A method and system for online monitoring of a gas polyethylene pipe production process
Patent Information
- Application Number
- CN202611092094.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-22
- Publication Date
- 2026-09-04
AI Technical Summary
然而，燃气聚乙烯管道挤出生产过程属于典型的长流程连续成型过程，不同工序之间存在明显的物料传输时滞、热传导时滞和质量响应滞后，不同过程参数对质量指标的影响在作用时间、作用强度和作用路径上存在差异
本申请提供的一种燃气聚乙烯管道生产过程在线监测方法，通过对燃气聚乙烯管道生产线各工序的过程参数信号和质量参数信号进行统一时间基准同步以及长度位置标定，能够建立过程参数与质量参数在时间和空间位置上的对应关系，提高多源数据对齐精度；根据各工序的物料停留时间和质量响应滞后特性，为不同过程参数分别确定基于工艺机理的非均匀滞后集合，而非采用统一的固定滞后，从而更准确地刻画不同过程参数对质量指标在作用时间、作用强度和作用路径上的差异化动态影响；根据模型求解得到综合关联评分，并将其与正常工况参考区间进行比较，输出预警信息；综合关联评分融合了多工序、多变量、多滞后时刻的动态关联信息，能够中反映生产过程的整体关联状态，实现了对燃气聚乙烯管道生产过程的综合关联评价和可靠在线预警，提升了生产过程在线监测的准确性和质量控制水平。
Smart Images
Figure FT_1 
Figure SMS_6 
Figure SMS_7
Abstract
Description
Technical Field
[0001] This application relates to the technical field of production monitoring, and in particular to an online monitoring system for the production process of gas-fired polyethylene pipelines. Background Technology
[0002] Currently, gas-fired polyethylene (PEG) pipelines are a crucial foundation material in urban gas transmission and distribution systems. Their quality indicators, such as outer diameter, wall thickness, out-of-roundness, and dimensional stability, directly affect their pressure-bearing capacity, connection reliability, and service safety. In the extrusion production of PEG pipelines, multiple processes—including raw material processing, melt extrusion, vacuum sizing, spray cooling, traction cutting, and online quality inspection—are interconnected. Process parameters such as barrel temperatures, die temperatures, melt pressure, screw speed, vacuum pressure, cooling water temperature, cooling water flow rate, and traction speed are interdependent and collectively influence the final product quality. Therefore, online monitoring of the production process and accurate identification of the correlation between key process parameters and quality indicators are critical issues in the quality control of PEG pipeline production.
[0003] In related technologies, gas-fired polyethylene pipeline production lines are typically equipped with industrial sensors for temperature, pressure, flow, and speed, as well as online monitoring devices such as outer diameter gauges and wall thickness gauges, to collect equipment operating status information and product dimensional and quality information. Existing monitoring methods mostly focus on upper and lower limit alarms for single process parameters, or online display and out-of-tolerance judgments for single quality indicators such as outer diameter and wall thickness. Some solutions also use empirical rules or simple statistical methods to analyze process parameters to assist operators in adjusting process conditions. However, the gas-fired polyethylene pipeline extrusion production process is a typical long-flow continuous molding process. Significant material transport time lags, heat conduction time lags, and quality response lags exist between different processes. The impact of different process parameters on quality indicators varies in terms of duration, intensity, and path. Existing technologies typically rely on single-moment data or single-variable threshold judgments, lacking a unified modeling of the dynamic coupling relationships between multiple processes, variables, and quality indicators, making it difficult to accurately reflect the time-varying correlation characteristics between various process parameters and quality indicators.
[0004] Therefore, there is an urgent need to provide an online monitoring technology suitable for the continuous extrusion production process of gas-fired polyethylene pipelines, so as to improve the accuracy of the dynamic correlation between process parameters and quality parameters, and enhance the reliability of online monitoring and early warning of production status. Summary of the Invention
[0005] The purpose of this application is to provide a method and system for online monitoring of the production process of gas-fired polyethylene pipelines, which can effectively improve the accuracy and reliability of the online monitoring process.
