Parameter optimization method and system in valve production

Abstract

The invention discloses a parameter optimization method and system in valve production, relates to the technical field of valve production processes, and aims to solve the problem of inaccurate optimization results in a parameter optimization process. By integrating process knowledge, the model can consider special process requirements and constraint conditions in the production process, disjunction of theory and practice is avoided, target analysis is carried out through conflict scene and collaborative scene recognition, contradictions and conformance relationships among different targets can be sensitively perceived, mild, moderate and severe anomaly levels are divided according to reasons, and the method is suitable for large-scale popularization and application. Differentiation processing strategies such as abnormal prompt, intervention prompt and automatic shutdown are matched, prompt can be given to guide production optimization in the case of light abnormity, and quick response can be made to guarantee production safety in the case of heavy abnormity.

Description

Technical Field

[0001] The present invention relates to the technical field of valve production technology, and in particular to a method and system for optimizing parameters in valve production. Background Art

[0002] The valve production process refers to the process of making raw materials through a series of processing, assembly, inspection and other processes to finally make valves that meet design requirements and quality standards.

[0003] Chinese patent publication number CN118552339B discloses a valve production parameter optimization method and system based on product requirements. The method mainly assigns weights to predetermined performance parameters based on the target application scenario of the target valve to obtain target weight assignment; screens predetermined performance parameters based on the target weight assignment to obtain key performance parameters; extracts a first historical record from a historical valve production parameter record, the first historical record including first historical production parameter information and first historical performance parameter information; uses the first historical key performance parameter information in a first traversal result obtained by traversing the first historical performance parameter information based on the key performance parameter as a dependent variable; uses the first historical production parameter information as an independent variable and performs a sensitivity analysis with the dependent variable to obtain a first sensitivity analysis result, and determines the key production parameters based on the first sensitivity analysis result; performs a global optimization on the key production parameters to obtain an optimal key production parameter strategy, and produces the target valve based on the optimal key production parameter strategy. Although the above patent solves the problem of production parameter optimization, the following problems still exist in actual operation:

[0004] 1. Before acquiring data on production parameters, each sensor was not effectively tested, and no targeted correlation analysis was performed on the data of each parameter, resulting in an inability to understand the close performance between the parameters.

[0005] 2. There is no targeted optimization target for obtaining production parameters and no optimization model is built, and there is no dynamic verification based on the optimization results, which leads to incomplete final optimization results.

[0006] 3. There is no effective real-time monitoring of the verification process, and no targeted abnormal analysis and warning level judgment of the monitoring results, which leads to the inability to handle abnormal situations in a timely manner. Summary of the Invention

[0007] The purpose of the present invention is to provide a parameter optimization method and system in valve production. By integrating process knowledge, the model can take into account the special process requirements and constraints in the production process, avoid the disconnection between theory and practice, and perform target analysis through conflict scenario and collaborative scenario identification. It can keenly perceive the contradictions and fit relationships between different goals, divide the abnormality levels into mild, moderate and severe levels according to the cause, match differentiated processing strategies such as abnormal prompts, intervention prompts and automatic shutdowns, and can give prompts to guide production optimization in the case of mild abnormalities, and can quickly respond to ensure production safety in the case of severe abnormalities, which can solve the problems in the existing technology.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A method for optimizing parameters in valve production, comprising:

[0010] First, the performance of each sensor in valve production is tested, and multi-source data is collected using the sensor. The collected multi-source data is preprocessed. After the data preprocessing is completed, the correlation analysis is performed on each parameter. Based on the results of the correlation analysis, multi-dimensional optimization targets are established, and an optimization model is constructed based on the established multi-dimensional optimization targets. Dynamic verification is performed based on the constructed optimization model.

[0011] Preferably, each sensor in the valve production is subjected to performance testing, including:

[0012] Sensors include temperature sensors, pressure sensors, flow sensors, displacement sensors, vision sensors, humidity sensors and liquid level sensors;

[0013] Then perform performance testing on each sensor. The performance testing standards for the sensors are:

[0014] The temperature sensor's error is within the rated accuracy range, and its response time meets the preset requirements. The pressure sensor's linear error is less than ±1% FS, and its hysteresis error is less than ±0.5% FS. The flow sensor's full-scale error is within the preset range, and its dynamic fluctuation is less than ±2%. The displacement sensor's linear error is less than ±0.1% FS. The visual sensor's image is distortion-free. The humidity sensor's full-scale error is within the preset range. The liquid level sensor's liquid level detection error is less than ±2mm, and its dynamic response time is less than 2 seconds.

[0015] Calibrate and adjust sensors according to performance test standards;

[0016] Finally, the performance test of the sensor is completed.

[0017] Preferably, using sensors to collect multi-source data includes:

[0018] First, set the acquisition parameters of the sensor, including the acquisition frequency, range and data unit;

[0019] Then synchronize the multi-source data acquisition scheme, including establishing a unified time base and data acquisition trigger conditions, where the data acquisition trigger conditions include periodic acquisition and time-triggered acquisition;

[0020] The sensor collects data according to the collection plan;

[0021] After data collection is completed, a transmission link is established through wireless transmission. After the transmission link is established, the data collected by each sensor is standardized in data format;

[0022] After the data format is unified, multi-source data of each sensor is obtained.

