Stainless steel water pipe production line maintenance method and system based on big data analysis

Abstract

The invention discloses a stainless steel water pipe production line maintenance method and system based on big data analysis, and relates to the technical field of water pipe production. The method comprises the following steps: locally calling a historical defect record based on a preset defect window; working procedure recording backtracking is conducted on the stainless steel water pipe production line, and N working procedure parameter sets are obtained; performing anomaly identification on the N process parameter sets, and performing process defect coupling analysis based on an identification result to obtain N groups of associated processes; deploying N edge nodes on the stainless steel water pipe production line; operating the N edge nodes in real time to carry out defect coupling trend analysis, and sending N operation and maintenance demand characteristics obtained through analysis to a production line maintenance cloud; and the production line maintenance cloud receives the N operation and maintenance demand features and performs operation and maintenance strategy analysis output according to the N operation and maintenance demand features. The technical problems that in the prior art, stainless steel water pipe production line maintenance strategies lack pertinence and are low in operation and maintenance efficiency are solved, and the technical effect of improving the maintenance strategy accuracy and the production line maintenance efficiency is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of water pipe production, and specifically to a maintenance method and system for a stainless steel water pipe production line based on big data analysis. Background Art

[0002] Stainless steel water pipes are widely used in multiple fields such as water supply and drainage, construction, etc. due to their strong corrosion resistance and long service life, and the automation degree of their production lines is also continuously improving. However, limited by complex technological processes, a variety of equipment types, and environmental factors, there are still a certain proportion of defect problems in the production process of stainless steel water pipes, such as cracks, depressions, uneven wall thickness, abnormal welds, etc. In the prior art, the operation and maintenance management of the production line mostly rely on regular inspections and manual experience judgment, lacking in-depth analysis and utilization of historical defect data, making it difficult to detect potential risks in a timely manner, resulting in insufficient pertinence of maintenance strategies and low response efficiency, and easily causing product quality fluctuations and production interruptions. Summary of the Invention

[0003] This application provides a maintenance method and system for a stainless steel water pipe production line based on big data analysis, which solves the technical problems of lack of pertinence in the maintenance strategy of the stainless steel water pipe production line and low operation and maintenance efficiency in the prior art.

[0004] In the first aspect of this application, a maintenance method for a stainless steel water pipe production line based on big data analysis is provided. The method includes: Locally invoking historical defect records based on a preset defect window, where the historical defect records include N defect batch sets of N defect types; performing process record backtracking on the stainless steel water pipe production line according to the N defect batch sets to obtain N process parameter sets; performing anomaly identification on the N process parameter sets, and performing process defect coupling analysis based on the identification results to obtain N groups of associated processes; deploying N edge nodes on the stainless steel water pipe production line according to the N groups of associated processes, where the N edge nodes are communicatively connected to N groups of associated production equipment of the N groups of associated processes; running the N edge nodes in real time for defect coupling trend analysis, and sending N operation and maintenance requirement features obtained from the analysis to the production line maintenance cloud, where the production line maintenance cloud is communicatively connected to the N edge nodes; the production line maintenance cloud receives and performs operation and maintenance strategy analysis and output according to the N operation and maintenance requirement features.

[0005] In the second aspect of this application, a maintenance system for a stainless steel water pipe production line based on big data analysis is provided. The system includes: A historical data calling module for locally calling historical defect records based on a preset defect window, where the historical defect records include N defect batch sets of N defect types; a backtracking module for backtracking the process records of the stainless steel water pipe production line according to the N defect batch sets to obtain N process parameter sets; a coupling analysis module for performing anomaly identification on the N process parameter sets and performing process defect coupling analysis based on the identification results to obtain N groups of associated processes; an edge node deployment module for deploying N edge nodes on the stainless steel water pipe production line according to the N groups of associated processes, where the N edge nodes are communicatively connected to N groups of associated production equipment of the N groups of associated processes; an analysis module for running the N edge nodes in real time to perform defect coupling trend analysis and sending N operation and maintenance requirement features obtained from the analysis to the production line maintenance cloud, where the production line maintenance cloud is communicatively connected to the N edge nodes; an operation and maintenance strategy output module for the production line maintenance cloud to receive and perform operation and maintenance strategy analysis and output according to the N operation and maintenance requirement features.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: First, locally call historical defect records based on a preset defect window, where the historical defect records include N defect batch sets of N defect types. Then, backtrack the process records of the stainless steel water pipe production line according to the N defect batch sets to obtain N process parameter sets. Next, perform anomaly identification on the N process parameter sets and perform process defect coupling analysis based on the identification results to obtain N groups of associated processes. Then, deploy N edge nodes on the stainless steel water pipe production line according to the N groups of associated processes, where the N edge nodes are communicatively connected to N groups of associated production equipment of the N groups of associated processes. Finally, run the N edge nodes in real time to perform defect coupling trend analysis and send N operation and maintenance requirement features obtained from the analysis to the production line maintenance cloud, where the production line maintenance cloud is communicatively connected to the N edge nodes; the production line maintenance cloud receives and performs operation and maintenance strategy analysis and output according to the N operation and maintenance requirement features. This solves the technical problems of the lack of pertinence of the maintenance strategy and low operation and maintenance efficiency in the existing stainless steel water pipe production line, and achieves the technical effect of improving the accuracy of the maintenance strategy and the production line maintenance efficiency. Description of the Drawings

[0007] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following described drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0008] Figure 1A schematic diagram of a stainless steel water pipe production line maintenance method based on big data analysis provided in an embodiment of the present application; Figure 2 A schematic diagram of the structure of a stainless steel water pipe production line maintenance system based on big data analysis provided in an embodiment of the present application.

