Monitoring and Warning Method and Device for Skid-mounted Equipment

Abstract

The present invention discloses a monitoring and early warning method and device for skid-mounted equipment, relating to the technical field of monitoring and early warning. The method includes: performing functional collaboration and fault impact causality analysis according to the collaborative functional blocks of the skid-mounted equipment to construct a causality association library; disassembling and clustering the causality association library to obtain a clustering result and corresponding relationship identifiers; searching for an early warning path based on the causality association library to determine hierarchical early warning information; and performing collaborative functional block early warning monitoring on the skid-mounted equipment according to the relationship identifiers and hierarchical early warning information to obtain an early warning identification result. The present invention solves the technical problems in the prior art that it is difficult to comprehensively and accurately master the causality of skid-mounted equipment, resulting in inaccurate monitoring and early warning, low safety and reliability, and poor management efficiency, and achieves the technical effect of realizing precise monitoring and early warning of the operating state of skid-mounted equipment, and improving the operating safety, reliability and management efficiency of skid-mounted equipment.

Description

Technical Field

[0001] The present invention relates to the technical field of monitoring and early warning, and particularly to a monitoring and early warning method and device for skid-mounted equipment. Background Art

[0002] Skid-mounted equipment plays a crucial role in many fields, such as petrochemical, energy extraction, industrial manufacturing and other industries. It usually integrates multiple functional blocks, and each functional block works together to achieve a specific production process or task. However, with the continuous improvement of the automation level and the increasing functional complexity of skid-mounted equipment, many challenges are faced during operation. Traditional monitoring technologies often only focus on local parameters or a single functional block of the equipment, lacking in-depth understanding and comprehensive analysis of the collaborative working relationship between functional blocks and the causal relationship of fault impacts. This limitation results in the inability to identify potential fault sources in a timely and accurate manner, and it is difficult to quickly locate the fault propagation path when a fault occurs, thus unable to achieve effective early warning. Due to the lack of a precise early warning mechanism, when a fault occurs in the equipment, it may quickly spread to the entire system, triggering a chain reaction, seriously affecting the operation safety and reliability of the equipment.

[0003] The prior art has technical problems such as difficulty in comprehensively and accurately grasping the causal relationship of skid-mounted equipment, resulting in inaccurate monitoring and early warning, low safety and reliability, and poor management efficiency. Summary of the Invention

[0004] The present application provides a monitoring and early warning method and device for skid-mounted equipment, aiming to solve the technical problems in the prior art that it is difficult to comprehensively and accurately grasp the causal relationship of skid-mounted equipment, resulting in inaccurate monitoring and early warning, low safety and reliability, and poor management efficiency.

[0005] In view of the above problems, the present application provides a monitoring and early warning method and device for skid-mounted equipment.

[0006] In the first aspect of the present application, a monitoring and early warning method for skid-mounted equipment is provided, and the method includes:

[0007] Conduct functional collaboration and fault impact causal relationship analysis according to the collaborative functional blocks of the skid-mounted equipment, and construct a causal association library; disassemble and cluster process the causal association library according to the input data, data analysis, and result output logic framework to obtain a clustering result and corresponding relationship identifiers, where the relationship identifiers include input relationship identifiers, processing relationship identifiers, and output relationship identifiers; based on the clustering result and corresponding relationship identifiers, search for an early warning path in the causal association library to determine hierarchical early warning information; and perform collaborative functional block early warning monitoring on the skid-mounted equipment according to the relationship identifiers and the hierarchical early warning information to obtain an early warning recognition result.

[0008] In the second aspect of the present application, a monitoring and early warning device for skid-mounted equipment is provided. The device includes:

[0009] A causal association library construction module for analyzing the functional collaboration and the causal relationship of fault impacts according to the collaborative functional blocks of the skid-mounted equipment, and constructing a causal association library; a relationship identifier acquisition module for disassembling and clustering the causal association library according to the input data, data analysis, and result output logic framework to obtain a clustering result and the corresponding relationship identifiers, where the relationship identifiers include input relationship identifiers, processing relationship identifiers, and output relationship identifiers; a hierarchical early warning information determination module for searching for an early warning path based on the causal association library according to the clustering result and the corresponding relationship identifiers to determine hierarchical early warning information; and an early warning identification result acquisition module for performing early warning monitoring on the collaborative functional blocks of the skid-mounted equipment according to the relationship identifiers and the hierarchical early warning information to obtain an early warning identification result.

[0010] One or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0011] Analyze the functional collaboration and the causal relationship of fault impacts according to the collaborative functional blocks of the skid-mounted equipment, and construct a causal association library; disassemble and cluster the causal association library according to the input data, data analysis, and result output logic framework to obtain a clustering result and the corresponding relationship identifiers; search for an early warning path based on the causal association library to determine hierarchical early warning information; and perform early warning monitoring on the collaborative functional blocks of the skid-mounted equipment according to the relationship identifiers and the hierarchical early warning information to obtain an early warning identification result. The technical effect of accurately monitoring and warning the operating state of the skid-mounted equipment is achieved, and the operating safety, reliability, and management efficiency of the skid-mounted equipment are improved. Description of the Drawings

[0012] 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 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.

[0013] Figure 1 It is a schematic flowchart of a monitoring and early warning method for skid-mounted equipment provided by an embodiment of the present application;

[0014] Figure 2 It is a schematic structural diagram of a monitoring and early warning device for skid-mounted equipment provided by an embodiment of the present application.

[0015] Description of the reference numerals: Causal association library construction module 10, relationship identifier acquisition module 20, hierarchical early warning information determination module 30, early warning identification result acquisition module 40. Detailed implementation mode

[0016] The present application provides a monitoring and early warning method and device for skid-mounted equipment, aiming to solve the technical problems in the prior art that it is difficult to comprehensively and accurately master the causal relationship of skid-mounted equipment, resulting in inaccurate monitoring and early warning, low safety and reliability, and poor management efficiency.

[0017] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0018] Embodiment 1, as Figure 1 shown, the present application provides a monitoring and early warning method for skid-mounted equipment, and the method includes:

[0019] Step S100: Conduct functional collaboration and failure impact causal relationship analysis according to the collaborative function blocks of the skid-mounted equipment, and construct a causal association library.

