Method and system for process parameter multi-process correlation modeling and adaptive control

Abstract

The application discloses a method and system for process parameter multi-process correlation modeling and adaptive control, and relates to the technical field of production process control. The method comprises the following steps: collecting process parameter time series data, equipment operation state time series data and quality detection result data associated with production batch identification, writing a timestamp and aligning the association, generating an associated data set containing a three-dimensional association identifier and a hash check code; extracting energy consumption stability characteristic quantities and fault concealment risk characteristic quantities within the evaluation period, and calculating the equipment health score; inputting the process time series characteristics and the equipment health score into the process optimization model, and outputting the target process parameter combination under the process constraint; before issuing, evaluating the risk level according to the adjustment range and the equipment health score and controlling the issuing mode; after taking effect, collecting the quality detection result data and the energy consumption index to update the process optimization model to form a closed loop. The application realizes the safe and adaptive optimization of process parameters under the perception of the equipment health state.

Description

Technical Field

[0001] This invention relates to the field of production process control technology, and in particular to a method and system for multi-process correlation modeling and adaptive control of process parameters. Background Technology

[0002] In industrial production, information such as process parameters, equipment operating status, and quality inspection results are typically generated in a time-series manner and stored in different production and testing systems. While process parameter optimization and process control based on this data are common, the following shortcomings still exist in practical applications.

[0003] On the one hand, the complex sources and numerous stages of multi-source data can easily lead to inconsistencies in timing, unclear relationships, and incomplete records, making process traceability and reproduction for production batches difficult and affecting the credibility and verifiability of optimization decisions. On the other hand, existing process optimizations often focus on single objectives such as quality or energy consumption, or fail to adequately consider equipment status, making it difficult to simultaneously reflect equipment health and potential risks during the optimization process. In cases of equipment performance fluctuations or risk accumulation, parameter adjustments without constraints on equipment status can easily trigger process instability, quality fluctuations, or abnormal energy consumption, affecting stable production operations. Process parameter adjustments typically need to simultaneously meet constraints such as safety, equipment capacity, and quality stability. Existing technologies often lack consistent and controllable governance mechanisms for risk assessment and execution control of parameter adjustments, making it difficult to adopt differentiated execution methods based on risk levels, resulting in biased adjustment strategies, erroneous issuance, or delayed responses. Furthermore, long-term operation of the production process involves equipment aging, material differences, and changes in operating conditions, potentially reducing the adaptability of optimization strategies over time. Without effective feedback and continuous updating mechanisms, optimization effects are prone to drift and are difficult to maintain long-term stability. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for multi-process correlation modeling and adaptive control of process parameters, enabling reliable correlation and traceability of multi-source data in industrial production, achieving optimized output and controllable execution of process parameters that take into account equipment status, and forming a continuous improvement mechanism based on production results, thereby improving the reliability of process optimization and the stability of the production process.

[0005] To achieve the above objectives, the present invention employs the following technical solution:

[0006] On the one hand, this invention provides a method for multi-stage correlation modeling and adaptive control of process parameters, including the following steps:

[0007] S1. Collect multi-source time-series data of the target production equipment during the production process. The multi-source time-series data includes process parameter time-series data, equipment operating status time-series data, and quality inspection result data associated with the production batch identifier. Perform alignment and association processing on the multi-source time-series data to generate an associated dataset. The associated dataset includes a three-dimensional association identifier and a hash checksum corresponding to the three-dimensional association identifier.

[0008] S2. Within a preset evaluation period, extract equipment health features based on the time-series data of the equipment's operating status, and calculate the equipment health score based on the equipment health features. The equipment health features include energy consumption stability features and fault concealment risk features.

[0009] S3. Construct the process parameter time series data into multi-process time series features according to process identifiers, and generate multi-process associated input features based on the parameter transfer relationship between different processes. Use the multi-process time series features, the multi-process associated input features and the equipment health score together as input to the process optimization model, and output the target process parameter combination under the preset process constraints.

[0010] S4. Before the target process parameter combination is issued, the risk level of the process parameter adjustment is assessed based on the adjustment range of the target process parameter combination relative to the currently effective process parameters and the equipment health score associated with the production batch identifier, and the issuance method of the target process parameter combination is controlled according to the risk level.

[0011] S5. After the target process parameter combination takes effect, collect the corresponding quality inspection result data and energy consumption index data, and update the process optimization model based on the quality inspection result data and energy consumption index data to form a closed loop of adaptive optimization of process parameters.

[0012] As a preferred embodiment of the present invention, the step of collecting multi-source time-series data of the target production equipment during the production process includes:

[0013] The same production batch cycle or continuous production cycle is used as the data collection cycle to collect time-series data of process parameters, time-series data of equipment operating status, and quality inspection results data associated with production batch identifiers.

[0014] Write a unified timestamp to the process parameter timing data, equipment operating status timing data, and quality inspection result data, and perform alignment processing on the multi-source timing data based on the unified timestamp;

[0015] A three-dimensional association identifier is generated based on the unified timestamp, material identifier, and equipment number, and the three-dimensional association identifier is bound to the aligned multi-source time-series data to form an association dataset;

[0016] A hash check code is calculated for the key binding fields in the associated dataset, including at least the three-dimensional association identifier, unified timestamp, device number, and production batch identifier. The hash check code is then associated with and stored with the corresponding associated data record to achieve data traceability and tamper-proof verification.

[0017] As a preferred embodiment of the present invention, the quality inspection result data is generated by a test management module, which includes centralized test processing and time-controlled test.

[0018] The centralized processing test includes: performing pre-parsing processing on test files associated with production batch identifiers, and generating quality indicator entries based on the pre-parsing results, wherein the quality indicator entries constitute at least a part of the quality inspection result data;

[0019] The time control test includes: triggering the start and end of the test according to preset time control conditions, and recording breakpoint information during the test; after the test is interrupted, continuing to execute the corresponding test process based on the breakpoint information to generate quality indicator items.

[0020] The steps for calculating the equipment health score based on the aforementioned equipment health characteristics include:

[0021] Based on the time-series data of equipment operation status, equipment energy consumption stability characteristics and equipment fault hidden risk characteristics are constructed respectively.

[0022] The energy consumption stability characteristics and hidden risk characteristics of the equipment are subjected to scale unification processing to obtain comparable health evaluation characteristics. The scale unification processing includes normalization processing or standardization processing.