[0009] Explanation of the accompanying drawings: historical data calling module 11, backtracking module 12, coupling analysis module 13, edge node deployment module 14, analysis module 15, operation and maintenance strategy output module 16. DETAILED DESCRIPTION

[0010] The present application solves the technical problems in the prior art of the lack of specificity in the maintenance strategy of the stainless steel water pipe production line and the low operation and maintenance efficiency by providing a stainless steel water pipe production line maintenance method and system based on big data analysis.

[0011] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0012] It should be noted that the terms "including" and "having" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or are inherent to these processes, methods, products or devices.

[0013] Embodiment 1, as Figure 1 As shown, the present application provides a stainless steel water pipe production line maintenance method based on big data analysis, wherein the method includes: The historical defect records are locally called based on a preset defect window, wherein the historical defect records include N defect batch sets of N defect types.

[0014] In the embodiment of the present application, a preset defect window is set to limit the scope and time interval of historical defect data calls. The preset defect window can be flexibly configured according to the actual production cycle, for example, the last 30 days, 60 days or a specific batch number is used as the time / batch boundary to avoid redundant data interfering with analysis efficiency. The historical defect record corresponding to the preset defect window is called in the local database. The historical defect record is the defect detection information recorded during the production process. The record format may include the time of defect occurrence, defect type, corresponding batch number, production equipment to which it belongs, defect level, location mark, etc.

[0015] The historical defect records include N sets of defect batches for N types of defects. Specifically, the system classifies according to the defect types (such as: cracks, deformations, corrosion, abnormal wall thickness, poor welding, etc.), and aggregates the historical records belonging to each type of defect according to the batch number, forming multiple independent sets of defect batches. Each set of defect batches corresponds to one type of defect and contains multiple historical batch defect data related to this type, which is used for subsequent process traceability and parameter correlation analysis.

[0016] Exemplarily, if the preset defect window is the most recent 30 days, the system retrieves the following example data from the local database: Defect type A (crack): includes crack records of batches P101, P106, and P109; Defect type B (corrosion): includes corrosion records of batches P102 and P108; Defect type C (weld anomaly): includes weld anomaly records of batches P103, P104, and P110; and so on. The system obtains N sets of defect batches corresponding to N types of defects.

[0017] Based on the N sets of defect batches, the process records of the stainless steel water pipe production line are traced back to obtain N sets of process parameters.

[0018] In the embodiment of the present application, for each set of defect batches, the production process record information corresponding to the batch number is extracted. The production process record information includes but is not limited to: raw material feeding information, key control parameters of each process section, operating status of processing equipment, operator information, environmental monitoring data (such as temperature and humidity), inspection and quality control records, etc.

[0019] Specifically, for a set of defect batches corresponding to a certain type of defect (such as cracks) (such as including batches P101, P106, and P109), the system traces back and queries the full-process process records of each batch in the production line one by one. For each batch, the key process parameters of each corresponding process are extracted, such as the temperature setting value, constant temperature time, and cooling rate of the annealing process; the rolling pressure and drawing speed of the forming process; the current, voltage, welding time, and weld temperature curve of the welding process; the acid concentration, treatment duration, and spray flow rate of the pickling process; the online inspection results and defect detection image numbers of the inspection process, etc.

[0020] For multiple batches in each set of defect batches, the system aggregates the above key parameters of all processes in time series, constructs a process parameter data set including multiple dimensions, and forms a set of process parameters corresponding to this type of defect. Finally, the above-mentioned traceability and aggregation operations are performed on all N sets of defect batches to obtain N sets of process parameters, which are respectively used for subsequent anomaly identification and defect coupling analysis.

[0021] Anomaly identification is performed on the N sets of process parameters, and process defect coupling analysis is performed based on the identification results to obtain N groups of associated processes.

[0022] By performing anomaly identification on N sets of process parameters, that is, comparing the actual parameter values in each process with the benchmark operation characteristics to analyze whether there are abnormal fluctuations or offsets in the parameters; if a certain parameter value exceeds the allowable deviation range or similar deviations repeatedly appear in multiple batches, it can be determined as an abnormal process parameter. Based on the anomaly identification results, perform process defect coupling analysis, that is, analyze the potential causal or statistical correlation relationships between the defect types and each abnormal process to obtain N sets of associated processes. An associated process refers to a combination of processes that cause the same defect, such as "rolling + welding" for pores and "flattening + polishing" for scratches.

[0023] Furthermore, performing anomaly identification on the N sets of process parameters and performing process defect coupling analysis based on the identification results to obtain N sets of associated processes, the method includes: Interactively obtain the benchmark operation characteristics of multiple production equipment in multiple processes of the stainless steel water pipe production line; use the multiple benchmark operation characteristics to traverse the first set of process parameters to identify abnormal operation parameters and obtain H sequences of abnormal process parameters, where the first set of process parameters includes H historical process parameter sequences; calculate the operation deviation of the H sequences of abnormal process parameters based on the multiple benchmark operation characteristics to obtain H sequences of abnormal deviation degrees; perform process defect coupling analysis based on the H sequences of abnormal process parameters and the H sequences of abnormal deviation degrees, and screen and locate the first set of associated production equipment among the multiple production equipment; and so on, perform anomaly identification on the N sets of process parameters and perform process defect coupling analysis based on the identification results to obtain the N sets of associated processes.