[0020] Specifically, a skid-mounted device is a combined device that integrates multiple modules with specific functions. These modules are pre-manufactured, assembled, and debugged in the factory, and then transported as a whole to the site for installation and use. For example, in the scenario of a natural gas filling station, the skid-mounted device includes a natural gas storage module, a filling module, a metering module, etc. Moreover, in some emergency rescue scenarios, communication modules, lighting modules, etc. can be added according to actual needs to enable rapid rescue work in case of emergencies. First, analyze the collaborative function blocks of the skid-mounted device to clarify its function collaborative relationships. For example, in a skid-mounted device for oil extraction, how the oil production module, separation module, and storage module work together to achieve the extraction, processing, and storage of crude oil. At the same time, sort out the connection relationships between the modules to determine the ways (such as pipeline connection, circuit connection, etc.) through which they are interconnected. Based on these module connection relationships, construct an adjacency matrix to present the connection situation between the modules. Precisely extract the type (such as data transmission connection, power transmission connection, etc.) and weight (reflecting the importance or influence of the connection) of each module connection, so as to reasonably configure the connection attributes, and further calculate the connection weights. Using these connection weights, convert the adjacency matrix into a weighted graph to obtain the connection association, where the modules are used as nodes and the connection edge weights clearly represent the connection strength between the connected modules. Evaluate the success and abnormality of the collaborative function relationships. After the evaluation, conduct an in-depth analysis of the influencing factors, and decompose them into a success factor set and an abnormality factor set. Calculate the factor contributions using the mean square error, and generate an initial causal diagram based on the factor contributions. Finally, make full use of historical sample data, apply the time series causal analysis method, carefully explore the influence correlation of each factor on the collaborative function result, and optimize the initial causal diagram based on this correlation to establish an accurate causal relationship. Organically integrate the constructed connection association and causal relationship to successfully construct a causal association library. By analyzing the functional collaboration and causal relationship of the failure impact of the collaborative function blocks of the skid-mounted device to construct a causal association library, the systematic sorting and integration of the complex internal relationships of the skid-mounted device are realized, effectively improving the accuracy and reliability of the operation status evaluation and potential fault prediction of the skid-mounted device.

[0021] Step S200: Decompose and cluster the causal association library according to the input data, data analysis, and result output logical framework to obtain the clustering results and corresponding relationship identifiers, where the relationship identifiers include input relationship identifiers, processing relationship identifiers, and output relationship identifiers.

[0022] Specifically, according to the logical framework of input data, data analysis, and result output, the causal relationships of each functional block in the causal association library are decomposed in detail. In this process, the causal relationships of input data - data processing - result output for each functional block are accurately determined. For example, for the monitoring data acquisition functional block, its input data is the original signal collected by the sensor, the data processing is operations such as filtering and amplifying the signal, and the result output is the effective monitoring data available for analysis. Based on these clear causal relationships, an attribute matrix is constructed, and the relationship identifiers of each node (functional block) are accurately recorded in the matrix, where the rows represent the modules and the columns represent the relationship identifier types (input, processing, output relationship identifiers). Then, according to the data sources of the input data and data processing, a dependency graph is constructed, and through calculation, the influence degree of the data source on the functional block is accurately evaluated, and the dependency influence weights are reasonably allocated. Next, based on the dependency influence weights of the data source on the functional block, a cross-weight matrix is constructed to clearly identify the cross-influence relationships between each functional block. Finally, a clustering is carried out based on the attribute matrix to initially aggregate the functional blocks with similar relationship identifiers together, and then a secondary clustering is performed according to the cross-weight matrix to further optimize and subdivide the results of the first clustering, so as to obtain more accurate and meaningful clustering results and the corresponding relationship identifiers. This achieves the further refinement, classification, and structuring of the complex causal association library, which helps to more clearly understand and grasp the data relationships and processing flows between the functional blocks of the skid-mounted equipment, provides strong support for the subsequent efficient early warning path search and accurate monitoring and early warning, and significantly improves the pertinence and effectiveness of the monitoring of the operating state of the skid-mounted equipment.

[0023] Step S300: Based on the clustering results and the corresponding relationship identifiers, search for the early warning path based on the causal association library to determine the hierarchical early warning information.

[0024] Specifically, perform standardization processing on the clustering results for the same data source to ensure that the data from the same data source is consistent in terms of format, unit, etc., facilitating subsequent unified analysis and processing. After completing the standardization processing, establish the input data connection between the data source and the functional block, enabling the data to be accurately transmitted between the data source and the functional block. Then, starting from the input end of the data source, conduct a comprehensive and in-depth path search based on the constructed causal association library. During the search process, determine the target warning functional blocks, which may be a single device or a functional unit composed of multiple devices. Calculate the influence weight value of the target warning functional block according to the relevance between the search path and the path in the causal association library. For example, if a certain search path is highly correlated with the high-risk fault path in the causal association library, the influence weight value of the corresponding target warning functional block will be relatively high. Finally, determine the hierarchical warning information based on the total influence weight value of the target warning functional blocks in the search path. The functional blocks corresponding to the paths with a higher total influence weight value will be given a higher warning level, thereby achieving hierarchical warning, enabling the operator to preferentially handle potential high-risk faults according to the warning level. It realizes the accurate positioning of potential risk propagation paths in the skid-mounted equipment complex system, quantifies the risk degree based on the path relevance, and thus generates targeted and hierarchical warning information, greatly improving the accuracy and effectiveness of the warning, enabling the operator to quickly identify high-risk areas and take measures preferentially, and effectively reducing the losses and impacts caused by equipment failures.

[0025] Step S400: Perform collaborative functional block warning monitoring on the skid-mounted equipment according to the relationship identifier and the hierarchical warning information, and obtain a warning recognition result.

