Composite monitoring system of petroleum pipe automatic production line

Abstract

The invention relates to the technical field of automatic production monitoring, and particularly discloses a composite monitoring system for an automatic production line of petroleum pipes, which realizes the stable processing and information identification process of the automatic production line of the petroleum pipes through the composite monitoring system. The system module comprises a sensor wiring module which is used for being connected with a sensor on an automatic production line to obtain the position layout of the sensor, determining wiring information and transmitting process signal parameters on the wiring information to a main control center; a key area selection module is arranged and used for determining key area information according to the process signal parameters obtained by the main control center; a composite monitoring module is arranged and used for importing the key area information into a composite numerical control platform to be matched with a corresponding process for monitoring, and an abnormal quality coefficient is obtained; and a monitoring identification module is arranged and is used for carrying out fault identification according to the abnormal quality coefficient and outputting fault early warning information.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic production monitoring, in particular to a composite monitoring system for an automatic production line of petroleum tubular goods. Background Art

[0002] The monitoring system of the automatic production line of oil pipes mainly performs automatic monitoring and analysis of non-destructive flaw detection at each processing stage of oil-specific pipes, and realizes the system to judge the production quality of oil pipes.

[0003] The various data involved in existing automatic oil pipe production lines are usually stored independently in each production unit, and therefore lack a unified platform for integration. For example, DCS system data can only be viewed at the local operation station and cannot be synchronized to the management system in real time, which easily forms a data barrier. In addition, additional processing programs are required on the automatic production line to obtain and store various multi-source data, which leads to poor correlation between multi-source data and affects the subsequent precise positioning of the oil pipe production status.

[0004] Since the data involved in the oil pipe monitoring system is diverse and lacks identification and positioning, a composite monitoring system is needed to achieve stable processing and information recognition in the oil pipe automatic production line. Summary of the Invention

[0005] The purpose of the present invention is to provide a composite monitoring system for an automatic production line of oil pipes to solve the following technical problems: How to achieve stable processing and information recognition process of oil pipe automatic production line through composite monitoring system.

[0006] The purpose of the present invention can be achieved through the following technical solutions: A composite monitoring system for an automatic production line of oil pipes, comprising: The sensor wiring module is used to connect the sensors on the automatic production line to obtain the sensor position layout, determine the wiring information, and transmit the process signal parameters on the wiring information to the main control center; The key area selection module is used to determine the key area information based on the process signal parameters obtained by the main control center; The composite monitoring module is used to import key area information into the composite CNC platform to match the corresponding process for monitoring and obtain abnormal quality coefficients; The monitoring and identification module is used to identify faults based on abnormal quality coefficients and output fault warning information.

[0007] Preferably, the sensor wiring module comprises: Sensors are used to collect real-time operating signals from various equipment on the automatic production line. These signals include temperature, corrosion, quality, and optical signals. The acquisition unit is used to convert the real-time operation signal into process signal parameters using multi-channel acquisition technology. The process signals include temperature mutation, corrosion rate, defect size, and strain mutation; the process signal parameters include temperature mutation parameters, corrosion rate parameters, defect size parameters, and strain mutation parameters; Wireless communication unit, used to transmit process signal parameters and ensure parameter data integrity.

[0008] Preferably, the process of determining the key area information includes: Extract features of process signal parameters to obtain process signal parameter features; Perform feature fusion on process signal parameter features and perform weighted analysis to obtain risk value; Based on K-means clustering, the risk value level is divided, and the abnormal information is matched with the historical fault database data. The process signal parameters with abnormal information are marked as negative samples, and the process signal parameters without abnormal information are marked as positive samples; A fault recognition model is built using negative and positive samples, and a feedback learning mechanism is established for the fault recognition model to identify high-fault process signals and obtain the weights of high-fault process signals; When the similarity between the high-fault process signal and the fault event feature is greater than 0.8 in combination with the historical fault database, the corresponding process operation area is recorded as the key area, and the key area is monitored to obtain key area information.

[0009] Preferably, the formula for obtaining the risk value by weighted analysis is: ; in, is the risk value, is the characteristic vector of temperature mutation, is the characteristic vector of corrosion rate, is the characteristic vector of defect size, is the characteristic vector of strain mutation; 、 、 、 The weights are obtained by training the historical data of the corresponding features.