[0023] The health evaluation features are comprehensively evaluated based on preset evaluation fusion rules to generate a device health score.

[0024] The evaluation fusion rules include: normalizing the energy consumption change coefficient, risk concealment coefficient, and temperature-related and vibration-related indicators extracted from the time-series data of the equipment's operating status, then inputting them into a fuzzy rule base, and performing defuzzification on the output of the fuzzy rule base to obtain an equipment health score that characterizes the health status of the equipment.

[0025] As a preferred embodiment of the present invention, the energy consumption stability characteristic includes the energy consumption variation coefficient. The energy consumption variation coefficient Calculate using the following formula:

[0026] ;

[0027] In the formula: It is the ratio of the standard deviation to the mean, calculated based on the energy consumption sampling sequence of the target production equipment within a preset evaluation period; It is the natural logarithm function;

[0028] The fault concealment risk characteristic includes the risk concealment coefficient. The risk concealment coefficient Calculate using the following formula:

[0029] ;

[0030] In the formula: The time difference between the current time and the center time of the high-incidence period of equipment failure is the time difference between the current time and the center time of the high-incidence period of equipment failure, which is obtained by DBSCAN density clustering based on the equipment failure occurrence time series. This is to preset the number of similar failures within a historical time window; It is a natural constant and The unit is hours, and 24 represents the 24-hour time base.

[0031] As a preferred embodiment of the present invention, the step of outputting the target process parameter combination includes:

[0032] Based on the process parameter time series data, the process parameter sampling sequence of each process is segmented according to the process identifier, and multi-process time series features reflecting the change trend, fluctuation amplitude and change rate of process parameters are extracted according to the preset time window.

[0033] The time-series features and the equipment health score are combined to construct joint features, forming a joint input feature that characterizes the current operating status of the equipment and its degree of process adaptability.

[0034] The joint input features are input into the process optimization model, and the target process parameter combination is output under the joint constraints of process parameter value range constraints, equipment operating capacity constraints, quality stability constraints and equipment coordination constraints.

[0035] The equipment coordination constraints are determined through the following steps: a scheduling fitness function is constructed based on equipment health score, equipment replacement time, equipment coordination penalty coefficient, and qualified deviation penalty coefficient; a genetic algorithm is used to search for the optimal production path with the goal of maximizing the scheduling fitness function; and the equipment coordination constraints are determined based on the optimal production path.

[0036] The scheduling fitness function satisfies the following formula:

[0037] ;

[0038] In the formula: This represents the scheduling fitness function value; This refers to the number of stages involved in the costume change process. For the first Equipment replacement time for each replacement stage; This is a very small amount of normal quantity; This is a comprehensive item encompassing collaboration and quality. This refers to the equipment coordination penalty coefficient. This is the penalty coefficient for acceptable deviation; and The preset weighting coefficients are used, and they satisfy the following conditions: .

[0039] As a preferred embodiment of the present invention, the method of assessing the risk level of process parameter adjustment and controlling the issuance of the target process parameter combination based on the risk level includes:

[0040] The rule engine executes the distribution control logic, and determines the corresponding distribution method based on the adjustment range of the target process parameter combination relative to the currently effective process parameter. The adjustment range is the absolute value or relative change ratio of the change of each process parameter in the target process parameter combination relative to the currently effective process parameter.

[0041] When the adjustment range does not exceed the first threshold, automatic issuance is performed; when the adjustment range is greater than the first threshold but not greater than the second threshold, issuance is performed after manual confirmation; when the adjustment range is greater than the second threshold, issuance is performed after approval.

[0042] When the equipment health score is lower than the preset health threshold or the risk concealment coefficient is higher than the preset risk threshold, the upper limit of at least one process parameter in the target process parameter combination will be lowered according to a preset ratio.

[0043] When there is an unclosed warning event status, the issuance method will be controlled to be issued after manual confirmation;

[0044] The early warning event processing includes: constructing a three-dimensional alarm identifier based on the device number, alarm type, and parameter name, and performing alarm deduplication within a preset sliding window; performing gradient escalation of alarm level based on the alarm unprocessed duration and a preset escalation threshold; and merging low-level alarms corresponding to the same device number into a summary alarm based on a preset merging time window. The alarm rules support online updates and version rollback.

[0045] As a preferred embodiment of the present invention, the process of updating the process optimization model includes:

[0046] Based on the production batch identifier, the quality inspection result data, energy consumption index data and joint input features are bound together to generate training samples for model training. The joint input features are feature vectors jointly constructed by process parameter time series features and equipment health scores.

[0047] When the number of new training samples reaches a preset threshold or the preset model update time period is reached, incremental training of the process optimization model is performed and the model parameters are updated.

[0048] Generate corresponding version identifiers for the model versions before and after the model update, and associate and store the version identifiers with the corresponding production batch identifiers.

[0049] As a preferred embodiment of the present invention, the method further includes a quality closed-loop control step and a packaging release verification step.

[0050] The quality closed-loop control steps include: locking the circulation permission of detected defective products, generating and issuing corresponding repair tasks; performing retesting after repair is completed, and unlocking the circulation permission when the retest is passed;

[0051] The packaging release verification step includes: verifying the completion status of mandatory tests and the status of defect handling of the product to be packaged during the product packaging process based on the production batch identifier and its associated quality inspection result data;

[0052] During the product packaging process, verify the consistency between the product model identification, production batch identification, and outer box identification;

[0053] When the number of boxes reaches a preset value, an outer box traceability code associated with the set of product serial numbers inside the box is generated. The outer box traceability code is an coded identifier generated based on the set of product serial numbers, and the outer box traceability code is written to the outer box label.

[0054] On the other hand, the present invention also provides a system for multi-stage correlation modeling and adaptive control of process parameters, applied to the method for multi-stage correlation modeling and adaptive control of process parameters as described above, including:

[0055] The multi-source data acquisition module is used to collect multi-source time-series data of the target production equipment during the production process. The multi-source time-series data includes process parameter time-series data, equipment operating status time-series data, and quality inspection result data associated with production batch identifiers.

[0056] The alignment and anti-tampering module is used to perform alignment and association processing on the multi-source time series data to generate an associated dataset containing a three-dimensional association identifier and a corresponding hash check code.