[0026] Specifically, according to the relationship identifiers, the monitoring system for the collaborative functional blocks of the skid-mounted equipment is configured specifically. The input relationship identifiers of each functional block are clarified, and the types and sources of data to be collected are determined. For example, for the functional block related to temperature, a temperature sensor is configured to collect real-time temperature data; according to the processing relationship identifiers, corresponding data processing modules are set up, such as data filtering and data conversion, to ensure that the collected data can be processed according to the predetermined logic; according to the output relationship identifiers, the output destinations and formats of the data are determined for subsequent early warning analysis. The monitoring system is started, and the data of each collaborative functional block is collected in real time. The data collection frequency is set according to the operating characteristics of the equipment and the early warning requirements. For example, for key parameters (such as pressure, flow rate, etc.), high-frequency collection (once per second or per millisecond) can be adopted, and for relatively less important parameters, the collection frequency can be appropriately reduced (once per minute or per hour). At the same time, a preliminary quality inspection is carried out on the collected data, and the obviously abnormal data points (such as data beyond the reasonable range) are removed to ensure the accuracy and reliability of the data. The real-time collected and processed data is compared with the hierarchical early warning information. For each collaborative functional block, according to its corresponding hierarchical early warning standard, it is judged whether the current data reaches or exceeds the corresponding early warning threshold. Considering the weight factors in the hierarchical early warning information, a comprehensive judgment is made on the early warning affected by multiple factors. If the early warning of a functional block is affected by multiple data sources or other functional blocks, according to the pre-determined weight distribution, the influence degree of each factor is comprehensively evaluated. For example, the operating state of a device is affected by three factors: pressure, flow rate, and temperature, and their weights are 0.3, 0.4, and 0.3 respectively. When the pressure data reaches the mild early warning level, the flow rate data is normal, and the temperature data reaches the moderate early warning level, the final early warning level is determined by weighted calculation (such as calculating the comprehensive early warning index = pressure early warning index × 0.3 + flow rate early warning index × 0.4 + temperature early warning index × 0.3). According to the comparison and judgment results, an early warning recognition result is generated. The early warning recognition result should include detailed information, such as the time when the early warning occurred, the collaborative functional block involved, the early warning level, and the possible reasons (analyzed based on the causal association library). The early warning recognition result is timely fed back to relevant personnel or systems, and is fed back in various ways, such as popping up a prominent early warning information window on the display screen of the monitoring center to display the detailed early warning content; at the same time, sending an early warning notice to the mobile terminal of the operator to ensure that the operator can obtain the early warning information in the first time and ensure the safe and stable operation of the skid-mounted equipment.

[0027] In a possible implementation manner, step S100 further includes:

[0028] Step S110: Obtain the module connection relationship and functional collaboration relationship of the skid-mounted equipment.

[0029] Step S120: Obtain connection associations according to the module connection relationship.

[0030] Step S130: Evaluate the success and abnormality of collaborative functions according to the functional collaboration relationship, and perform causal relationship analysis based on the influencing factors of the success and abnormality of collaborative functions to establish a causal relationship.

[0031] Step S140: Integrate the connection associations with the causal relationship to construct the causal association library.

[0032] Specifically, for skid-mounted equipment, through modular assembly, different functional devices are integrated onto one or more skids, and their module connection relationships are comprehensively sorted out. This requires analyzing the physical structure and electrical connections of the skid-mounted equipment to clarify how each module is connected through pipelines, wires, etc. For example, in an oil extraction skid-mounted equipment, determine the pipeline connection direction and connection interface type between the oil production module and the oil-water separation module, as well as the circuit connection layout between the electrical control module and each execution module. At the same time, for the functional collaboration relationship, during the operation of the skid-mounted equipment, deploy a multi-type sensor network to monitor the operation parameters (such as temperature, pressure, flow rate, voltage, current, etc.) of each module in real time and accurately. Transmit the collected massive operation data to the data processing center, and discover the internal associations between the changes in the operation parameters of different modules through association rule mining algorithms to infer the functional collaboration relationship; use clustering analysis algorithms to group data points with similar functional collaboration characteristics into one category to further clarify the functional collaboration mode. At the same time, combined with an expert system, integrate industry experience and knowledge into the analysis process to verify and optimize the analysis results to ensure that the obtained functional collaboration relationship is accurate and reliable.

[0033] Construct an adjacency matrix based on the actual situation of module connections to visually present the connection status between each module in the form of a matrix. During the construction process, extract the type (such as hydraulic connection, signal connection, etc.) and weight (reflecting the importance of the connection to the overall operation of the equipment) of each module connection, and reasonably configure the connection attributes accordingly. Then, accurately calculate the connection weights according to the connection attributes to quantify various characteristics of the connection. Finally, relying on these connection weights, convert the adjacency matrix into a weighted graph. In the weighted graph, each module becomes a node, and the weight of the connection edge clearly indicates the connection strength between the connected modules. In this way, the originally complex module connection relationship is transformed into a connection association model that is convenient for analysis and processing.

[0034] Based on the clearly defined functional collaboration relationships, conduct evaluations on the success and abnormality of collaborative functions. By collecting a large amount of actual operation data, use data analysis algorithms to evaluate the operation results of collaborative functions, and judge whether the expected functions are successfully achieved and whether there are abnormal situations. For example, for the water quality purification collaborative function of the sewage treatment skid-mounted equipment, monitor whether the water quality indicators after treatment meet the standards to evaluate the success. If the water quality fluctuates abnormally, it is regarded as having abnormality. After the evaluation, deeply analyze various factors affecting the success and abnormality of collaborative functions, and decompose them into a set of success factors and a set of abnormal factors. Then, based on these factor sets, use the mean square error calculation method to quantify the contribution degree of each factor, thereby generating an initial causal diagram to preliminarily show the relationship between factors and the results of collaborative functions. Subsequently, make full use of historical sample data and, with the help of time series causal analysis technology, deeply explore the influence correlation of each factor on the results of collaborative functions, and further clarify the causal relationship between factors. Finally, optimize the initial causal diagram according to the discovered influence correlation, remove unreasonable causal relationships, and supplement missing key relationships, thereby establishing an accurate and reliable causal relationship model.

[0035] When integrating connection associations and causal relationships to construct a causal association library, first use a database management system to create a special storage structure for the causal association library, which has the ability to store module connection information, causal relationship logic, and the association mapping between the two. Then, through a data mapping algorithm, map the module nodes in the connection association one-to-one with the factors in the causal relationship to ensure the accurate association of physical connections and functional influences. For information such as connection attributes (such as connection type, weight, etc.) in the connection association and the contribution degree of influence factors in the causal relationship, use a data fusion algorithm to calculate and generate a fused association weight to quantitatively represent the mutual influence degree after the two are fused. During the fusion process, adopt a data verification mechanism to verify the fusion results based on historical operation data and actual failure cases to ensure that the fused causal association conforms to the actual operation logic of the equipment. Finally, input the fused complete information into the causal association library according to the predetermined data format and storage rules, and establish a comprehensive, accurate, and interrelated database containing module connection relationships and causal relationships. It realizes a deep analysis of the internal structure and functional collaboration of the skid-mounted equipment, transforms complex module connection relationships and functional collaboration relationships into a causal association library, provides a comprehensive, accurate, and structured data basis and logical basis for the fault diagnosis, performance optimization, and precise monitoring and early warning of the skid-mounted equipment, and effectively improves the intelligent level and reliability of the operation management of the skid-mounted equipment.

