Valve fitting online quality regulation method and system

Abstract

This invention relates to the field of intelligent manufacturing and industrial automation testing technology, specifically to an online quality control method and system for valve components. The invention monitors the production line's quality data stream in real time, identifies deviation trends and generates defect awareness pulses, and combines this with a time-series prediction model to predict defect outbreak times, thus mapping the process path in reverse. The system employs an intelligent compensation mechanism weighted by fuzzy logic and historical cases to dynamically fine-tune processing parameters, and introduces collaborative compensation and circuit breaker mechanisms to ensure safe and effective control. Successful control cases are ultimately saved to a knowledge base. This achieves advanced prediction and real-time precise control of quality defects during valve component production, effectively reducing scrap rates and improving the intelligence level of the production line and product quality stability.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing and industrial automation testing technology, specifically to an online quality control method and system for valve components, applicable to real-time quality monitoring, defect prediction and proactive compensation control in automated production lines for valve components. Background Technology

[0002] Valve fittings are widely used in key industrial sectors such as petrochemicals, power generation, and water supply and drainage. Their manufacturing precision and reliability directly affect the safety and stability of the entire fluid control system. With the development of Industry 4.0 and intelligent manufacturing, valve fitting production has gradually shifted from traditional single-machine operations to highly automated assembly line production. However, due to the complex structure and numerous processing steps involved in valve fittings, including casting, machining, and surface treatment, the quality-influencing factors during production are highly coupled and nonlinear. This makes real-time monitoring and precise control of product quality a pressing problem that needs to be solved in the industry.

[0003] Existing quality control technologies for valve component manufacturing primarily rely on post-production sampling or threshold alarm mechanisms at single workstations. This passive quality control model often suffers from delays; by the time a genuine defect is detected, a large number of scrap products may have already been generated, resulting in significant economic losses. Furthermore, traditional adjustment methods for detecting quality deviations often depend on manual experience or fixed rule bases, making it difficult to cope with the complex and ever-changing interference factors in the production environment. The lack of early prediction and dynamic compensation for quality fluctuation trends prevents effective intervention before defects occur, and when single-point compensation is ineffective, there is a lack of systematic multi-process collaborative control strategies. Moreover, existing production management systems often neglect the temporal characteristics and correlation between equipment status in the data stream when processing quality data, making it difficult to accurately identify early signs of defect awareness. When faced with complex processing parameter disturbances, existing technologies lack an adaptive adjustment mechanism based on fuzzy matching of historical success cases and real-time data, resulting in inaccurate generation of compensation parameters, and potentially exacerbating production instability due to erroneous adjustments. At the same time, the existing system lacks an effective fault tolerance and rollback mechanism. Once the adjustment instruction fails to be executed, it is impossible to stop the loss in time, which affects the continuity of the production line and the yield rate of the final product. Summary of the Invention

[0004] To overcome the problems of lagging quality inspection, lack of predictability, and insufficient dynamic response capability to complex production disturbances in existing technologies, this invention provides an online quality control method and system for valve accessories. By automatically generating defect awareness pulses containing fluctuation characteristics and equipment status, using a time-series prediction model to deduce the actual defect outbreak time and reverse map the process path, and combining the technical solutions of fuzzy-historical weighted compensation principle, collaborative compensation, and circuit breaker backoff mechanism, it realizes the advanced prediction and real-time precise control of quality defects in the valve accessory production process, effectively reducing the scrap rate and improving the intelligence level of the production line and the stability of product quality.

[0005] The technical solution of this application specifically includes:

[0006] According to one aspect of this application, an online quality control method for valve fittings is provided, comprising:

[0007] When any inspection station on the automated valve parts production line detects a deviation trend in the quality data stream, it automatically generates a defect awareness pulse that includes fluctuation characteristics and equipment status.

[0008] After receiving a defect awareness pulse, the production management system uses a time-series prediction model to predict the time when the defect awareness pulse will erupt into a real defect in the future workpiece cycle. Using the time of the real defect as the endpoint, it reverse maps the process path that the defect awareness pulse will pass through. On the process path, it generates compensation instructions according to the fuzzy-history weighted compensation principle to fine-tune the processing parameters to suppress defect formation.

[0009] Monitor the quality data stream of the workpiece after compensation. If the quality data stream still shows a deviation trend, it is determined that the local compensation is insufficient. Perform collaborative compensation according to the process path and processing node status to be traversed by the compensation instruction.

[0010] During the collaborative compensation process, the convergence trend of the verification parameters and the attenuation rate of the defect pulse are verified in real time. If two consecutive verifications fail, the current collaborative compensation is determined to be invalid, triggering the circuit breaker rollback mechanism to re-execute the collaborative compensation until the verification is successful.

[0011] When the quality data stream recovers to the steady-state range and continues for no less than three workpiece cycles, the parameters, paths, and collaborative strategies of this control process are automatically saved as an adaptive control case. At the same time, a processing report of this quality event is pushed to the production management system.

[0012] As a further option of the method of the present invention, the defect awareness pulse generation step includes:

[0013] Deploy multi-source heterogeneous detection nodes and standardized data acquisition, and maintain a fixed length through edge computing modules at each detection station. A sliding time window is used to fit the time series of data within the window using a linear regression algorithm or to identify deviation trends in the quality data stream using a cumulative sum control chart algorithm.

[0014] If a deviation trend is detected, the pulse generation mechanism is triggered to extract the amplitude, direction, frequency and acceleration of the fluctuation as the fluctuation feature vector. The feed rate multiplier, spindle speed setpoint and remaining tool life percentage are obtained through industrial Ethernet as a snapshot of the equipment status.

[0015] The deviation degree is normalized and mapped to a confidence level based on the extent to which the deviation point exceeds the control limit. The pulse generation time, fluctuation feature vector, equipment status snapshot, and confidence level are then encapsulated into a defect awareness pulse with a four-tuple structure.

[0016] As a further option of the method of the present invention, the step of confirming the time point of the actual defect includes:

[0017] The production management system retrieves historical time period data similar to the current defect awareness pulse characteristics from the historical database, inputs the deviation sequence within the sliding window into a pre-trained long short-term memory network or gated recurrent unit deep learning model, and outputs the deviation prediction sequence for future workpiece cycles.

[0018] Traverse the deviation prediction sequence, find the first index that satisfies the deviation amount being greater than or equal to the scrap threshold, and add the number of cycles corresponding to the index to the current time to determine the predicted outbreak time of the actual defect.

[0019] If the deviation prediction sequence does not exceed the prediction range and the prediction curve shows a regression trend, then the predicted outbreak time of the actual defect will be set to a null value and deep regulation will be temporarily suspended.