[0006] To achieve the above objectives, this application provides the following solution: In a first aspect, this application provides an online monitoring method for the production process of gas-fired polyethylene pipelines. The method includes: collecting process parameter signals and quality parameter signals from each stage of the gas-fired polyethylene pipeline production line, synchronizing them with a unified time reference, and calibrating their length and position to obtain a multivariate data stream; constructing a process parameter data set and a quality parameter data set based on the multivariate data stream; determining a non-uniform lag set for each process parameter in the process parameter data set according to the material residence time and quality response lag characteristics of each stage; constructing a process parameter lag block vector based on the non-uniform lag set, and confirming the quality parameters in conjunction with the quality parameter data set. Vector; A dynamic multivariate correlation analysis model with sparse constraints on variable groups is established and solved for the process parameter lag block vector and the quality parameter vector to obtain the load weights and correlation strength coefficients under each typical mode; wherein, the sparse constraints on variable groups treat the weights of the same process parameter at different lag times as a group, and treat the weights of multiple process parameters in the same process segment as a group; a comprehensive correlation score is calculated based on the load weights and the correlation strength coefficients; a normal operating condition reference interval is determined based on historical stable production data, and the normal operating condition reference interval is compared with the comprehensive correlation score to obtain a comparison result, and a warning information is output based on the comparison result.
[0007] Secondly, this application provides an online monitoring system for the production process of gas-fired polyethylene pipelines. The system includes: a sensing unit for real-time acquisition of process parameter signals and quality parameter signals at each stage of the gas-fired polyethylene pipeline production line; a data acquisition and synchronization unit for synchronizing the process parameter signals and quality parameter signals using a unified time reference and calibrating their length and position to obtain a multivariate data stream; a correlation analysis unit for constructing a process parameter data set and a quality parameter data set based on the multivariate data stream; and determining a non-uniform lag set for each process parameter in the process parameter data set based on the material residence time and quality response lag characteristics of each stage; and further for constructing a process parameter lag block vector based on the non-uniform lag set. The quality parameter vector is confirmed by combining the quality parameter data set; a dynamic multivariate correlation analysis model with sparse constraints of variable groups is established and solved for the process parameter lag block vector and the quality parameter vector to obtain the load weights and correlation strength coefficients under each typical mode; a comprehensive correlation score is calculated based on the load weights and correlation strength coefficients; wherein, the sparse constraints of variable groups treat the weights of the same process parameter at different lag times as a group, and treat the weights of multiple process parameters in the same process segment as a group; a status display and early warning unit is used to determine the normal operating condition reference interval based on historical stable production data, compare the normal operating condition reference interval with the comprehensive correlation score to obtain the comparison result, and output early warning information based on the comparison result.
[0008] According to the specific embodiments provided in this application, the following technical effects are disclosed: This application provides an online monitoring method for the production process of gas-fired polyethylene pipelines. By synchronizing process parameter signals and quality parameter signals of each process in the gas-fired polyethylene pipeline production line with a unified time reference and calibrating their length and position, it can establish a correspondence between process parameters and quality parameters in time and space, improving the alignment accuracy of multi-source data. Based on the material residence time and quality response lag characteristics of each process, a non-uniform lag set based on the process mechanism is determined for different process parameters, rather than using a uniform fixed lag. This more accurately characterizes the differentiated dynamic impact of different process parameters on quality indicators in terms of action time, intensity, and path. A comprehensive correlation score is obtained by solving the model and compared with a reference interval under normal operating conditions, outputting early warning information. The comprehensive correlation score integrates dynamic correlation information from multiple processes, multiple variables, and multiple lag times, reflecting the overall correlation status of the production process. This achieves comprehensive correlation evaluation and reliable online early warning for the gas-fired polyethylene pipeline production process, improving the accuracy of online monitoring and the level of quality control. Attached Figure Description
[0009] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0010] Figure 1 This is a flowchart illustrating the online monitoring method for the production process of gas-fired polyethylene pipelines in this embodiment of the application. Detailed Implementation
[0011] 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.
[0012] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0013] like Figure 1 The diagram shown is a schematic flow chart of an online monitoring method for the production process of gas-fired polyethylene pipelines according to an embodiment of this application. It includes steps S110 to S170: S110: Collect process parameter signals and quality parameter signals of each process in the gas-fired polyethylene pipeline production line, and perform unified time reference synchronization and length position calibration to obtain a multivariate data stream.