[0024] The system interacts with the process knowledge platform or historical data warehouse to retrieve the operation benchmark characteristics of the production equipment in multiple key processes of the stainless steel water pipe production line. The benchmark operation characteristics include but are not limited to: equipment set values, standard operation ranges, stable output indicators, process fluctuation tolerances, process coupling rules, etc. The system uses the above multiple benchmark operation characteristics to compare and judge the parameters one by one for the H historical process parameter sequences included in the first set of process parameters, and identifies the parameter sequences that are inconsistent with the benchmark characteristics and deviate from the set range. The identification result is H sequences of abnormal process parameters, where each sequence corresponds to a specific parameter offset event at a process link or equipment point.

[0025] The system further performs a running deviation quantification analysis on the above-mentioned H abnormal process parameter sequences to form H abnormal deviation degree sequences. The deviation degree calculation can be based on various methods, such as Euclidean distance, Z-score deviation, moving average deviation degree, interval deviation integral, etc., which are used to measure the severity and persistence of parameter abnormalities. According to the H abnormal process parameter sequences and the H abnormal deviation degree sequences, a process defect coupling analysis is carried out, and a group of associated production equipment that has a significant impact on the current defect is screened and located among multiple process production equipment, and then the first group of associated processes to which the equipment belongs is mapped. The first group of associated processes is the key link causing the current defect type and is suitable for subsequent edge deployment and key operation and maintenance. The above process is sequentially executed for N process parameter sets, and finally N groups of associated processes corresponding to N defect types are obtained. Each group of associated processes represents a high coupling relationship between the current defect and certain specific links in the production process.

[0026] Furthermore, according to the H abnormal process parameter sequences and the H abnormal deviation degree sequences, a process defect coupling analysis is carried out, and the first group of associated production equipment is screened and located among the multiple process production equipment. The method includes: By aggregating the H abnormal process parameters for production equipment, multiple abnormal frequency characteristics of the multiple process production equipment are obtained; by aggregating the H abnormal deviation degree sequences for production equipment, multiple abnormal probability contributions of the multiple process production equipment are obtained; according to the multiple abnormal probability contributions and the multiple abnormal frequency characteristics, a defect abnormal association quantification is performed, and multiple defect abnormal contribution degrees are output; according to a preset contribution threshold and the multiple defect abnormal contribution degrees, the first group of associated production equipment is screened from the multiple process production equipment; based on process mapping, the first group of associated processes of the first group of associated production equipment is extracted.

[0027] The system performs an aggregation statistics on the H abnormal process parameter sequences according to the process equipment dimension, and counts the frequency of each process production equipment in all H abnormal parameters, that is, the number of times of parameter abnormalities occurring on each equipment, to obtain multiple abnormal frequency characteristics of the multiple process production equipment.

[0028] The system performs an aggregation operation on the H abnormal deviation degree sequences in the equipment dimension. For example, weighted average can be used to evaluate the comprehensive influence of each process production equipment in all abnormal deviations, that is, to perform unified normalization and contribution quantification on the deviation data of each equipment, and obtain an abnormal probability contribution index for each equipment, reflecting its weight role in the overall deviation.

[0029] By fusing the abnormal frequency characteristics and the abnormal probability contribution, a unified defect abnormal association index is formed. Specifically, an equipment abnormal contribution function is constructed Ci=α · Fi+β · Pi ,Fi is the abnormal frequency of the i-th device, Pi is the abnormal probability contribution, and α and β are weighting factors obtained through empirical tuning or adaptive learning. After calculation, the defect abnormal contribution degree of each device is obtained, reflecting its coupling strength with the defect.

[0030] The system presets a contribution threshold (such as 1.5 times the average contribution degree or an empirically set value), sorts the contribution degrees of all devices, and filters out the set of devices that exceed the threshold as the high-coupling devices of the current defect type, which are the first group of associated production devices. Map the first group of associated production devices to their respective processes in the process mapping table, extract all the process links to which this group of devices belong, and obtain the first group of associated processes corresponding to this defect type.

[0031] Furthermore, by aggregating production devices for the H abnormal deviation degree sequences, multiple abnormal probability contributions of the production devices of the multiple processes are obtained. The method includes: Through production device aggregation, multiple abnormal deviation degrees and multiple abnormal deviation time windows of the production devices of the first process are obtained; call the first deviation weight and the first time weight of the production devices of the first process from the process characteristic weight library; perform weighted fusion of the multiple abnormal deviation degrees and the multiple abnormal deviation time windows according to the first deviation weight and the first time weight, and output the first abnormal probability contribution; and so on. By aggregating production devices for the H abnormal deviation degree sequences, the multiple abnormal probability contributions are obtained.

[0032] Specifically, using the device number or process number as the aggregation index, classify and organize each deviation data involved in the H abnormal deviation degree sequences. Taking the production devices of the first process as an example, extract the deviation degrees in all the abnormal process parameter sequences related to it to form a set of numerical values; at the same time, extract the corresponding abnormal occurrence time windows to form the multiple abnormal deviation degrees and abnormal deviation time windows of this device.

