of material flow directions affected by quality risks based on the destination information.

[0108] Extract the destination information of the transferred materials at each time and space point in the material flow time and space trajectory, group and summarize them according to the destination, and generate a material flow list affected by quality risks. Each entry in the material flow list includes the destination, material quantity, material batch number and current status.

[0109] Step S660: Based on the material flow list affected by quality risks and the risk intensity parameters of the risk transmission nodes, generate a traceability interception instruction for the materials that have been transferred. The traceability interception instruction includes the batch identifier of the materials to be intercepted, the interception priority marker, and the list of interception destinations.

[0110] The traceability interception command extracts the batch number of the material to be intercepted from the material flow list affected by quality risks. The interception priority flag is determined according to the risk intensity parameter of the risk transmission node. The interception destination list is a collection of all flow destinations in the material flow list affected by quality risks. The traceability interception command is sent to the warehouse control terminal of each flow destination through the logistics management module of the enterprise information management system.

[0111] Step S710: Extract the sequence of process nodes that each risk diffusion path passes through in the quality risk diffusion path diagram, convert the sequence of process nodes into an ordered node chain with process nodes as elements, and obtain the material flow time interval between adjacent process nodes in each ordered node chain. The material flow time interval is the time consumed for material to flow from the preceding process node to the following process node.

[0112] The material flow time interval is extracted from the production planning and scheduling data by measuring the difference between the planned completion time and the planned start time of adjacent process nodes. For each pair of adjacent process nodes in an ordered node chain, the material flow time interval is recorded as the flow duration between the pair of adjacent process nodes.

[0113] Step S720: Based on the material flow time interval and the risk retention time corresponding to the preceding process node, calculate the risk cross-node transmission time in each ordered node chain from the preceding process node to the subsequent process node, correlate and compare the risk cross-node transmission time with the fluctuation amplitude span in the time-series fluctuation characteristics of the quality indicators corresponding to the subsequent process node, and determine the attenuation influence coefficient of the risk cross-node transmission time on the quality fluctuation amplitude of the subsequent process node.

[0114] The risk propagation time across nodes is Tcond = max(Tflow, Tret), where Tflow is the material flow time interval and Tret is the risk retention time of the preceding process node. The risk propagation time Tcond across nodes is correlated with the fluctuation amplitude span Aflu in the time-series fluctuation characteristics of the quality indicators of the subsequent process node, and the attenuation influence coefficient η = exp(-κ × Tcond / Aflu) is calculated, where κ is the attenuation rate constant.

[0115] Step S730: Extract the risk activation time of the first process node in each ordered node chain, and calculate the delay offset of the risk arrival time and the residual amount of risk intensity decay along the ordered node chain node by node based on the risk cross-node propagation time and the attenuation influence coefficient.

[0116] For the k-th process node in an ordered node chain, the delay offset of the risk arrival time is ΔTk = ΔTk-1 + Tcond_k, where ΔTk-1 is the delay offset of the previous node, and Tcond_k is the cross-node propagation time of the risk from the (k-1)-th node to the k-th node. The residual attenuation of the risk intensity is Sk = Sk-1 × ηk, where Sk-1 is the risk intensity of the previous node, and ηk is the attenuation influence coefficient from the (k-1)-th node to the k-th node. Initial conditions: ΔT1 = 0, S1 = Sinit.

[0117] Step S740: Based on the delay offset and the attenuation residual, perform time axis compression or stretching on the risk state evolution sequence of each process node in the quality risk diffusion path diagram to generate a corrected risk state evolution sequence. Based on the corrected risk state evolution sequence, extract the corrected risk saturation time of each process node. Compare the deviation between the corrected risk saturation time and the original risk saturation time to obtain the time deviation of each process node.

[0118] The time axis of the risk state evolution sequence for each process node is shifted by ΔTk. The risk value is multiplied by the ratio of the decay residual Sk to the original risk value to generate a corrected risk state evolution sequence. The corrected risk saturation time tsat_new is extracted from the corrected risk state evolution sequence. The time deviation Δtsat = tsat_new - tsat_ori.

[0119] Step S750: Construct a time deviation distribution pattern of the production process topology based on the time deviation of all process nodes. Identify process nodes whose time deviation exceeds the deviation tolerance range in the time deviation distribution pattern as time-sensitive risk nodes. Mark the upstream process nodes of the material flow corresponding to the time-sensitive risk nodes in the production process topology as time control bottleneck process nodes. Generate material flow speed control instructions for the time control bottleneck process nodes based on the time deviation.

[0120] The time deviations of all process nodes are displayed as a heatmap on the production process topology, forming a time deviation distribution pattern. A deviation tolerance range [δmin, δmax] is set, and process nodes with time deviations exceeding this range are marked as time-sensitive risk nodes. The upstream process nodes of the material flow of these time-sensitive risk nodes are extracted from the production process topology and marked as time control bottleneck process nodes. Material flow speed control instructions include the target process node identifier, the current material flow speed, and the suggested adjustment speed.

[0121] Step S810: Extract the risk state evolution sequence of each process node in the quality risk diffusion path diagram, and extract the risk growth stage data segment from the risk activation time to the risk saturation time in the risk state evolution sequence.

[0122] For the risk state evolution sequence of each process node, starting from the risk activation time and ending at the risk saturation time, the risk value sequence within this interval is extracted as a data segment of the risk growth stage.

[0123] Step S820: Perform growth curve shape classification processing on the data segments of the risk growth stage, and divide the data segments of the risk growth stage into gradual cumulative growth curves, sudden leap growth curves, or fluctuating climbing growth curves.

[0124] Calculate the first-order difference sequence of the data segment during the risk growth phase. If the relative standard deviation of the first-order difference sequence is less than a preset stability threshold and the mean is small, it is classified as a gradual cumulative growth curve. If the first-order difference sequence has a maximum point exceeding a preset jump threshold and the first-order difference values ​​before that point are small, it is classified as a sudden jump growth curve. If the first-order difference sequence shows alternating positive and negative values ​​and the absolute values ​​are all large, it is classified as a fluctuating upward growth curve.

[0125] Step S830: Obtain the process node corresponding to the gradual cumulative growth curve as the gradual risk node, obtain the process node corresponding to the sudden leap growth curve as the sudden risk node, and obtain the process node corresponding to the fluctuating climbing growth curve as the fluctuation risk node.

[0126] The growth curve shape classification results are mapped to the corresponding process nodes, and then classified into the gradual risk node set, the sudden risk node set, and the fluctuation risk node set according to the three types.