[0014] S120. Construct process parameter data sets and quality parameter data sets based on multivariate data streams.
[0015] S130. Based on the material residence time and quality response lag characteristics of each process, determine the non-uniform lag set for each process parameter in the process parameter data set.
[0016] S140. Construct the process parameter lag block vector based on the non-uniform lag set, and confirm the quality parameter vector by combining it with the quality parameter data set.
[0017] S150. Establish and solve a dynamic multivariate correlation analysis model with sparse constraints on the process parameter lag block vector and the quality parameter vector to obtain the load weights and correlation strength coefficients under each typical mode. Among them, the sparse constraints on the variable groups regard the weights of the same process parameter at different lag times as a group, and regard the weights of multiple process parameters belonging to the same process segment as a group.
[0018] S160. Calculate the comprehensive correlation score based on the load weight and the relevant strength coefficient.
[0019] S170. Determine the reference range for normal operating conditions based on historical stable production data, compare the reference range for normal operating conditions with the comprehensive correlation score to obtain the comparison result, and output early warning information based on the comparison result.
[0020] Through steps S110 to S170, the multi-source process parameter signals and quality parameter signals dispersed in various processes of the gas-fired polyethylene pipeline production line can be aligned in a unified time and length position. Based on this, a process parameter lag block vector is constructed using a non-uniform lag set based on the process mechanism. The load weight and correlation strength coefficient under each typical mode are solved by a dynamic multivariate correlation analysis model with sparse constraints on variable groups. This enables comprehensive correlation score calculation and online early warning, thereby improving the accuracy, timeliness and quality control level of online monitoring of the gas-fired polyethylene pipeline production process.
[0021] In the above S110, process parameter signals and quality parameter signals of each process in the gas-fired polyethylene pipeline production line are collected, and a unified time reference synchronization and length position calibration are performed to obtain a multivariate data stream.
[0022] In the signal acquisition stage, corresponding sensing units and acquisition nodes are deployed near the raw material processing, melt extrusion, vacuum sizing, spray cooling, traction cutting, and online quality inspection processes. The process parameter signals include density, moisture, and thermogravimetric analysis signals for the raw material processing process; barrel temperature, die temperature, melt temperature, melt pressure, and screw speed signals for each heating zone in the melt extrusion process; vacuum pressure signals for the vacuum sizing process; cooling water temperature and flow rate signals for the spray cooling process; and traction speed encoder and traction speed signals for the traction cutting process. Quality parameter signals include online measurement signals for outer diameter, wall thickness, and out-of-roundness for the online quality inspection process.
[0023] The first m The process variables at the sampling time The observed values are denoted as , will the The quality variables at the sampling time The observed values are denoted as Thus, the original signal vectors for the process parameters and the original signal vectors for the quality parameters are constructed as follows: in, M Indicates the number of process parameters. N Indicates the number of quality parameters.
[0024] Data acquisition can be achieved using a combination of distributed acquisition nodes and a primary synchronization node. Corresponding sensing units and acquisition nodes are deployed near each process step. The primary synchronization node manages the unified time reference, while the distributed acquisition nodes handle sampling, buffering, and uploading from their nearest location, thereby reducing sampling deviations caused by long-distance transmission.
[0025] After signal acquisition is completed, a unified time base synchronization and length / position calibration are performed. The master synchronization node outputs a unified clock signal, and the calibrated local timestamps of each distributed acquisition node are represented as follows: in, Indicates the first The local time of each data collection node. Indicates the first The correction amount of each acquisition node relative to the master synchronization clock. This indicates the timestamp after the unified time base synchronization is completed.
[0026] Traction speed encoder output pulse count Based on the encoder resolution and the circumference of the traction roller, the corresponding length coordinates are defined as follows: in, This represents the increment in pipe length corresponding to a single encoder pulse.
[0027] A single sampling record, after synchronization with a unified time reference and length and position calibration, is recorded as: This results in a multivariate data stream with time and length coordinate markers. Since pipeline production proceeds continuously at a constant traction speed, the positional calibration along the pipeline length can precisely correlate the product quality data of a certain cross-section detected online with the process parameters of each preceding process when it passed through that cross-section.