[0057] The equipment health scoring module is used to extract equipment health feature quantities based on equipment operating status time-series data within a preset evaluation period, and to calculate the equipment health score based on the equipment health feature quantities.

[0058] The process parameter optimization module is used to construct multi-process time sequence features from process parameter time sequence data according to process identifiers, and generate multi-process associated input features based on the parameter transfer relationship between different processes. The multi-process time sequence features, the multi-process associated input features and equipment health score are used as inputs to the process optimization model, and the target process parameter combination is output under the condition of satisfying preset process constraints.

[0059] The risk control module is used to assess the risk level of process parameter adjustment based on the adjustment range of the target process parameter combination relative to the currently effective process parameters and the corresponding equipment health score before the target process parameter combination is issued, and to control the issuance method of the target process parameter combination according to the risk level.

[0060] The model update module is used to collect the corresponding quality inspection results data and energy consumption index data after the target process parameter combination takes effect, and update the process optimization model based on the data.

[0061] The beneficial effects of this invention are as follows: This invention collects time-series data of process parameters, equipment operating status, and quality inspection results associated with production batch identifiers during the production process. It then performs alignment and association processing on the multi-source time-series data to form an associated dataset containing three-dimensional associated identifiers and their corresponding hash check codes. This makes the correspondence between multi-source data clearer and provides a foundation for consistency verification and process traceability of production data, thereby improving the traceability and reliability of production process data. Within a preset evaluation period, it extracts energy consumption stability features and fault concealment risk features based on equipment operating status time-series data and calculates equipment health scores. This allows process parameter optimization to comprehensively reflect equipment operating status and potential risk levels while utilizing the time-series features of process parameters, improving the robustness and executability of optimization decisions. After outputting the target process parameter combination, this invention assesses the adjustment risk level and controls the distribution method before distribution, combining the adjustment range of the target process parameter combination relative to the currently effective process parameters and the equipment health score associated with the production batch identifier. This provides hierarchical control capabilities for the process parameter adjustment process, improving the controllability and consistency of the parameter adjustment execution process. After the target process parameter combination takes effect, the present invention collects the corresponding quality inspection result data and energy consumption index data and uses them to update the process optimization model, forming an adaptive optimization closed loop for process parameters, which is beneficial to maintain the adaptability of the optimization strategy and the stability of the optimization effect. Attached Figure Description

[0062] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0063] Figure 1 This is a flowchart of a method according to an embodiment of the present invention;

[0064] Figure 2 This is a schematic diagram of the modular structure of the system according to an embodiment of the present invention. Detailed Implementation

[0065] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention are within the scope of protection of the present invention.

[0066] like Figure 1 As shown, this is an embodiment of the present invention. This embodiment provides a method for multi-process correlation modeling and adaptive control of process parameters. In order to ensure that the data required for subsequent health assessment and process optimization have a basis for correlation, traceability and consistency verification, this embodiment first performs data acquisition and alignment correlation steps.

[0067] S1. Collect multi-source time-series data of the target production equipment during the production process. The multi-source time-series data includes process parameter time-series data, equipment operating status time-series data, and quality inspection result data associated with the production batch identifier. Perform alignment and association processing on the multi-source time-series data to generate an associated dataset. The associated dataset includes a three-dimensional association identifier and a hash checksum corresponding to the three-dimensional association identifier.

[0068] In this embodiment, multi-source time-series data can be provided by the equipment control system, the field acquisition unit, and the quality inspection system respectively. A unified data recording format is preferably used during acquisition to ensure that data from different sources can be managed and accessed within the same timeline and batch context. The steps for acquiring multi-source time-series data of the target production equipment during the production process include:

[0069] The same production batch cycle or continuous production cycle is used as the data collection cycle to collect time-series data of process parameters, time-series data of equipment operating status, and quality inspection results data associated with production batch identifiers.

[0070] In this embodiment, the collection period can be determined based on the start and end of the batch or a preset time window to ensure that the collection range matches the production process.

[0071] Write a unified timestamp to the process parameter timing data, equipment operating status timing data, and quality inspection result data, and perform alignment processing on the multi-source timing data based on the unified timestamp;

[0072] Furthermore, a unified timestamp can adopt a standard time with a unified time zone and a unified precision; alignment processing can resample and aggregate data with different sampling frequencies according to a preset alignment granularity, so that the data are comparable in the time dimension.

[0073] A three-dimensional association identifier is generated based on the unified timestamp, material identifier, and equipment number, and the three-dimensional association identifier is bound to the aligned multi-source time-series data to form an association dataset;

[0074] Material identification and equipment number are provided by the production system or equipment management system. The three-dimensional associated identification can be used as an index field and bound to the aligned process parameter record, equipment status record and quality indicator record for storage, so as to support the retrieval and traceability of data from different sources.

[0075] A hash check code is calculated for the key binding fields in the associated dataset, including at least the three-dimensional association identifier, unified timestamp, device number, and production batch identifier. The hash check code is then associated with and stored with the corresponding associated data record to achieve data traceability and tamper-proof verification.

[0076] In this embodiment, the key binding fields can be serialized in a fixed field order, and then the hash check code can be calculated and stored for subsequent consistency verification and traceability verification.

[0077] In this embodiment, quality inspection result data is generated by the test management module and correlated with the production batch identifier to facilitate correlation analysis with process parameters and equipment status data. The quality inspection result data is generated by the test management module, which includes centralized test processing and time-controlled test processing.

[0078] The centralized processing test includes: performing pre-parsing processing on test files associated with production batch identifiers, and generating quality indicator entries based on the pre-parsing results, wherein the quality indicator entries constitute at least a part of the quality inspection result data;

[0079] Pre-parsing is used to identify the structure and index mapping relationship of test files, and to convert test output into structured quality index entries for storage and retrieval.

[0080] The time control test includes: triggering the start and end of the test according to preset time control conditions, and recording breakpoint information during the test; after the test is interrupted, continuing to execute the corresponding test process based on the breakpoint information to generate quality indicator items.

[0081] Breakpoint information is used to support process recovery after interruption, ensuring the continuity and traceability of the quality indicator generation process.

[0082] In this embodiment, after the construction of the associated dataset described in S1 is completed, the device health score calculation stage begins.