[0010] Preferably, the risk value level classification method is: Number of clusters Determine: The silhouette coefficient method is used to evaluate clustering density, and the cluster corresponding to the silhouette coefficient peak is selected. value; Clustering feature input: risk value of process signal in historical fault database As input; Definition of level boundaries: After clustering, the threshold is divided according to the location of the cluster center: when When ≤0.07, it is judged as a low-risk cluster. The value is low risk; When 0.07< When ≤0.58, it is judged as a medium-risk cluster. The value is medium risk; when When it is greater than 0.58, it is judged as a high-risk cluster. The value is high risk.

[0011] When the real-time signal risk value falls into the high-risk cluster, it is marked as a negative sample. When the real-time signal risk value falls into the low-risk cluster and the medium-risk cluster, it is marked as a positive sample.

[0012] Preferably, the high-fault process signal weight is obtained in the following manner: Constructing feature vectors , , and used as input to the logistic regression model; Logistic regression model outputs abnormal probability ;in Indicates abnormality; By the abnormal probability Make a judgment, when the abnormal probability When <0.01, the weight ranking of the process signals is output, and the feature vectors in the top 20% of the weight ranking are identified as high-fault signals; and the weights of the high-fault process signals are obtained.

[0013] Preferably, the composite monitoring module is used to import key area information into the composite numerical control platform to match the corresponding process for monitoring and obtain abnormal quality coefficients, and the method includes: Obtain key area coordinates and input them into the CNC platform process library for monitoring and analysis; The abnormal quality coefficient is obtained as follows:

[0014] in, is the abnormal quality coefficient; is the weight of abnormal frequency, is the weight of the degree of deviation, The number of points where the risk value exceeds the limit per unit time for the abnormal point; To count the number of risk values ​​per unit time for all points, including those that exceed the limit and those that do not exceed the limit; is the total number of real-time input feature vectors, and ∈ ; For the The measured values ​​of the input eigenvectors; Input feature vector threshold for the system.

[0015] Preferably, the monitoring and identification module is used to identify faults based on the abnormal quality coefficient and output fault warning information. The process is as follows: Will With preset interval To compare: like <0.3, the current process signal is judged to be normal and the operation log is recorded; If 0.3≤ If the value is less than 0.7, the current process signal is judged to be at risk, an early warning signal is generated, and an operation and maintenance inspection is performed; like If ≥0.7, it is judged that the current process signal has a fault and the machine is stopped for inspection.

[0016] Beneficial effects of the present invention: (1) The present invention uses a sensor wiring module to connect to the automatic production line and obtain the position layout of various sensors. By integrating the sensor network, the signal transmission path is optimized to transmit real-time process parameters to the main control center, thereby improving the response speed and reliability of the entire production system data.

[0017] (2) The present invention realizes the determination process of key area information through the key area selection module. Through intelligent processing of process signal parameters, key area information is determined based on the process signal parameters obtained by the main control center to accurately locate high-risk areas that need priority monitoring in the production line. In addition, the key area selection module realizes multi-source data analysis, dynamic threshold optimization and incorporates spatial topology algorithm to realize the accurate selection process of key areas.

[0018] (3) After the composite monitoring module determines the deployment of the selected module for the key area, the present invention accelerates the monitoring process of the composite CNC platform for high-risk areas, imports the key area information into the composite CNC platform to match the corresponding process for monitoring, obtains the coordinates of the key area through spatial mapping technology, associates the coordinate area with the corresponding process parameters, realizes the matching of spatial coordinate conversion with the production process sequence, synchronously loads the monitoring threshold, and completes the real-time monitoring analysis process and the abnormality identification process.

[0019] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0021] Figure 1 This is a module diagram of a composite monitoring system for an automatic production line of oil pipes according to the present invention; Figure 2 This is a unit diagram of the sensor wiring module of the present invention. DETAILED DESCRIPTION

[0022] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0023] See also Figure 1 As shown, the present invention is a composite monitoring system for an automatic production line of oil pipes, comprising: The sensor wiring module is used to connect the sensors on the automatic production line to obtain the sensor position layout, determine the wiring information, and transmit the process signal parameters on the wiring information to the main control center; The key area selection module is used to determine the key area information based on the process signal parameters obtained by the main control center; The composite monitoring module is used to import key area information into the composite CNC platform to match the corresponding process for monitoring and obtain abnormal quality coefficients; The monitoring and identification module is used to identify faults based on abnormal quality coefficients and output fault warning information.