[0127] Step S840: Extract the set of upstream process nodes of the progressive risk node in the production process topology, analyze the time lag correlation between the risk retention time of each process node in the upstream process node set and the risk activation time of the progressive risk node, and determine the upstream continuous pressure source process node that drives the accumulation of progressive risk.

[0128] Calculate the Pearson correlation coefficient between the risk retention time of each process node in the upstream process node set and the risk activation time of the progressive risk node, and take the upstream process node with the maximum correlation coefficient that exceeds the preset correlation threshold as the upstream continuous pressure source process node.

[0129] Step S850: Extract the set of upstream process nodes of the sudden risk node in the production process topology, analyze the mutation of the risk intensity parameter of each process node in the upstream process node set before the risk activation time of the sudden risk node, and determine the upstream mutation shock source process node that triggers the sudden risk leap.

[0130] Extract the change sequence of risk intensity parameters of each process node in the upstream process node set within a preset time window before the risk activation time of the sudden risk node, and calculate the peak change rate of this change sequence. Take the upstream process node corresponding to the maximum peak change rate as the upstream sudden shock source process node.

[0131] Step S860: Extract the set of upstream process nodes of the fluctuation risk node in the production process topology, analyze the synchronicity between the alternating change frequency of the risk intensity parameter of each process node in the upstream process node set and the risk growth fluctuation frequency of the fluctuation risk node, and determine the upstream oscillation transmission source process node that causes the fluctuation risk.

[0132] Calculate the coherence coefficient between the power spectral density of the risk intensity parameter sequence of each process node in the upstream process node set and the power spectral density of the risk growth data segment of the volatility risk node. Take the upstream process node corresponding to the maximum coherence coefficient as the upstream oscillation propagation source process node.

[0133] Step S870: Based on the risk transmission characteristics of the upstream continuous pressure source process node, the upstream sudden shock source process node, and the upstream oscillation transmission source process node, assign differentiated upstream risk blocking strategy labels to each type of source process node, and generate risk parameter reconfiguration instructions for each upstream source process node.

[0134] For upstream processes experiencing continuous pressure, the upper limit of their risk retention time is lowered to a preset percentage of the original value. For upstream processes experiencing sudden shocks, the instantaneous peak value limit of their risk intensity parameter is lowered to a preset percentage of the original value. For upstream processes experiencing oscillation propagation, the upper limit of their risk intensity parameter fluctuation frequency is lowered to a preset percentage of the original value. A risk parameter reconfiguration command is generated based on the lowered parameter values ​​and sent to the corresponding control terminal.

[0135] Step S910: The nodes of each process in the quality risk diffusion path diagram are layered according to their hierarchical position in the production process topology to obtain the vertical layered structure of the risk diffusion path diagram.

[0136] Starting with the first process node in the production process topology with an in-degree of 0 as the 0th layer, the layer depth increases by 1 for each directed edge along the material flow direction, and all process nodes are assigned to the corresponding layers to obtain a vertical layered structure.

[0137] Step S920: Within the same level of the vertical hierarchical structure, extract the risk activation time of all process nodes within that level, and sort the process nodes within that level according to the chronological order of their risk activation times to obtain the risk activation time sequence of the same level.

[0138] All process nodes within the same level are sorted in ascending order by risk activation time to form a risk activation time sequence at the same level.

[0139] Step S930: Calculate the time interval between the risk activation times of adjacent process nodes in the same-layer risk activation time sequence, compare the time interval with the preset same-layer risk transmission time benchmark value, and select process node pairs with a time interval less than the same-layer risk transmission time benchmark value as same-layer risk coupling candidate pairs.

[0140] Calculate the time interval between the risk activation times of adjacent process nodes and compare this time interval with the risk propagation time benchmark Tbase at the same level. If the time interval is less than Tbase, these two process nodes are considered as candidate pairs for risk coupling at the same level.

[0141] Step S931: Extract the risk inflow path of the two process nodes in the same-layer risk coupling candidate pair in the quality risk diffusion path diagram, and determine whether there is a shared upstream process node in the risk inflow path of the two process nodes. If there is a shared upstream process node, then the same-layer risk coupling candidate pair is determined as a same-layer common-source coupling risk node pair.

[0142] Extract all risk inflow paths of the two process nodes in the quality risk diffusion path graph from the candidate risk coupling pairs at the same level. Take the intersection of the sets of upstream process nodes traversed by the risk inflow paths of the two process nodes. If the intersection is not empty, there are shared upstream process nodes, and the candidate risk coupling pairs at the same level are identified as a common source coupled risk node pair at the same level.

[0143] Step S932: Extract the risk inflow path of the two process nodes in the same-layer risk coupling candidate pair in the quality risk diffusion path diagram, and determine whether there are intersecting path nodes in the risk inflow paths of the two process nodes. If there are intersecting path nodes, then the same-layer risk coupling candidate pair is determined as the same-layer cross-coupled risk node pair.

[0144] Extract the set of all process nodes on the risk inflow path of the two process nodes in the same-layer risk coupling candidate pair. If the two sets have other cross process nodes in addition to sharing the upstream process node, then the same-layer risk coupling candidate pair is determined as the same-layer cross-coupled risk node pair.

[0145] Step S940: For the same-layer common-source coupled risk node pair, calculate the length of the overlap time interval of the risk retention time of each of the two process nodes, and generate the superposition amplification coefficient of the same-layer common-source coupled risk node based on the length of the overlap time interval; for the same-layer cross-coupled risk node pair, calculate the Pearson correlation coefficient of the risk state evolution sequence of each of the two process nodes, and generate the resonance amplification coefficient of the same-layer cross-coupled risk node based on the Pearson correlation coefficient.

[0146] The superposition amplification factor Aoverlap = 1 + (Toverlap / min(Tdur1, Tdur2)), where Toverlap is the length of the overlap time interval, and Tdur1 and Tdur2 are the risk dwell times of the two process nodes, respectively. The resonance amplification factor Ares = 1 + |rpearson|, where rpearson is the Pearson correlation coefficient of the risk state evolution sequences of the two process nodes.

[0147] Step S950: Adjust the risk saturation time of the same-layer common-source coupled risk node pair forward according to the superposition amplification factor, adjust the risk saturation time of the same-layer cross-coupled risk node pair forward according to the resonance amplification factor, update the forward-adjusted risk saturation time to the quality risk diffusion path diagram, and recalculate the risk state evolution sequence of the affected subsequent process nodes.