[0023] Preferably, the collected multi-source data is subjected to data preprocessing, including:

[0024] Perform data cleaning on the collected multi-source data. Data cleaning involves identifying and processing missing data in the multi-source data, and then identifying and correcting outliers in the multi-source data.

[0025] After data cleaning, data conversion is performed to convert the data into the unit and format of unified multi-source data;

[0026] After data conversion, data integration is performed. Data integration involves establishing data association identifiers for multi-source data and handling conflicting data. The association identifier is established to generate a unique identifier for each valve artifact in the multi-source data. At the same time, the data from different sensors in the same time period are arranged in chronological order.

[0027] After data integration, data reduction is performed. Data reduction is to aggregate multi-source data according to timestamps and eliminate redundant data.

[0028] Finally, the data preprocessing of multi-source data is completed.

[0029] Preferably, after data preprocessing is completed, correlation analysis is performed on each parameter, including:

[0030] Screening the analysis parameters in the pre-processed multi-source data, including process parameters, technological parameters and output parameters;

[0031] The screened analysis parameters are grouped according to the process stage, where the process stage is divided into parameter groups according to the valve production process, which includes casting, machining, welding, assembly and testing;

[0032] After the process stages are grouped, the parameter data of the same workpiece at each process stage are concatenated according to the timestamp. After the concatenation is completed, a complete production record of the parameters, time and results in the same batch is obtained;

[0033] Perform correlation analysis on the complete production records using visualization tools, including drawing scatter plots and heat maps. Analyze the changing trends in the scatter plots and the correlation strength between parameters in the heat map.

[0034] After the correlation analysis is completed, the parameter combination characteristics under different production conditions are identified, and the parameter distribution of each group is compared based on the parameter combination characteristics;

[0035] According to the parameter distribution, the output results of the valve production parameters are grouped into groups, including a high qualified rate group and a low qualified rate group;

[0036] The output results are grouped and correlated, including a key parameter list, an impact path diagram, and a pattern library. The key parameter list is the parameter with the greatest impact on the target; the impact path diagram is the causal relationship between parameters; and the pattern library is the parameter combination corresponding to different output results.

[0037] Finally, the correlation analysis data between each parameter in the multi-source data is obtained.

[0038] Preferably, a multi-dimensional optimization target is established based on the results of the correlation analysis, including:

[0039] Before establishing multi-dimensional optimization goals, a multi-dimensional goal system should be established first. The multi-dimensional goal system includes quality dimension, efficiency dimension, cost dimension and safety dimension.

[0040] Then set the optimization target value, including setting it based on historical data, industry standards and parameter adjustable range;

[0041] Target identification is performed based on a multi-dimensional target system and optimized target values, including conflict scenario identification and collaborative scenario identification;

[0042] Construct a target association graph based on the identified targets;

[0043] Finally, the establishment of multi-dimensional optimization goals is completed.

[0044] Preferably, the optimization model is constructed according to the established multi-dimensional optimization objectives, including:

[0045] Before building the optimization model, confirm the input and output variables of the model;

[0046] Among them, the parameters that have the greatest impact on the target in the correlation analysis are used as input variables; the quantitative indicators in the multi-dimensional optimization target system are used as output variables;

[0047] The basic model is selected according to the characteristics of the multi-dimensional optimization objectives. The multi-dimensional optimization objectives include single-objective optimization and multi-objective optimization. Among them, the basic model is the linear regression model for single-objective optimization and the multi-objective genetic algorithm model for multi-objective optimization.

[0048] Retrieve the production process knowledge model from the database and embed the production process knowledge model into the basic model respectively;

[0049] After embedding is completed, the basic model is built, including weight allocation, objective function setting and optimal solution search;

[0050] After the model is built, it is verified. Model verification involves using the model to predict parameter combinations in historical production and comparing them with the actual results. If the deviation exceeds 5%, check whether key parameters are missed.

[0051] After the model verification is completed and qualified, the construction of the optimization model is completed.

[0052] Preferably, dynamic verification is performed based on the constructed optimization model, including:

[0053] Before dynamic verification, the control group and test group in valve production are first confirmed. The control group uses the original production parameters as a benchmark reference; the test group uses the optimized parameter combination recommended by the model.

[0054] Then confirm the sample size of each group, where each group produces at least 200 valves;

[0055] A trial run was conducted based on the control group and the experimental group. The trial run was as follows:

[0056] S1: Enter the model recommended parameters in the test group equipment and lock the parameter modification permission;

[0057] S2: For the same batch of workpieces in the control group and the experimental group, sensor data and quality data are collected synchronously;

[0058] S3: Set the warning threshold, and automatically suspend production and issue an alarm when it is triggered;

[0059] After the trial run is completed, the data of the control group and the experimental group will be compared, and the difference analysis will be carried out based on the index comparison results;

[0060] Finally, the dynamic verification results of the optimization model are obtained based on the difference analysis results.

[0061] A valve production parameter optimization system, comprising:

[0062] Dynamic verification data monitoring unit, used to:

[0063] Before dynamic verification, build a real-time monitoring system architecture, including sensor deployment and edge computing device deployment;

[0064] After the real-time monitoring system architecture is built, confirm the real-time monitoring indicators, which include input parameter indicators and output result indicators;

[0065] After the real-time monitoring indicators are confirmed, the collection frequency of dynamic verification data is confirmed, including high-frequency collection, medium-frequency collection and low-frequency collection;

[0066] After dynamic verification data collection is completed, data synchronization and calibration are performed;

[0067] Finally, the real-time monitoring of dynamic verification data is completed.