[0028] In step S120 above, a process parameter data set and a quality parameter data set are constructed based on the multivariate data stream. The multivariate data stream obtained in step S110 is preprocessed to construct the process parameter data set and the quality parameter data set. Preprocessing includes outlier detection, short-term missing value imputation, long-term missing value marking, and dimensional unification.
[0029] For outlier detection, a discrimination method based on the absolute deviation of the median is adopted. For the... The median of the process variables within the sliding window is denoted as . The absolute deviation of the median is denoted as If a sampled value satisfies: The sampled value is then marked as an outlier, where, This is the preset threshold coefficient.
[0030] For short-term missing points, linear interpolation is used for compensation. If at time... and There are missing sample points. Then the interpolated value can be expressed as: For consecutive missing lengths exceeding a preset threshold In cases where the time window is invalid, it will be marked as an invalid window and will not participate in subsequent model training or online inference. The dimensional unification process uses a standardized method. For the first... The process variable and the first The standardized results of the three quality variables are expressed as follows: in, , They represent the first The mean and standard deviation of each process variable. , They represent the first The mean and standard deviation of each quality variable.
[0031] After preprocessing, the process parameter data set can be denoted as: The set of quality parameter data is denoted as: in, Indicates the total sampling length.
[0032] In step S130 above, based on the material residence time and quality response lag characteristics of each process, a non-uniform lag set is determined for each process parameter in the process parameter data set. In the extrusion production of gas-fired polyethylene pipelines, there are significant material transport lags, heat conduction lags, and quality response lags between different processes. The impact of different process parameters on quality indicators varies in terms of duration, intensity, and path. If a uniform fixed lag is applied to all process parameters, the actual impact path of each process parameter on quality indicators cannot be accurately reflected. Therefore, it is necessary to determine differentiated lag time sets for different process parameters.
[0033] For the The process parameters are defined as follows: their non-uniform lag set is: in, Indicates the first The first process parameter One effective lag, Indicates the first The number of lags for each process parameter. The non-uniform lag set can be determined jointly based on prior knowledge of the process mechanism and data-driven analysis. As an embodiment of the present invention, it can be determined first based on the material residence time range. After obtaining the candidate lag intervals, the lag corresponding to the peak value is selected based on the cross-correlation function between the process variable and the quality variable. The cross-correlation function can be expressed as: when When a local peak is reached within the candidate lag interval, the corresponding lag can be... Included in the non-uniform hysteresis set By using the above methods, differentiated lag time sets can be determined for different process parameters such as barrel temperature in the melt extrusion process, vacuum pressure in the vacuum sizing process, and cooling water flow rate in the spray cooling process, thereby more accurately depicting the actual impact path of different parameters on quality indicators.
[0034] In step S140 above, a process parameter lag patch vector is constructed based on the non-uniform lag set, and a quality parameter vector is confirmed by combining the quality parameter data set. After determining the non-uniform lag set for each process parameter, a process parameter lag patch vector and a quality parameter vector aligned with the process parameter lag patch vector on the time and length coordinates are constructed using their respective non-uniform lag sets.
[0035] For the Each process parameter, at time... The process parameter lag subvector is defined as: Furthermore, all process parameters at time... The process parameter lag patch vector is defined as: The mass parameter vector, aligned with the process parameter hysteresis patch vector on the time and length coordinates, is defined as: In this embodiment, since step S110 above has completed the synchronization of the unified time reference and the calibration of the length position, therefore The mass parameter vector represented by The process parameter lag block vector represents the quality response of the same pipe spatial location or the same pipe cross section.
[0036] In S150 above, a dynamic multivariate correlation analysis model with sparse constraints on variable groups is established and solved for the process parameter lag block vector and the quality parameter vector to obtain the load weight and correlation strength coefficient under each typical mode.
[0037] Within the sliding time window, the process parameter lag block vector samples are assembled into a matrix. The mass parameter vector sample forms a matrix Then we have: in, This indicates the number of samples within the current sliding time window.