[0148] The risk saturation time is shifted forward and adjusted to tsat_new=tsat_ori / (1+A), where A is the superposition amplification factor or resonance amplification factor. The adjusted risk saturation time is updated in the quality risk diffusion path diagram, and step S135 is re-executed to recalculate the risk state evolution sequence of the affected process nodes.

[0149] Step S1010: Extract the fluctuation direction conversion frequency from the temporal fluctuation characteristics of the quality indicators corresponding to each process node in the quality risk diffusion path diagram, and convert the fluctuation direction conversion frequency into the quality state reversal activity parameter of that process node; extract the inter-item transmission delay duration fluctuation range from the correlation transmission characteristics of the non-conforming items corresponding to each process node in the quality risk diffusion path diagram, and convert the inter-item transmission delay duration fluctuation range into the defect transmission instability parameter of that process node.

[0150] The quality state reversal activity parameter is equal to the frequency of fluctuation direction reversal. The defect propagation instability parameter is equal to the difference between the upper and lower limits of the fluctuation range of the inter-item propagation delay time, divided by the mean of the fluctuation range.

[0151] Step S1020: Construct a two-dimensional coordinate system for risk self-sustaining capacity assessment of each process node based on the quality state reversal activity parameter and the defect propagation instability parameter, and generate coordinate positioning points for each process node in the two-dimensional coordinate system for risk self-sustaining capacity assessment.

[0152] The horizontal axis of the two-dimensional coordinate system for risk self-sufficiency assessment represents the activity parameter of quality state reversal, and the vertical axis represents the instability parameter of defect propagation. The two parameter values ​​of each process node are mapped to coordinate positioning points in the coordinate system.

[0153] Step S1030: Divide the region in the two-dimensional coordinate system for risk self-sufficiency assessment according to the distribution density of coordinate positioning points to obtain the high-stability zone, the fluctuating zone, and the vulnerable zone.

[0154] A clustering method based on kernel density estimation was used to divide the coordinate positioning points into three regions. The region near the origin with smaller values ​​on both the horizontal and vertical axes is the high-stability risk self-sufficiency region, the region far from the origin with larger values ​​on both the horizontal and vertical axes is the risk-vulnerable region, and the middle region is the risk-vulnerability fluctuation region.

[0155] Step S1040: Extract process nodes falling into the risk self-sufficiency vulnerable zone as self-sufficiency weak nodes, trace the self-sufficiency weak nodes back to the first risk source node on the risk inflow path of the quality risk diffusion path diagram to obtain the risk vulnerability transmission link; extract process nodes falling into the risk self-sufficiency fluctuation zone as self-sufficiency oscillation nodes, analyze the time change trend of the fluctuation direction conversion frequency of the self-sufficiency oscillation nodes, and identify the self-sufficiency decay direction of the self-sufficiency oscillation nodes; extract process nodes falling into the risk self-sufficiency high stability zone as self-sufficiency stable nodes, and mark the self-sufficiency stable nodes as quality risk blocking anchor points in the production process topology.

[0156] For nodes with weak self-sustaining capacity, trace back along the reverse risk diffusion direction of the quality risk diffusion path to the first risk source node to obtain the risk vulnerability transmission link. Calculate the slope of the time series linear fitting of the fluctuation direction switching frequency for nodes with oscillating self-sustaining capacity; if the slope is positive, the direction of self-sustaining capacity decay is the direction of deterioration. Mark nodes with stable self-sustaining capacity as quality risk blocking anchor points.

[0157] Step S1050: Using the quality risk blocking anchor point, the quality risk diffusion path map is divided into multiple risk isolation domains. Within each risk isolation domain, a risk isolation reinforcement strategy is generated based on the risk vulnerability transmission link of the weak self-sustaining node and the self-sustaining attenuation direction of the self-sustaining oscillation node. According to the risk isolation reinforcement strategy, a quality enhancement inspection checkpoint is set at the quality risk blocking anchor point, and the implementation time and inspection frequency of the quality enhancement inspection checkpoint are sent to the control terminal corresponding to the quality risk blocking anchor point.

[0158] The quality risk propagation path map is divided into multiple risk isolation domains using quality risk blocking anchor points as dividing boundaries. For each risk isolation domain, additional quality inspection frequencies are added to the risk vulnerability transmission links of nodes with weak self-sustainability within the domain, and quality parameter monitoring and early warning thresholds are set for nodes with oscillating self-sustainability within the domain in the direction of self-sustainability decay. Enhanced quality inspection checkpoints are set at the quality risk blocking anchor points, with the inspection frequency proportional to the average of the number of nodes with weak self-sustainability and the self-sustainability decay rate of nodes with oscillating self-sustainability within the isolation domain. The implementation period covers the production time window corresponding to the risk vulnerability transmission links of nodes with weak self-sustainability within the isolation domain.

[0159] Step S1110: Extract the risk state evolution sequence of each process node in the quality risk diffusion path diagram, and transform the risk state evolution sequence of each process node into a continuous change function of risk intensity over time.

[0160] For the risk state evolution sequence of each process node, cubic spline interpolation is used to interpolate the discrete-time step risk state values ​​into a continuous risk intensity variation function R(t) over time. Cubic spline interpolation constructs a cubic polynomial within each adjacent discrete-time step interval. The polynomial satisfies the conditions of equal function values ​​and continuous first derivative at the endpoints of the interval. The domain of R(t) is the interval from the risk activation time to the last time step of the risk state evolution sequence.

[0161] Step S1120: Solve the first derivative of the continuous change function of risk intensity over time for each process node to obtain the rate of change function of risk intensity, and solve the second derivative of the rate of change function of risk intensity to obtain the acceleration function of risk intensity change.

[0162] The rate of change function of risk intensity, V(t), is the first derivative of R(t), V(t) = dR(t) / dt. The acceleration function of risk intensity change, A(t), is the first derivative of V(t), which is also the second derivative of R(t), A(t) = dV(t) / dt = d²R(t) / dt². The numerical solution of the derivatives uses the finite difference method, calculating the difference ratio between adjacent sampling points at densely sampled points on the interpolation curve.

[0163] Step S1130: Identify the acceleration growth inflection point and deceleration growth inflection point in the risk growth process of each process node according to the risk intensity change rate function and the risk intensity change acceleration function. The acceleration growth inflection point is the moment when the risk intensity change acceleration changes from negative to positive, and the deceleration growth inflection point is the moment when the risk intensity change acceleration changes from positive to negative.