[0020] As a further option of the method of the present invention, the compensation instruction generation step includes:

[0021] Query the real-time process scheduling table and material flow tracking information in the manufacturing execution system to determine the unique identification code of the workpiece at the processing position at the actual defect outbreak time, and trace the digital process list based on the workpiece's unique identification code to determine the process node corresponding to the actual defect outbreak time.

[0022] Starting from the current inspection station node in the process topology diagram, a reverse search is performed to reach the process node corresponding to the actual defect outbreak time point. All process nodes on the reverse path are determined as the set of nodes to be compensated.

[0023] The current fluctuation feature vector is input into the fuzzy logic controller. The preliminary compensation amount is obtained by querying the predefined fuzzy rule base and defuzzifying. At the same time, the similarity between the current fluctuation feature vector and the best matching case in the historical successful control case base is calculated to obtain the historical reference compensation amount.

[0024] The fuzzy weights and historical weights are dynamically adjusted based on the stability of the operating conditions, and the final compensation amount is calculated using the following formula: ;in, This is the real-time membership value calculated based on a fuzzy rule base. For current features With historical case library Similarity to the best-matching cases; and These are fuzzy weights and historical weights, respectively, satisfying... .

[0025] As a further option of the method of the present invention, the determination and generation of the collaborative compensation includes:

[0026] Establish a post-compensation monitoring window, extract key process quality characteristic data of the workpiece after compensation operation to form a post-compensation data stream, and compare the post-compensation data stream with the data stream before compensation.

[0027] If the mean of the post-compensation data stream does not return to the vicinity of the target centerline or the fluctuation amplitude does not decrease significantly, it is determined that the quality data stream still has a deviation trend.

[0028] Traverse the set of nodes to be compensated and collect the real-time status information of each node, including the remaining time of the processing cycle, the equipment load rate, and the adjustable range of the process parameters.

[0029] Based on the preset progressive priority control protocol, the control priority of each node is calculated by comprehensively considering the influence of the process on the final defect, the flexibility of parameter adjustment, and the real-time load of the current node. The collaborative compensation instructions of each node are generated in order of control priority from high to low.

[0030] As a further option of the method of the present invention, the execution of collaborative compensation also includes performing collaborative correction of execution parameters and process rebalancing according to a progressive priority control protocol:

[0031] When adjusting the processing parameters of a certain process, the impact on the processing allowance of subsequent processes is evaluated simultaneously. If the cutting amount of a certain process is increased, the allowance of the subsequent finishing process is automatically reduced to maintain the balance of the total process allowance and avoid the process from being overloaded due to excessive allowance or the subsequent process from being undercut due to insufficient allowance.

[0032] By using a global process parameter optimization algorithm, the energy consumption or processing time on the path where the node set to be compensated is located is minimized while meeting the final quality constraints.

[0033] As a further option of the method of the present invention, the method for simultaneously verifying the parameter convergence trend and the defect pulse attenuation rate is as follows:

[0034] The monitoring thread runs in parallel to plot the time response curves of key control parameters in real time and calculate the rate of change of the response curves. If the absolute value of the rate of change gradually decreases and approaches 0, and the curve gradually approaches the target value, then the parameter is determined to show a convergence trend.

[0035] Temporary defect awareness pulses are continuously generated based on real-time data, and the current confidence level is extracted. The defect pulse decay rate is calculated using the following formula: ;in, Represents the defect pulse attenuation rate. To adjust the initial pulse confidence level, The confidence level at the current moment;

[0036] If the defect pulse decay rate remains greater than 0 and the value gradually increases over time, it indicates that the defect pulse is decaying rapidly and the control direction is correct.

[0037] As a further option of the method of the present invention, the triggering method of the circuit breaker rollback mechanism is as follows:

[0038] Set a verification cycle. If the parameters diverge or the defect pulse decay rate falls below the lower limit threshold in a certain evaluation, it is recorded as a verification failure.

[0039] Maintain a consecutive failure counter. If two consecutive verification results are both failures, the current collaborative compensation strategy is deemed unsuitable for the current working conditions, and the collaborative compensation is deemed to have failed.

[0040] Immediately stop the output of all currently executing collaborative compensation commands, issue emergency rollback commands to relevant equipment, and force all modified processing parameters to be restored to their original values ​​before adjustment or the system's default safe values;

[0041] Based on the experience gained from this failure, adjust the fuzzy weights and historical weight coefficients, or select a different combination of collaborative compensation strategies and re-execute the control process starting from generating compensation instructions until verification is successful.

[0042] As a further option of the method of the present invention, the adaptive control case generation step includes:

[0043] Construct a feature index for the case, which includes the original defect awareness impulse feature vector, deviation type, and the sequence of process nodes involved;

[0044] Record the control decisions adopted this time, including the allocation values ​​of fuzzy weights and historical weights, the specific compensation amount of each process, the priority sequence of collaborative compensation, and the process rebalancing strategy;

[0045] Record the evaluation indicators of the regulation effect, including regulation response time, overshoot and steady-state error, and store the above structured information in the historical successful regulation case database and establish a multi-dimensional index.

[0046] Another aspect of this application provides an online quality control system for valve fittings, the system comprising:

[0047] The defect awareness pulse generation module is used to automatically generate a defect awareness pulse containing fluctuation characteristics and equipment status when any inspection station on the automated production line of valve parts identifies a deviation trend in the quality data stream.

[0048] The prediction and compensation instruction generation module is used to, after receiving the defect awareness pulse, deduce the time point when the defect awareness pulse will explode into a real defect in the future workpiece cycle based on the time series prediction model; with the time point of the real defect as the endpoint, reverse map the process path that the defect awareness pulse will pass through; generate compensation instructions on the process path according to the fuzzy-history weighted compensation principle, and fine-tune the processing parameters to suppress defect formation.

[0049] The collaborative compensation module is used to monitor the quality data stream of the workpiece after compensation. If the quality data stream still shows a deviation trend, it is determined that the local compensation is insufficient. Based on the process path and processing node status to be traversed by the compensation instruction, collaborative compensation is performed.

[0050] The safety control and circuit breaker module is used to verify the convergence trend of parameters and the attenuation rate of defect pulses in real time during the collaborative compensation process. If two consecutive verifications fail, the current collaborative compensation is determined to be invalid, triggering the circuit breaker rollback mechanism to re-execute the collaborative compensation until the verification is successful.