[0083] S2. Within a preset evaluation period, extract equipment health features based on the time-series data of the equipment's operating status, and calculate the equipment health score based on the equipment health features. The equipment health features include energy consumption stability features and fault concealment risk features.

[0084] The steps for calculating the equipment health score based on the aforementioned equipment health characteristics include:

[0085] Based on the time-series data of equipment operation status, equipment energy consumption stability characteristics and equipment fault hidden risk characteristics are constructed respectively.

[0086] In this embodiment, the preset evaluation cycle can be set according to the production cycle or process stage, such as the duration of key sections in a batch or a rolling evaluation within a fixed time window. Equipment operating status time-series data may include energy consumption sampling sequences, temperature and vibration-related sampling sequences, and fault / alarm time records, enabling health assessments to cover both equipment operating stability and potential risks. The step of calculating the equipment health score based on the equipment health characteristics includes: constructing equipment energy consumption stability characteristics and equipment fault hidden risk characteristics based on the equipment operating status time-series data;

[0087] In terms of constructing energy consumption stability characteristics, this embodiment preferably calculates the ratio of standard deviation to mean from the equipment energy consumption sampling sequence within the evaluation period to obtain the basic quantity reflecting the degree of energy consumption fluctuation, and forms the energy consumption change coefficient accordingly.

[0088] The energy consumption stability characteristic includes the energy consumption variation coefficient. The energy consumption variation coefficient Calculate using the following formula:

[0089] ;

[0090] In the formula: It is the ratio of the standard deviation to the mean, calculated based on the energy consumption sampling sequence of the target production equipment within a preset evaluation period; It is the natural logarithm function;

[0091] In this embodiment, the energy consumption sampling sequence can be obtained from the output of the electricity meter / energy meter or the equipment control system. The sampling frequency can be consistent with the equipment status sampling frequency or calculated after being aligned with a preset granularity to ensure consistent calculation caliber.

[0092] In terms of constructing the risk characteristics of fault concealment, this embodiment identifies high-incidence intervals of faults based on the time series of equipment fault occurrences, and obtains the risk concealment coefficient by combining the current time with historical statistics.

[0093] The fault concealment risk characteristic includes the risk concealment coefficient. The risk concealment coefficient Calculate using the following formula:

[0094] ;

[0095] The fault occurrence time series can be obtained from equipment alarm logs, maintenance records, or fault code reporting records. In order to obtain the center time of the high fault occurrence period, this embodiment uses density clustering to cluster the fault occurrence time series to obtain the center time of the high fault occurrence period, and calculates the time difference between this and the current time.

[0096] In the formula: The time difference between the current time and the center time of the high-incidence period of equipment failure is the time difference between the current time and the center time of the high-incidence period of equipment failure, which is obtained by DBSCAN density clustering based on the equipment failure occurrence time series. This is to preset the number of similar failures within a historical time window; It is a natural constant and The unit is hours, and 24 represents the 24-hour time base.

[0097] In implementation, the preset historical time window can be set according to the maintenance cycle or management cycle, such as the last 7 days, the last 30 days, etc.; the number of similar faults can be classified and counted according to fault type or fault code to ensure consistent statistical standards.

[0098] After obtaining the energy consumption stability characteristics and fault concealment risk characteristics, to ensure the comparability of indicators with different dimensions, this embodiment performs scale unification processing on the health characteristics, and can simultaneously perform caliber-based processing on the temperature-related indicators and vibration-related indicators extracted from the equipment operating status time-series data. Scale unification processing is performed on the equipment energy consumption stability characteristics and equipment fault concealment risk characteristics to obtain comparable health evaluation characteristics. The scale unification processing includes normalization processing or standardization processing.

[0099] A comprehensive evaluation of health assessment features is performed based on preset evaluation fusion rules to obtain an equipment health score. The evaluation fusion rules include: normalizing energy consumption variation coefficients, risk concealment coefficients, temperature-related indicators extracted from the equipment's operating status time-series data, and vibration-related indicators, then inputting these into a fuzzy rule base. The output of the fuzzy rule base is then defuzzified to obtain an equipment health score ranging from 0 to 100. The fuzzy rule base can be pre-configured to describe the mapping relationship between energy consumption fluctuations, risk levels, temperature and vibration states, and health scores. The defuzzification process can use common methods such as the centroid method to convert the fuzzy inference results into numerical scores, allowing the scores to be directly used for subsequent process optimization input construction and risk control deployment.

[0100] In this embodiment, after the equipment health score calculation described in S2 is completed, the process parameter optimization stage begins.

[0101] S3. Construct the process parameter time series data into multi-process time series features according to process identifiers, and generate multi-process associated input features based on the parameter transfer relationship between different processes. Use the multi-process time series features, the multi-process associated input features and the equipment health score as inputs to the process optimization model, and output the target process parameter combination under preset process constraints.

[0102] In this embodiment, the time-series data of process parameters can be sourced from the equipment control system or process acquisition system, covering key process parameters that affect quality and energy consumption. To facilitate model processing, it is preferable to segment and characterize the time-series data of process parameters using a preset time window. Based on the time-series data of process parameters, the sampling sequence of process parameters for each process is segmented according to the process identifier, and multi-process time-series features reflecting the changing trend, fluctuation amplitude, and rate of change of process parameters are extracted according to the preset time window. For example, the trend can be characterized by the change in mean within the window, the slope of the linear fit, or the cumulative change; the fluctuation amplitude can be characterized by the standard deviation, peak-to-peak value, or quantile difference; and the rate of change can be characterized by the difference between adjacent sampling points, the maximum rate of change within the window, or derivative statistics. The above feature extraction method can be configured according to the sampling frequency and process cycle of different process parameters to reflect the dynamic change characteristics of the parameters under the current operating conditions.

[0103] After obtaining the time-series features, they are jointly constructed with the equipment health score to characterize the current operating state of the equipment and its degree of process adaptation. The time-series features and the equipment health score are combined to form a joint input feature that characterizes the current operating state of the equipment and its degree of process adaptation. This joint input feature can be generated through feature concatenation, combining the time-series features of each process parameter with the equipment health score within the corresponding evaluation period into a single input vector; alternatively, it can be segmented and concatenated based on process stage information to enhance the ability to express differences across process stages.