[0164] Traverse the time axis of A(t) and detect the point where A(t) changes sign. The moment when A(t) changes from a negative value to a positive value after crossing zero is the inflection point of accelerated growth, tacc. The moment when A(t) changes from a positive value to a negative value after crossing zero is the inflection point of decelerated growth, tdec. If there are multiple points of sign change, take the first inflection point of accelerated growth and the last inflection point of decelerated growth, respectively.

[0165] Step S1140: Extract the risk intensity value corresponding to the inflection point of accelerated growth for each process node as the risk warning trigger threshold, and extract the risk intensity value corresponding to the inflection point of decelerated growth for each process node as the risk intervention saturation threshold.

[0166] The risk warning trigger threshold Rwarn = R(tacc), which is the risk intensity function value at the inflection point of accelerated growth (tacc). The risk intervention saturation threshold Rsat = R(tdec), which is the risk intensity function value at the inflection point of decelerated growth (tdec).

[0167] Step S1150: Obtain the risk transmission chain composed of multiple consecutive process nodes in the quality risk diffusion path diagram. Arrange the risk warning trigger thresholds of each process node in the risk transmission chain according to the upstream and downstream order of the process to obtain a risk warning trigger threshold sequence. Compare the risk warning trigger thresholds of adjacent process nodes in the risk warning trigger threshold sequence and calculate the difference between the risk warning trigger thresholds of adjacent process nodes. If the risk warning trigger threshold of the subsequent process node is less than the risk warning trigger threshold of the preceding process node, then mark the subsequent process node as a risk warning forward sensitive node.

[0168] The risk transmission chain is a complete directed path from the starting process node to the ending process node in the quality risk diffusion path diagram. The Rwarn values ​​of each process node are arranged sequentially along the risk transmission chain according to the material flow direction, forming a risk warning trigger threshold sequence. For the k-th and (k+1)-th process nodes in the sequence, the difference ΔRwarn = Rwarn(k+1) - Rwarn(k) is calculated. If ΔRwarn < 0, meaning the risk warning trigger threshold of the subsequent process node is lower than that of the preceding process node, then the subsequent process node is more sensitive to risk, and is marked as a risk warning forward-moving sensitive node.

[0169] Step S1160: Arrange the risk intervention saturation thresholds of each process node in the risk transmission chain according to the upstream and downstream order of the process to obtain a risk intervention saturation threshold sequence. Compare the risk intervention saturation thresholds of adjacent process nodes in the risk intervention saturation threshold sequence and calculate the difference between the risk intervention saturation thresholds of adjacent process nodes. If the risk intervention saturation threshold of the subsequent process node is greater than the risk intervention saturation threshold of the preceding process node, then mark the subsequent process node as a risk intervention lag node.

[0170] The Rsat values ​​of each process node are arranged sequentially along the risk transmission chain in the direction of material flow, forming a risk intervention saturation threshold sequence. For the k-th and (k+1)-th process nodes in the sequence, the difference ΔRsat = Rsat(k+1) - Rsat(k) is calculated. If ΔRsat > 0, meaning the risk intervention saturation threshold of the subsequent process node is greater than that of the preceding process node, then the subsequent process node requires a higher risk intensity to reach intervention saturation, and this subsequent process node is marked as a risk intervention lag node.

[0171] Step S1170: Generate an advance warning time correction instruction for the risk warning forward sensitive node, generate an intervention time lag correction instruction for the risk intervention lagging node, and send the advance warning time correction instruction and the intervention time lag correction instruction to the corresponding control terminal respectively.

[0172] The warning time advance correction instruction includes the target process node identifier, the original warning trigger threshold, and the corrected warning trigger threshold. The corrected warning trigger threshold is the average of the risk warning trigger threshold of the preceding process node and the original risk warning trigger threshold of the sensitive node whose warning has been moved forward. The intervention time lag correction instruction includes the target process node identifier, the original intervention saturation threshold, and the corrected intervention saturation threshold. The corrected intervention saturation threshold is the average of the risk intervention saturation threshold of the preceding process node and the original risk intervention saturation threshold of the node whose intervention lags behind. Both types of correction instructions are sent to the corresponding control terminal according to the target process node identifier.

[0173] Step S1210: Obtain process monitoring video stream data of the batch of products to be controlled at each production stage, wherein the process monitoring video stream data includes a sequence of process operation video frames continuously collected during the production operation process.

[0174] The video monitoring data of the process is extracted from the video monitoring module of the enterprise information management system according to the process node number and time range. The process operation video frame sequence is a set of continuous image frames sorted by the acquisition timestamp. Each frame is a two-dimensional pixel matrix, and the matrix elements store the RGB three-channel color values.

[0175] Step S1220: Extract key operation action frames from the video frame sequence of the process operation, and perform operation posture skeleton point extraction processing on the key operation action frames to obtain the coordinate set of the operator's operation posture skeleton points.

[0176] The inter-frame difference method is used to detect frames in the video frame sequence of the operation process where the change in the operation action exceeds a preset threshold as key operation action frames. For each key operation action frame, a pre-trained human pose estimation model is invoked to extract skeleton points. This human pose estimation model adopts the Hourglass architecture based on a convolutional neural network. The input is the key operation action frame image, and the output is a set of joint coordinates of the operator, including the two-dimensional pixel coordinates of a preset number of skeleton points such as the shoulder joint, elbow joint, wrist joint, and finger joint.

[0177] Step S1230: Arrange the set of operation posture skeleton point coordinates of continuous key operation action frames corresponding to the same process node in chronological order to construct the operation posture skeleton point trajectory sequence. Perform action decomposition processing on the operation posture skeleton point trajectory sequence to divide the operation posture skeleton point trajectory sequence into multiple basic operation action segments. Each basic operation action segment corresponds to an indivisible operation action unit.

[0178] The set of skeleton point coordinates of consecutive key operation action frames within the same process node is arranged in ascending order by timestamp to form an operation posture skeleton point trajectory sequence. An action segmentation algorithm based on velocity zero-crossing points is used to calculate the motion velocity sequence of the skeleton points in the operation posture skeleton point trajectory sequence. The moment when the velocity in the velocity sequence drops below a preset velocity threshold is used as the action segmentation boundary. The skeleton point trajectory segment between two adjacent action segmentation boundaries constitutes a basic operation action segment.