[0051] The adaptive learning and feedback module is used to automatically save the parameters, paths and collaborative strategies of this control process as an adaptive control case when the quality data stream recovers to the steady state range and continues for no less than three workpiece cycles. At the same time, it pushes the processing report of this quality event to the production management system.

[0052] The beneficial effects of this application are as follows:

[0053] This invention, through real-time quality data stream analysis and defect outbreak time prediction, can shorten the response time to quality anomalies from the traditional manual hours to the automatic millisecond / second level, enabling early intervention for deviations. Combined with fuzzy-historical weighted compensation and multi-process collaborative correction, it can reduce the batch defect rate caused by a single parameter malfunction by 30%-50%, significantly improving the first-pass yield.

[0054] The circuit breaker rollback and real-time verification mechanism introduced in this invention ensures the safety and reliability of the control process, and can reduce the risk of equipment abnormalities or chain processing failures caused by improper compensation strategies by more than 70%, thus achieving precise control and risk circuit breaker in the control process.

[0055] Practice has shown that this method can reduce reliance on the experience of senior process engineers by more than 75%, and reduce the time for quality incident analysis, decision-making and report generation from several hours manually to within minutes automatically completed by the system, thus achieving efficient quality control in the manufacturing process. Attached Figure Description

[0056] Figure 1 A schematic diagram of the overall online quality control method for valve components;

[0057] Figure 2 S100 flowchart of online quality control method for valve fittings;

[0058] Figure 3 S200 flowchart for online quality control method of valve accessories;

[0059] Figure 4 S300 flowchart for online quality control method of valve accessories;

[0060] Figure 5 S400 flowchart for online quality control method of valve accessories;

[0061] Figure 6 S500 flowchart for online quality control method of valve accessories. Detailed Implementation

[0062] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0063] The theoretical foundation of this invention is built upon three pillars: time-series prediction and deduction theory, reverse process mapping theory, and fuzzy-history weighted compensation theory. By constructing defect awareness impulses, implementing multi-parameter collaborative correction, and designing a controlled circuit breaker backoff mechanism, the adaptive control and continuous learning of valve component production process quality are ultimately achieved.

[0064] The definitions of the core variables and the derivation of the formulas are as follows:

[0065] Quality deviation and deviation trend: defined at time... The first collection Each quality characteristic value is The corresponding process standard is Then at that time quality deviation Defined as: ; in length of Within the time window, the deviation sequences of all features constitute a deviation trend flow. Through the analysis of Statistical analysis is performed, and if the value exceeds a preset threshold, it is determined that there is a deviation trend.

[0066] Defect awareness impulse: A defect awareness impulse is generated when a deviation from the trend is detected. .pulse Defined as a quadruple: ;in, The pulse generation time; The extracted fluctuation feature vector; This is a snapshot of the device state at the time the pulse was generated; The confidence level represents the likelihood that a deviation from the trend will evolve into a real defect. The value increases with the degree of deviation exceeding the limit.

[0067] Predicting the actual time of defect outbreak: Production management system receives pulses Subsequently, the timing of defect outbreaks was extrapolated based on a time-series prediction model. .definition From the current moment Starting from, deviation First time exceeding the scrap threshold The moment: The prediction process uses a nonlinear time series model to extrapolate the future deviation trajectory.

[0068] Fuzzy-historical weighted compensation model: for generating the optimal compensation amount A fuzzy-historical weighted model was established. It is a weighted combination of real-time fuzzy matching and historical successful cases: ;in, This is the real-time membership value calculated based on a fuzzy rule base. For current features With historical case library Similarity to the best-matching cases; and These are fuzzy weights and historical weights, respectively, satisfying... .

[0069] Defect pulse attenuation rate: Defined during the collaborative compensation process. Used to quantify the effect of regulation. The decay rate characterizing pulse intensity: ;in, To adjust the initial pulse confidence level, This represents the confidence level at the current moment. A higher decay rate indicates more effective regulation.

[0070] The specific embodiments of the present invention will be described in detail below.

[0071] Example 1:

[0072] Please see Figure 1 The diagram illustrates the overall flowchart of an online quality control method for valve fittings provided by an embodiment of the present invention. The method includes:

[0073] S100: Quality data stream deviation trend identification and defect awareness pulse generation stage;

[0074] S200: Defect outbreak timing prediction, path reverse mapping and fuzzy-history weighted compensation stage;

[0075] S300: Compensation effect monitoring and progressive collaborative compensation triggering stage;

[0076] S400: Collaborative compensation process verification and circuit breaker rollback control phase;

[0077] S500: Steady-state recovery confirmation and adaptive control case library update phase.

[0078] The specific plan is as follows:

[0079] In an online quality control method for valve components, S100 achieves the transformation from raw physical quantity measurement to defect awareness pulse generation through intelligent detection terminals and edge computing gateways deployed at key workstations on the production line.

[0080] Please refer to Figure 2 It illustrates a flowchart of step S100 in an exemplary online quality control method for valve fittings according to this application, the contents of which include:

[0081] S110: In this invention, high-precision acquisition equipment is deployed at various key quality control points in the automated production line for valve components. Key locations include, but are not limited to: laser measuring instruments at the discharge port of CNC lathes, inter-process infrared probes in multi-axis machining centers, and three-dimensional vision scanning systems at the final inspection station for finished products.

[0082] In one possible implementation, the content collected includes:

[0083] For the dimensional inspection station, continuous measurements of valve body diameter, flange thickness, and sealing surface roughness are collected, with the sampling frequency set to once per piece or high-frequency continuous sampling.

[0084] For the surface quality station, surface texture image data and diffraction spot characteristics are collected.

[0085] For machining equipment, the real-time spindle speed, feed axis ratio, tool wear compensation value, hydraulic system pressure, and motor load current waveform are collected through the PLC interface.

[0086] In one possible implementation, the detection node has a built-in edge computing module that performs analog-to-digital conversion, filtering, noise reduction, and outlier removal on the original analog signal to ensure that the data uploaded to the production management system has high reliability.

[0087] S120: In this invention, the edge computing module of each detection station maintains a fixed length. A sliding time window that stores the most recent data. One workpiece or the most recent The module continuously calculates the statistical characteristics of the data within a given time frame to determine if any deviations from the trend exist.

[0088] In one possible implementation, the deviation trend identification logic includes:

[0089] Calculate the arithmetic mean of the quality data within the sliding window. and standard deviation .

[0090] If the current data point satisfy ,in The sensitivity coefficient is preset and appears continuously. If a point shows a unidirectional drift, it is determined that there is a deviation trend.