[0024] In the above technical solution, the problem of inaccurate sensor positioning of data processed by the automatic production line is solved through the sensor wiring module. Multiple sensors are set up, and by connecting to the automatic production line and obtaining the position layout of various sensors, the sensor network is integrated and the signal transmission path is optimized to transmit real-time process parameters to the main control center, thereby improving the response speed and reliability of the entire production system data. In the factory's integrated wiring process, the port status is identified by an intelligent patch panel, and the fiber optic backbone network is integrated with gigabit access switches to reduce cable redundancy and lower signal attenuation. The key area selection module realizes intelligent processing of process signal parameters, determines key area information based on process signal parameters obtained by the main control center, and accurately locates high-risk areas that require priority monitoring in the production line. The key area selection module also realizes multi-source data analysis, dynamic threshold optimization, and incorporates spatial topology algorithms to achieve accurate key area selection. After the deployment of the selected modules for key areas is determined through the composite monitoring module, the monitoring process of the composite CNC platform is accelerated for high-risk areas. Specifically, the key area information is imported into the composite CNC platform to match the corresponding process for monitoring. The coordinates of the key areas are obtained through spatial mapping technology, and the coordinate areas are associated with the corresponding process parameters to achieve matching of spatial coordinate conversion with the production process sequence. The monitoring threshold is loaded synchronously to complete the real-time monitoring and analysis process. The synchronization process of multiple sensor data (time series alignment technology) is also realized to obtain the abnormal quality coefficient. Furthermore, the abnormal quality coefficient level can be identified through the monitoring and identification module, and fault identification can be performed based on the abnormal quality coefficient, and fault warning information can be output to improve the fault response efficiency of the automatic production line.

[0025] As an embodiment of the present invention, please refer to Figure 2 As shown, the sensor wiring module includes: Sensors are used to collect real-time operating signals from various equipment on the automatic production line. These signals include temperature, corrosion, quality, and optical signals. The acquisition unit is used to convert the real-time operation signal into process signal parameters using multi-channel acquisition technology. The process signals include temperature mutation, corrosion rate, defect size, and strain mutation; the process signal parameters include temperature mutation parameters, corrosion rate parameters, defect size parameters, and strain mutation parameters; Wireless communication unit, used to transmit process signal parameters and ensure parameter data integrity.

[0026] In the above technical solution, the sensor wiring module consists of a sensor, an acquisition unit, and a wireless communication unit. The sensors are used to collect various real-time operating signals, including temperature signals, corrosion signals, quality signals, and optical signals. Temperature signals are usually collected through temperature sensors, such as thermocouples, to monitor the surface and internal temperature of the equipment in a non-contact infrared manner. Corrosion signals are collected through electrochemical sensors to monitor the concentration of corrosion particles in the liquid environment in real time, and the corrosion rate parameters are calculated using ice cleaning and corrosion rate algorithms. Quality signals are captured through machine vision sensors to capture the quality of the surface of the oil pipe, usually including surface defects such as scratches and pits, and the defect size parameters are quantified through image processing algorithms. Optical signals are detected through fiber grating sensors to detect sudden changes in material strain, allowing for high-sensitivity health monitoring of the internal structure of the equipment. The acquisition unit converts the above-mentioned real-time operation signal into a quantitative conversion process of converting the analog signal into a digital signal; the technologies involved in the conversion process mainly include multi-channel synchronous acquisition technology and signal conditioning and conversion process technology; multi-channel synchronous acquisition technology is used to meet the high-density sensor layout, and real-time signal acquisition is achieved through synchronous sampling. It is triggered by FPGA control and the sampling rate is set at 100kHz to eliminate the time difference of dynamic signals and ensure the accuracy of transient signals such as strain mutations; the signal conditioning and conversion process technology mainly involves the processing of analog signals and the digital conversion process: first, it has a built-in programmable gain amplifier PGA and anti-aliasing filter to suppress power frequency interference and realize the smooth acquisition of analog signals; then, a 24-bit high-precision ADC (signal-to-noise ratio ≥110dB) is used to convert the original signal into a process signal and obtain the corresponding process signal parameters. The obtained process signals include temperature mutation, corrosion rate, defect size, and strain mutation; the process signal parameters include temperature mutation parameters, corrosion rate parameters, defect size parameters, and strain mutation parameters; A wireless communication unit is set up to transmit process signal parameters through a transmission protocol, ensure the timing of multi-node data transmission, and ensure the integrity of parameter data by adding redundant check bits.