[0068] Preferably, including:

[0069] The monitoring data adjustment early warning unit is used to:

[0070] First, define abnormal data, which includes data where parameters deviate from the model's recommended values, data where indicators fail to meet standards, and data where safety parameters exceed the bottom line.

[0071] According to the defined abnormal data, the abnormal data in the dynamic verification data monitored in real time is confirmed;

[0072] The confirmed abnormal data is subjected to abnormality location, cause analysis, and abnormality level identification. Abnormality location is to locate the abnormal point in production corresponding to the abnormal data; cause analysis is to check the status of the sensor according to the located abnormal point and perform cause analysis based on the sensor status; abnormality level identification is to confirm the abnormality level based on the cause analysis results, and the abnormality level is divided into mild abnormality, moderate abnormality, and severe abnormality;

[0073] Finally, according to the abnormal level and cause, the control terminal is used to adjust the valve production process parameters;

[0074] Among them, mild abnormalities are abnormal prompts; moderate abnormalities are intervention prompts; severe abnormalities are automatic shutdowns.

[0075] Compared with the prior art, the present invention has the following beneficial effects:

[0076] 1. The present invention provides a method and system for optimizing parameters in valve production, which classifies process parameters, technological parameters, and output parameters according to production links such as casting and machining, ensuring that the analysis is carried out around the core process of valve production. It not only comprehensively covers all production links, but also focuses on key parameters, avoids data redundancy interference, and makes the analysis more systematic and targeted.

[0077] 2. The present invention provides a parameter optimization method and system for valve production. By clearly dividing the control group and the experimental group, the original production parameters are used as a benchmark reference, and the optimization parameter combination recommended by the model is used as the experimental object to form an intuitive comparison system. It can clearly show the actual impact of the optimization model on production and eliminate other interference factors. By incorporating process knowledge, the model can take into account the special process requirements and constraints in the production process, avoid the disconnection between theory and practice, and perform target analysis through conflict scenario and collaborative scenario identification, and can keenly perceive the contradictions and fit relationships between different goals.

[0078] 3. The present invention provides a method and system for optimizing parameters in valve production, which accurately locks the production abnormality points through abnormality positioning, conducts cause analysis in combination with the sensor status, avoids blind troubleshooting, and improves fault diagnosis efficiency; divides abnormality levels into mild, moderate, and severe levels according to the causes, and matches differentiated processing strategies such as abnormal prompts, intervention prompts, and automatic shutdowns. It can not only give prompts to guide production optimization in the case of mild abnormalities, but also respond quickly to ensure production safety in the case of severe abnormalities. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 Schematic diagram of parameter optimization steps in valve production of the present invention. DETAILED DESCRIPTION

[0080] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0081] In order to solve the problem in the existing technology that before obtaining the production parameter data, each sensor is not effectively tested, and the correlation analysis of each parameter data is not carried out, which leads to the inability to understand the close performance between the parameters, please refer to Figure 1 , this embodiment provides the following technical solutions:

[0082] A method for optimizing parameters in valve production, comprising:

[0083] First, the performance of each sensor in valve production is tested, and multi-source data is collected using the sensor. The collected multi-source data is preprocessed. After the data preprocessing is completed, the correlation analysis is performed on each parameter. Based on the results of the correlation analysis, multi-dimensional optimization targets are established, and an optimization model is constructed based on the established multi-dimensional optimization targets. Dynamic verification is performed based on the constructed optimization model.

[0084] Specifically, the performance of the sensor is first tested to eliminate equipment with insufficient accuracy or failure, and to avoid the interference of "junk data" on the optimization process. Multi-source data covers the entire valve production process, and can characterize the production process from multiple dimensions such as equipment operating status, process parameters, and environmental variables, providing a complete data base for subsequent analysis. The preprocessing link can eliminate random errors in the data collection process, so that subsequent analysis can focus on the real production laws. Correlation analysis can identify the implicit connection between parameters. Multi-dimensional targets can be weighted through hierarchical analysis or entropy weight method to ensure that the optimization direction meets the actual needs of the enterprise. Dynamic monitoring can promptly detect deviations between the model and actual production, trigger model iteration, and form a closed loop of "optimization-verification-re-optimization".

[0085] Each sensor in valve production is tested for performance, including:

[0086] Sensors include temperature sensors, pressure sensors, flow sensors, displacement sensors, vision sensors, humidity sensors and liquid level sensors;

[0087] Then perform performance testing on each sensor. The performance testing standards for the sensors are:

[0088] The temperature sensor's error is within the rated accuracy range, and its response time meets the preset requirements. The pressure sensor's linear error is less than ±1% FS, and its hysteresis error is less than ±0.5% FS. The flow sensor's full-scale error is within the preset range, and its dynamic fluctuation is less than ±2%. The displacement sensor's linear error is less than ±0.1% FS. The visual sensor's image is distortion-free. The humidity sensor's full-scale error is within the preset range. The liquid level sensor's liquid level detection error is less than ±2mm, and its dynamic response time is less than 2 seconds.

[0089] Calibrate and adjust sensors according to performance test standards;

[0090] Finally, the performance test of the sensor is completed.