[0033] Retrieve the first deviation weight and the first time weight corresponding to the production devices of the first process from the preset process characteristic weight library. The first deviation weight is used to measure the sensitivity and weight of the severity of the device deviation in the overall abnormal recognition, and the first time weight is used to measure the influence degree of the abnormal deviation occurrence time in the overall trend. The process characteristic weight library is formed based on historical statistics or expert experience and has a unique mapping relationship with the device. Based on the above weight parameters, the system performs weighted fusion of the multiple abnormal deviation degrees and the multiple abnormal deviation time windows to generate the first abnormal probability contribution. According to the above steps, perform the aggregation processing of the abnormal deviation degrees and time windows for the production devices of other processes in the same way, call their respective deviation weights and time weights, and calculate their corresponding abnormal probability contribution values one by one, and finally obtain the abnormal probability contributions of each device.

[0034] Furthermore, based on the first deviation weight and the first time weight, a weighted fusion of the multiple abnormal deviation degrees and the multiple abnormal deviation time windows is performed to output a first abnormal probability contribution. The method includes: Calculating the average deviation degree of the multiple abnormal deviation degrees; solving the cumulative deviation of the multiple abnormal deviation time windows to obtain a cumulative deviation time; performing a weighted fusion of the average deviation degree and the cumulative deviation time based on the first deviation weight and the first time weight to output the first abnormal probability contribution.

[0035] In an embodiment of the present application, based on the first deviation weight and the first time weight, a weighted fusion of the multiple abnormal deviation degrees and the multiple abnormal deviation time windows of the production equipment of the first process is performed to calculate the first abnormal probability contribution of the equipment. Specifically, all abnormal deviation degree values corresponding to the production equipment of the first process are statistically processed to obtain its average deviation degree; the abnormal time windows corresponding to all abnormal events of the equipment are cumulatively summed to obtain the total abnormal duration of the equipment; the above average deviation degree and the cumulative deviation time are weighted and fused to calculate the first abnormal probability contribution.

[0036] Deploy N edge nodes in the stainless steel water pipe production line according to the N groups of associated processes, where the N edge nodes are communicatively connected to N groups of associated production equipment of the N groups of associated processes.

[0037] In an embodiment of the present application, N edge nodes are deployed in the stainless steel water pipe production line according to N groups of associated processes determined by process defect coupling analysis. Each edge node establishes a local communication connection with a corresponding group of associated production equipment. Specifically, based on process control logic or production management system (MES) records, the specific production equipment set corresponding to each group of associated processes is determined. A set of edge computing units is deployed in the nearby industrial control network node or equipment control cabinet of each group of associated production equipment.

[0038] Furthermore, the method of deploying N edge nodes in the stainless steel water pipe production line according to the N groups of associated processes includes: Locally calling the M device fault data sets of M associated production equipment in the first group of associated production equipment; performing fault level - type identification on the M device fault data sets to obtain M device fault level - type sets; screening and aggregating the H abnormal process parameter sequences according to the M associated production equipment to obtain M defect abnormal parameter sets; using the M device fault data sets, M device fault level - type sets and M defect abnormal parameter sets to perform hierarchical modeling to obtain M hierarchical fault detection models; completing the local deployment of the first edge node by paralleling the M hierarchical fault detection models.

[0039] Specifically, for each set of associated processes, the system locally invokes the historical fault data sets of M associated devices in the corresponding first set of associated production devices. The historical fault data sets contain fault records, operating status parameters, and time information during the operation of the devices; the fault data of each device is processed for identifying the fault level and fault type, and the faults are classified into multiple levels (such as critical / serious / minor) and types (such as electrical fault / thermal control fault / mechanical jamming, etc.) to form a standardized device fault level - type set; according to the device numbers, the parameter information matching the M associated devices is screened out from the previously obtained H abnormal process parameter sequences and aggregated to obtain M defective abnormal parameter sets. The system inputs the above three data bases, namely the device fault data set, the fault level - type set, and the defective abnormal parameter set, into the modeling module to construct hierarchical fault detection models for the M devices respectively. This model includes a single - device fault identification module, an abnormal feature detection module, and a fault level judgment module, which can achieve information fusion and identification at different levels. Finally, the system deploys the M hierarchical fault detection models in a parallel structure to the target edge node to complete the local modeling deployment of the node, enabling it to have the ability to perceive and preliminarily diagnose the operating status of the associated production devices in real - time, providing data support for subsequent operation and maintenance requirement analysis.

[0040] Furthermore, using the M device fault data sets, the M device fault level - type sets, and the M defective abnormal parameter sets for hierarchical modeling to obtain M hierarchical fault detection models, the method includes: Based on the knowledge graph, map - associate and store the M device fault data sets and the M device fault level - type sets to complete the construction of the M single - device fault identification models; perform fluctuation analysis on the M defective abnormal parameter sets to output M defective fluctuation features; construct M operation abnormal detection engines based on the M defective fluctuation features; according to the M associated production devices, map - cascade the M single - device fault identification models and the M operation abnormal detection engines to complete the construction of the M hierarchical fault detection models.