[0104] Based on the aforementioned joint input features, the data is input into a process optimization model for solution, and the target process parameter combination is output under the premise of satisfying constraints. The joint input features are input into the process optimization model, and the target process parameter combination is output under the joint constraints of process parameter value range constraints, equipment operating capacity constraints, quality stability constraints, and equipment coordination constraints. The process parameter value range constraints can be determined by process specifications or equipment safety boundaries; equipment operating capacity constraints can be determined by equipment rated capacity, allowable load range, or protection thresholds; and quality stability constraints can be determined by the acceptable range or fluctuation limit of key quality indicators. These constraints ensure that the optimized output meets actual executable requirements and avoids parameter combinations that exceed equipment capabilities or affect quality stability.

[0105] To further ensure the feasibility of production line collaboration and scheduling, this embodiment introduces equipment collaboration constraints and determines them through scheduling optimization. The equipment collaboration constraints are determined through the following steps: a scheduling fitness function is constructed based on equipment health scores, equipment changeover times, equipment collaboration penalty coefficients, and qualification deviation penalty coefficients. A genetic algorithm is then used to search for the optimal production path with the goal of maximizing the scheduling fitness function. The equipment collaboration constraints are then determined based on the optimal production path. In implementation, candidate production paths can be encoded as equipment / process sequences. A genetic algorithm is used to perform selection, crossover, and mutation operations, and the fitness function is combined to evaluate the path's merits, thereby searching for the optimal or near-optimal path that meets the collaboration requirements. Equipment collaboration constraints, such as collaboration matching priority, available equipment set, or collaboration threshold, are then derived from this path.

[0106] The scheduling fitness function satisfies the following formula:

[0107] ;

[0108] In the formula: This represents the scheduling fitness function value; This refers to the number of stages involved in the costume change process. The sequence number of the costume change process and ; For the first Equipment replacement time for each replacement stage; This is a very small amount of normal quantity; This is a comprehensive item encompassing collaboration and quality. This is the equipment coordination penalty coefficient; This is the penalty coefficient for acceptable deviation; and The preset weighting coefficients are used, and they satisfy the following conditions: .

[0109] Through the above methods, this embodiment can output the target process parameter combination based on a comprehensive consideration of equipment status and process dynamic characteristics, and ensure that the optimization results match the feasibility of production scheduling through collaborative constraints, thereby providing a stable optimization output for subsequent risk control and closed-loop updates.

[0110] In this embodiment, after the process optimization model outputs the target process parameter combination, in order to avoid unnecessary disturbance to the production process caused by parameter adjustment and to improve the controllability and traceability of parameter changes, a risk assessment and issuance control are performed before the parameters are issued.

[0111] S4. Before the target process parameter combination is issued, the risk level of the process parameter adjustment is assessed based on the adjustment range of the target process parameter combination relative to the currently effective process parameters and the equipment health score associated with the production batch identifier, and the issuance method of the target process parameter combination is controlled according to the risk level.

[0112] In this embodiment, risk level assessment and distribution method control can be implemented by a rule engine. The rule engine can be deployed on the process management system or the host computer of the control system to execute unified governance logic for parameter changes. The assessment to obtain the risk level of process parameter adjustment and the control of the distribution method of the target process parameter combination based on the risk level include: executing the distribution control logic through the rule engine, determining the corresponding distribution method based on the adjustment range of the target process parameter combination relative to the currently effective process parameter, where the adjustment range is the absolute value or relative change ratio of the change of each process parameter in the target process parameter combination relative to the currently effective process parameter; in implementation, the adjustment range can be calculated item by item for each parameter: for each process parameter, its absolute change and relative change ratio are calculated, for example, the percentage change based on the currently effective value, and the rule engine can use the maximum change, the maximum relative change ratio, or the change summed by weight as the judgment input to ensure that a identifiable risk level can still be obtained when multiple parameters are adjusted simultaneously.

[0113] Based on a tiered rule of adjustment range and preset thresholds, the method for issuing target process parameter combinations is determined. When the adjustment range does not exceed the first threshold, automatic issuance is performed; when the adjustment range is greater than the first threshold but not greater than the second threshold, issuance is performed after manual confirmation; when the adjustment range is greater than the second threshold, issuance is performed after approval. In specific implementation, the first and second thresholds can be configured according to process specifications, equipment safety boundaries, or historical parameter adjustment experience, and the version and effective scope can be maintained in the production management system to ensure that different production lines or different product models have differentiated threshold configuration capabilities.

[0114] In addition to the adjustment range, this embodiment further incorporates equipment status risk into the control strategy. When the equipment health score or risk concealment coefficient reaches a preset trigger condition, the constraints on the target parameter combination are tightened. When the equipment health score is lower than a preset health threshold or the risk concealment coefficient is higher than a preset risk threshold, the upper limit of at least one process parameter in the target process parameter combination is lowered according to a preset ratio. In implementation, the upper limit reduction can be applied to key process parameters related to equipment load, energy consumption, or quality sensitivity, and the reduction ratio can be configured according to a preset ratio table to make parameter adjustments more conservative, thereby improving the operational safety margin under low health or high risk conditions.

[0115] To ensure that alarm and warning status effectively constrain parameter distribution, this embodiment sets up warning linkage rules: when an unclosed warning event exists, the distribution method is upgraded to manual confirmation before distribution. When an unclosed warning event status exists, the distribution method is controlled to manual confirmation before distribution; the unclosed warning event status can be provided by the alarm management module or the device status monitoring module, and a corresponding relationship is established with the device number, thereby realizing linkage control for distribution to the same device.

[0116] To improve the stability and maintainability of alarm event handling, this embodiment standardizes the alarm event handling process. The alarm event handling includes: constructing a three-dimensional alarm identifier based on device number, alarm type, and parameter name, and performing alarm deduplication within a preset sliding window; performing a gradient escalation of alarm levels based on the alarm's unprocessed duration and a preset escalation threshold; and merging low-level alarms corresponding to the same device number into a summary alarm based on a preset merging time window. Alarm rules support online updates and version rollback. In implementation, the three-dimensional alarm identifier is used to uniquely identify alarm events of the same type and parameter dimension for the same device, thereby avoiding alarm storms caused by duplicate reporting within the sliding window; gradient escalation is used to increase the alarm's attention level when an alarm has been unprocessed for a long time; the merging strategy is used to aggregate multiple low-level alarms from the same device into a summary alarm to improve alarm handling efficiency; online updates and version rollback are used to maintain the alarm strategy during production and maintain traceable version management, reducing the impact of strategy changes on production stability.