[0091] When using a linear regression algorithm to fit a time series of data within a window, if the absolute value of the slope of the fitted line is... If the drift threshold is exceeded and the intercept value is moving away from the center of the tolerance zone, it is determined that there is a drift trend.

[0092] In one possible implementation, the system employs a cumulative sum control chart algorithm to cumulatively monitor small process deviations. Once the cumulative sum exceeds a decision threshold, the system will take action. This triggers deviation trend identification, and the algorithm has extremely high sensitivity to low-frequency, slowly changing deviations.

[0093] S130: In this invention, once S120 determines that there is a deviation trend, the system immediately triggers the pulse generation mechanism to encapsulate the full abnormal information at the current moment.

[0094] In one possible implementation, the process of generating a defect awareness impulse includes:

[0095] Extracting wave feature vectors This includes the magnitude, direction, frequency of fluctuation, and acceleration of the deviation.

[0096] Read the current snapshot of the processing equipment status. The system obtains the current feed rate multiplier, spindle speed setting, tool number and remaining tool life percentage, coolant temperature and flow rate via industrial Ethernet or fieldbus.

[0097] Calculate confidence level The confidence level is normalized based on the degree to which the deviation exceeds the control limits and the value of the cumulative sum statistic. For example, if the deviation reaches 90% of the tolerance band, then... If the deviation reaches the scrap limit, then .

[0098] The above information is packaged to generate a defect awareness pulse. The data is then tagged with a timestamp and workstation ID, and uploaded in real time to the central processing engine of the production management system via a message queue.

[0099] In an online quality control method for valve fittings, S200 receives defect awareness pulses as input and generates precise fine-tuning compensation instructions through prediction, mapping, and weighted calculation.

[0100] Please refer to Figure 3 It illustrates a flowchart of step S200 in an exemplary online quality control method for valve fittings according to this application, the contents of which include:

[0101] S210: In this invention, after receiving a pulse, the production management system first analyzes the urgency of the defect development and predicts when the quality will deteriorate to an unacceptable scrap level if no intervention is made.

[0102] In one possible implementation, the execution of the time series prediction model includes:

[0103] Retrieve from historical databases the current pulse Historical data with similar characteristics, including deviation evolution curves under similar operating conditions.

[0104] The deviation sequence within the current sliding window Historical feature vectors are input into a pre-trained time series prediction model. Optionally, the prediction model may employ deep learning models such as Long Short-Term Memory networks or gated recurrent units, which can capture non-linear time dependencies.

[0105] Model output future Deviation prediction sequence for each workpiece cycle .

[0106] Traverse the prediction sequence and find the first one that satisfies the condition. index Then The predicted outbreak time point was determined to be the actual defect.

[0107] If the predicted sequence is in If the range is within the limit and the prediction curve shows a regression trend, then... Setting it to null indicates that the risk is controllable, and in-depth regulation can be temporarily suspended, with only routine monitoring performed.

[0108] S220: In this invention, based on the predicted outbreak time point, the physical location where the defect will occur is determined, and the process range that needs to be controlled is locked by tracing back from the current location.

[0109] In one possible implementation, the execution of reverse mapping of the process path includes:

[0110] Query the real-time process scheduling table and material flow tracking information in the Manufacturing Execution System to determine the process. A unique identifier for a workpiece that is always in the processing position.

[0111] Based on the workpiece ID, trace the digital process list of the workpiece to determine where... The specific process nodes that are currently being executed or are about to be executed. For example, if defects are predicted to occur during the finishing grinding process, then This is the precision grinding station.

[0112] From the current testing station nodes Starting from the process topology diagram Search results in The process path is defined in the process topology diagram. The process topology diagram predefines the sequential relationships of each process and the material transport path.

[0113] Considering the timing of processing, if If it is downstream of the current process, then the reverse path is used. Including the current process and All intermediate processes between; if If it is the current process, then the path only includes the current process.

[0114] Let the set of all process nodes on the reverse path be denoted as the node set to be compensated. This set is the target range for subsequent compensation commands.

[0115] S230: In this invention, for For each node, the optimal parameter adjustment amount is calculated by combining the fuzzy matching results of real-time data with historical success cases.

[0116] In one possible implementation, the calculation of fuzzy-history weighted compensation includes:

[0117] Current fluctuation characteristics The input is fed into the fuzzy logic controller. Fuzzy sets are defined for the input variables: deviation magnitude and deviation rate of change. A fuzzy set is also defined for the output variable: compensation strength. The initial compensation amount is obtained by defuzzifying the data by querying a predefined fuzzy rule base. .

[0118] Calculate the current characteristics from a database of historically successful regulatory cases. Similarity to feature vectors of historical cases. Similarity can be calculated using Euclidean distance or cosine similarity. Select the top-K cases with the highest similarity. Based on the actual compensation amounts used in these cases, calculate the historical reference compensation amount using a weighted average. .

[0119] The weighting coefficients are dynamically adjusted based on the stability of the operating conditions. If the operating parameters are in a stable range, the historical weights are increased. It relies on mature experience; if there are sudden changes in the operating conditions, the fuzzy weights are increased. To enhance real-time response.

[0120] Calculate the final compensation instruction: The instructions contain specific parameters to be adjusted, as well as the direction and value of adjustment.

[0121] S240: In this invention, the generated compensation command is sent to the corresponding machine tool controller or PLC to perform fine-tuning operations.

[0122] In one possible implementation, the strategy for fine-tuning execution includes:

[0123] For the feed parameters, a superimposed fine-tuning method is adopted, that is, increasing or decreasing the feed rate based on the original CNC program feed rate. Specify the percentage to avoid directly modifying the underlying CNC program.

[0124] For tool wear compensation, the value of the corresponding tool compensation register in the CNC system is directly modified.

[0125] The system records the time when the command is issued, the parameters included, and the expected correction effect, which serves as benchmark data for subsequent evaluation of the control effect.

[0126] In an online quality control method for valve fittings, S300 is responsible for monitoring the effect of local compensation and activating a multi-process collaborative compensation mechanism when the effect is insufficient.

[0127] Please refer to Figure 4It illustrates a flowchart of step S300 in an exemplary online quality control method for valve fittings according to this application, the contents of which include:

[0128] S310: In this invention, after the compensation instruction is executed, the production line continues to run, and subsequent inspection stations or the next inspection cycle of the same inspection station continue to collect the workpiece quality data after compensation processing.

[0129] In one possible implementation, the monitoring process includes:

[0130] Establish a new post-compensation monitoring window specifically for marking and tracking workpieces that have undergone compensation operations throughout their entire lifecycle.