[0041] Based on knowledge graph technology, semantic mapping and structured association are carried out between the historical fault data of each device and its corresponding fault level and type, and M fault identification models for single devices are constructed. This model has the ability to identify and classify common faults of a single device; multi-dimensional time series fluctuation analysis is performed on the set of defect abnormal parameters corresponding to each device, and the fluctuation characteristics under different operating states are extracted to obtain M defect fluctuation characteristics, which are used to characterize the dynamic change trend during the operation of the device; based on each defect fluctuation characteristic, the system constructs M operation anomaly detection engines to identify potential cross-condition anomaly behaviors; finally, according to the process positions and functional relationships of M associated production devices, the above-mentioned single-fault identification models and operation anomaly detection engines are correspondingly mapped and cascaded to achieve information sharing and multi-level linkage analysis, and finally the construction of M-level fault detection models is completed, providing a model basis for edge node deployment.

[0042] The N edge nodes are operated in real time for defect coupling trend analysis, and the N operation and maintenance requirement characteristics obtained from the analysis are sent to the production line maintenance cloud, where the production line maintenance cloud is communicatively connected to the N edge nodes.

[0043] In the embodiment of the present application, each edge node intermittently receives the operation data sequence uploaded in real time by the associated production devices connected to it based on a preset operation and maintenance time window. The data includes working condition parameters in multiple dimensions such as temperature, current, pressure, vibration amplitude, and product quality index. After receiving the data, the edge node parallelly loads the corresponding operation data into the hierarchical fault detection model deployed locally; the associated production device is independently fault-identified through the single-fault identification model to obtain a single-identification result; when there is an abnormal flag in the single-identification result, the corresponding operation anomaly detection engine is further activated to perform associated defect trend detection on the operation data of the corresponding device to obtain an operation anomaly detection result. The edge node performs operation and maintenance backtracking analysis on the corresponding device according to the set 1 single-identification result and its corresponding set 1 operation anomaly detection result, extracts the device combination, risk level, and trend factor that constitute the operation and maintenance focus object of this node, and constructs the operation and maintenance requirement characteristics representing the operation and maintenance object, operation and maintenance priority, and response strategy. After the edge node completes the construction of the operation and maintenance requirement characteristics, through the communication connection established with the production line maintenance cloud, the N operation and maintenance requirement characteristics generated by the N edge nodes are synchronously sent to the production line maintenance cloud.

[0044] Furthermore, the method for operating the N edge nodes in real time for defect coupling trend analysis and sending the N operation and maintenance requirement characteristics obtained from the analysis to the production line maintenance cloud includes: The first edge node intermittently receives M sets of device operation sequences transmitted back by the M associated production devices based on a preset operation and maintenance time window; parallelly loads the M sets of device operation sequences into the M hierarchical fault detection models, performs single-fault identification through the M single-fault identification models, and outputs M single-identification results; maps and activates Q operation anomaly detection engines according to Q single-identification results with a value of 1 in the M single-identification results; parallelly loads Q sets of device operation sequences corresponding to the Q single-identification results into the Q operation anomaly detection engines for associated defect detection, and outputs Q operation anomaly detection results; performs device backtracking based on P operation anomaly detection results with a value of 1 in the Q operation anomaly detection results and the Q single-identification results to obtain the first operation and maintenance requirement feature.

[0045] Preferably, each edge node (taking the first edge node as an example) intermittently receives M sets of device operation sequences transmitted back by the M associated production devices corresponding to it based on a preset operation and maintenance time window; the edge node parallelly loads the received operation data into the M hierarchical fault detection models deployed for further analysis; each fault detection model performs preliminary fault identification through a single-fault identification model and outputs M single-identification results. Then, according to Q single-identification results marked as "1" in the single-identification results, Q corresponding operation anomaly detection engines are activated to parallelly load the operation data sequences of these devices into the corresponding Q operation anomaly detection engines for associated defect detection, and Q operation anomaly detection results are output. Then, based on P anomaly detection results marked as "1" in these operation anomaly detection results, a backtracking analysis is performed together with the corresponding single-fault identification results, and finally the first operation and maintenance requirement feature is determined and output.

[0046] Furthermore, the method further includes: If all the M single-identification results are set to 0, the M operation anomaly detection engines are not activated and the operation and maintenance time window is updated; after updating the operation and maintenance time window, the first edge node is run to perform defect coupling trend analysis on the first group of associated production devices.

[0047] If the recognition results of M monomers are all set to 0, that is, no monomer faults are detected, then the corresponding M abnormal operation detection engines are not activated, and no abnormal operation detection is performed. At this time, the first edge node will update its operation and maintenance time window according to a preset policy, usually adjusted based on a specific time interval or device state change. The updated operation and maintenance time window will be used as a new time window to start collecting and analyzing the operation data of the device again. After updating the operation and maintenance time window, the first edge node will continue to monitor and analyze the operation status of the first group of associated production devices and perform defect coupling trend analysis. Specifically, the edge node will re-evaluate the operation data and historical defect records of relevant devices based on the updated time window, perform defect pattern analysis, and identify potential operation deviations or abnormal trends.

[0048] The production line maintenance cloud receives and performs operation and maintenance strategy analysis and output according to the N operation and maintenance requirement characteristics.