[0036] In a possible implementation manner, step S120 further includes:

[0037] Step S121: Construct an adjacency matrix according to the module connection relationship.

[0038] Step S122: Extract the types and weights of the connections of each module and configure the connection attributes.

[0039] Step S123: Configure the connection weights according to the connection attributes.

[0040] Step S124: Based on the connection weights, convert the adjacency matrix into a weighted graph to obtain the connection associations, where the modules are nodes and the connection edge weights represent the connection strength between the connected modules.

[0041] Specifically, when constructing an adjacency matrix based on the module connection relationships of a skid-mounted device, it is first necessary to number all the modules in the skid-mounted device to uniquely identify each module. For example, for a skid-mounted device that includes an oil production module, a separation module, a storage module, etc., they can be numbered 1, 2, 3, etc. in sequence. Then, create a two-dimensional matrix where the number of rows and columns of the matrix is equal to the total number of modules. For the elements in the matrix, if there is a connection relationship between module i and module j, then mark 1 at the position of the i-th row and j-th column and the j-th row and i-th column of the matrix (because the connection relationship is bidirectional), indicating the existence of a connection; if there is no connection, then mark 0. In this way, an adjacency matrix that can intuitively reflect the module connection status is constructed, and through this matrix, it can be quickly determined whether there is a direct connection between any two modules.

[0042] After constructing the adjacency matrix, extract the types and weights of the connections of each module to configure the connection attributes. Through on-site investigation and referring to the equipment technical documents, determine the types of module connections, such as different types of hydraulic connections, electrical connections, data transmission connections, etc. existing in the skid-mounted device. For determining the connection weights, multiple factors need to be considered comprehensively, such as the importance of the connection, the frequency of data or energy transmission, the degree of influence of the connection on the overall function of the equipment, etc. For example, for the pipeline connection (belonging to the hydraulic connection type) between the oil production module and the separation module, if this pipeline is responsible for transporting a large amount of crude oil and is crucial for the subsequent separation process, then assign it a relatively high weight value, such as 0.8; while for the electrical connection between some auxiliary devices, if it is only used to transmit control signals and has a relatively small impact on the overall function, then assign a relatively low weight value, such as 0.3. According to these extracted and calculated connection types and weights, configure the corresponding attribute information for each connection relationship to more accurately describe the characteristics of the connection in the subsequent process.

[0043] Determine the connection weight according to the configured connection attributes. For different types of connection attributes, different calculation methods are used to quantify the connection weight. For numerical connection attributes (such as connection bandwidth, pipeline diameter, etc.), directly use them as part of the connection weight; for categorical attributes (such as connection type being hydraulic, electrical, etc.), convert them into corresponding numerical weights through a pre-set weight mapping table. For example, set the weight mapping value for hydraulic connection to 0.6, and the weight mapping value for electrical connection to 0.4, etc. At the same time, the influence of factors such as connection stability and reliability on the weight also needs to be considered, and the weight is adjusted by introducing a correction factor. For example, if a certain connection has poor stability and is prone to failures, then multiply by a correction factor less than 1 (such as 0.8) when calculating the weight to reduce its weight. After comprehensive calculation, accurately configure the connection weight for each module connection to more precisely reflect the actual influence of the connection.

[0044] Based on the calculated connection weights, convert the adjacency matrix into a weighted graph to obtain connection associations. In the weighted graph, each module serves as a node, and the node number corresponds to the module number in the previous adjacency matrix. The connection edges between modules are represented by line segments with weights, and the weight of the connection edge is the previously configured connection weight, which clearly indicates the connection strength between the connected modules. In this way, the module connection relationship originally represented in matrix form is transformed into a connection association in the form of a weighted graph that is more intuitive and can better reflect the difference in connection strength. In the weighted graph, the connection tightness between each module can be intuitively observed. For example, a connection edge with a larger weight indicates a tight connection between the corresponding two modules, with frequent and important data or energy transmission; while a connection edge with a smaller weight indicates a relatively weak connection. This connection association provides an important basis for further analyzing the overall structure and functional relationship of the skid-mounted equipment. It realizes the transformation of the abstract module connection relationship of the skid-mounted equipment into an intuitive and quantifiable connection association model. Based on the adjacency matrix, integrating attribute information such as connection type and weight to generate a weighted graph, clearly presenting the difference in connection strength between modules, providing a structured, digital and more expressive data basis and analysis perspective for subsequent accurate analysis of equipment structure, functional relationship and fault impact, etc., and improving the accuracy and effectiveness of understanding and processing the internal connection logic of the skid-mounted equipment.

[0045] In a possible implementation manner, step S130 further includes:

[0046] Step S131: Decompose the influence factors according to the evaluation results of the success and abnormality of the collaborative function to obtain a set of success factors and a set of abnormality factors.

[0047] Step S132: Calculate the factor contribution based on the set of success factors and the set of abnormality factors through the mean square error.

[0048] Step S133: Produce an initial causal diagram based on the factor contributions.

[0049] Step S134: Through time series causal analysis based on historical sample data, explore the impact correlation of each factor on the collaborative function result.

[0050] Step S135: Optimize the initial causal diagram according to the impact correlation to obtain the causal relationship.

[0051] Specifically, use a data acquisition system to comprehensively collect various data during the operation of the skid-mounted equipment, including equipment operation status parameters (such as temperature, pressure, flow rate, rotation speed, etc.), environmental monitoring data (such as ambient temperature, humidity, air pressure, etc.), and operation record data (such as valve opening changes, equipment start-stop times, etc.). Then, use the principal component analysis (PCA) algorithm to perform dimensionality reduction on these data, extract the main data features, and reduce data redundancy. Next, input the dimensionality-reduced data into a pre-trained classification model, which is constructed based on a large amount of historical data and expert experience and can accurately judge the correlation degree between the data and the success or abnormality of the collaborative function according to the data features. Through the classification results of the model, screen out the data features highly correlated with the success of the collaborative function as success factors to form a success factor set; screen out the data features closely related to the abnormality of the collaborative function as abnormal factors to form an abnormal factor set. At the same time, conduct manual review and adjustment on the initially screened factors to ensure the accuracy and integrity of the factor set.