[0038] As an embodiment of the present invention, the covariance matrix and the cross-covariance matrix are defined as follows: As an embodiment of the present invention, the objective function of the dynamic multivariate correlation analysis model with sparse constraints on variable groups is: And satisfy the constraints: in, For process-side load vectors, For the mass-side load vector, Let be the cross-covariance matrix between the process parameter lag patch vector and the quality parameter vector. Let covariance be the process parameter lag patch vector. Let covariance be the vector of quality parameters. For the first The load subvectors corresponding to each group of process variables For the first Load subvectors corresponding to each mass variable group and The group norm penalty coefficient.
[0039] In the objective function, the first term The first term is the correlation enhancement term, used to increase the correlation strength between typical combination variables on the process side and typical combination variables on the quality side under structured sparsity constraints. The last two terms are group norm penalty terms, i.e., variable group sparsity constraint terms. The variable group sparsity constraint has a two-layer structure: the first layer treats the weights of the same process parameter at different lag times in its non-uniform lag set as a variable group for constraint; the second layer treats the weights of multiple process parameters belonging to the same process segment as a variable group for constraint. Through the above two layers of structured sparsity constraints, the model can promote the overall selection or suppression of multiple weights of the same variable or the same process segment during model solving, thereby shrinking the weights of irrelevant variable groups or weakly correlated variable groups to zero.
[0040] In terms of solution, the alternating proximal gradient algorithm is used to iteratively solve the objective function. Let the process-side load vector and the mass-side load vector be respectively at the r-th iteration. and Then it can be updated as follows: in, Indicates the step size parameter. and These are the normalization adjustment coefficients during iteration. This represents the proximal operator corresponding to the group norm penalty. and These represent the normalized projections that satisfy the constraints.
[0041] After iteratively solving the above process, the following can be obtained: The process-side load vector, mass-side load vector, and related strength coefficients corresponding to each typical mode are denoted as follows: in, Indicates the first The correlation strength coefficients corresponding to each typical model. The number of typical models retained. It can be determined based on the cumulative contribution rate of the relevant intensity coefficients of the candidate typical modes and in combination with the process interpretability.
[0042] Furthermore, the first The process-side load vector under a typical mode is represented as follows: in, Indicates the first The process parameter in the first... The first typical model The load weight corresponding to each lag time.
[0043] In S160 above, a comprehensive correlation score is calculated based on the load weight and the relevant intensity coefficient.
[0044] Before calculating the comprehensive correlation score, the correlation importance index of each process parameter and the correlation contribution index of each process segment are calculated based on the load weight and correlation strength coefficient, so as to reflect the dynamic correlation between process parameters and quality parameters from multiple perspectives.
[0045] The first The correlation importance index of each process parameter is marked as The calculation formula is as follows: in, The number of typical patterns to be retained. For the first The number of lags in the non-uniform lag set corresponding to each process parameter.
[0046] Those belonging to the same process segment The importance index of multiple process parameters is determined by preset weights. We perform a weighted summation to obtain the process segment correlation contribution index. : in, .
[0047] All process parameters are associated with importance indicators according to preset weights We perform a weighted summation to obtain the total importance of the process side. : in, .
[0048] All process segment correlation contribution indicators are assigned to preset weights We perform a weighted summation to obtain the total contribution of the process side. : in, The number of process segments, .
[0049] All correlation strength coefficients corresponding to typical patterns are weighted according to preset weights. A weighted average is used to obtain the basic score of the relevant intensity. : in, .
[0050] Furthermore, comprehensive correlation score It can be calculated using the following formula: in, , and The preset weighting coefficients satisfy the following conditions: Optionally, as an embodiment of the present invention, in the calculation , and Before that, you can first... , and Interval normalization was performed to reduce the impact of different dimensions and scales on the comprehensive correlation score.
[0051] In step S170 above, a normal operating condition reference range is determined based on historical stable production data, the normal operating condition reference range is compared with the comprehensive correlation score to obtain a comparison result, and a warning information is output based on the comparison result.
[0052] Let the comprehensive correlation score sequence obtained during the historical stable production phase be: Their mean and standard deviation are as follows: in, The length of historical stable production data.
[0053] For example, the normal operating condition reference range can be represented as: in, The coefficients are preset interval coefficients.