[0049] The production line maintenance cloud receives the operation and maintenance requirement characteristics sent by each edge node through the communication connection with multiple edge nodes; through operation and maintenance strategy analysis, the cloud system outputs corresponding maintenance strategies and decisions, including maintenance suggestions for specific devices or processes, such as regular inspections, fault prevention measures, equipment replacement cycles, etc.

[0050] In summary, the embodiments of the present application have at least the following technical effects: First, based on a preset defect window, the historical defect records are locally called, where the historical defect records include N defect batch sets of N defect types. Then, the process records of the stainless steel water pipe production line are traced back according to the N defect batch sets to obtain N process parameter sets. Next, the N process parameter sets are abnormally identified, and process defect coupling analysis is performed based on the identification results to obtain N groups of associated processes. Then, N edge nodes are deployed on the stainless steel water pipe production line according to the N groups of associated processes, where the N edge nodes are communicatively connected to the N groups of associated production devices of the N groups of associated processes. Finally, the N edge nodes are run in real time for defect coupling trend analysis, and the N operation and maintenance requirement characteristics obtained from the analysis are sent to the production line maintenance cloud, where the production line maintenance cloud is communicatively connected to the N edge nodes; the production line maintenance cloud receives and performs operation and maintenance strategy analysis and output according to the N operation and maintenance requirement characteristics. This solves the technical problems of the lack of pertinence of the maintenance strategy of the stainless steel water pipe production line and low operation and maintenance efficiency in the prior art, and achieves the technical effects of improving the accuracy of the maintenance strategy and the maintenance efficiency of the production line.

[0051] Embodiment 2, based on the same inventive concept as the method for maintaining a stainless steel water pipe production line based on big data analysis in the foregoing embodiment, as Figure 2 shown, the present application provides a system for maintaining a stainless steel water pipe production line based on big data analysis, where the system includes: A historical data calling module 11, configured to locally call historical defect records based on a preset defect window, where the historical defect records include N defect batch sets of N defect types; a backtracking module 12, configured to backtrack the process records of the stainless steel water pipe production line according to the N defect batch sets to obtain N process parameter sets; a coupling analysis module 13, configured to perform anomaly identification on the N process parameter sets and perform process defect coupling analysis based on the identification results to obtain N groups of associated processes; an edge node deployment module 14, configured to deploy N edge nodes on the stainless steel water pipe production line according to the N groups of associated processes, where the N edge nodes are communicatively connected to N groups of associated production equipment of the N groups of associated processes; an analysis module 15, configured to run the N edge nodes in real time for defect coupling trend analysis and send N operation and maintenance requirement features obtained from the analysis to the production line maintenance cloud, where the production line maintenance cloud is communicatively connected to the N edge nodes; an operation and maintenance strategy output module 16, configured to receive and perform operation and maintenance strategy analysis and output according to the N operation and maintenance requirement features by the production line maintenance cloud.

[0052] Further, the coupling analysis module 13 is configured to execute the following method: Interactively obtain multiple baseline operation characteristics of multiple process production equipment in the stainless steel water pipe production line; use the multiple baseline operation characteristics to traverse the first process parameter set for abnormal operation parameter identification to obtain H abnormal process parameter sequences, where the first process parameter set includes H historical process parameter sequences; calculate the operation deviation of the H abnormal process parameter sequences according to the multiple baseline operation characteristics to obtain H abnormal deviation degree sequences; perform process defect coupling analysis according to the H abnormal process parameter sequences and H abnormal deviation degree sequences, and screen and locate the first group of associated production equipment among the multiple process production equipment; and so on, perform anomaly identification on the N process parameter sets and perform process defect coupling analysis based on the identification results to obtain the N groups of associated processes.

[0053] Further, the coupling analysis module 13 is configured to execute the following method: Aggregate the production equipment of the H abnormal process parameters to obtain multiple abnormal frequency characteristics of the multiple process production equipment; aggregate the H abnormal deviation degree sequences to obtain multiple abnormal probability contributions of the multiple process production equipment; perform defect anomaly association quantification according to the multiple abnormal probability contributions and multiple abnormal frequency characteristics, and output multiple defect anomaly contribution degrees; screen and obtain the first group of associated production equipment from the multiple process production equipment according to a preset contribution threshold and the multiple defect anomaly contribution degrees; and extract the first group of associated processes of the first group of associated production equipment based on process mapping.

[0054] Further, the coupling analysis module 13 is used to execute the following method: Through the aggregation of production equipment, obtain multiple abnormal deviation degrees and multiple abnormal deviation time windows of the production equipment in the first process; call the first deviation weight and the first time weight of the production equipment in the first process from the process characteristic weight library; perform weighted fusion of the multiple abnormal deviation degrees and the multiple abnormal deviation time windows according to the first deviation weight and the first time weight, and output the first abnormal probability contribution; and so on, through the aggregation of production equipment for the H abnormal deviation degree sequences, obtain the multiple abnormal probability contributions.

[0055] Further, the edge node deployment module 14 is used to execute the following method: Locally call the M device fault data sets of M associated production equipment in the first group of associated production equipment; perform fault level - type identification on the M device fault data sets to obtain M device fault level - type sets; screen and aggregate the H abnormal process parameter sequences according to the M associated production equipment to obtain M defect abnormal parameter sets; use the M device fault data sets, the M device fault level - type sets, and the M defect abnormal parameter sets to perform hierarchical modeling to obtain M hierarchical fault detection models; complete the local deployment of the first edge node by paralleling the M hierarchical fault detection models.