[0052] For the success factor set, for each success factor X in it i , construct a univariate linear model to predict the success score S, where β0 and β1 are model parameters obtained by fitting data through methods such as the least squares method. According to the formula calculate the predicted mean squared error (MSE), where N is the number of samples, is the actual success score, is the success score predicted by the model. Finally, according to the formula contribution(X i ) = 1 - MES i calculate the factor contribution value of the success factor X i . For the abnormal factor set, use a similar method. Use the univariate model to predict the abnormality score A, contribution , where is the actual abnormality score, is the predicted abnormality score. After obtaining the contribution values of the success factors and abnormal factors, perform normalization processing. Normalize the contribution value through the formula contribution to make the contribution value fall between 0 and 1, where is the total number of factors. This ensures that the contribution values ​​of different factors are comparable, which facilitates subsequent analysis and processing. For example, when generating a causal diagram or evaluating the importance of factors, it can more accurately reflect the relative contribution of each factor.

[0053] Each factor in the success factor set and the abnormal factor set is taken as a node and laid out in the graph according to its category (success factor or abnormal factor). Then, the weight and direction of the connection edge between nodes are determined according to the calculated factor contribution value. For factors with large factor contribution values, they are taken as the main causal starting point or key node, and directed connection edges are established with other factors that may be affected by them. The weight of the connection edge can be directly quantified by the size of the factor contribution value. The larger the contribution value, the higher the weight of the connection edge, indicating that the factor has a stronger influence on other factors. For example, if a success factor has a high contribution value, and it is judged to have a positive impact on several other factors based on domain knowledge and experience, then a directed edge is drawn from the factor node to other related factor nodes, and the corresponding weight is marked. In this way, all factors and their interrelationships are presented in a graphical way to form an initial causal graph, which preliminarily shows the possible causal relationship and influence between the factors, and provides a basic framework for subsequent causal relationship optimization and analysis.

[0054] With the help of historical sample data, the time series causal analysis method is used to deeply explore the correlation of the impact of each factor on the synergistic function results. Collect historical data during the long-term operation of the skid-mounted equipment, which contains the values ​​of each factor at different time points and the corresponding synergistic function results. For example, for chemical production skid-mounted equipment, collect historical data on factors such as reaction temperature, raw material flow, catalyst activity, and final product quality data in the past few months or even years. Then, through the time series analysis algorithm, analyze the order and correlation between the changes in each factor in the time dimension and the changes in the synergistic function results. Determine which factor changes will have a significant impact on the synergistic function results after a certain time lag, as well as the direction (positive correlation or negative correlation) and intensity of this impact, so as to obtain the correlation information of the impact of each factor or factor combination on the synergistic function results.

[0055] Optimize the initial causal diagram based on the mined impact correlation. According to the results of the impact correlation analysis, correct and improve the causal relationships in the initial causal diagram. If it is found that there is no connection between two factors in the initial causal diagram, but a significant causal relationship is determined between them through the impact correlation analysis, then add the corresponding directed line segment in the causal diagram; if it is found that the causal relationship direction between a certain factor and the collaborative function result in the initial causal diagram is incorrect, then correct it; at the same time, for the causal relationships with weak correlation, appropriately adjust the thickness or weight of the connection line segment to more accurately reflect the actual situation. Through such an optimization process, remove unreasonable causal relationships and supplement missing key relationships, and finally obtain a model that can accurately reflect the true causal relationships of each factor in the collaborative function of the skid-mounted equipment. It realizes the decomposition of influencing factors from the collaborative function evaluation results, quantifies their contributions, constructs and optimizes the causal diagram, thereby accurately mining the causal relationships between each factor and the collaborative function result, providing reliable technical support for equipment operation analysis, fault diagnosis, and performance optimization.

[0056] In a possible implementation manner, step S134 further includes:

[0057] Step S1341: Set a dynamic time window and a factor variable screening gradient. The dynamic time window is used for impact correlation analysis of multiple time lengths, and the factor variable screening gradient is used to control the continuous increase and change of the number of factors.

[0058] Step S1342: Perform variable setting according to the dynamic time window and the factor variable screening gradient, and identify the dependence relationship of each factor on the collaborative function result through a time series analysis network model to obtain the impact correlation of each factor or factor combination on the collaborative function result.

[0059] Specifically, first, determine the range of the time window. For example, for the collaborative function data of a certain device, set the shortest time window to 1 day and the longest time window to 30 days. Then, set dynamic time windows at a certain step size (such as 1 day). Starting from the shortest time window, increase the step size in sequence to form a series of time windows of different lengths, such as 1 day, 2 days, 3 days, etc., until the longest time window of 30 days is reached. In this way, for the same set of data, the influence correlation analysis can be carried out at different time lengths to comprehensively capture the short-term, medium-term, and long-term influence effects. For example, when analyzing the influence of the device temperature factor on the collaborative function, through time windows of different lengths, the immediate influence of the temperature fluctuation within one day on the collaborative function and the long-term influence of the temperature change trend within one month on the collaborative function are observed. For the factor variable screening gradient, first determine all the factors that may affect the collaborative function. Suppose there are 10 factors in total, such as device operation parameters (pressure, flow rate, rotation speed, etc.), environmental factors (temperature, humidity, etc.), and operation factors (operation frequency, operation duration, etc.). Then, start from a single factor and increase the number of factors in sequence to form a gradient. First, analyze the influence of a single factor (such as pressure) on the collaborative function, then analyze the influence of the combination of two factors (such as pressure and flow rate), then analyze the influence of the combination of three factors (such as pressure, flow rate, and rotation speed), and so on, until the combination including all 10 factors is analyzed. In this way, it is possible to gradually and deeply explore the influence of different combinations of factors on the collaborative function results, from simple single-factor analysis to complex multi-factor interaction analysis, comprehensively excavate the relationship between each factor and the collaborative function, and provide a rich and detailed data basis and analysis dimension for subsequent correlation analysis and causal diagram construction.

[0060] Precisely set variables according to the set dynamic time window and factor variable screening gradient. The dynamic time window changes sequentially from short to long, such as from several hours to several weeks. After each adjustment of the window length, combined with the factor variable screening gradient, data with different numbers of factor combinations are segmented according to the time window. Using a long short-term memory network (LSTM) time series analysis network model, input this processed data into the model. LSTM has a special memory cell structure and can well capture long-term dependencies in the time series. When processing multivariate time series data, the model will simultaneously consider the changes of multiple factor variables at different time points. Through repeated learning and training of the input data, the interaction patterns between various factors are mined. For example, when analyzing the collaborative function of chemical production skid-mounted equipment, the input factor variables include reaction temperature, raw material flow rate, catalyst concentration, etc. With the change of the time window and different factor combinations, the LSTM model learns the variation laws of these variables over time and identifies the dependencies between each factor or factor combination and the collaborative function results (such as product quality, production efficiency, etc.). Finally, the model outputs the influence correlation of each factor or factor combination on the collaborative function result, thus clearly revealing which factors or factor combinations have the most significant impact on the collaborative function at different time scales, providing a key basis for subsequent causal relationship analysis and optimization. It realizes the comprehensive and accurate mining of the dependencies and influence correlations between each factor or factor combination and the collaborative function result under multiple time scales and multiple factor combination dimensions.