[0054] If the current time is a comprehensive correlation score In continuous Within a sliding time window, the following conditions must be met: The current operating condition is then determined to be an abnormal associated state, and an early warning message is triggered.
[0055] Process parameters can be sorted from largest to smallest according to their correlation importance index, and the top ones can be selected. The process parameters are a set of key parameters to be adjusted. For sets The first in For each process parameter, calculate its current window mean: And compare it with the reference values corresponding to historical stable production stages. By comparing the values, we can obtain the adjustment direction amount: when At that time, the output will be "Improve the first Adjustment suggestions for "each process parameter"; when At that time, the output will be "reduce the first". Adjustment suggestions for "each process parameter"; when If the value is less than the preset tolerance, output a suggestion to "keep the current parameters".
[0056] Furthermore, when the overall correlation score is within the normal operating condition reference range, for a certain process segment... If its process segment correlation contribution index The rate continues to rise and exceeds the process optimization threshold within the preset observation period. That is, satisfying: and Generate operating condition optimization prompts to remind operators to prioritize the key process parameters corresponding to the aforementioned process segment.
[0057] The method in this embodiment, through steps S110 to S170, establishes a temporal and spatial correspondence between process parameters and quality parameters by synchronizing process parameter signals and quality parameter signals of each process in the gas-fired polyethylene pipeline production line with a unified time reference and calibrating length and position. By constructing a non-uniform lag set based on the process mechanism according to the material residence time and quality response lag characteristics of each process, the differentiated dynamic influence characteristics of different process parameters on quality indicators are more accurately characterized. On this basis, a dynamic multivariate correlation analysis model with sparse constraints on variable groups is established, realizing unified modeling of the dynamic coupling relationship of multiple processes and multiple variables. Finally, by comparing the comprehensive correlation score with the reference interval of normal operating conditions, a comprehensive correlation evaluation and reliable online early warning of the gas-fired polyethylene pipeline production process are realized.
[0058] This application also provides an online monitoring system for the production process of gas-fired polyethylene pipelines. This system can be installed in electronic equipment and includes multiple sensing units, a data acquisition and synchronization unit, a correlation analysis unit, and a status display and early warning unit.
[0059] Multiple types of sensing units are used to collect process parameter signals that characterize the process status and quality parameter signals that characterize the product characteristics in real time at each stage of the gas-fired polyethylene pipeline production line.
[0060] The data acquisition and synchronization unit is used to acquire process parameter signals and quality parameter signals, synchronize a unified time reference, and calibrate length and position, forming a multivariable data stream with time markers and length coordinate markers.
[0061] The correlation analysis unit is used to preprocess the multivariate data stream, construct process parameter data sets and quality parameter data sets, and construct process parameter lag block vectors and quality parameter vectors within a sliding time window based on the non-uniform lag set based on the process mechanism. It uses a dynamic multivariate correlation analysis model with sparse constraints on variable groups to calculate the dynamic correlation between process parameters and quality parameters, and generates key process parameter correlation importance index, process segment correlation contribution index, and comprehensive correlation score.
[0062] The status display and early warning unit is used to display the importance index of key process parameters, the contribution index of process segments, and the comprehensive correlation score in real time, and to issue early warning prompts and process parameter adjustment suggestions when the comprehensive correlation score deviates from the preset normal range.
[0063] As one implementation of this embodiment, the correlation analysis unit may further include a data preprocessing module, a dynamic patch construction module, a process structure constraint sparse dynamic multivariate correlation modeling module, and a correlation index calculation module.
[0064] The data preprocessing module is used to perform outlier detection, short-term missing data imputation, long-term missing data marking, and dimensional unification on multivariate data streams.
[0065] The dynamic block construction module is used to determine non-uniform lag sets based on the process parameter data set and the quality parameter data set, according to the material residence time and quality response lag characteristics of each process, for different process parameters. Within the sliding time window, it uses the respective non-uniform lag sets to construct process parameter lag block vectors and quality parameter vectors aligned with the process parameter lag block vectors on the time and length coordinates.
[0066] The process structure constraint sparse dynamic multivariate correlation modeling module is used to establish a dynamic multivariate correlation analysis model with sparse constraints on variable groups based on the process parameter lag block vector and the mass parameter vector, and solve for the process side load vector, the mass side load vector and the corresponding correlation strength coefficient.