[0131] Extract the quality characteristic data of these workpieces at key processes to form a post-compensation data stream. .

[0132] contrast Data stream before compensation execution Observe whether the deviation trend is alleviated, whether the mean of the exemplary data moves closer to the center line, and whether the variance of the fluctuation decreases.

[0133] S320: In this invention, the system uses strict judgment logic to confirm the effectiveness of local compensation in order to determine whether it is necessary to upgrade the control measures.

[0134] In one possible implementation, the determination logic includes:

[0135] If the data stream is subsequently compensated If the mean value returns to near the target center line, and the proportion of data points falling within the control limits returns to normal levels, and no new abnormal patterns appear, then the compensation is deemed effective, and the process can be switched to regular monitoring.

[0136] If the data stream is subsequently compensated If the fluctuation range decreases, but the mean still shows a significant systematic bias; or if the fluctuation range does not decrease significantly, or even shows a divergent trend, then it is determined that the quality data stream still has a deviation trend, that is, insufficient local compensation.

[0137] S330: In this invention, when local compensation is insufficient, it means that the fine-tuning of a single process is no longer able to correct the current quality deviation, and it is necessary to mobilize the power of multiple processes on the path for coordinated correction.

[0138] In one possible implementation, the execution of collaborative compensation includes:

[0139] Traverse the set of nodes to be compensated The system collects real-time status information of each node through an industrial network, including the remaining time of the current processing cycle, equipment load rate, current tool life, and the adjustable range of the process parameters.

[0140] Based on a pre-defined progressive priority control protocol, the control priority of each node is calculated. The priority calculation comprehensively considers the impact of the process on the final defect, the flexibility of parameter adjustment, and the real-time load of the current node.

[0141] Coordinated compensation instructions are generated sequentially for each node, from highest to lowest priority. High-priority nodes handle the main corrections, while low-priority nodes handle minor adjustments, ensuring a smooth distribution of the correction process across the entire process chain.

[0142] Perform coordinated compensation by simultaneously or in time-sharing the parameters to the controllers of all relevant nodes via the network.

[0143] S340: In this invention, collaborative compensation is not merely a simple superposition of parameters, but also involves the rebalancing of process capabilities to ensure optimal coordination between processes.

[0144] In one possible implementation, process rebalancing includes:

[0145] When adjusting the processing parameters of a certain process, simultaneously assess its impact on the processing allowance of subsequent processes.

[0146] If a certain process increases the cutting amount, the allowance of the subsequent finishing process will be automatically reduced to maintain the balance of the total process allowance and avoid the process from being overloaded due to excessive allowance or the subsequent process from being undercut due to insufficient allowance.

[0147] By using a global process parameter optimization algorithm, the energy consumption or processing time along the entire path is minimized while meeting the final quality constraints, thereby achieving process rebalancing and improving overall production efficiency.

[0148] In an online quality control method for valve fittings, S400 ensures the safety of the collaborative compensation process by preventing erroneous control from causing production accidents through real-time verification and a circuit breaker mechanism.

[0149] Please refer to Figure 5 It illustrates a flowchart of step S400 in an exemplary online quality control method for valve fittings according to this application, the contents of which include:

[0150] S410: In this invention, while performing collaborative compensation, the system runs a monitoring thread in parallel to analyze the dynamic response after parameter adjustment in real time.

[0151] In one possible implementation, verification of the parameter convergence trend includes:

[0152] Real-time plotting of time response curves for key control parameters.

[0153] Calculate the rate of change of the response curve. If the absolute value of the rate of change gradually decreases and approaches 0, and the curve gradually approaches the target value, then the parameter is considered to be converging.

[0154] Set a convergence threshold and a convergence window. If the change in the values ​​of several consecutive sampling points is less than the preset dead zone threshold and the values ​​remain within the target tolerance zone, then the convergence state is confirmed.

[0155] S420: In this invention, in addition to focusing on the convergence of the quality value, it is also necessary to focus on the change in the intensity of the defect awareness pulse that characterizes the abnormal signal.

[0156] In one possible implementation, verification of the defective pulse attenuation rate includes:

[0157] Continuously generate temporary defect awareness pulses based on current real-time data. And extract its confidence level. .

[0158] Using formula Calculate the attenuation rate.

[0159] like A value consistently greater than 0, and gradually increasing over time, indicates that the defect pulse is rapidly decaying, the abnormal signal is weakening, and the control direction is correct and effective.

[0160] S430: In this invention, a strict multi-verification mechanism is set to eliminate occasional interference and ensure the rigor of the circuit breaker mechanism activation.

[0161] In one possible implementation, the determination of verification failure includes:

[0162] Define a verification cycle, for example, a comprehensive evaluation is performed every 5 workpiece cycles or every minute.

[0163] If, in a certain evaluation, the parameter divergence or the defect pulse decay rate is lower than the set lower threshold, it is recorded as a verification failure.

[0164] The system maintains a consecutive failure counter. If two consecutive verification results are both failures, the current collaborative compensation strategy is determined to be unsuitable for the current working conditions, and the collaborative compensation is deemed to have failed.

[0165] The requirement of two consecutive measurements is to prevent misjudgments caused by noise from a single measurement, environmental vibration, or momentary external interference, and to ensure that the protection mechanism is triggered only after confirmation of invalidity.

[0166] S440: Trigger the circuit breaker rollback mechanism and re-execute the collaborative compensation until the verification is successful.

[0167] In this invention, once a failure is determined, a protective circuit breaker is immediately activated to prevent erroneous parameters from continuing to act on the production line, causing batch scrap or equipment damage.

[0168] In one possible implementation, the circuit breaker rollback mechanism includes:

[0169] Immediately stop the output of all currently executing collaborative compensation commands.

[0170] An emergency rollback command is issued to the relevant equipment to forcibly restore all modified processing parameters to their original values ​​before adjustment or the system's default safe values.

[0171] After the rollback is completed, the system enters a short cooling or stabilization period to wait for the production status to stabilize and logs the failure.

[0172] Based on the experience of this failure, the system re-analyzes the deviation trend, adjusts the weight coefficients in the fuzzy-historical weighted principle, or selects different combinations of collaborative compensation strategies, restarts the control process from S200 or S300, until it is verified and forms a cyclical optimization.

[0173] In an online quality control method for valve fittings, the S500 performs knowledge accumulation and feedback after successful control, forming a complete optimization.

[0174] Please refer to Figure 6 It illustrates a flowchart of step S500 in an exemplary online quality control method for valve fittings according to this application, the contents of which include:

[0175] S510: In this invention, the control is confirmed to be truly successful only when the quality data stream remains stable for a sufficiently long period of time, thus avoiding misjudgment due to short-term fluctuations.