[0117] Through the aforementioned risk assessment and control measures, this embodiment enables the issuance of target process parameter combinations to possess hierarchical control, status linkage, and alarm management capabilities, thereby providing controllable execution guarantees for the stable effectiveness and closed-loop updates of subsequent target process parameter combinations.

[0118] In this embodiment, after completing the risk control described in S4 and making the target process parameter combination effective, the process enters the result feedback and model update stage to achieve an adaptive closed loop for process parameter optimization.

[0119] S5. After the target process parameter combination takes effect, collect the corresponding quality inspection result data and energy consumption index data, and update the process optimization model based on the quality inspection result data and energy consumption index data to form a closed loop of adaptive optimization of process parameters.

[0120] In this embodiment, quality inspection result data can be output by the test management module and bound to the production batch identifier; energy consumption index data can be obtained from energy consumption metering devices or equipment operation data statistics, and collected according to the same production batch identifier or the same effective period, thereby ensuring that the updated data corresponds to the current parameter issuance action. The process of updating the process optimization model can adopt an incremental training method to introduce the latest production feedback while maintaining the stability of the model. The process of updating the process optimization model includes: binding the quality inspection result data, energy consumption index data and joint input features based on the production batch identifier to generate training samples for model training, wherein the joint input features are feature vectors jointly constructed by the time series features of process parameters and equipment health scores;

[0121] In terms of implementation, training samples can be organized at the batch level. Each sample contains at least: joint input features as model input, quality detection result data and energy consumption index data as supervision signals for model training or feedback quantities for optimization objectives, and production batch identifiers are retained in the samples for traceability and auditing.

[0122] Incremental training and model parameter updates are performed when preset trigger conditions are met. Incremental training and model parameter updates are performed when the number of new training samples reaches a preset threshold or when a preset model update time period is reached.

[0123] The preset quantity threshold and model update time period can be configured according to the production rhythm, model training resources, and business needs. For example, for high-frequency production scenarios, training can be triggered after a certain number of new batch samples have been accumulated; for stable production scenarios, training can also be triggered at fixed intervals to ensure the model's adaptability to equipment state drift and changes in operating conditions.

[0124] To ensure the model update process is traceable and supports backtracking analysis, this embodiment identifies and stores the model versions before and after the update in a linked manner. A corresponding version identifier is generated for each model version before and after the update, and this version identifier is then linked and stored with the corresponding production batch identifier.

[0125] In terms of implementation, the version identifier can be generated in the form of version number or hash digest, and a relationship can be established with the batch identifier of the sample set used for training, so as to quickly locate the model version and the corresponding training data range when quality anomalies or model performance fluctuations occur, thereby supporting operation and maintenance actions such as rollback or retraining.

[0126] In addition to the closed-loop model update, this embodiment can also be configured with a closed-loop control process integrated with production quality management to ensure a consistent traceability and verification mechanism for the handling and release of defective products. The method also includes a quality closed-loop control step and a packaging release verification step.

[0127] Regarding closed-loop quality control, the steps include: locking the circulation permission of detected defective products, generating and issuing corresponding repair tasks; performing retesting after repair is completed, and unlocking the circulation permission when the retest is passed.

[0128] Among them, the circulation permission lock can be achieved by setting a status mark on the corresponding product serial number or batch record through the production management system, so that it cannot enter the subsequent packaging or outbound process; the maintenance task can be pushed to the maintenance terminal or work order system, and the retest can be regenerated by the test management module and the defect handling status can be updated.

[0129] Regarding packaging release verification, the packaging release verification steps include: based on the production batch identifier and its associated quality inspection result data, verifying the completion status of mandatory tests and the status of defect handling of the product to be packaged during the product packaging process;

[0130] The packaging process includes verification to ensure that the products to be packaged have completed the required tests and that there are no unresolved defects. Furthermore, the consistency between the product model identification, production batch identification, and outer box identification is verified during the product packing process.

[0131] Consistency verification can prevent traceability disruptions or quality risks caused by incorrect or mixed packaging.

[0132] When the required number of boxes is reached, an outer box traceability master code is generated and written to the label to achieve box-level traceability. When the number of boxes reaches a preset value, an outer box traceability master code associated with the set of product serial numbers inside the box is generated. The outer box traceability master code is an coded identifier generated based on the set of product serial numbers, and the outer box traceability master code is written to the outer box label.

[0133] In terms of implementation, the outer box traceability master code can be generated by encoding or summarizing the set of product serial numbers inside the box in a fixed order and establishing a binding relationship with the outer box identification. This allows for quick location of the set of product serial numbers inside the box and its corresponding batch quality information in subsequent warehousing, logistics, or after-sales scenarios, thereby improving traceability efficiency and consistency.

[0134] Through the above steps, this embodiment forms a closed-loop process of parameter issuance, result collection, model update, quality handling, and release verification. This enables the process parameter optimization strategy to be continuously revised based on production results, while working in conjunction with quality control and traceability links to enhance the stability and engineering feasibility of the solution in long-term operation.

[0135] like Figure 2 As shown, another embodiment of the present invention provides a system for multi-stage correlation modeling and adaptive control of process parameters, applied to the method for multi-stage correlation modeling and adaptive control of process parameters as described above, including:

[0136] The multi-source data acquisition module is used to collect multi-source time-series data of the target production equipment during the production process. The multi-source time-series data includes process parameter time-series data, equipment operating status time-series data, and quality inspection result data associated with production batch identifiers.

[0137] The alignment and anti-tampering module is used to perform alignment and association processing on the multi-source time series data to generate an associated dataset containing a three-dimensional association identifier and a corresponding hash check code.

[0138] The equipment health scoring module is used to extract equipment health feature quantities based on equipment operating status time-series data within a preset evaluation period, and to calculate the equipment health score based on the equipment health feature quantities.

[0139] The process parameter optimization module is used to construct multi-process time sequence features from process parameter time sequence data according to process identifiers, and generate multi-process associated input features based on the parameter transfer relationship between different processes. The multi-process time sequence features, the multi-process associated input features and equipment health score are used as inputs to the process optimization model, and the target process parameter combination is output under the condition of satisfying preset process constraints.