[0061] In a possible implementation manner, step S200 further includes:

[0062] Step S210: Decompose the causal relationship of each functional block according to the input data - data analysis - result output logic framework, and determine the causal relationship of input data - data processing - result output of each functional block.

[0063] Step S220: Based on the causal relationship of input data - data processing - result output of each functional block, construct an attribute matrix to record the relationship identifiers of each node, where the rows represent modules and the columns represent relationship identifier types.

[0064] Step S230: According to the input data and the data sources of data processing, construct a dependency graph, calculate the influence degree of the data sources on the functional blocks, and assign dependency influence weights.

[0065] Step S240: According to the dependency influence weights of the data sources on the functional blocks, construct a cross-weight matrix to identify cross-influence relationships.

[0066] Step S250: Perform a first clustering based on the attribute matrix and a second clustering based on the cross-weight matrix to obtain the clustering result.

[0067] Specifically, comprehensively collect the input and output data of each functional block of the skid-mounted equipment, including sensor data, operation instructions, and equipment status, etc., and preprocess to ensure the availability of the data. Then use the Granger causality test algorithm to analyze the causality of the input and output data, construct a Bayesian network model to describe the causal relationship of data processing, and use the decision tree algorithm to analyze the logic of the processing flow. Then use the cross-validation method to evaluate the accuracy of the mined causal relationship, optimize the model according to the verification results, and finally accurately determine the causal relationship of input data - data processing - result output of each functional block by adjusting the lag order of the Granger causality test, improving the conditional probability table of the Bayesian network, or improving the decision tree branch rules.

[0068] Use the feature engineering algorithm to process the information of each functional block, extract the key features as the column identification types of the attribute matrix, and use the principal component analysis (PCA) algorithm to screen out the input data types, key steps of data processing, and key features of result output that are of important representativeness to the functional block. Determine the number of rows of the attribute matrix according to the number of functional blocks, scan each functional block one by one. For each functional block, fill in the corresponding input data source and type into the corresponding column according to the predefined coding system (for example, assign a unique digital code to different types of input data), record the name of the algorithm used (which can be the function name in the algorithm library or the unique identifier of the custom algorithm) or the description of the processing flow in detail in the data processing column, and fill in the information such as the format and destination of the output data in the result output column, so as to construct the attribute matrix and accurately record the relationship identifiers of each node.

[0069] Construct a dependency graph with data sources (such as sensors, external data interfaces, etc.) and functional blocks as nodes. For each functional block, determine the directly dependent data sources and draw directed edges from the data source nodes to the functional block nodes in the dependency graph. For example, in a chemical reaction skid-mounted device, a directed edge is drawn from the temperature sensor data source node to the reaction kettle temperature control functional block node, indicating that the reaction kettle temperature control depends on the data from the temperature sensor. Consider the data transfer and interaction relationships between functional blocks. If the output of one functional block is used as the input of another functional block, draw the corresponding directed edge. For example, since the product output of the reaction kettle is used as the input of the subsequent separation and purification functional block, a directed edge is drawn between the reaction kettle node and the separation and purification node, with the direction from the reaction kettle to the separation and purification. Use data structures such as graph databases (e.g., Neo4j) or adjacency matrices to store the dependency graph for convenient subsequent querying, modification, and analysis operations. In the graph database, attributes can be added to nodes and edges, such as the type of data source and the functional description of the functional block, to enrich the information of the dependency graph. Comprehensively consider various factors of data sources to construct an index system, including data accuracy (error rate compared with standard values or reliable data sources), stability (calculating variance or standard deviation), real-time performance (update frequency or latency), reliability (based on historical failure or maintenance records), and importance to functional blocks (data analysis). Use the Analytic Hierarchy Process (AHP) to construct a judgment matrix to determine the weights of each index. Assume the weight of data accuracy is 0.3, stability is 0.2, real-time performance is 0.2, reliability is 0.1, and importance is 0.2. Taking the temperature sensor data source as an example, its accuracy score is 0.98, the standard deviation is 0.5, i.e., the stability score is 0.8, the real-time performance score is 0.9, the reliability score is 0.95, and the importance score for the reaction kettle temperature control is 0.9. The influence degree score = 0.3×0.98 + 0.2×0.8 + 0.2×0.9 + 0.1×0.95 + 0.2×0.9 = 0.899. Normalize the calculated influence degree scores so that the sum of the influence degree weights of all data sources on a specific functional block is 1. Assign the normalized weights to the corresponding edges in the dependency graph, i.e., the edges from the data source nodes to the functional block nodes. In this way, each edge in the dependency graph has a weight, representing the degree of dependence influence of the data source on the functional block.

[0070] The number of rows and columns of the cross-weight matrix is equal to the number of functional blocks plus the number of data sources. For example, if there are 5 functional blocks and 3 data sources, the cross-weight matrix is an 8×8 square matrix. The elements in the matrix are classified and defined. The elements on the main diagonal are used to represent the importance or stability weights of the data sources or functional blocks themselves (initialized to 1 or determined according to other methods). The elements off the main diagonal are used to represent the cross-influence weights between data sources and functional blocks or between functional blocks, and are filled according to the weights in the dependency graph. For the cross-influence weights between data sources and functional blocks, directly fill the weights on the corresponding edges in the dependency graph into the corresponding positions in the cross-weight matrix. For example, if the dependency influence weight of the temperature sensor data source on the reactor temperature control functional block in the dependency graph is 0.397, then fill 0.397 at the intersection of the row corresponding to the temperature sensor data source and the column corresponding to the reactor temperature control functional block in the cross-weight matrix. For the cross-influence weights between functional blocks, they are determined by analyzing the data transfer relationship and dependency degree between functional blocks. If there is a data dependency between two functional blocks and there is a corresponding directed edge connecting them in the dependency graph, the cross-influence weight is determined according to the weight of the edge and the importance of data transfer. For example, there is data transfer between the reactor and the separation and purification functional block, and according to the analysis of the dependency graph and actual situation, the influence of the reactor on separation and purification is relatively large, with a weight of 0.6. Then fill 0.6 at the intersection of the row corresponding to the reactor functional block and the column corresponding to the separation and purification functional block in the cross-weight matrix. At the same time, fill a symmetric value (such as 0.4, determined according to the specific situation) at the intersection of the row corresponding to the separation and purification functional block and the column corresponding to the reactor functional block to represent the cross-influence relationship between them. Check and adjust the constructed cross-weight matrix to ensure the symmetry (if there is a symmetric relationship) and consistency of the matrix. For example, check the logical rationality of the cross-influence weights between functional blocks. If it is found that the influence weight of a certain functional block on another functional block does not conform to the actual situation, re-analyze the data transfer relationship and dependency degree and correct the weights. Verify the cross-weight matrix through simulated data or actual case data, and observe whether the weights in the matrix can accurately reflect the cross-influence relationship between data sources and functional blocks and between functional blocks. For example, change the data value of a certain data source, calculate the influence on the relevant functional blocks according to the weights in the matrix, and compare it with the changes in the actual operation of the equipment. If there is a large deviation, further optimize the weights in the matrix.