[0067] The correlation index calculation module is used to calculate the correlation importance index of each process parameter and the correlation contribution index of each process segment based on the correlation strength coefficient and the load weight of each variable in the process-side load vector under different typical modes and different lag times, and synthesize the comprehensive correlation score according to preset rules.
[0068] The status display and early warning unit can display the importance index of key process parameters, the contribution index of process segments, and the comprehensive correlation score in real time in the form of trend curves, bar charts, sorted lists, or status indicator lights. Based on the comparison results of the comprehensive correlation score and the reference range of normal operating conditions, it can output abnormal correlation early warning information and process parameter adjustment suggestions.
[0069] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.
[0070] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).
[0071] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.
[0072] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0073] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.
Claims
1. A method for online monitoring of the production process of gas-fired polyethylene pipelines, characterized in that, The method includes: The process parameter signals and quality parameter signals of each process in the gas-fired polyethylene pipeline production line are collected, and the time reference is synchronized and the length and position are calibrated to obtain a multivariate data stream. Construct a process parameter data set and a quality parameter data set based on the multivariate data stream; Based on the material residence time and quality response lag characteristics of each process, a non-uniform lag set is determined for each process parameter in the process parameter data set; The process parameter lag block vector is constructed based on the non-uniform lag set, and the quality parameter vector is confirmed by combining it with the quality parameter data set. A dynamic multivariate correlation analysis model with sparse constraints on variable groups is established and solved for the process parameter lag block vector and the quality parameter vector to obtain the load weights and correlation strength coefficients under each typical mode; wherein, the sparse constraints on variable groups regard the weights of the same process parameter at different lag times as a group, and regard the weights of multiple process parameters in the same process segment as a group. Calculate the comprehensive correlation score based on the load weight and the relevant intensity coefficient; Based on historical stable production data, a reference range for normal operating conditions is determined. The reference range for normal operating conditions is compared with the comprehensive correlation score to obtain a comparison result. Based on the comparison result, a warning message is output.
2. The online monitoring method for the production process of gas-fired polyethylene pipelines according to claim 1, characterized in that, The process parameter signals include: density signal, moisture signal, and thermogravimetric analysis signal of the raw material processing step; barrel temperature signal, die temperature signal, melt temperature signal, melt pressure signal, and screw speed signal of each heating zone in the melt extrusion step; vacuum pressure signal of the vacuum sizing step; cooling water temperature signal and cooling water flow rate signal of the spray cooling step; and traction speed encoder signal and traction speed signal of the traction cutting step. The quality parameter signals include online measurement signals of outer diameter, online measurement signals of wall thickness, and online measurement signals of out-of-roundness in the online quality inspection process.
3. The online monitoring method for the production process of gas-fired polyethylene pipelines according to claim 2, characterized in that, The process of synchronizing a unified time base and calibrating length and position includes: The pulse signal from the traction speed encoder is received by the main synchronization node, and the traction pulse count is converted into a length coordinate along the pipe length direction. The master synchronization node distributes a unified time reference to multiple distributed acquisition nodes, so that the local clock of each distributed acquisition node is synchronized with the unified time reference. Each distributed acquisition node locally acquires and caches process parameter signals and quality parameter signals near the corresponding process, and adds a uniform time stamp and a length coordinate stamp converted from traction pulse count to each acquired data.
4. The online monitoring method for the production process of gas-fired polyethylene pipelines according to claim 1, characterized in that, The process parameter data set and the quality parameter data set are constructed based on the preprocessed multivariate data stream. The preprocessing includes outlier detection, short-term missing data imputation, long-term missing data marking, and dimensional unification processing.
5. The online monitoring method for the production process of gas-fired polyethylene pipelines according to claim 1, characterized in that, The determination of the non-uniform lag set for each process parameter includes: Candidate lag intervals are determined based on the material residence time range of the specified process parameters in the process. Calculate the cross-correlation function between the specified process parameter and each quality parameter within the candidate lag interval, and select the lag time corresponding to the local peak value of the cross-correlation function to include it in the non-uniform lag set of the specified process parameter; wherein, the cross-correlation function is calculated by the following formula: in, Indicates the first m The process variables at the sampling time t Preprocessed observations Indicates the first n The preprocessed observation values of each quality variable at the sampling time. Lag time, This represents the total sampling length.