[0140] The risk control module is used to assess the risk level of process parameter adjustment based on the adjustment range of the target process parameter combination relative to the currently effective process parameters and the corresponding equipment health score before the target process parameter combination is issued, and to control the issuance method of the target process parameter combination according to the risk level.

[0141] The model update module is used to collect the corresponding quality inspection results data and energy consumption index data after the target process parameter combination takes effect, and update the process optimization model based on the data.

[0142] In this embodiment, the system for multi-process parameter correlation modeling and adaptive control can be deployed in a collaborative computing environment at the factory edge and center. The edge side is used for data acquisition, preprocessing, and some real-time control, while the center side is used for model training, optimization calculation, strategy management, and version management. Modules can interact with each other via industrial bus, message queues, or interface services. Key index fields such as production batch identifiers, equipment numbers, and material identifiers can be uniformly managed at the storage layer, ensuring the system's scalability and portability across different production lines and product types. During system operation, the multi-source data acquisition module collects process parameters, equipment status, and quality inspection results in batches. The alignment and anti-tampering module performs time alignment and association identification on the multi-source data to generate a verifiable associated dataset. The equipment health scoring module outputs equipment health scores within a preset evaluation period to characterize equipment status and potential risks. The process parameter optimization module outputs target process parameter combinations based on process time sequence characteristics and health scores. The risk control module assesses the risk level of the target parameter combinations and controls the distribution method. The model update module updates the process optimization model based on the quality and energy consumption feedback after it takes effect. Thus, at the system level, a closed-loop optimization process with reliable data, controllable risks, and continuous updates, corresponding to the method implementation, is formed.

[0143] In summary, this invention constructs a comprehensive solution for optimizing process parameters in the production process. It integrates the correlation and traceability of multi-source production data, equipment status assessment, process optimization output, execution process control, and subsequent updates into a closed-loop process, forming an engineering-ready implementation path. This solution can be deployed in conjunction with existing production equipment, testing systems, and production management platforms. It is applicable to different organizational methods such as batch production and continuous production, and its application can be easily expanded to different equipment types and process scenarios. It can provide general technical support for enterprises to carry out intelligent process management and production process optimization, and has high application value and promising prospects for promotion.

[0144] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0145] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process. Furthermore, the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functionality involved.

[0146] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in this application, and these should all be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for multi-stage correlation modeling and adaptive control of process parameters, characterized in that, Includes the following steps: S1. Collect multi-source time-series data of the target production equipment during the production process. The multi-source time-series data includes process parameter time-series data, equipment operating status time-series data, and quality inspection result data associated with the production batch identifier. Alignment and correlation processing are performed on the multi-source time-series data to generate a correlated dataset; The associated dataset includes a 3D association identifier and a hash checksum corresponding to the 3D association identifier; S2. Within a preset evaluation period, extract equipment health features based on the time-series data of the equipment's operating status, and calculate the equipment health score based on the equipment health features. The equipment health features include energy consumption stability features and fault concealment risk features. S3. Construct the process parameter time series data into multi-process time series features according to process identifiers, and generate multi-process associated input features based on the parameter transfer relationship between different processes. Use the multi-process time series features, the multi-process associated input features and the equipment health score as inputs to the process optimization model, and output the target process parameter combination under preset process constraints. The steps for combining the output target process parameters include: Based on the time-series data of the process parameters, the sampling sequence of process parameters for each process is segmented according to the process identifier, and multi-process time-series features reflecting the changing trend, fluctuation amplitude, and rate of change of process parameters are extracted according to a preset time window. These time-series features are then combined with the equipment health score to construct a joint input feature characterizing the current operating state of the equipment and its degree of process adaptability. This joint input feature is input into the process optimization model, and under the joint constraints of process parameter value range, equipment operating capacity, quality stability, and equipment coordination, the target combination of process parameters is output. The equipment coordination constraint is determined through the following steps: a scheduling fitness function is constructed based on the equipment health score, equipment changeover time, equipment coordination penalty coefficient, and qualification deviation penalty coefficient. A genetic algorithm is then used to search for the optimal production path with the goal of maximizing the scheduling fitness function. The equipment coordination constraint is determined based on the optimal production path. The scheduling fitness function satisfies the following formula: ； In the formula: This represents the scheduling fitness function value; This refers to the number of stages involved in the costume change process. For the first Equipment replacement time for each replacement stage; This is a very small amount of normal quantity; This is a comprehensive item encompassing collaboration and quality. This is the equipment coordination penalty coefficient; This is the penalty coefficient for acceptable deviation; and The preset weighting coefficients are used, and they satisfy the following conditions: ; S4. Before the target process parameter combination is issued, the risk level of the process parameter adjustment is assessed based on the adjustment range of the target process parameter combination relative to the currently effective process parameters and the equipment health score associated with the production batch identifier, and the issuance method of the target process parameter combination is controlled according to the risk level. S5. After the target process parameter combination takes effect, collect the corresponding quality inspection result data and energy consumption index data, and update the process optimization model based on the quality inspection result data and energy consumption index data to form a closed loop of adaptive optimization of process parameters.

2. The method for multi-process correlation modeling and adaptive control of process parameters according to claim 1, characterized in that, The steps for collecting multi-source time-series data of the target production equipment during the production process include: The same production batch cycle or continuous production cycle is used as the data collection cycle to collect time-series data of process parameters, time-series data of equipment operating status, and quality inspection results data associated with production batch identifiers. Write a unified timestamp to the process parameter timing data, equipment operating status timing data, and quality inspection result data, and perform alignment processing on the multi-source timing data based on the unified timestamp; A three-dimensional association identifier is generated based on the unified timestamp, material identifier, and equipment number, and the three-dimensional association identifier is bound to the aligned multi-source time-series data to form an association dataset; A hash check code is calculated for the key binding fields in the associated dataset, including at least the three-dimensional association identifier, unified timestamp, device number, and production batch identifier. The hash check code is then associated with and stored with the corresponding associated data record to achieve data traceability and tamper-proof verification.

3. The method for multi-process correlation modeling and adaptive control of process parameters according to claim 2, characterized in that, The quality inspection result data is generated by the test management module, which includes centralized test processing and time-controlled test processing. The centralized processing test includes: performing pre-parsing processing on test files associated with production batch identifiers, and generating quality indicator entries based on the pre-parsing results, wherein the quality indicator entries constitute at least a part of the quality inspection result data; The time control test includes: triggering the start and end of the test according to preset time control conditions, and recording breakpoint information during the test; after the test is interrupted, continuing to execute the corresponding test process based on the breakpoint information to generate quality indicator items.