[0027] As an embodiment of the present invention, the process of determining key area information includes: Extract features of process signal parameters to obtain process signal parameter features; Perform feature fusion on process signal parameter features and perform weighted analysis to obtain risk value; Based on K-means clustering, the risk value level is divided, and the abnormal information is matched with the historical fault database data. The process signal parameters with abnormal information are marked as negative samples, and the process signal parameters without abnormal information are marked as positive samples; A fault recognition model is built using negative and positive samples, and a feedback learning mechanism is established for the fault recognition model to identify high-fault process signals and obtain the weights of high-fault process signals; When the similarity between the high-fault process signal and the fault event feature is greater than 0.8 in combination with the historical fault database, the corresponding process operation area is recorded as the key area, and the key area is monitored to obtain key area information.

[0028] In the above technical solution, the key areas are divided by combining industrial intelligent control and fault diagnosis through the process of determining key area information. First, feature extraction is performed. Specifically, through time domain / frequency domain feature extraction, the mean (stability), variance (volatility), and effective value (energy intensity) of the process signal are calculated to extract the process temperature mutation parameter characteristics, corrosion rate parameter characteristics, defect size parameter characteristics, and strain mutation parameter characteristics. The process signal characteristics are fused through multi-sensor feature fusion technology to eliminate the timing deviation between sensors. Then, the risk value calculation process is carried out to determine the proximity of each process operation to the ideal fault, and then output the risk value. Furthermore, risk value levels are divided based on K-means clustering. By dividing the risk value levels, a fault identification model is introduced: specifically, by constructing a fault identification model to implement cluster analysis of samples, and combining the abnormal information with the historical fault database data, the process signal parameters with abnormal information are marked as negative samples, and the process signal parameters without abnormal information are marked as positive samples; the fault identification model is built with negative samples and positive samples, and a feedback learning mechanism is established for the fault identification model to identify high-fault process signals and obtain the weights of high-fault process signals; Finally, when the similarity between the high-fault process signal and the fault event feature is greater than 0.8 based on the historical fault database, the corresponding process operation area is recorded as the key area, and the key area is monitored to obtain key area information; The above-mentioned method improves the efficiency of fault determination and reduces the false alarm rate by mining the feature of process signal parameters, especially deploying the fusion nodes of priority sensors in the high-fault processes in the oil pipe processing, and combining it with edge computing to achieve response.

[0029] As an embodiment of the present invention, the formula for obtaining the risk value by weighted analysis calculation is:

[0030] in, is the risk value, is the characteristic vector of temperature mutation, is the characteristic vector of corrosion rate, is the characteristic vector of defect size, is the characteristic vector of strain mutation; 、 、 、 The weights are obtained by training the historical data of the corresponding features.

[0031] In the above technical solution, the risk value is quantified by dividing the risk level based on K-means clustering, and the temperature mutation parameter characteristics, corrosion rate parameter characteristics, defect size parameter characteristics, and strain mutation parameter characteristics are vectorized through weighted calculation to obtain the risk values ​​of these feature vectors. The random forest feature importance analysis is used to ensure that each parameter feature has a corresponding weight; therefore, the risk value obtained after weighted calculation can effectively determine the high-probability failure areas in the historical fault database.