[0179] Step S1240: Extract the displacement velocity sequence and joint angle change sequence of the operation posture skeleton points in each basic operation action segment, and construct the action normative feature vector of the basic operation action segment based on the displacement velocity sequence and the joint angle change sequence.

[0180] For each skeleton point in each basic operation motion segment, the displacement of that skeleton point between adjacent frames is calculated and divided by the inter-frame time interval to obtain the displacement velocity sequence. The change in the angle between the lines connecting adjacent skeleton points in the basic operation motion segment is calculated to obtain the joint angle change sequence. The mean and standard deviation of the displacement velocity sequence and the mean and standard deviation of the joint angle change sequence are concatenated into a feature vector, which is the motion normativity feature vector.

[0181] Step S1250: Compare the motion standard feature vectors of all basic operation motion segments within the same process node with the preset standard operation motion standard feature vector library to obtain the motion deviation score of each basic operation motion segment.

[0182] A pre-defined standard operation motion specification feature vector library stores the standard operation motion specification feature vectors corresponding to each standard operation motion unit at this process node. The Euclidean distance is calculated between the motion specification feature vector of each basic operation motion segment and the corresponding standard operation motion specification feature vector in the library. This Euclidean distance is the motion deviation score. A larger Euclidean distance indicates a more severe deviation from the standard specification.

[0183] Step S1260: Construct an operation standardization decay curve for the same process node based on the deviation scores of all basic operation action segments within the same process node. The operation standardization decay curve reflects the changing trend of the standardization of operation actions as production time progresses.

[0184] Using the time sequence number of the basic operation action segments as the horizontal axis and the action deviation score as the vertical axis, the action deviation scores of all basic operation action segments within the same process node are connected in chronological order to form an operation standardization decay curve. A linear fit is performed on the operation standardization decay curve, and the slope of the fitted line is extracted as the trend direction parameter, while the coefficient of determination of the fitted line is extracted as the trend change rate parameter.

[0185] Step S1270: Convert the trend direction parameter and trend change rate parameter of the operation standardization decay curve into the operation quality decay feature vector of the process node, and perform feature splicing and fusion of the operation quality decay feature vector with the risk intensity parameter of the corresponding process node in the quality risk diffusion path diagram to generate a multi-source risk fusion feature vector of the process node.

[0186] The operational quality degradation feature vector is a two-dimensional vector containing a trend direction parameter and a trend change rate parameter. This two-dimensional vector is concatenated along its feature dimension with the risk intensity parameter of the corresponding process node in the quality risk diffusion path diagram to generate a multi-source risk fusion feature vector. The dimension of the concatenated multi-source risk fusion feature vector is equal to the dimension of the operational quality degradation feature vector plus the dimension of the risk intensity parameter.

[0187] Step S1280: Replace the original risk intensity parameter of the corresponding process node in the quality risk diffusion path diagram with the multi-source risk fusion feature vector, update the quality risk diffusion path diagram and re-execute the risk transmission path deduction.

[0188] The multi-source risk fusion feature vector replaces the original risk intensity parameter as the new risk intensity representation for the corresponding process node, updating the risk state value of that process node in the quality risk diffusion path diagram. Based on the updated quality risk diffusion path diagram, steps S135 to S136 are re-executed for risk state iterative update and path diagram generation to obtain a new quality risk diffusion path diagram that integrates operational quality degradation information.

[0189] Step S1310: Obtain a set of appearance inspection images of sample products that have been sampled and inspected in the batch of products to be controlled. The set of appearance inspection images includes multi-view appearance images of the sample product surface taken from multiple angles. Perform defect area segmentation processing on each multi-view appearance image and extract the set of contour boundary pixels of the appearance defect area in the multi-view appearance image.

[0190] The set of appearance inspection images is extracted from the quality inspection module of the enterprise information management system. For each multi-view appearance image, a semantic segmentation model based on deep learning is used to segment defect regions. This semantic segmentation model adopts the U-Net architecture. The encoder extracts multi-scale image features through continuous convolution and pooling operations, and the decoder restores spatial resolution through upsampling and skip connections. The output is a binary segmentation mask of the same size as the input image. Pixels with a value of 1 in the mask represent appearance defect regions. The set of contour boundary pixels is extracted from the connected components with a value of 1 in the segmentation mask.

[0191] Step S1320: Calculate the geometric morphological features of the appearance defect region based on the set of outline boundary pixels. The geometric morphological features include the area-to-perimeter ratio parameter of the defect region and the fractal dimension parameter of the defect region outline.

[0192] The area of ​​the defect region is equal to the total number of pixels in the connected region enclosed by the set of pixels representing the contour boundaries. The perimeter of the defect region is equal to the sum of the distances between adjacent boundary pixels in the set of pixels representing the contour boundaries. The area-to-perimeter ratio is the area divided by the square of the perimeter. The fractal dimension parameter of the defect region contour is calculated using the box-counting dimension method. The contour image is divided into grids of different scales, and the logarithmic ratio of the number of grids required to cover the contour as a function of the grid scale is calculated; this ratio is the fractal dimension parameter.

[0193] Step S1330: Perform three-dimensional spatial mapping reconstruction processing on the geometric morphological features of the same defect area in different multi-view appearance images to generate a three-dimensional morphological reconstruction model of the defect area. Extract the depth direction extension parameters and the structural layering features inside the defect area from the three-dimensional morphological reconstruction model. Determine the defect penetration level of the defect area based on the depth direction extension parameters and the structural layering features.

[0194] The geometric features of the same defect region from at least two different viewpoints are input into a stereo vision-based 3D reconstruction algorithm, which outputs a 3D morphological reconstruction model of the defect region. The depth of the defect extending from the surface to the interior is measured along the normal direction of the defect region from the 3D morphological reconstruction model and used as a depth extension parameter. Within the extension depth range, the layering variation of material density within the defect region is analyzed, and the number of density abrupt interfaces is used as a structural layering feature. The defect penetration level is determined jointly by the depth extension parameter and the structural layering feature; the larger both are, the higher the penetration level.

[0195] Step S1340: Obtain the production process batch information corresponding to the sample product, and trace the production process node sequence that the sample product has gone through based on the production process batch information.

[0196] The production traceability module of the enterprise information management system retrieves the production process sequence of a sample product based on its batch number. The production process sequence lists all process node identifiers and processing time ranges for each process node, arranged chronologically by production time, from raw materials to finished product.