[0176] In one possible implementation, the steady-state determination includes:

[0177] Set a steady-state range, which is usually a tighter narrower range of the target tolerance zone, to ensure that the product quality has sufficient margin.

[0178] Monitor the quality data stream in real time and check whether all data points fall within the steady-state range.

[0179] Start the steady-state duration counter. Whenever a workpiece's data falls into the steady-state range, the counter increments by 1. If a data point exceeds the range, the counter is reset to zero and counting restarts.

[0180] Only when the counter value reaches 3 or above is it confirmed that the quality data stream has returned to a steady state, and the control is considered successful.

[0181] S520: In this invention, each successful adjustment is a valuable experience, which the system transforms into part of the knowledge base for future reuse.

[0182] In one possible implementation, the storage of adaptive control cases includes:

[0183] Construct a feature index for the case, including the original defect awareness impulse feature vector. Deviation type, sequence of process nodes involved .

[0184] Record the regulatory decisions adopted this time, including the allocation values ​​of fuzzy weights and historical weights. , Specific compensation amounts for each process Priority sequence of collaborative compensation and process rebalancing strategy.

[0185] Record the evaluation indicators of the control effect, such as control response time, overshoot, and steady-state error.

[0186] The structured information is stored in a database of successful historical control cases and a multi-dimensional index is created for quick access by the fuzzy-historical weighted compensation principle in the future.

[0187] S530: In this invention, the system automatically summarizes the basic information of the quality event, including the time of occurrence, the equipment involved, and the type of defect. It records a complete timeline of the handling process, including the pulse generation time, the compensation command issuance time, the collaborative compensation initiation time, and the steady-state recovery time. It generates a comparison chart of quality data before and after the event, intuitively displaying the control effect, and includes equipment status analysis.

[0188] The report is converted into PDF, HTML, or web page format and pushed to the digital dashboard of the production management system and the mobile terminals of relevant managers and maintenance personnel through a messaging middleware to complete the control of this quality incident.

[0189] Example 2:

[0190] This invention was deployed and validated over a six-month period in an automated production workshop for ball valve components. This workshop primarily produces valve bodies and discs for gate valves used in oil pipelines, and the production line includes five precision CNC lathes, three vertical machining centers, and two automated assembly lines. Key dimensions involved include valve body bore diameter tolerances. 0.02mm, sealing surface roughness 0.4 and coaxiality 0.01mm. The monitoring range covers more than 800 quality data points and 500 equipment status points.

[0191] During implementation, the following specific configuration was adopted, and the specific parameter configuration table is shown in Table 1:

[0192] Table 1 System Implementation Configuration Parameters

[0193]

[0194] During the trial operation, the system successfully handled several complex quality fluctuation events, as detailed in the following examples:

[0195] Case 1: Drift in inner hole dimensions caused by tool wear;

[0196] During the machining process of the valve body's inner bore, the uneven peeling of the tool coating causes the inner bore diameter to show a monotonically increasing trend.

[0197] In step S100, a dimensional deviation exceeding 3 was detected at the 15th workpiece. Control limits, generating defect awareness impulses, confidence level It is 0.85.

[0198] After receiving the pulse in step S200, the LSTM model predicts that without intervention, the size will exceed the scrap limit at the 22nd workpiece. The system uses the finishing process as the endpoint and reverse mapping reveals that both roughing and finishing processes are on the path.

[0199] Since similar tool wear patterns occur frequently in the historical case database, the fuzzy-history weighted model will... Dynamically adjust to 0.8. Based on historical successful cases, it is recommended to reserve a 0.01mm allowance in the roughing process and perform tool compensation fine-tuning in the finishing process.

[0200] After performing local compensation, the S300 monitoring system detected that although the subsequent workpiece dimensions did not deteriorate, they did not completely return to the center value, indicating insufficient local compensation. The system then initiated collaborative compensation, slightly reducing the feed rate of the roughing operation to minimize cutting force deformation.

[0201] S400 real-time verification showed that the size rapidly converged to the centerline, and the defect pulse attenuation rate... Achieving 80% means passing the verification on the first try.

[0202] After S500 confirms that five consecutive products are qualified, the wear compensation strategy for this specific tool coating will be stored in the case library. Specific data comparisons are shown in Table 2.

[0203] Table 2 Comparison of data before and after adjustment in Case 1 (tool wear)

[0204]

[0205] Case 2: Coaxiality error caused by thermal deformation;

[0206] During the high-temperature period in summer, due to changes in ambient temperature, the spindle of the machining center undergoes slight thermal expansion, causing the valve disc coaxiality to regularly exceed the tolerance during the afternoon.

[0207] The S100 identified fluctuations in coaxiality data and temperature sensor data that showed a strong correlation.

[0208] The S200 predicts the defect outbreak time to be 40 minutes later. The reverse mapping path only includes the final finishing process.

[0209] Due to the nonlinear nature of thermal deformation, fuzzy matching becomes dominant. Adjusted to 0.7. The system generates time-segmented linear thermal compensation coefficients based on fuzzy rules.

[0210] After the collaborative compensation was implemented, the S400, during its first verification, found that although the parameters began to converge, the convergence speed was extremely slow, and the defect pulse decay rate was high. If the value is below the threshold, it is recorded as a first verification failure. The system did not trigger the circuit breaker, but minor adjustments were made.

[0211] During the second verification, a sudden drop in ambient temperature caused the coaxiality data to fluctuate in the opposite direction, disrupting the convergence trend and resulting in verification failure. Two consecutive failures triggered the circuit breaker mechanism, resetting all compensation parameters to zero.

[0212] The system was reinitialized, and the latest temperature and dimensional data were reacquired. The regenerated compensation strategy incorporated a negative feedback term for the rate of temperature change. After re-executing the collaborative compensation, the coaxiality quickly stabilized, and continuous verification was successful.

[0213] Case 3: Abnormal waviness of sealing surface in multi-process coupling;

[0214] A certain batch of raw materials had a higher hardness, which caused high-frequency chatter marks to appear when processing the sealing surface, affecting the sealing performance.

[0215] S100 detected abnormal fluctuations in the roughness Ra value, accompanied by high-frequency oscillations in the spindle load current.

[0216] S200 predictions will lead to batch seal failures. The reverse mapping path covers four processes from rough milling to finish grinding.

[0217] This is a typical multi-process coupling problem. S300 determined that adjusting the local fine grinding process could not eliminate the vibration marks left by rough milling.