[0071] For the first clustering based on the attribute matrix, the K-means or hierarchical clustering method is used to cluster the attribute matrix. First, the feature vectors of each module are extracted from the attribute matrix, and these feature vectors reflect the attributes of the module in terms of input data type, data processing method, result output form, etc. Then, the feature vectors are divided into different clusters through the K-means algorithm, or a clustering tree is constructed using hierarchical clustering, and the clustering result is obtained by cutting the clustering tree, so as to identify modules with similar functions. For the second clustering based on the cross-weight matrix, the spectral clustering method is used to cluster the cross-relationship matrix. The cross-influence relationship information between modules is obtained from the cross-weight matrix. By constructing a graph structure (regarding modules as nodes and cross-influences as edges), the spectral clustering algorithm is used to segment the graph to identify modules or variables with strong cross-influences. The results of the attribute matrix clustering and the cross-weight matrix clustering are fused by constructing a joint clustering model , where α and β are weight parameters. Adjust the values of α and β according to the actual requirements and data characteristics, and comprehensively consider the contributions of the attribute matrix and the cross-relationship matrix in clustering to obtain the final clustering result.

[0072] In a possible implementation manner, step S300 further includes:

[0073] Step S310: Perform same-data-source standardization processing according to the clustering result, and establish an input data connection between the data source and the functional block.

[0074] Step S320: Based on the clustering result, starting from the input end of the data source, perform path search according to the causal association library to determine the target warning functional block, and obtain the influence weight value of the target warning functional block according to the path relevance in the search path and the causal association library. The target warning functional block includes a single device or a combination of devices.

[0075] Step S330: Determine the hierarchical warning information based on the total amount of the influence weight values of the target warning functional block in the search path.

[0076] Specifically, for the data in the clustering result, perform same-data-source standardization processing. For example, for data such as pump flow, compressor inlet pressure, and heat exchanger temperature difference, normalize them all to the [0, 1] interval to eliminate the differences caused by inconsistent ports, protocols, and units between different devices, which is convenient for subsequent warning analysis. After completing the standardization processing, establish an input data connection between the data source and the functional block according to the functional logic and data flow of the device to ensure that the data can be accurately transmitted to the corresponding functional block for processing and analysis.

[0077] The input end of the data source is determined based on the clustering results. For example, in the existing standardized processing, the device port corresponding to the data such as pump flow, compressor inlet pressure, and heat exchanger temperature difference is the data source input end. Taking these data source input ends as the starting point, the path search is carried out according to the causal association library. The causal association library records in detail the causal association information between devices and between the functional blocks inside the equipment. For example, the path is pump, compressor, heat exchanger, and desulfurization tower. Starting from the pump data source input end, use a suitable path search algorithm (such as depth-first search or breadth-first search) to search in the causal association library. Assuming that the depth-first search is adopted, firstly, the connection relationship between the pump and the compressor is deeply explored, and it is found that the operating state of the pump will directly affect the operation of the compressor, so the path segment "pump to compressor" is determined. Then, continue to explore from the compressor to find its causal relationship with the heat exchanger, and further expand the path to "pump to compressor to heat exchanger". By analogy, the complete path "pump to compressor to heat exchanger to desulfurization tower" is finally obtained. In this process, the nodes and path information passed through each step are recorded in detail. These records will provide a key basis for the subsequent determination of the target early warning functional blocks and the calculation of the impact weights. According to the searched path, the pumps, compressors, heat exchangers, and desulfurization towers are all target early warning functional blocks. They each have important functions and are interconnected through data and operation logic. For example, changes in the flow rate of the pump will affect the inlet pressure of the compressor, which in turn affects the heat exchange effect of the heat exchanger, and ultimately affects the desulfurization efficiency of the desulfurization tower. Therefore, these devices are in a critical position during the operation of the entire skid-mounted equipment. Once a problem occurs in a certain link, it may trigger a chain reaction and affect the normal operation of the entire system. Therefore, they are identified as target early warning functional blocks. In order to obtain the impact weights of the target early warning functional blocks, a dynamic weight matrix W is constructed. dyn , this matrix is ​​constructed based on the strength of the causal relationship between the equipment in the system. The construction process is as follows: Determine the number of rows and columns of the matrix: For example, since there are four types of equipment in the system, namely pumps, compressors, heat exchangers, and desulfurization towers, the matrix is ​​a 4×4 square matrix. The strength of the causal relationship between the equipment is analyzed from the causal association library to determine the weight value. For example, from pumps to compressors, based on the understanding of equipment operation and the analysis of historical data, it is found that the operating status of the pump has an important impact on the operation of the compressor. After quantitative evaluation (possibly through data correlation analysis, expert experience judgment, etc.), its weight is determined to be 0.8, so 0.8 is filled in the first row and second column of the matrix to represent the weight of the pump's impact on the compressor. Similarly, the weight of the compressor to the heat exchanger is 0.6, and 0.6 is filled in the second row and third column of the matrix; the weight of the heat exchanger to the desulfurization tower is 0.9, and 0.9 is filled in the third row and fourth column of the matrix. For other elements in the matrix that do not have a direct causal relationship, fill in 0 according to the actual situation (indicating that there is no direct impact). In this way, a dynamic weight matrix is ​​obtained. , in this matrix, the value of an element represents the weight relationship between adjacent nodes. The weight from the pump to the compressor is 0.8, which indicates that the influence weight of the pump on the compressor is 0.8; the weight from the compressor to the heat exchanger is 0.6, meaning that the influence weight of the compressor on the heat exchanger is 0.6; the weight from the heat exchanger to the desulfurization tower is 0.9, that is, the influence weight of the heat exchanger on the desulfurization tower is 0.9. In the dynamic weight matrix W dyn , the corresponding weight relationships are used to accurately determine the influence weights of each target warning function block in the entire causal chain. These weights will provide key quantitative bases for hierarchical warning in subsequent steps, helping to more accurately evaluate the importance and affected degree of different devices or device combinations in the system, thus providing strong support for the effective operation of the warning system.