[0197] Step S1350: Associate the defect penetration level of the sample product with each process node in the production process node sequence that the sample product has passed through, and determine the contribution probability weight of each process node to the final product appearance defect.

[0198] For each process node in the production process sequence, a contribution probability weight is determined based on the correlation between the processing time range of that process node and the defect generation mechanism. The contribution probability weight is proportional to the influence coefficient of that process node on the product surface processing, which is obtained from the process characteristic configuration table in the enterprise information management system. The defect penetration level is then allocated to each process node according to the contribution probability weight.

[0199] Step S1360: Introduce the contribution probability weight into the risk inflow path weight calculation of the corresponding process node in the quality risk diffusion path diagram, adjust the risk diffusion probability of each process node transmitting risk to its downstream process node, and re-execute the risk state evolution calculation of the quality risk diffusion path diagram using the adjusted risk diffusion probability to update the risk state evolution sequence of each process node in the quality risk diffusion path diagram.

[0200] The adjusted risk inflow path weights are obtained by multiplying the contribution probability weights of each process node by the original risk inflow path weights. Based on these adjusted risk inflow path weights, the risk diffusion probability of each process node transmitting risk to its downstream process nodes is recalculated. The risk state iterative update process of step S135 is then re-executed using the adjusted risk diffusion probability to generate an updated risk state evolution sequence.

[0201] Step S1370: Based on the updated risk state evolution sequence, redetermine the risk retention time of the risk transmission node, and use the redetermined risk retention time as the basis for generating targeted quality intervention instructions.

[0202] Based on the updated risk state evolution sequence, step S143 is re-executed to extract the risk activation time and risk saturation time of the risk transmission nodes, and the risk retention time is recalculated. The newly determined risk retention time replaces the original risk retention time and is used as the input parameter for generating targeted quality intervention instructions in steps S152 to S155.

[0203] Figure 2 This is a visual representation of the production process risk diffusion situation diagram generated in steps S130 to S140 of this embodiment of the invention. It corresponds to the quality risk diffusion path diagram generated in step S130 and the risk transmission node information determined in step S140. The statistics card at the top of the interface displays "8 active process nodes", "3 risk transmission nodes", "longest risk retention time 38 minutes", and "2 sets of same-layer coupled risk pairs". The topology diagram shows the material flow direction between process nodes in the order of "incoming material inspection" → "raw material pretreatment" → "injection molding #3" → "assembly workshop" → "finished product inspection". The "injection molding #3" node is marked as a risk transmission node, showing a risk retention time of 38 minutes and a risk amplification indicator; the "finished product inspection" node is marked as a risk saturation state. This diagram intuitively reflects the analysis results of the risk transmission nodes and their risk retention time determined based on the quality risk diffusion path diagram in step S140.

[0204] Figure 3This is the configuration interface for the targeted quality intervention instruction generated in step S150 of this embodiment of the invention. It corresponds to the process in steps S152 to S155 of generating targeted quality intervention instructions containing the timing of control intervention based on the risk retention time and risk amplification index of the risk transmission node. The figure shows the control intervention timing configuration options for the risk transmission node "Injection Molding #3", including the control start time (2026-06-10 14:30:00) and the control duration (45 minutes). It also shows the control measure intensity level parameter (Level 3 Enhancement), corresponding to specific control measures such as a 3-fold increase in sampling frequency, a 50% increase in sample size, and automatic parameter retesting. The bottom shows the target control terminals, including "Injection Molding Workshop - No. 3 Machine Terminal (MQTT)" and "Quality Inspection Group - Mobile Terminal (WebSocket)," corresponding to the process in step S156 of encapsulating the protocol according to the communication protocol type of the control terminal and sending the instruction.

[0205] Figure 4 This is the traceability interception instruction interface generated in step S660 of this embodiment of the invention. It corresponds to the process in steps S640 to S660 where the quantity of materials already transferred is calculated based on the risk retention time and material flow speed of the risk transmission node, and traceability interception instructions are generated according to the material flow list affected by quality risks. The figure shows the traceability interception list of affected materials in tabular form, including material batch ID (BATCH-A12, BATCH-A15, BATCH-A23), flow destination (East China finished product temporary storage warehouse, assembly workshop buffer area, South China distribution center), current location (in transit, production line temporary storage, already out of the warehouse), priority (urgent, high), and status (instruction issued, pending confirmation, logistics notified). This figure intuitively reflects the batch identifier of the materials to be intercepted, the interception priority mark, and the list of interception destinations contained in the traceability interception instruction.

[0206] Figure 5This refers to the multi-source risk fusion configuration interface in steps S1010 to S1050 of this embodiment of the invention. It corresponds to step S1010, which extracts the quality state reversal activity parameter and defect propagation instability parameter to construct a two-dimensional coordinate system for risk self-sufficiency assessment, and steps S1020 to S1050, which identify weak nodes in self-sufficiency based on the risk self-sufficiency vulnerability zone, mark quality risk blocking anchor points, and generate risk isolation and reinforcement strategies. The figure shows three types of multi-source risk fusion modules: operational standardization analysis (current score 92.5, action deviation meets standard), equipment anomaly overlap (overlap 87%, risk diffusion probability automatically increased by 1.5 times), and production environment disturbance (recent impact window 13:45~14:15, temperature transient +3.2°C, 2 nodes already associated). Each module is equipped with switches such as "Merge to Risk Intensity", "Automatic Linkage Adjustment", and "Environmental Parameter Linkage" to control whether multi-source data participates in risk intensity calculation. This corresponds to the process in step S1270 where the operation quality degradation feature vector and the risk intensity parameter are spliced ​​and fused to generate the process node multi-source risk fusion feature vector.

[0207] In an exemplary embodiment, a product quality intelligent control system based on an information management system is provided. This system can be a terminal, server, etc., and its internal structure includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, near-field communication, or other technologies. When the computer program is executed by the processor, it implements a product quality intelligent control method based on an information management system. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device can be a touch layer covering the display screen, or a button, trackball, or touchpad set on the shell of the product quality intelligent control system based on the information management system, or an external keyboard, touchpad, or mouse, etc.