[0056] Further, the edge node deployment module 14 is used to execute the following method: Perform mapping - association storage of the M device fault data sets and the M device fault level - type sets based on the knowledge graph to complete the construction of M single - entity fault recognition models; perform fluctuation analysis on the M defect abnormal parameter sets and output M defect fluctuation characteristics; construct M operation abnormal detection engines based on the M defect fluctuation characteristics; map and cascade the M single - entity fault recognition models and the M operation abnormal detection engines according to the M associated production equipment to complete the construction of the M hierarchical fault detection models.

[0057] Further, the analysis module 15 is used to execute the following method: The first edge node intermittently receives M sets of device operation sequences transmitted back by the M associated production devices based on a preset operation and maintenance time window; parallelly loads the M sets of device operation sequences into the M hierarchical fault detection models, performs single-fault identification via the M single-fault identification models, and outputs M single-identification results; maps and activates Q operation anomaly detection engines according to Q single-identification results with a value of 1 in the M single-identification results; parallelly loads Q sets of device operation sequences corresponding to the Q single-identification results into the Q operation anomaly detection engines for associated defect detection, and outputs Q operation anomaly detection results; performs device backtracking based on P operation anomaly detection results with a value of 1 in the Q operation anomaly detection results and the Q single-identification results to obtain the first operation and maintenance requirement characteristics.

[0058] Further, the coupling analysis module 13 is used to execute the following method: Calculate the average deviation degree of the multiple anomaly deviation degrees; perform cumulative deviation solution on the multiple anomaly deviation time windows to obtain the cumulative deviation time; perform weighted fusion of the average deviation degree and the cumulative deviation time according to the first deviation weight and the first time weight, and output the first anomaly probability contribution.

[0059] Further, the analysis module 15 is used to execute the following method: If all the M single-identification results are set to 0, the M operation anomaly detection engines are not activated and the operation and maintenance time window is updated; after updating the operation and maintenance time window, the first edge node is run to perform defect coupling trend analysis on the first group of associated production devices.

[0060] It should be noted that the above order of the embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above describes specific embodiments of this specification. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0061] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0062] This specification and the drawings are only exemplary descriptions of the present application and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. A method for maintaining a stainless steel water pipe production line based on big data analysis, characterized in that The method includes: Locally invoking historical defect records based on a preset defect window, where the historical defect records include N defect batch sets of N defect types; Retrospectively recording the processes of the stainless steel water pipe production line according to the N defect batch sets to obtain N process parameter sets; Identifying anomalies in the N process parameter sets and performing process defect coupling analysis based on the identification results to obtain N sets of associated processes; Deploying N edge nodes on the stainless steel water pipe production line according to the N sets of associated processes, where the N edge nodes are communicatively connected to N sets of associated production equipment of the N sets of associated processes; Running the N edge nodes in real time for defect coupling trend analysis and sending the N operation and maintenance requirement characteristics obtained from the analysis to the production line maintenance cloud, where the production line maintenance cloud is communicatively connected to the N edge nodes; The production line maintenance cloud receives and performs operation and maintenance strategy analysis and output according to the N operation and maintenance requirement characteristics.

2. The method for maintaining a stainless steel water pipe production line based on big data analysis according to claim 1, wherein, Identifying anomalies in the N process parameter sets and performing process defect coupling analysis based on the identification results to obtain N sets of associated processes, the method includes: Interactively obtaining multiple reference operation characteristics of multiple process production equipment in the stainless steel water pipe production line; Using the multiple reference operation characteristics to traverse the first process parameter set for identifying abnormal operation parameters, to obtain H abnormal process parameter sequences, where the first process parameter set includes H historical process parameter sequences; Calculating operation deviations for the H abnormal process parameter sequences according to the multiple reference operation characteristics to obtain H abnormal deviation degree sequences; Performing process defect coupling analysis according to the H abnormal process parameter sequences and H abnormal deviation degree sequences, and screening and positioning the first set of associated production equipment among the multiple process production equipment; And so on, identifying anomalies in the N process parameter sets and performing process defect coupling analysis based on the identification results to obtain the N sets of associated processes.

3. The maintenance method of the stainless steel water pipe production line based on big data analysis according to claim 2, characterized in that, Performing process defect coupling analysis according to the H abnormal process parameter sequences and H abnormal deviation degree sequences, and screening and positioning the first set of associated production equipment among the multiple process production equipment, the method includes: Aggregating the H abnormal process parameters by production equipment to obtain multiple abnormal frequency characteristics of the multiple process production equipment; Aggregating the H abnormal deviation degree sequences by production equipment to obtain multiple abnormal probability contributions of the multiple process production equipment; Performing defect anomaly association quantification according to the multiple abnormal probability contributions and multiple abnormal frequency characteristics, and outputting multiple defect anomaly contribution degrees; Selecting the first set of associated production equipment from the multiple process production equipment according to a preset contribution threshold and the multiple defect anomaly contribution degrees; Extracting the first set of associated processes of the first set of associated production equipment based on process mapping.