[0218] The system initiates full-path collaborative compensation. Following a progressive priority, the cutting depth of the rough milling operation is first reduced to decrease the excitation force; secondly, the rotational speed of the finish milling operation is adjusted to avoid the resonant frequency; and finally, the feed rate of the finish grinding operation is optimized for smoothing.

[0219] S400 was closely monitored during the migration process. Although the initial adjustment of rough milling parameters led to a decrease in efficiency, S500 ultimately confirmed that the product qualification rate increased from 60% to 99%. Specific parameter adjustments for each process are shown in Table 3.

[0220] Table 3. Case Study 3 (Multi-Process Coupling) Collaborative Compensation Strategy Table

[0221]

[0222] Implementation Results Statistics:

[0223] After six months of continuous operation and data analysis, the method of this invention has achieved significant technical results on the valve parts production line. Detailed statistical data is shown in Table 4:

[0224] Table 4. Overall System Performance Statistics

[0225]

[0226] As can be seen from the above data, the online quality control method for valve accessories provided by this invention can significantly improve the response speed and processing accuracy of quality anomalies. Through intelligent collaborative compensation and a safe fuse mechanism, it can greatly improve product yield and the adaptability of the production line while ensuring production safety.

[0227] Example 3:

[0228] An online quality control system for valve fittings includes:

[0229] The defect awareness pulse generation module is used to automatically generate a defect awareness pulse containing fluctuation characteristics and equipment status when any inspection station on the automated production line of valve parts identifies a deviation trend in the quality data stream.

[0230] The prediction and compensation instruction generation module is used to, after receiving the defect awareness pulse, deduce the time point when the defect awareness pulse will explode into a real defect in the future workpiece cycle based on the time series prediction model; with the time point of the real defect as the endpoint, reverse map the process path that the defect awareness pulse will pass through; generate compensation instructions on the process path according to the fuzzy-history weighted compensation principle, and fine-tune the processing parameters to suppress defect formation.

[0231] The collaborative compensation module is used to monitor the quality data stream of the workpiece after compensation. If the quality data stream still shows a deviation trend, it is determined that the local compensation is insufficient. Based on the process path and processing node status to be traversed by the compensation instruction, collaborative compensation is performed.

[0232] The safety control and circuit breaker module is used to verify the convergence trend of parameters and the attenuation rate of defect pulses in real time during the collaborative compensation process. If two consecutive verifications fail, the current collaborative compensation is determined to be invalid, triggering the circuit breaker rollback mechanism to re-execute the collaborative compensation until the verification is successful.

[0233] The adaptive learning and feedback module is used to automatically save the parameters, paths and collaborative strategies of this control process as an adaptive control case when the quality data stream recovers to the steady state range and continues for no less than three workpiece cycles. At the same time, it pushes the processing report of this quality event to the production management system.

[0234] Those skilled in the art will understand that the embodiments of this application are provided as methods, systems, or computer program products. Therefore, this application takes the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application takes the form of a computer program product implemented on one or more computer storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer program code. The solutions in the embodiments of this application are implemented using various computer languages, exemplified by the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0235] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, are implemented by computer program instructions. These computer program instructions are provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart illustrations and / or block diagrams.

[0236] These computer program instructions are also stored in a computer read-memory that can direct a computer or other programmed data processing device to operate in a particular manner, such that the instructions stored in the computer read-memory produce an article of manufacture including instruction means that implement the functions specified in the flowchart or multiple flowcharts and / or block diagram blocks or multiple block diagrams.

[0237] These computer program instructions are also loaded onto a computer or other programming data processing device to cause a series of operational steps to be performed on the computer or other programming device to produce a computer-implemented process, such that the instructions, which execute on the computer or other programming device, provide steps for implementing the functions specified in the flowchart flow or multiple flows and / or the block diagram blocks or multiple blocks.

[0238] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0239] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method of online quality control of valve fittings, characterized in that, include: When any inspection station on the automated valve parts production line detects a deviation trend in the quality data stream, it automatically generates a defect awareness pulse that includes fluctuation characteristics and equipment status. After receiving a defect awareness pulse, the production management system uses a time-series prediction model to predict the time when the defect awareness pulse will erupt into a real defect in the future workpiece cycle. Using the time of the real defect as the endpoint, it reverse maps the process path that the defect awareness pulse will pass through. On the process path, it generates compensation instructions according to the fuzzy-history weighted compensation principle to fine-tune the processing parameters to suppress defect formation. Monitor the quality data stream of the workpiece after compensation. If the quality data stream still shows a deviation trend, it is determined that the local compensation is insufficient. Perform collaborative compensation according to the process path and processing node status to be traversed by the compensation instruction. During the collaborative compensation process, the convergence trend of parameters and the attenuation rate of defect pulses are verified in real time. If two consecutive verifications fail, the current collaborative compensation is deemed invalid, triggering the circuit breaker rollback mechanism to re-execute the collaborative compensation until the verification passes. When the quality data stream recovers to the steady-state range and continues for no less than three workpiece cycles, the parameters, paths and collaborative strategies of this control process are automatically saved as an adaptive control case. At the same time, a processing report of this quality event is pushed to the production management system. The defect awareness pulse generation step includes: The deployment of multi-source heterogeneous detection nodes and standardized data collection maintains fixed-length sliding time windows through edge computing modules of various detection stations , and linear regression algorithms are used to fit the time series of data within the windows or cumulative sum control chart algorithms are used to identify deviation trends of quality data streams. If a deviation trend is detected, the pulse generation mechanism is triggered to extract the amplitude, direction, frequency and acceleration of the fluctuation as the fluctuation feature vector. The feed rate multiplier, spindle speed setpoint and remaining tool life percentage are obtained through industrial Ethernet as a snapshot of the equipment status. Based on the degree to which the deviation exceeds the control limit, the deviation degree is normalized and mapped to a confidence level, and the pulse generation time, fluctuation feature vector, equipment status snapshot and confidence level are encapsulated into a defect awareness pulse with a four-tuple structure. The compensation instruction generation step includes: Query the real-time process scheduling table and material flow tracking information in the manufacturing execution system to determine the unique identification code of the workpiece at the processing position at the actual defect outbreak time, and trace the digital process list based on the workpiece's unique identification code to determine the process node corresponding to the actual defect outbreak time. Starting from the current inspection station node in the process topology diagram, a reverse search is performed to reach the process node corresponding to the actual defect outbreak time point. All process nodes on the reverse path are determined as the set of nodes to be compensated. The current fluctuation feature vector is input into the fuzzy logic controller. The preliminary compensation amount is obtained by querying the predefined fuzzy rule base and defuzzifying. At the same time, the similarity between the current fluctuation feature vector and the best matching case in the historical successful control case base is calculated to obtain the historical reference compensation amount. The fuzzy weights and historical weights are dynamically adjusted based on the stability of the operating conditions, and the final compensation amount is calculated using the following formula: ;in, This is the real-time membership value calculated based on a fuzzy rule base. For current features With historical case library Similarity to the best-matching cases; and These are fuzzy weights and historical weights, respectively, satisfying... .