[0208] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A product quality intelligent control method based on an information management system, characterized in that, The method includes: Obtain the set of historical quality sampling inspection records corresponding to the batch of products to be controlled, wherein the set of historical quality sampling inspection records contains multiple quality indicator detection data units collected at different sampling time nodes; The historical quality sampling record set is subjected to quality fluctuation pattern mining processing to obtain the time-series fluctuation characteristics of the quality indicators of the batch of products to be controlled and the correlation transmission characteristics of the non-conforming items in the sampling inspection. Based on the time-series fluctuation characteristics of the quality indicators and the correlation transmission characteristics of the non-conforming items in the sampling inspection, a preset quality risk diffusion deduction logic is invoked to generate a quality risk diffusion path diagram for the batch of products to be controlled. Based on the quality risk diffusion path diagram, determine the risk transmission nodes and the risk retention time of each production link in the batch of products to be controlled; Based on the risk transmission node and the risk retention time of the risk transmission node, a targeted quality intervention instruction containing the timing of control intervention is generated, and the targeted quality intervention instruction is distributed to the control terminal of the corresponding production link.

2. The product quality intelligent control method based on an information management system according to claim 1, characterized in that, The historical quality inspection record set is subjected to quality fluctuation pattern mining processing to obtain the time-series fluctuation characteristics of the quality indicators of the batch of products to be controlled and the correlation transmission characteristics of the non-conforming items in the inspection, including: Extract the timestamp of each quality indicator detection data unit in the historical quality sampling record set, construct the time series arrangement structure of the multiple quality indicator detection data units based on the timestamp, identify the numerical fluctuation trajectory of each quality indicator in the time series arrangement structure, and separate the periodically repeating quality indicator fluctuation segments from the numerical fluctuation trajectory. The fluctuation range and direction conversion frequency of the quality index fluctuation segments are analyzed to obtain the fluctuation range and direction conversion frequency of each quality index between adjacent sampling time nodes. The time-series fluctuation characteristics of the quality index of the batch of products to be controlled are constructed based on the fluctuation range and the fluctuation direction conversion frequency. Extract the non-conforming item markers from each sampling record in the historical quality sampling record set, count the co-occurrence frequency of different non-conforming item markers in the same sampling record, construct a non-conforming item co-occurrence frequency distribution table based on the co-occurrence frequency, and determine non-conforming item groups with stable co-occurrence relationships in the non-conforming item co-occurrence frequency distribution table; The first occurrence time of each non-conforming item in the non-conforming item group is arranged in chronological order to obtain the triggering time chain of the non-conforming item group. The inter-item propagation delay duration is calculated based on the time interval between the occurrence times of adjacent non-conforming items in the triggering time chain, and the fluctuation range of the inter-item propagation delay duration is extracted. The correlation transmission characteristics of the non-compliant items in the random inspection are constructed based on the fluctuation range of the triggering timing chain and the transmission delay duration between the items.

3. The product quality intelligent control method based on an information management system according to claim 1, characterized in that, Based on the time-series fluctuation characteristics of the quality indicators and the correlation transmission characteristics of the non-conforming items in the sampling inspection, a preset quality risk diffusion deduction logic is invoked to generate a quality risk diffusion path diagram for the batch of products to be controlled, including: Obtain the production process topology structure corresponding to the batch of products to be controlled, wherein the production process topology structure includes multiple process nodes and the material flow direction relationship between the multiple process nodes; Map and associate each quality indicator in the time-series fluctuation characteristics of the quality indicators with the multiple process nodes to determine the quality monitoring process node corresponding to each quality indicator; The triggering sequence chain in the associated transmission characteristics of the non-conforming items in the sampling inspection is expanded into a project transmission network containing an initial triggering project node and subsequent transmission project nodes. Based on the transmission direction relationship between each non-conforming item in the project transmission network, the diffusion flow direction of the non-conforming items is marked in the production process topology. The fluctuation amplitude span and fluctuation direction conversion frequency in the time-series fluctuation characteristics of the quality indicators are respectively converted into the risk intensity parameter and risk change trend parameter of the quality monitoring process node. The risk diffusion probability of the risk spreading from the quality monitoring process node to its adjacent process nodes is calculated based on the risk intensity parameter and the risk change trend parameter. The risk diffusion probability and the diffusion direction of the non-conforming items are input into a preset quality risk diffusion deduction logic, and risk state iterative update processing is performed on the production process topology to generate a risk state evolution sequence for each process node in the production process topology. Based on the risk state evolution sequence, the risk activation time and risk saturation time of each process node are extracted. Based on the material flow direction relationship of the production process topology, the risk activation time and risk saturation time of each process node are spliced ​​together in time and space to generate the quality risk diffusion path diagram.

4. The product quality intelligent control method based on an information management system according to claim 1, characterized in that, Based on the quality risk diffusion path diagram, the risk transmission nodes in each production stage of the batch of products to be controlled and the risk retention time of the risk transmission nodes are determined, including: The quality risk diffusion path diagram is processed to identify path branch points. Process nodes where the risk flow in the quality risk diffusion path diagram branches are extracted as candidate risk transmission nodes. The number of upstream risk inflow paths and the number of downstream risk outflow paths of the candidate risk transmission nodes in the quality risk diffusion path diagram are obtained. When the number of downstream risk outflow paths of the candidate risk transmission node is greater than the number of upstream risk inflow paths of the candidate risk transmission node, the candidate risk transmission node is determined as a risk transmission node. Extract the risk activation time and risk saturation time corresponding to the risk transmission node in the quality risk diffusion path diagram, calculate the time difference between the risk saturation time and the risk activation time, use the time difference as the risk retention time of the risk transmission node, and associate and store the process identifier of the risk transmission node in the production process topology with the risk retention time. Based on the path connection relationship between each risk transmission node in the quality risk diffusion path diagram, the risk transmission level of the risk transmission node is determined, the risk retention time of other risk transmission nodes at the same risk transmission level as the risk transmission node is extracted, and the distribution central tendency parameter of the risk retention time within the same risk transmission level is calculated. The degree of abnormality in the retention of the risk at the risk transmission node is determined based on the degree of deviation between the risk retention duration at the risk transmission node and the distribution central tendency parameter.