[0032] As an embodiment of the present invention, the risk value level classification method is as follows: Number of clusters Determine: The silhouette coefficient method is used to evaluate clustering density, and the cluster corresponding to the silhouette coefficient peak is selected. value; Clustering feature input: risk value of process signal in historical fault database As input; Definition of level boundaries: After clustering, the threshold is divided according to the location of the cluster center: when When ≤0.07, it is judged as a low-risk cluster. The value is low risk; When 0.07< When ≤0.58, it is judged as a medium-risk cluster. The value is medium risk; when When it is greater than 0.58, it is judged as a high-risk cluster. The value is high risk.

[0033] When the real-time signal risk value falls into the high-risk cluster, it is marked as a negative sample. When the real-time signal risk value falls into the low-risk cluster and the medium-risk cluster, it is marked as a positive sample.

[0034] In the above technical solution, the number of clusters needs to be determined in the clustering algorithm. , by quantifying the intra-cluster compactness (cohesion) and inter-cluster separation (separation) through the silhouette coefficient method, select the cluster that maximizes the silhouette coefficient value to ensure the optimal cluster structure; the silhouette coefficient calculation formula is:

[0035] in, For samples The average distance to other samples in the same cluster (cohesion), For samples Average distance to the nearest cluster (separation) Overall silhouette coefficient For all samples The average value of The closer it is to 1, the better the clustering effect. In the actual verification process, by randomly initializing the clusters multiple times and calculating the fluctuation of the silhouette coefficient, the stability of the clustering results can be verified to avoid local optimal solutions.

[0036] In addition, cluster center positioning: after the number of clusters is determined, the location of the cluster center (center of mass) directly determines the risk threshold; judgment: when When ≤0.07, it is judged as a low-risk cluster. The value is low risk; when 0.07< When ≤0.58, it is judged as a medium-risk cluster. The value is medium risk; when When it is greater than 0.58, it is judged as a high-risk cluster. The value is high risk; Boundary rationality: When the silhouette coefficient is optimal, the similarity of samples within the cluster is high and the difference between clusters is significant, ensuring that the risk boundary can accurately distinguish different risk patterns. If the number of clusters is not selected properly, it may cause the cluster center to shift, causing the threshold to lose its discrimination; and the number of clusters directly affects the identification range of high-risk clusters; if If the value is too small, the high-risk cluster may contain too many medium / low-risk samples, resulting in mislabeling of negative samples (high-risk); If the value is too large, the high-risk cluster may miss some real high-risk samples, reducing the sensitivity of the model.

[0037] As an embodiment of the present invention, the method for obtaining the signal weight of the high-fault process is as follows: Constructing feature vectors , , and used as input to the logistic regression model; Logistic regression model outputs abnormal probability ;in Indicates abnormality; By the abnormal probability Make a judgment, when the abnormal probability When <0.01, the weight ranking of the process signal is output, and the feature vectors in the top 20% of the weight ranking are identified as high-fault signals; and the weight of the high-fault process signal is obtained.

[0038] In the above technical solution, the process of further weight analysis of the process signal is to extract multi-dimensional features from the process signal and construct a feature matrix ∈ ,in is the number of samples, is the number of features; then preprocess the extracted multidimensional features; build the model goal: predict the abnormal probability of process signals Constructing feature vectors , , and used as input to the logistic regression model; and logistic regression analysis was performed, when the abnormal probability When <0.01, the process signal is judged to be normal, triggering subsequent weight analysis; The weight sorting rules are used for judgment. The specific process is as follows: Get the weight vector of each process from the trained logistic regression model ; The sorting rules are: one is in descending order: by absolute value of weight Sorting (the larger the absolute value, the more significant the impact on the abnormality probability); and screening the top 20%: select the features with the top 20% weight ranking as high-fault signals (for example, when there are 100 features in total, take the top 20). The top 20 features are high-fault processes, and their weights are determined to be 20% or above.

[0039] As an embodiment of the present invention, the composite monitoring module is used to import key area information into the composite numerical control platform to match the corresponding process for monitoring and obtain the abnormal quality coefficient. The method includes: Obtain key area coordinates and input them into the CNC platform process library for monitoring and analysis; The abnormal quality coefficient is obtained as follows:

[0040] in, is the abnormal quality coefficient; is the weight of abnormal frequency, is the weight of the degree of deviation, The number of points where the risk value exceeds the limit per unit time for the abnormal point; To count the number of risk values ​​per unit time for all points, including those that exceed the limit and those that do not exceed the limit; is the total number of real-time input feature vectors, and ∈ ; For the The measured values ​​of the input eigenvectors; The characteristic vector threshold is input to the system, and is obtained through noise suppression technology. The adaptive window is determined through moving average filtering to obtain the corresponding characteristic vector threshold points.