6. The online monitoring method for the production process of gas-fired polyethylene pipelines according to claim 1, characterized in that, The objective function of the dynamic multivariate correlation analysis model with sparse constraints on variable groups is shown below: The constraints include: in, For process-side load vectors, For the mass-side load vector, Let be the cross-covariance matrix between the process parameter lag patch vector and the quality parameter vector. Let covariance be the process parameter lag patch vector. Let covariance be the vector of quality parameters. For the first g The load subvectors corresponding to each group of process variables For the first h Load subvectors corresponding to each mass variable group and The group norm penalty coefficient; The process variable group includes: a variable group consisting of the weights of the same process parameter at different lag times in its non-uniform lag set, and a variable group consisting of the weights of multiple process parameters within the same process segment.
7. The online monitoring method for the production process of gas-fired polyethylene pipelines according to claim 1, characterized in that, Before calculating the comprehensive correlation score, the method further includes: Based on the load weight and the relevant strength coefficient, calculate the correlation importance index of each process parameter and the correlation contribution index of each process segment; The importance index of the process parameters is calculated using the following formula: in, Let m be the correlation importance index for the m-th process parameter. The number of typical patterns to be retained. For the first k The correlation strength coefficients corresponding to each typical model For the first m The number of lags in the non-uniform lag set corresponding to each process parameter. For the first The process parameter in the first... The first typical model The load weight corresponding to each lag time; The correlation contribution index of the aforementioned process segment is calculated using the following formula: In the formula, For process segment s The correlation contribution index For the first m The preset weights of each process parameter, and satisfying .
8. The online monitoring method for the production process of gas-fired polyethylene pipelines according to claim 7, characterized in that, The step of outputting early warning information based on the comparison results includes: When the comprehensive correlation score falls outside the normal operating condition reference range within a series of sliding time windows, the current operating condition is determined to be an abnormal correlation state, and an abnormal correlation warning message is output. When the comprehensive correlation score is within the normal working condition reference range, and the correlation contribution index of a certain process segment continues to rise and exceeds the optimization threshold within the preset observation time, the working condition optimization prompt information is output.
9. The online monitoring method for the production process of gas-fired polyethylene pipelines according to claim 8, characterized in that, The step of outputting early warning information based on the comparison results also includes: The process parameters are sorted from largest to smallest according to the importance index associated with the process parameters to determine the set of key process parameters; Calculate the current window mean of each process parameter in the set of key process parameters, compare it with the reference value of the historical stable production stage, generate process parameter adjustment suggestions based on the comparison results, and output them.
10. An online monitoring system for the production process of gas-fired polyethylene pipelines, characterized in that, The system includes: The sensing unit is used to collect process parameter signals and quality parameter signals in real time at each stage of the gas-fired polyethylene pipeline production line. The data acquisition and synchronization unit is used to synchronize the process parameter signals and quality parameter signals with a unified time reference and to calibrate the length and position to obtain a multivariable data stream. The correlation analysis unit is used to construct process parameter data sets and quality parameter data sets based on the multivariate data stream; and to determine non-uniform lag sets for each process parameter in the process parameter data set based on the material residence time and quality response lag characteristics of each process; it is also used to construct process parameter lag block vectors based on the non-uniform lag sets, and to confirm quality parameter vectors in conjunction with the quality parameter data set; to establish and solve a dynamic multivariate correlation analysis model with variable group sparse constraints on the process parameter lag block vectors and the quality parameter vectors, obtaining the load weights and correlation strength coefficients under each typical mode; and to calculate a comprehensive correlation score based on the load weights and correlation strength coefficients; wherein, the variable group sparse constraints treat the weights of the same process parameter at different lag times as a group, and treat the weights of multiple process parameters in the same process segment as a group; The status display and early warning unit is used to determine the normal operating condition reference range based on historical stable production data, compare the normal operating condition reference range with the comprehensive correlation score to obtain a comparison result, and output early warning information based on the comparison result.