5. The intelligent product quality control method based on an information management system according to claim 1, characterized in that, Based on the risk transmission nodes and the risk retention duration at the risk transmission nodes, a targeted quality intervention instruction containing the timing of control intervention is generated, and the targeted quality intervention instruction is distributed to the control terminals of the corresponding production links, including: Obtain the process identifier of the risk transmission node in the production process topology, and query the control terminal address and control terminal communication protocol type corresponding to the risk transmission node based on the process identifier; The control intervention sequence of the risk transmission node is determined based on the risk retention time of the risk transmission node, and the control intervention sequence includes the control initiation time and the control duration. Obtain the number of risk inflow paths and the number of risk outflow paths corresponding to the risk transmission node, and determine the risk amplification index of the risk transmission node based on the number of risk inflow paths and the number of risk outflow paths. The control measure intensity level parameter of the risk transmission node is generated based on the risk retention time and the risk amplification index; Call the preset control instruction template library, match the corresponding control instruction template according to the control measure intensity level parameter, fill the control intervention sequence and the process identifier of the risk transmission node into the control instruction template, and generate a targeted quality intervention instruction; The targeted quality intervention instruction is encapsulated according to the communication protocol type of the control terminal to generate an encapsulated instruction data packet that matches the communication protocol type of the control terminal. The encapsulated instruction data packet is sent to the control terminal corresponding to the control terminal address through a message distribution channel. The encapsulated instruction data packet is used to trigger the control terminal to perform a quality intervention operation corresponding to the intensity level parameter of the control measure. The system receives control execution feedback information returned by the control terminal and updates the risk status marker of the risk transmission node in the quality risk diffusion path diagram based on the control execution feedback information.

6. The intelligent product quality control method based on an information management system according to claim 1, characterized in that, The method further includes: Obtain the historical supply quality data set of the raw material suppliers corresponding to the batch of products to be controlled. The historical supply quality data set includes the supply batch records of each raw material supplier and the quality inspection pass rate data of each batch of supply. Based on the quality inspection pass rate data of each batch of supply in the historical supply quality data set, construct the time series change curve of the quality pass rate of each raw material supplier, perform trend inflection point detection processing on the time series change curve of the quality pass rate, and identify the trend inflection point moment when the quality pass rate changes from a stable state to a declining state in the time series change curve of the quality pass rate. Extract the quality inspection pass rate data of multiple consecutive batches of raw material suppliers after the trend inflection point, and calculate the quality degradation rate parameter of the raw material supplier based on the quality inspection pass rate data of multiple consecutive batches of raw material suppliers. Obtain the raw material supplier identifier corresponding to the raw material batch used in the batch of products to be controlled; match the corresponding quality degradation rate parameter according to the raw material supplier identifier; and deduce the contribution of the raw material quality in the batch of products to be controlled to the diffusion of finished product quality risk based on the quality degradation rate parameter and the production time node of the batch of products to be controlled. The contribution is superimposed on the initial risk node of the quality risk diffusion path diagram, and the risk intensity parameter of each process node in the quality risk diffusion path diagram is updated.

7. The intelligent product quality control method based on an information management system according to claim 1, characterized in that, The method further includes: Obtain the set of production equipment operation status records corresponding to the batch of products to be controlled. The set of production equipment operation status records contains time series data of operation parameters of each production equipment during the production period of the batch of products to be controlled. Perform abnormal pattern segment extraction processing on the time series data of operation parameters to obtain the set of abnormal segment data of operation parameters of the production equipment during the production period of the batch of products to be controlled. Based on the duration and frequency of occurrence of abnormal segments in the set of abnormal operating parameters, the operational stability deviation feature of the production equipment is constructed. The operational stability deviation feature of the production equipment is then associated with the process nodes in the production process topology to determine the affected process nodes that are affected by the operating status of the production equipment. Obtain the risk activation time corresponding to the affected process node in the quality risk diffusion path diagram, compare the overlap between the risk activation time of the affected process node and the time range of the abnormal operating parameters of the production equipment, adjust the risk diffusion probability of the affected process node according to the overlap comparison result, and increase the risk diffusion probability of the affected process node when the overlap exceeds a preset overlap threshold.

8. The intelligent product quality control method based on an information management system according to claim 1, characterized in that, The method further includes: Obtain the production environment monitoring data set corresponding to the batch of products to be controlled. The production environment monitoring data set includes the temperature and humidity monitoring value sequence and the air cleanliness monitoring value sequence of the production environment during the production period of the batch of products to be controlled. Perform environmental parameter fluctuation pattern analysis on the temperature and humidity monitoring value sequence and the air cleanliness monitoring value sequence to extract the transient events and gradual drift trends of environmental parameters in the production environment during the production period of the batch of products to be controlled. The environmental disturbance impact window of the production environment is determined based on the occurrence time and duration of the transient event of the environmental parameters. The cumulative direction of environmental parameter offset in the production environment is determined based on the direction and rate of the gradual drift trend of the environmental parameters. Obtain a list of process nodes in production state within the environmental disturbance impact window, and align the risk status of each process node in the process node list with the environmental disturbance impact window in the time dimension of the quality risk diffusion path diagram. Obtain the set of process nodes corresponding to the sensitive quality indicators in the cumulative direction of the environmental parameter offset, and scale up the risk diffusion probability of each process node in the set of process nodes proportionally according to the rate of the gradual drift trend of the environmental parameter.

9. The intelligent product quality control method based on an information management system according to claim 1, characterized in that, The method further includes: Obtain a set of quality complaint records for historical similar product batches corresponding to the batch of products to be controlled. The set of quality complaint records for historical similar product batches includes the batch number of the complained product, the type of quality problem complained about, and the time when the complaint occurred. Based on the types of quality complaints, the historical quality complaint records of similar product batches are classified and aggregated to obtain complaint event sequences corresponding to different types of quality complaints. The time interval distribution of each type of quality complaint event sequence is analyzed to obtain the central tendency parameter and dispersion parameter of the time interval. Based on the central tendency parameter and dispersion parameter of the complaint time interval, a distribution model of the latent period of quality hazards corresponding to each type of complaint quality problem is constructed. Match the non-conforming item types corresponding to each risk transmission node in the quality risk diffusion path diagram of the batch of products to be controlled with the complaint quality problem types, obtain the risk activation time of the successfully matched risk transmission node, and deduce the complaint outbreak prediction time window of the quality hidden danger corresponding to the risk transmission node after the batch of products to be controlled is put into the market based on the quality hidden danger incubation period distribution model. Based on the predicted complaint outbreak time window, an after-sales quality proactive intervention instruction is generated for the batch of products to be controlled. The after-sales quality proactive intervention instruction includes the proactive recall trigger time and the proactive recall coverage.

10. A product quality intelligent control system based on an information management system, characterized in that, include: processor; A machine-readable storage medium for storing machine-executable instructions of the processor; The processor is configured to execute the product quality intelligent control method based on an information management system as described in any one of claims 1 to 9 by executing the machine-executable instructions.

Images