[0091] Specifically, it covers seven core sensors, including temperature and pressure, enabling full-category testing. This ensures the integrity of monitoring of all key parameters in valve production, avoids production risks caused by missing sensor types, and provides comprehensive protection for production process control. In terms of performance testing standards, precise and strict indicators are formulated based on the characteristics of different sensors. This differentiated standard can accurately locate sensor performance defects and ensure its measurement accuracy in corresponding functions, laying a solid foundation for valve product quality. Calibration and adjustment have changed the simple testing model. Once problems are discovered, sensor performance can be corrected in a timely manner, forming a closed loop of testing and optimization, significantly improving the reliability and stability of sensors. Calibrated sensors can maintain high performance over the long term, reducing the risk of failure in the production process and minimizing downtime maintenance costs and production delays caused by sensor failure.

[0092] Leverage sensors to collect data from multiple sources, including:

[0093] First, set the acquisition parameters of the sensor, including the acquisition frequency, range and data unit;

[0094] Then synchronize the multi-source data acquisition scheme, including establishing a unified time base and data acquisition trigger conditions, where the data acquisition trigger conditions include periodic acquisition and time-triggered acquisition;

[0095] The sensor collects data according to the collection plan;

[0096] After data collection is completed, a transmission link is established through wireless transmission. After the transmission link is established, the data collected by each sensor is standardized in data format;

[0097] After the data format is unified, multi-source data of each sensor is obtained.

[0098] Specifically, clarifying the acquisition frequency, measurement range, and data units can ensure the accuracy and adaptability of data collection. Setting an appropriate acquisition frequency based on the actual application scenario can avoid data redundancy or loss; accurately defining the measurement range can prevent data overflow or invalid collection; unifying data units can eliminate conversion errors in subsequent data processing, laying the foundation for high-quality data collection; establishing a unified time base allows data collected by different sensors to be temporally comparable, facilitating subsequent data correlation analysis; diversified data collection trigger conditions, such as periodic collection, can meet routine monitoring needs, while time-triggered collection is suitable for data acquisition in emergencies or at specific time points, improving the flexibility and practicality of data collection. The establishment of wireless transmission links and the unified design of data formats enable efficient data circulation and convenient processing. Wireless transmission breaks free from the constraints of cables, reduces deployment costs and complexity, and improves system flexibility; unified data formats break down data barriers between different sensors, allowing data to be quickly identified and processed, greatly improving data processing efficiency and accelerating the transformation of data from collection to application.

[0099] The collected multi-source data is preprocessed, including:

[0100] Perform data cleaning on the collected multi-source data. Data cleaning involves identifying and processing missing data in the multi-source data, and then identifying and correcting outliers in the multi-source data.

[0101] After data cleaning, data conversion is performed to convert the data into the unit and format of unified multi-source data;

[0102] After data conversion, data integration is performed. Data integration involves establishing data association identifiers for multi-source data and handling conflicting data. The association identifier is established to generate a unique identifier for each valve artifact in the multi-source data. At the same time, the data from different sensors in the same time period are arranged in chronological order.

[0103] After data integration, data reduction is performed. Data reduction is to aggregate multi-source data according to timestamps and eliminate redundant data.

[0104] Finally, the data preprocessing of multi-source data is completed.

[0105] Specifically, missing value processing effectively avoids analytical biases caused by incomplete data. For example, filling in missing valve temperature data collected by sensors can prevent misjudgment of equipment failures. Outlier correction eliminates noise interference, avoids false warnings caused by occasional sensor anomalies, and ensures that the data truly reflects the operating status of the valve workpiece. It unifies the units and formats of pressure, flow, and other data collected by different sensors, breaking down data barriers and facilitating subsequent analysis and modeling. For example, unifying time formats of different precisions allows each source data to "communicate" under the same standard, laying the foundation for data mining and machine learning. Unique identification and time sorting connect discrete data in series to form a complete valve workpiece operation data chain, which is convenient for traceability and correlation analysis. Processing conflicting data eliminates contradictory information, such as coordinating the measurement differences of the same parameter by different sensors to ensure data consistency and provide a reliable basis for comprehensive decision-making. Aggregation by timestamp reduces the data volume, eliminating redundancy to reduce storage costs and computing burden, while retaining key features.

[0106] After data preprocessing is completed, correlation analysis is performed on each parameter, including:

[0107] Screening the analysis parameters in the pre-processed multi-source data, including process parameters, technological parameters and output parameters;

[0108] The screened analysis parameters are grouped according to the process stage, where the process stage is divided into parameter groups according to the valve production process, which includes casting, machining, welding, assembly and testing;

[0109] After the process stages are grouped, the parameter data of the same workpiece at each process stage are concatenated according to the timestamp. After the concatenation is completed, a complete production record of the parameters, time and results in the same batch is obtained;

[0110] Perform correlation analysis on the complete production records using visualization tools, including drawing scatter plots and heat maps. Analyze the changing trends in the scatter plots and the correlation strength between parameters in the heat map.

[0111] After the correlation analysis is completed, the parameter combination characteristics under different production conditions are identified, and the parameter distribution of each group is compared based on the parameter combination characteristics;

[0112] According to the parameter distribution, the output results of the valve production parameters are grouped into groups, including a high qualified rate group and a low qualified rate group;

[0113] The output results are grouped and correlated, including a key parameter list, an impact path diagram, and a pattern library. The key parameter list is the parameter with the greatest impact on the target; the impact path diagram is the causal relationship between parameters; and the pattern library is the parameter combination corresponding to different output results.