4. The method for multi-process correlation modeling and adaptive control of process parameters according to claim 1, characterized in that, The steps for calculating the equipment health score based on the aforementioned equipment health characteristics include: Based on the time-series data of equipment operation status, equipment energy consumption stability characteristics and equipment fault hidden risk characteristics are constructed respectively. The energy consumption stability characteristics and hidden risk characteristics of the equipment are subjected to scale unification processing to obtain comparable health evaluation characteristics. The scale unification processing includes normalization processing or standardization processing. The health evaluation features are comprehensively evaluated based on preset evaluation fusion rules to generate a device health score. The evaluation fusion rules include: normalizing the energy consumption change coefficient, risk concealment coefficient, and temperature-related and vibration-related indicators extracted from the time-series data of the equipment's operating status, then inputting them into a fuzzy rule base, and performing defuzzification on the output of the fuzzy rule base to obtain an equipment health score that characterizes the health status of the equipment.

5. The method for multi-process correlation modeling and adaptive control of process parameters according to claim 4, characterized in that, The energy consumption stability characteristic includes the energy consumption variation coefficient. The energy consumption variation coefficient Calculate using the following formula: ； In the formula: It is the ratio of the standard deviation to the mean, calculated based on the energy consumption sampling sequence of the target production equipment within a preset evaluation period; It is the natural logarithm function; The fault concealment risk characteristic includes the risk concealment coefficient. The risk concealment coefficient Calculate using the following formula: ； In the formula: The time difference between the current time and the center time of the high-incidence period of equipment failure is the time difference between the current time and the center time of the high-incidence period of equipment failure, which is obtained by DBSCAN density clustering based on the equipment failure occurrence time series. This is to preset the number of similar failures within a historical time window; It is a natural constant and The unit is hours, and 24 represents the 24-hour time base.

6. The method for multi-process correlation modeling and adaptive control of process parameters according to claim 1, characterized in that, The assessment obtains the risk level of process parameter adjustment, and controls the distribution method of the target process parameter combination based on the risk level, including: executing distribution control logic through a rule engine, and determining the corresponding distribution method based on the adjustment range of the target process parameter combination relative to the currently effective process parameter, wherein the adjustment range is the absolute value or relative change ratio of the change of each process parameter in the target process parameter combination relative to the currently effective process parameter; When the adjustment range does not exceed the first threshold, automatic issuance is performed; when the adjustment range is greater than the first threshold but not greater than the second threshold, issuance is performed after manual confirmation; when the adjustment range is greater than the second threshold, issuance is performed after approval. When the equipment health score is lower than the preset health threshold or the risk concealment coefficient is higher than the preset risk threshold, the upper limit of at least one process parameter in the target process parameter combination will be lowered according to a preset ratio. When there is an unclosed warning event status, the issuance method will be controlled to be issued after manual confirmation; The early warning event processing includes: constructing a three-dimensional alarm identifier based on the device number, alarm type, and parameter name, and performing alarm deduplication within a preset sliding window; performing gradient escalation of alarm level based on the alarm unprocessed duration and a preset escalation threshold; and merging low-level alarms corresponding to the same device number into a summary alarm based on a preset merging time window. The alarm rules support online updates and version rollback.

7. The method for multi-process correlation modeling and adaptive control of process parameters according to claim 1, characterized in that, The process of updating the process optimization model includes: Based on the production batch identifier, the quality inspection result data, energy consumption index data and joint input features are bound together to generate training samples for model training. The joint input features are feature vectors jointly constructed by process parameter time series features and equipment health scores. When the number of new training samples reaches a preset threshold or the preset model update time period is reached, incremental training of the process optimization model is performed and the model parameters are updated. Generate corresponding version identifiers for the model versions before and after the model update, and associate and store the version identifiers with the corresponding production batch identifiers.

8. The method for multi-process correlation modeling and adaptive control of process parameters according to claim 1, characterized in that, The method also includes a quality closed-loop control step and a packaging release verification step; the quality closed-loop control step includes: locking the circulation permission of the detected defective products, generating and issuing the corresponding repair task; performing a retest after the repair is completed, and releasing the circulation permission lock when the retest is passed; The packaging release verification step includes: verifying the completion status of mandatory tests and the status of defect handling of the product to be packaged during the product packaging process based on the production batch identifier and its associated quality inspection result data; During the product packaging process, verify the consistency between the product model identification, production batch identification, and outer box identification; When the number of boxes reaches a preset value, an outer box traceability code associated with the set of product serial numbers inside the box is generated. The outer box traceability code is an coded identifier generated based on the set of product serial numbers, and the outer box traceability code is written to the outer box label.

9. A system for multi-stage correlation modeling and adaptive control of process parameters, applied to the method for multi-stage correlation modeling and adaptive control of process parameters as described in any one of claims 1 to 7, characterized in that, include: The multi-source data acquisition module is used to collect multi-source time-series data of the target production equipment during the production process. The multi-source time-series data includes process parameter time-series data, equipment operating status time-series data, and quality inspection result data associated with production batch identifiers. The alignment and anti-tampering module is used to perform alignment and association processing on the multi-source time series data to generate an associated dataset containing a three-dimensional association identifier and a corresponding hash check code. The equipment health scoring module is used to extract equipment health feature quantities based on equipment operating status time-series data within a preset evaluation period, and to calculate the equipment health score based on the equipment health feature quantities. The process parameter optimization module is used to construct multi-process time sequence features from process parameter time sequence data according to process identifiers, and generate multi-process associated input features based on the parameter transfer relationship between different processes. The multi-process time sequence features, the multi-process associated input features and equipment health score are used as inputs to the process optimization model, and the target process parameter combination is output under the condition of satisfying preset process constraints. The risk control module is used to assess the risk level of process parameter adjustment based on the adjustment range of the target process parameter combination relative to the currently effective process parameters and the corresponding equipment health score before the target process parameter combination is issued, and to control the issuance method of the target process parameter combination according to the risk level. The model update module is used to collect the corresponding quality inspection results data and energy consumption index data after the target process parameter combination takes effect, and update the process optimization model based on the data.

Images