[0041] In the above technical solution, the abnormal frequency term in the calculation of the abnormal quality coefficient includes the frequency density used to quantify the abnormality, and then the abnormal quality coefficient reflects the process stability, high-frequency abnormality It indicates the existence of systemic risks, thus ensuring timely abnormal feedback and adjustment based on frequency density. It is also used to reflect the severity of single-point deviation based on the deviation degree item, thereby capturing sudden high-risk events; the weight of abnormal frequency and the weight of the degree of deviation It is dynamically allocated according to the process characteristics. The oil pipe automatic production line designed in this paper mainly has low frequency and high loss. The response value can be set to 、 During the actual dynamic weight adjustment process, historical fault data can also be analyzed based on the LSTM model, and the weight coefficient can be updated weekly to reduce the false alarm rate.

[0042] Real-time acquisition of feature vectors through multi-sensor collaboration , using the trigger (encoder Z pulse) to align multiple source signals, eliminate time offset, and achieve timing synchronization; through edge computing to count overruns, embedded PLC (such as Siemens S7-1200) real-time statistics (Exceeding limit points) and (Total points), counting error ≤ 0.1%: In the stamping process, if the vibration exceeds the limit within 1 minute , total sampling points (sampling frequency 10Hz), then .

[0043] As an embodiment of the present invention, the monitoring and identification module is used to identify faults based on the abnormal quality coefficient and output fault warning information. The process is as follows: Will With preset interval To compare: like <0.3, the current process signal is judged to be normal and the operation log is recorded; If 0.3≤ If the value is less than 0.7, the current process signal is judged to be at risk, an early warning signal is generated, and an operation and maintenance inspection is performed; like If ≥0.7, it is judged that the current process signal has a fault and the machine is stopped for inspection.

[0044] In the above technical solution, further, the abnormal quality coefficient is performed in the monitoring and identification module Further fault identification, by setting thresholds to achieve different risk levels, when generating When the fault identification logic is triggered, <0.3, the current process signal is judged to be normal and the operation log is recorded; when 0.3≤ <0.7, the current process signal is judged to be at risk, an early warning signal is generated, and an operation and maintenance inspection is carried out; when If ≥0.7, it is judged that the current process signal has a fault and the machine is stopped for inspection.

[0045] The various embodiments in this specification are described in a progressive manner. Similar portions between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from the other embodiments. In particular, the device, apparatus, and non-volatile computer storage medium embodiments are generally similar to the method embodiments, so their descriptions are relatively simplified. For relevant details, refer to the descriptions of the method embodiments.

[0046] The foregoing description is of specific embodiments of this specification. Other embodiments are within the scope of the accompanying documents. In some cases, the actions or steps described in this application can be performed in an order different from that shown in the embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order shown or the sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0047] The above content is merely an example and explanation of the concept of the present invention. Those skilled in the art may make various modifications or additions to the described specific embodiments or replace them in a similar manner. As long as they do not deviate from the concept of the invention or exceed the scope defined in this application, they should all fall within the scope of protection of the present invention.

Claims

1. A composite monitoring system for an automatic production line of oil pipes, characterized in that: include: The sensor wiring module is used to connect the sensors on the automatic production line to obtain the sensor position layout, determine the wiring information, and transmit the process signal parameters on the wiring information to the main control center; The key area selection module is used to determine the key area information based on the process signal parameters obtained by the main control center; The process of determining the key area information includes: Extract features of process signal parameters to obtain process signal parameter features; Perform feature fusion on process signal parameter features and perform weighted analysis to obtain risk value; Based on K-means clustering, the risk value level is divided, and the abnormal information is matched with the historical fault database data. The process signal parameters with abnormal information are marked as negative samples, and the process signal parameters without abnormal information are marked as positive samples; A fault recognition model is built using negative and positive samples, and a feedback learning mechanism is established for the fault recognition model to identify high-fault process signals and obtain the weights of high-fault process signals; When the similarity between the high-fault process signal and the fault event feature is greater than 0.8 based on the historical fault database, the corresponding process operation area is recorded as the key area, and the key area is monitored to obtain key area information; The composite monitoring module is used to import key area information into the composite CNC platform to match the corresponding process for monitoring and obtain abnormal quality coefficients; The monitoring and identification module is used to identify faults based on abnormal quality coefficients and output fault warning information.