[0114] Finally, the correlation analysis data between each parameter in the multi-source data is obtained.

[0115] Specifically, process parameters, process parameters, and output parameters are classified according to production links such as casting and machining to ensure that the analysis is carried out around the core process of valve production. This layered processing method not only comprehensively covers all aspects of production, but also focuses on key parameters, avoids data redundancy interference, makes the analysis more systematic and targeted, accurately captures the parameter change patterns in the production process, and connects the parameter data of the same workpiece in each process stage in series by timestamp, integrates discrete data into a complete production record, and truly restores the production process sequence. This operation closely links parameters with time and results, and can deeply explore the dynamic impact of parameter changes on production results, providing an accurate time series basis for tracing the root causes of production problems and optimizing process flows. By drawing scatter plots and heat maps for correlation analysis, abstract data relationships are intuitively presented. Scatter plots clearly show parameter change trends, and heat maps quantify the correlation strength between parameters, helping technicians quickly identify key parameters and potential associations and avoid subjective judgment bias. Visualization lowers the barrier to data analysis, improves analysis efficiency and accuracy, and provides strong support for decision-making. It identifies parameter combination characteristics under different production conditions, compares the distribution of parameters across groups, and then categorizes high-qualified and low-qualified groups, clarifying the mechanisms by which parameters influence production results. Based on this, it generates a list of key parameters, impact path diagrams, and a model library, providing specific guidance for production optimization, effectively improving product quality and efficiency while reducing production costs.

[0116] In order to solve the problem in the existing technology that there is no targeted optimization target and optimization model for obtaining production parameters, and no dynamic verification based on the optimization results, which leads to the imperfect final optimization results, please refer to Figure 1 , this embodiment provides the following technical solutions:

[0117] Establish multi-dimensional optimization goals based on the results of correlation analysis, including:

[0118] Before establishing multi-dimensional optimization goals, a multi-dimensional goal system should be established first. The multi-dimensional goal system includes quality dimension, efficiency dimension, cost dimension and safety dimension.

[0119] Then set the optimization target value, including setting it based on historical data, industry standards and parameter adjustable range;

[0120] Target identification is performed based on a multi-dimensional target system and optimized target values, including conflict scenario identification and collaborative scenario identification;

[0121] The target association graph is constructed based on the identified targets. The target association graph is as follows:

[0122]

[0123] Finally, the establishment of multi-dimensional optimization goals is completed.

[0124] Specifically, a multi-dimensional target system covering quality, efficiency, cost and safety is constructed to comprehensively cover the core elements of valve production. The quality dimension ensures that product performance meets standards, the efficiency dimension shortens the production cycle, the cost dimension controls resource consumption, and the safety dimension ensures stable production. This comprehensive target setting avoids the limitations of single-dimensional optimization, provides systematic guidance for production operations, and sets optimization target values ​​based on historical data, industry standards, and adjustable parameter ranges, so that the targets have solid data support and practical feasibility. Historical data reflects the company's past production levels, industry standards provide external reference benchmarks, and the adjustable parameter range is combined with actual production capacity. The combination of these three ensures that the targets are both challenging and realistic, avoids targets that are too high or too low, and improves the feasibility of optimization solutions. Target analysis is performed through the identification of conflict scenarios and collaborative scenarios, which can keenly perceive the contradictions and compatibility between different targets. Conflict scenario identification can predict potential trade-offs that may arise during target optimization. Collaborative scenario identification explores the potential for synergy between targets, providing a basis for scientific decision-making, helping to balance the needs of all parties and maximize production efficiency. A target association diagram visualizes the logical relationships between targets, directly illustrating the interplay between quality, efficiency, cost, and safety. This allows companies to quickly grasp the overall situation, accurately identify key targets and influencing factors, and develop collaborative optimization strategies to avoid compromising overall interests due to partial optimization, effectively improving the systematicity and coordination of production management.

[0125] Optimization model construction is carried out based on the established multi-dimensional optimization objectives, including:

[0126] Before building the optimization model, confirm the input and output variables of the model;

[0127] Among them, the parameters that have the greatest impact on the target in the correlation analysis are used as input variables; the quantitative indicators in the multi-dimensional optimization target system are used as output variables;

[0128] The basic model is selected according to the characteristics of the multi-dimensional optimization objectives. The multi-dimensional optimization objectives include single-objective optimization and multi-objective optimization. Among them, the basic model is the linear regression model for single-objective optimization and the multi-objective genetic algorithm model for multi-objective optimization.

[0129] Retrieve the production process knowledge model from the database and embed the production process knowledge model into the basic model respectively;

[0130] After embedding is completed, the basic model is built, including weight allocation, objective function setting and optimal solution search;

[0131] After the model is built, it is verified. Model verification involves using the model to predict parameter combinations in historical production and comparing them with the actual results. If the deviation exceeds 5%, check whether key parameters are missed.

[0132] After the model verification is completed and qualified, the construction of the optimization model is completed.