2. The online quality control method for valve fittings according to claim 1, characterized in that, The steps for confirming the time point of the actual defect include: The production management system retrieves historical time period data similar to the current defect awareness pulse characteristics from the historical database, inputs the deviation sequence within the sliding window into a pre-trained long short-term memory network or gated recurrent unit deep learning model, and outputs the deviation prediction sequence for future workpiece cycles. Traverse the deviation prediction sequence, find the first index that satisfies the deviation amount being greater than or equal to the scrap threshold, and add the number of cycles corresponding to the index to the current time to determine the predicted outbreak time of the actual defect. If the deviation prediction sequence does not exceed the prediction range and the prediction curve shows a regression trend, then the predicted outbreak time of the actual defect will be set to a null value and deep regulation will be temporarily suspended.

3. The online quality control method for valve fittings according to claim 1, characterized in that, The determination and generation of the collaborative compensation includes: Establish a post-compensation monitoring window, extract key process quality characteristic data of the workpiece after compensation operation to form a post-compensation data stream, and compare the post-compensation data stream with the data stream before compensation. If the mean of the post-compensation data stream does not return to the vicinity of the target centerline or the fluctuation amplitude does not decrease significantly, it is determined that the quality data stream still has a deviation trend. Traverse the set of nodes to be compensated and collect real-time status information of each node, including the remaining time of the processing cycle, equipment load rate and the adjustable range of process parameters; Based on the preset progressive priority control protocol, the control priority of each node is calculated by comprehensively considering the influence of the process on the final defect, the flexibility of parameter adjustment, and the real-time load of the current node. The collaborative compensation instructions of each node are generated in order of control priority from high to low.

4. The online quality control method for valve fittings according to claim 3, characterized in that, The execution coordination compensation also includes parameter coordination correction and process rebalancing according to a progressive priority control protocol: When adjusting the processing parameters of a certain process, the impact on the processing allowance of subsequent processes is evaluated simultaneously. If the cutting amount of a certain process is increased, the allowance of the subsequent finishing process is automatically reduced to maintain the balance of the total process allowance and avoid the process from being overloaded due to excessive allowance or the subsequent process from being undercut due to insufficient allowance. By using a global process parameter optimization algorithm, the energy consumption or processing time on the path where the node set to be compensated is located is minimized while meeting the final quality constraints.

5. The online quality control method for valve fittings according to claim 1, characterized in that, The method for simultaneously verifying the parameter convergence trend and defect pulse attenuation rate is as follows: The monitoring thread runs in parallel to plot the time response curves of key control parameters in real time and calculate the rate of change of the response curves. If the absolute value of the rate of change gradually decreases and approaches 0, and the curve gradually approaches the target value, then the parameter is determined to show a convergence trend. Temporary defect awareness pulses are continuously generated based on real-time data, and the current confidence level is extracted. The defect pulse decay rate is calculated using the following formula: ;in, Represents the defect pulse attenuation rate. To adjust the initial pulse confidence level, The confidence level at the current moment; If the defect pulse decay rate remains greater than 0 and the value gradually increases over time, it indicates that the defect pulse is decaying rapidly and the control direction is correct.

6. The online quality control method for valve fittings according to claim 5, characterized in that, The circuit breaker rollback mechanism is triggered as follows: Set a verification cycle. If the parameters diverge or the defect pulse decay rate falls below the lower limit threshold in a certain evaluation, it is recorded as a verification failure. Maintain a consecutive failure counter. If two consecutive verification results are both failures, the current collaborative compensation strategy is deemed unsuitable for the current working conditions, and the collaborative compensation is deemed to have failed. Immediately stop the output of all currently executing collaborative compensation commands, issue emergency rollback commands to relevant equipment, and force all modified processing parameters to be restored to their original values ​​before adjustment or the system's default safe values; Based on the experience gained from this failure, adjust the fuzzy weights and historical weight coefficients, or select a different combination of collaborative compensation strategies and re-execute the control process starting from generating compensation instructions until verification is successful.

7. The online quality control method for valve fittings according to claim 1, characterized in that, The adaptive control case generation steps include: Construct a feature index for the case, which includes the original defect awareness impulse feature vector, deviation type, and the sequence of process nodes involved; Record the control decisions adopted, including the allocation values ​​of fuzzy weights and historical weights, the specific compensation amount of each process, the priority sequence of collaborative compensation, and the process rebalancing strategy; Record the evaluation indicators of the regulation effect, including regulation response time, overshoot and steady-state error, and store them in the historical successful regulation case library and establish a multi-dimensional index.

8. An online quality control system for valve fittings for performing the method as described in any one of claims 1-7, characterized in that, The system includes: The defect awareness pulse generation module is used to automatically generate a defect awareness pulse containing fluctuation characteristics and equipment status when any inspection station on the automated production line of valve parts identifies a deviation trend in the quality data stream. The prediction and compensation instruction generation module is used to, after receiving the defect awareness pulse, deduce the time point when the defect awareness pulse will explode into a real defect in the future workpiece cycle based on the time series prediction model; with the time point of the real defect as the endpoint, reverse map the process path that the defect awareness pulse will pass through; generate compensation instructions on the process path according to the fuzzy-history weighted compensation principle, and fine-tune the processing parameters to suppress defect formation. The collaborative compensation module is used to monitor the quality data stream of the workpiece after compensation. If the quality data stream still shows a deviation trend, it is determined that the local compensation is insufficient. Based on the process path and processing node status to be traversed by the compensation instruction, collaborative compensation is performed. The safety control and fuse module is used to verify the parameter convergence trend and defect pulse decay rate in real time during the collaborative compensation process. If two consecutive verifications fail, the current collaborative compensation is deemed invalid, triggering the circuit breaker rollback mechanism to re-execute the collaborative compensation until the verification passes. The adaptive learning and feedback module is used to automatically save the parameters, paths and collaborative strategies of this control process as an adaptive control case when the quality data stream recovers to the steady state range and continues for no less than three workpiece cycles. At the same time, it pushes the processing report of this quality event to the production management system.

Images