4. The method for maintaining a stainless steel water pipe production line based on big data analysis according to claim 3, wherein, Aggregating the H abnormal deviation degree sequences by production equipment to obtain multiple abnormal probability contributions of the multiple process production equipment, the method includes: Aggregating by production equipment to obtain multiple abnormal deviation degrees and multiple abnormal deviation time windows of the first process production equipment; Invoke the first deviation weight and the first time weight of the first process production equipment from the process characteristic weight library; Based on the first deviation weight and the first time weight, perform weighted fusion of the multiple abnormal deviation degrees and the multiple abnormal deviation time windows, and output the first abnormal probability contribution; And so on, by aggregating the production equipment for the H abnormal deviation degree sequences, obtain the multiple abnormal probability contributions.

5. The method for maintaining a stainless steel water pipe production line based on big data analysis according to claim 2, characterized in that, Deploy N edge nodes according to the N groups of associated processes in the stainless steel water pipe production line, and the method includes: Locally invoke the M device fault data sets of the M associated production devices in the first group of associated production devices; Perform fault level - type identification on the M device fault data sets to obtain M device fault level - type sets; According to the M associated production devices, screen and aggregate the H abnormal process parameter sequences to obtain M defective abnormal parameter sets; Use the M device fault data sets, the M device fault level - type sets, and the M defective abnormal parameter sets to perform hierarchical modeling to obtain M hierarchical fault detection models; Complete the local deployment of the first edge node by paralleling the M hierarchical fault detection models.

6. The maintenance method of the stainless steel water pipe production line based on big data analysis according to claim 5, characterized in that, Use the M device fault data sets, the M device fault level - type sets, and the M defective abnormal parameter sets to perform hierarchical modeling to obtain M hierarchical fault detection models, and the method includes: Based on the knowledge graph, perform mapping - associated storage of the M device fault data sets and the M device fault level - type sets to complete the construction of M single - entity fault recognition models; Perform fluctuation analysis on the M defective abnormal parameter sets and output M defective fluctuation characteristics; Construct M operation abnormal detection engines based on the M defective fluctuation characteristics; According to the M associated production devices, map - cascade the M single - entity fault recognition models and the M operation abnormal detection engines to complete the construction of the M hierarchical fault detection models.

7. The method for maintaining a stainless steel water pipe production line based on big data analysis according to claim 6, wherein Run the N edge nodes in real - time for defect coupling trend analysis, and send the N operation and maintenance requirement characteristics obtained from the analysis to the production line maintenance cloud, and the method includes: The first edge node intermittently receives the M groups of device operation sequences transmitted back by the M associated production devices based on a preset operation and maintenance time window; Parallel - load the M groups of device operation sequences into the M hierarchical fault detection models, and perform single - entity fault recognition through the M single - entity fault recognition models to output M single - entity recognition results; Map - activate Q operation abnormal detection engines according to the Q single - entity recognition results with a value of 1 in the M single - entity recognition results; Parallel - load the Q groups of device operation sequences corresponding to the Q single - entity recognition results into the Q operation abnormal detection engines for associated defect detection and output Q operation abnormal detection results; Perform device backtracking according to the P operation abnormal detection results with a value of 1 in the Q operation abnormal detection results and the Q single - entity recognition results to obtain the first operation and maintenance requirement characteristic.

8. The maintenance method of a stainless steel water pipe production line based on big data analysis according to claim 4, characterized in that, Based on the first deviation weight and the first time weight, perform weighted fusion of the multiple abnormal deviation degrees and the multiple abnormal deviation time windows, and output the first abnormal probability contribution, and the method includes: Calculate the average deviation degree of the multiple abnormal deviation degrees; Solve the cumulative deviation of the multiple abnormal deviation time windows to obtain the cumulative deviation time; Perform weighted fusion of the average deviation degree and the cumulative deviation time according to the first deviation weight and the first time weight, and output the first abnormal probability contribution.

9. The maintenance method of a stainless steel water pipe production line based on big data analysis according to claim 7, characterized in that The method further includes: If the M single-body recognition results are all set to 0, deactivate the M running anomaly detection engines and update the operation and maintenance time window; After updating the operation and maintenance time window, run the first edge node to perform defect coupling trend analysis on the first group of associated production equipment.

10. A maintenance system for a stainless steel water pipe production line based on big data analysis, characterized in that, For implementing the stainless steel water pipe production line maintenance method based on big data analysis according to any one of claims 1-9, the system includes: A historical data calling module, configured to locally call historical defect records based on a preset defect window, where the historical defect records include N defect batch sets of N defect types; A backtracking module, configured to backtrack the process records of the stainless steel water pipe production line according to the N defect batch sets to obtain N process parameter sets; A coupling analysis module, configured to perform anomaly recognition on the N process parameter sets, and perform process defect coupling analysis based on the recognition results to obtain N groups of associated processes; An edge node deployment module, configured to deploy N edge nodes on the stainless steel water pipe production line according to the N groups of associated processes, where the N edge nodes are communicatively connected to N groups of associated production equipment of the N groups of associated processes; An analysis module, configured to run the N edge nodes in real time to perform defect coupling trend analysis, and send the N operation and maintenance requirement features obtained from the analysis to the production line maintenance cloud, where the production line maintenance cloud is communicatively connected to the N edge nodes; An operation and maintenance strategy output module, configured to receive and perform operation and maintenance strategy analysis and output according to the N operation and maintenance requirement features by the production line maintenance cloud.

Images