[0133] Specifically, the parameters with the greatest impact on the target in the correlation analysis are used as input variables to ensure that the model focuses on key influencing factors and avoids interference from redundant data; the quantitative indicators in the multi-dimensional optimization target system are used as output variables so that the model output directly corresponds to the core production target. This precise variable selection method allows the model to efficiently capture key information, improve the targetedness and effectiveness of optimization, and flexibly select linear regression models or multi-objective genetic algorithm models based on the single-objective or multi-objective characteristics of the multi-dimensional optimization target. When optimizing for a single objective, the linear regression model is simple and efficient, and can quickly establish a linear relationship between the variable and the target; when optimizing for multiple objectives, the multi-objective genetic algorithm model can effectively handle multiple conflicting objectives and find the optimal solution set. This adaptive selection ensures that the model is highly consistent with actual needs, improves the model's solution capabilities, retrieves the production process knowledge model from the database and embeds it into the basic model, combines actual production experience with the theoretical model, and makes the model more suitable for actual production scenarios. By incorporating process knowledge, the model can account for the unique process requirements and constraints of the production process, eliminating the disconnect between theory and practice, enhancing the model's practicality and reliability, and making the optimization results more practical and instructive. A rigorous model validation process promptly identifies model deviations by predicting historical production parameter combinations and comparing them with actual results. When deviations exceed 5%, the model checks for missing key parameters to ensure model accuracy. This validation mechanism effectively eliminates model errors, ensures model reliability, and ensures that optimization strategies based on the model are effective and reliable, reducing production decision-making risks.

[0134] Dynamic verification is performed based on the constructed optimization model, including:

[0135] Before dynamic verification, the control group and test group in valve production are first confirmed. The control group uses the original production parameters as a benchmark reference; the test group uses the optimized parameter combination recommended by the model.

[0136] Then confirm the sample size of each group, where each group produces at least 200 valves;

[0137] A trial run was conducted based on the control group and the experimental group. The trial run was as follows:

[0138] S1: Enter the model recommended parameters in the test group equipment and lock the parameter modification permission;

[0139] S2: For the same batch of workpieces in the control group and the experimental group, sensor data and quality data are collected synchronously;

[0140] S3: Set the warning threshold, and automatically suspend production and issue an alarm when it is triggered;

[0141] After the trial run is completed, the data of the control group and the experimental group will be compared, and the difference analysis will be carried out based on the index comparison results;

[0142] Finally, the dynamic verification results of the optimization model are obtained based on the difference analysis results.

[0143] Specifically, by clearly dividing the control group into a control group and a test group, using the original production parameters as a baseline reference and the optimized parameter combinations recommended by the model as the test subjects, an intuitive comparison system was formed. This design clearly demonstrates the actual impact of the optimization model on production, eliminates other interfering factors, makes the verification results more convincing, and accurately evaluates the effectiveness and feasibility of the optimization model. The sample size of at least 200 valves per group meets statistical requirements and effectively reduces the bias in the results caused by too few samples. Sufficient sample data can more realistically reflect the actual production situation, enhance the stability and credibility of the verification results, ensure that the data-based difference analysis and final verification conclusions have a solid data foundation, and provide a reliable basis for the subsequent application of the model. During the trial operation, the parameter modification permissions of the test group are locked to avoid human interference and ensure the purity of the test data; the control group and test group data are collected simultaneously to ensure data consistency and comparability; set warning thresholds and realize automatic pause and alarm to effectively prevent production accidents caused by inappropriate model parameters, which not only ensures production safety, but also can timely discover potential problems and improve the safety and controllability of the verification process. By comparing indicators and performing difference analysis on the data of the control group and the test group, it is possible to quantify the changes brought about by the optimization model in terms of quality, efficiency, cost and other dimensions. Accurately locate the advantages and disadvantages of model optimization, help enterprises clearly understand the actual application effect of the model, provide intuitive and accurate information support for further optimization of the model and actual production decisions, and promote the optimization model to better serve valve production.

[0144] In order to solve the problem in the existing technology that there is no effective real-time monitoring of the verification process, and no targeted abnormal analysis and warning level judgment of the monitoring results, which leads to the inability to handle abnormal situations in a timely manner, please refer to Figure 1 , this embodiment provides the following technical solutions:

[0145] A valve production parameter optimization system, comprising:

[0146] Dynamic verification data monitoring unit, used to:

[0147] Before dynamic verification, build a real-time monitoring system architecture, including sensor deployment and edge computing device deployment;

[0148] After the real-time monitoring system architecture is built, confirm the real-time monitoring indicators, which include input parameter indicators and output result indicators;

[0149] After the real-time monitoring indicators are confirmed, the collection frequency of dynamic verification data is confirmed, including high-frequency collection, medium-frequency collection and low-frequency collection;

[0150] After dynamic verification data collection is completed, data synchronization and calibration are performed;

[0151] Finally, the real-time monitoring of dynamic verification data is completed.

[0152] The monitoring data adjustment early warning unit is used to:

[0153] First, define abnormal data, which includes data where parameters deviate from the model's recommended values, data where indicators fail to meet standards, and data where safety parameters exceed the bottom line.

[0154] According to the defined abnormal data, the abnormal data in the dynamic verification data monitored in real time is confirmed;

[0155] The confirmed abnormal data is subjected to abnormality location, cause analysis, and abnormality level identification. Abnormality location is to locate the abnormal point in production corresponding to the abnormal data; cause analysis is to check the status of the sensor according to the located abnormal point and perform cause analysis based on the sensor status; abnormality level identification is to confirm the abnormality level based on the cause analysis results, and the abnormality level is divided into mild abnormality, moderate abnormality, and severe abnormality;

[0156] Finally, according to the abnormal level and cause, the control terminal is used to adjust the valve production process parameters;

[0157] Among them, mild abnormalities are abnormal prompts; moderate abnormalities are intervention prompts; severe abnormalities are automatic shutdowns.