2. The composite monitoring system for an automatic production line of oil pipes according to claim 1, characterized in that: The sensor wiring module includes: Sensors are used to collect real-time operating signals of various equipment on the automatic production line, including temperature signals, corrosion signals, quality signals, and optical signals; An acquisition unit, configured to convert real-time operation signals into process signal parameters using a multi-channel acquisition technology, wherein the process signals include temperature mutation, corrosion rate, defect size, and strain mutation; and the process signal parameters include temperature mutation parameters, corrosion rate parameters, defect size parameters, and strain mutation parameters; Wireless communication unit, used to transmit process signal parameters and ensure parameter data integrity.

3. The composite monitoring system for an automatic production line of oil pipes according to claim 1, characterized in that: The formula for obtaining the risk value through weighted analysis is: ； in, is the risk value, is the characteristic vector of temperature mutation, is the characteristic vector of corrosion rate, is the characteristic vector of defect size, is the characteristic vector of strain mutation; 、 、 、 The weights are obtained by training the historical data of the corresponding features.

4. The composite monitoring system for an automatic production line of oil pipes according to claim 3, characterized in that: The risk value level classification method is as follows: Number of clusters Determine: The silhouette coefficient method is used to evaluate clustering density, and the cluster corresponding to the silhouette coefficient peak is selected. value; Clustering feature input: risk value of process signal in historical fault database As input; Definition of level boundaries: After clustering, the threshold is divided according to the location of the cluster center: when When ≤0.07, it is judged as a low-risk cluster. The value is low risk; When 0.07< When ≤0.58, it is judged as a medium-risk cluster. The value is medium risk; when When it is greater than 0.58, it is judged as a high-risk cluster. The value is high risk. When the real-time signal risk value falls into the high-risk cluster, it is marked as a negative sample. When the real-time signal risk value falls into the low-risk cluster and the medium-risk cluster, it is marked as a positive sample.

5. According to the composite monitoring system for an automatic production line of oil pipes in claim 3, the method for obtaining the signal weight of the high-fault process is: Constructing feature vectors , , and used as input to the logistic regression model; Logistic regression model outputs abnormal probability ;in Indicates abnormality; By the abnormal probability Make a judgment, when the abnormal probability When <0.01, the weight ranking of the process signals is output, and the feature vectors in the top 20% of the weight ranking are identified as high-fault signals; and the weights of the high-fault process signals are obtained.

6. The composite monitoring system for an automatic production line of oil pipes according to claim 5, characterized in that: The composite monitoring module is used to import key area information into the composite numerical control platform to match the corresponding process for monitoring and obtain abnormal quality coefficients. The method includes: Obtain the coordinates of the key areas and input them into the CNC platform process library for monitoring and analysis; The abnormal quality coefficient is obtained as follows: in, is the abnormal quality coefficient; is the weight of abnormal frequency, is the weight of the degree of deviation, The number of points where the risk value exceeds the limit per unit time for the abnormal point; To count the number of risk values ​​per unit time for all points, including those that exceed the limit and those that do not exceed the limit; is the total number of real-time input feature vectors, and ∈ ; For the The measured values ​​of the input eigenvectors; Enter the feature vector threshold for the system.

7. The composite monitoring system for an automatic production line of oil pipes according to claim 6, characterized in that: The monitoring and identification module is used to identify faults based on the abnormal quality coefficient and output fault warning information. The process is as follows: Will With preset interval To compare: like <0.3, the current process signal is judged to be normal and the operation log is recorded; If 0.3≤ If the value is less than 0.7, the current process signal is judged to be at risk, an early warning signal is generated, and an operation and maintenance inspection is performed; like If ≥0.7, it is judged that the current process signal has a fault and the machine is stopped for inspection.

Images