[0158] Specifically, by building a real-time monitoring system architecture and deploying sensors and edge computing equipment, we can achieve instant perception and preliminary processing of production data, ensuring the timeliness and accuracy of data collection. Clarify real-time monitoring indicators such as input parameters and output results, focus on key production factors, and avoid interference from invalid data. Flexibly set high-frequency, medium-frequency, and low-frequency acquisition frequencies to reasonably allocate computing resources while ensuring data quality based on different data characteristics and production process requirements. Data synchronization and calibration mechanisms eliminate data errors, provide a reliable data foundation for subsequent analysis and decision-making, and improve the overall efficiency and accuracy of production monitoring. Abnormal data is clearly defined in multiple dimensions, covering key production risk points such as parameter deviation, indicator substandardness, and safety bottom line violations, so that abnormal judgments have clear standards. In the exception handling process, production anomalies are accurately identified through anomaly positioning, and the cause analysis is conducted in combination with the sensor status to avoid blind investigation and improve fault diagnosis efficiency. According to the cause, the anomaly levels are divided into mild, moderate, and severe levels, and differentiated processing strategies such as abnormal prompts, intervention prompts, and automatic shutdowns are matched. In the case of mild anomalies, prompts can be given to guide production optimization, and in the case of severe anomalies, rapid responses can be made to ensure production safety, realizing precise and intelligent anomaly management and control. The dynamic verification data monitoring unit and the monitoring data adjustment and early warning unit work closely together. The former provides accurate real-time data, and the latter handles anomalies in a timely manner based on the data, forming a closed loop from data collection, monitoring to anomaly response. This collaborative mechanism can monitor the production status in real time, quickly discover and solve potential problems, effectively reduce production risks, ensure the stable operation of the valve production process, and help enterprises continuously optimize production parameters and improve production quality and efficiency.

[0159] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0160] While the embodiments of the present invention have been shown and described, it will be apparent to those skilled in the art that various changes, modifications, substitutions, and alterations can be made to the embodiments without departing from the principles and spirit of the invention.

Claims

1. A method for optimizing parameters in valve production, characterized in that: include: First, the performance of each sensor in valve production is tested, and multi-source data is collected using the sensor. The collected multi-source data is preprocessed. After the data preprocessing is completed, the correlation analysis is performed on each parameter. Based on the results of the correlation analysis, multi-dimensional optimization targets are established, and an optimization model is constructed based on the established multi-dimensional optimization targets. Dynamic verification is performed based on the constructed optimization model.

2. A valve production parameter optimization method according to claim 1, characterized in that: Each sensor in valve production is subjected to performance testing, including: Sensors include temperature sensors, pressure sensors, flow sensors, displacement sensors, vision sensors, humidity sensors and liquid level sensors; Then perform performance testing on each sensor. The performance testing standards for the sensors are: The temperature sensor's error is within the rated accuracy range, and its response time meets the preset requirements. The pressure sensor's linear error is less than ±1% FS, and its hysteresis error is less than ±0.5% FS. The flow sensor's full-scale error is within the preset range, and its dynamic fluctuation is less than ±2%. The displacement sensor's linear error is less than ±0.1% FS. The visual sensor's image is distortion-free. The humidity sensor's full-scale error is within the preset range. The liquid level sensor's liquid level detection error is less than ±2mm, and its dynamic response time is less than 2 seconds. Calibrate and adjust sensors according to performance test standards; Finally, the performance test of the sensor is completed.

3. A valve production parameter optimization method according to claim 1, characterized in that: Leverage sensors to collect data from multiple sources, including: First, set the acquisition parameters of the sensor, including the acquisition frequency, range and data unit; Then synchronize the multi-source data acquisition scheme, including establishing a unified time base and data acquisition trigger conditions, where the data acquisition trigger conditions include periodic acquisition and time-triggered acquisition; The sensor collects data according to the collection plan; After data collection is completed, a transmission link is established through wireless transmission. After the transmission link is established, the data collected by each sensor is standardized in data format; After the data format is unified, multi-source data of each sensor is obtained.

4. A valve production parameter optimization method according to claim 1, characterized in that: The collected multi-source data is preprocessed, including: Perform data cleaning on the collected multi-source data. Data cleaning involves identifying and processing missing data in the multi-source data, and then identifying and correcting outliers in the multi-source data. After data cleaning, data conversion is performed to convert the data into the unit and format of unified multi-source data; After data conversion, data integration is performed. Data integration involves establishing data association identifiers for multi-source data and handling conflicting data. The association identifier is established to generate a unique identifier for each valve artifact in the multi-source data. At the same time, the data from different sensors in the same time period are arranged in chronological order. After data integration, data reduction is performed. Data reduction is to aggregate multi-source data according to timestamps and eliminate redundant data. Finally, the data preprocessing of multi-source data is completed.

5. The method for optimizing parameters in valve production according to claim 1, characterized in that: After data preprocessing is completed, correlation analysis is performed on each parameter, including: Screening the analysis parameters in the pre-processed multi-source data, including process parameters, technological parameters and output parameters; The screened analysis parameters are grouped according to the process stage, where the process stage is divided into parameter groups according to the valve production process, which includes casting, machining, welding, assembly and testing; After the process stages are grouped, the parameter data of the same workpiece at each process stage are concatenated according to the timestamp. After the concatenation is completed, a complete production record of the parameters, time and results in the same batch is obtained; Perform correlation analysis on the complete production records using visualization tools, including drawing scatter plots and heat maps. Analyze the changing trends in the scatter plots and the correlation strength between parameters in the heat map. After the correlation analysis is completed, the parameter combination characteristics under different production conditions are identified, and the parameter distribution of each group is compared based on the parameter combination characteristics; According to the parameter distribution, the output results of the valve production parameters are grouped into groups, including a high qualified rate group and a low qualified rate group; The output results are grouped and correlated, including a key parameter list, an impact path diagram, and a pattern library. The key parameter list is the parameter with the greatest impact on the target; the impact path diagram is the causal relationship between parameters; and the pattern library is the parameter combination corresponding to different output results. Finally, the correlation analysis data between each parameter in the multi-source data is obtained.

6. A valve production parameter optimization method according to claim 1, characterized in that: Establish multi-dimensional optimization goals based on the results of correlation analysis, including: Before establishing multi-dimensional optimization goals, a multi-dimensional goal system should be established first. The multi-dimensional goal system includes quality dimension, efficiency dimension, cost dimension and safety dimension. Then set the optimization target value, including setting it based on historical data, industry standards and parameter adjustable range; Target identification is performed based on a multi-dimensional target system and optimized target values, including conflict scenario identification and collaborative scenario identification; Construct a target association graph based on the identified targets; Finally, the establishment of multi-dimensional optimization goals is completed.

7. A valve production parameter optimization method according to claim 1, characterized in that: Optimization model construction is carried out based on the established multi-dimensional optimization objectives, including: Before building the optimization model, confirm the input and output variables of the model; Among them, the parameters that have the greatest impact on the target in the correlation analysis are used as input variables; the quantitative indicators in the multi-dimensional optimization target system are used as output variables; The basic model is selected according to the characteristics of the multi-dimensional optimization objectives. The multi-dimensional optimization objectives include single-objective optimization and multi-objective optimization. Among them, the basic model is the linear regression model for single-objective optimization and the multi-objective genetic algorithm model for multi-objective optimization. Retrieve the production process knowledge model from the database and embed the production process knowledge model into the basic model respectively; After embedding is completed, the basic model is built, including weight allocation, objective function setting and optimal solution search; After the model is built, it is verified. Model verification involves using the model to predict parameter combinations in historical production and comparing them with the actual results. If the deviation exceeds 5%, check whether key parameters are missed. After the model verification is completed and qualified, the construction of the optimization model is completed.

8. A valve production parameter optimization method according to claim 1, characterized in that: Dynamic verification is performed based on the constructed optimization model, including: Before dynamic verification, the control group and test group in valve production are first confirmed. The control group uses the original production parameters as a benchmark reference; the test group uses the optimized parameter combination recommended by the model. Then confirm the sample size of each group, where each group produces at least 200 valves; A trial run was conducted based on the control group and the experimental group. The trial run was as follows: S1: Enter the model recommended parameters in the test group equipment and lock the parameter modification permission; S2: For the same batch of workpieces in the control group and the experimental group, sensor data and quality data are collected synchronously; S3: Set the warning threshold, and automatically suspend production and issue an alarm when it is triggered; After the trial run is completed, the data of the control group and the experimental group will be compared, and the difference analysis will be carried out based on the index comparison results; Finally, the dynamic verification results of the optimization model are obtained based on the difference analysis results.

9. A valve production parameter optimization system, applied to a valve production parameter optimization method according to any one of claims 1 to 8, characterized in that: include: Dynamic verification data monitoring unit, used to: Before dynamic verification, build a real-time monitoring system architecture, including sensor deployment and edge computing device deployment; After the real-time monitoring system architecture is built, confirm the real-time monitoring indicators, which include input parameter indicators and output result indicators; After the real-time monitoring indicators are confirmed, the collection frequency of dynamic verification data is confirmed, including high-frequency collection, medium-frequency collection and low-frequency collection; After dynamic verification data collection is completed, data synchronization and calibration are performed; Finally, the real-time monitoring of dynamic verification data is completed.

10. A valve production parameter optimization system according to claim 9, characterized in that: include: The monitoring data adjustment early warning unit is used to: First, define abnormal data, which includes data where parameters deviate from the model's recommended values, data where indicators fail to meet standards, and data where safety parameters exceed the bottom line. According to the defined abnormal data, the abnormal data in the dynamic verification data monitored in real time is confirmed; The confirmed abnormal data is subjected to abnormality location, cause analysis, and abnormality level identification. Abnormality location is to locate the abnormal point in production corresponding to the abnormal data; cause analysis is to check the status of the sensor according to the located abnormal point and perform cause analysis based on the sensor status; abnormality level identification is to confirm the abnormality level based on the cause analysis results, and the abnormality level is divided into mild abnormality, moderate abnormality, and severe abnormality; Finally, according to the abnormal level and cause, the control terminal is used to adjust the valve production process parameters; Among them, mild abnormalities are abnormal prompts; moderate abnormalities are intervention prompts; severe abnormalities are automatic shutdowns.

Images