[0078] Calculate the total influence weight of the target warning function block in the search path. According to the dynamic weight matrix W dyn , the weight from the pump to the compressor is 0.8, the weight from the compressor to the heat exchanger is 0.6, and the weight from the heat exchanger to the desulfurization tower is 0.9. Then the total influence weight P weight= 0.8 + 0.6 + 0.9 = 2.3. Then, based on the total influence weight calculated, the classified early warning information is determined. According to the pre-set classification rules, when the total weight reaches or exceeds a certain threshold (for example, it may be set here that when the total weight is greater than or equal to 2, the early warning level is severe), the early warning level is determined to be severe. If the data source only affects a single module, then the corresponding level of early warning information is issued for that single device, prompting the operator of the possible risks of the device and the measures to be taken. For example, if a certain data source only affects a single device, the pump, and the calculated total weight corresponds to a medium early warning level, then a medium early warning for the pump is issued, reminding the operator to pay attention to the operating status of the pump, and may need to check the relevant parameters of the pump, maintenance conditions, etc. When the data source affects multiple devices, according to the access and operation path of the data source, its influence on each device is analyzed in detail. For example, in the above path, the pump, compressor, heat exchanger, and desulfurization tower are all affected. According to factors such as the weights involved in the path for each device and the importance of the device itself, the early warning levels of different devices are determined respectively. For devices with higher importance and larger affected weights, such as the desulfurization tower (the sum of the affected weights from the previous devices is relatively high), a higher level of early warning is given; for relatively less important devices with smaller affected weights, such as the compressor (only affected by the pump with a weight of 0.8), a relatively lower level of early warning is given, but still requires the operator's attention to promptly discover potential problems and take effective preventive and treatment measures to ensure the safe and stable operation of the entire system. Through this classification early warning method based on the total influence weight, the risk status of different devices or device combinations in the system can be more accurately prompted to the operator, improving the effectiveness and pertinence of the early warning, and providing strong support for taking timely countermeasures.

[0079] Embodiment 2. Based on the same inventive concept as the monitoring and early warning method for skid-mounted equipment in the foregoing embodiment, as Figure 2 shown, the present application provides a monitoring and early warning device for skid-mounted equipment. The device in the embodiment of the present application and the method embodiment are based on the same inventive concept. Among them, the device includes:

[0080] A causal association library construction module 10, configured to perform function collaboration and failure influence causal relationship analysis according to the collaborative function blocks of the skid-mounted equipment, and construct a causal association library.

[0081] A relationship identifier acquisition module 20, configured to disassemble and cluster process the causal association library according to the input data, data analysis, and result output logic framework, obtain a clustering result and the corresponding relationship identifiers, and the relationship identifiers include input relationship identifiers, processing relationship identifiers, and output relationship identifiers.

[0082] The hierarchical warning information determination module 30 is configured to search for a warning path based on the clustering result and the corresponding relationship identifier, and determine hierarchical warning information based on the causal association library.

[0083] The warning recognition result acquisition module 40 is configured to perform collaborative function block warning monitoring on the skid-mounted equipment according to the relationship identifier and the hierarchical warning information, and obtain a warning recognition result.

[0084] Furthermore, the causal association library construction module 10 further includes:

[0085] The connection relationship acquisition unit is configured to obtain the module connection relationship and the function collaboration relationship of the skid-mounted equipment.

[0086] The connection association acquisition unit is configured to obtain a connection association according to the module connection relationship.

[0087] The causal relationship establishment unit is configured to evaluate the success and abnormality of the collaborative function according to the function collaboration relationship, and perform causal relationship analysis based on the influencing factors of the success and abnormality evaluations of the collaborative function, and establish a causal relationship.

[0088] The causal relationship fusion unit is configured to fuse the connection association and the causal relationship to construct the causal association library.

[0089] Furthermore, the connection association acquisition unit further includes:

[0090] The adjacency matrix construction unit is configured to construct an adjacency matrix according to the module connection relationship.

[0091] The connection attribute configuration unit is configured to extract the type and weight of each module connection and configure connection attributes.

[0092] The connection weight configuration unit is configured to configure connection weights according to the connection attributes.

[0093] The weighted graph conversion unit is configured to convert the adjacency matrix into a weighted graph based on the connection weights to obtain the connection association, where the module is a node and the connection edge weight represents the connection strength between the connected modules.

[0094] Furthermore, the causal relationship establishment unit further includes:

[0095] The factor set acquisition unit is configured to perform influence factor decomposition according to the success and abnormality evaluation results of the collaborative function to obtain a success factor set and an abnormality factor set.

[0096] The factor contribution calculation unit is configured to calculate factor contributions based on the success factor set and the abnormality factor set through mean square error.

[0097] An initial causal graph generation unit for generating an initial causal graph according to the factor contribution.

[0098] An influence correlation mining unit for mining the influence correlation of each factor on the collaborative function result through time series causal analysis based on historical sample data.

[0099] An initial causal graph optimization unit for optimizing the initial causal graph according to the influence correlation to obtain the causal relationship.

[0100] Further, the influence correlation mining unit includes:

[0101] A factor variable screening gradient setting unit for setting a dynamic time window and a factor variable screening gradient. The dynamic time window is used for influence correlation analysis of multiple time lengths, and the factor variable screening gradient is used to control the continuous increase and change of the number of factors.

[0102] A dependency relationship identification unit for setting variables according to the dynamic time window and the factor variable screening gradient, and identifying the dependency relationship of each factor on the collaborative function result through a time series analysis network model to obtain the influence correlation of each factor or factor combination on the collaborative function result.

[0103] Further, the relationship identification acquisition module 20 includes:

[0104] A causal relationship determination unit for decomposing the causal relationship of each functional block according to the input data, data analysis, and result output logic framework, and determining the causal relationship of input data - data processing - result output of each functional block.

[0105] A relationship identification recording unit for constructing an attribute matrix based on the causal relationship of input data - data processing - result output of each functional block, and recording the relationship identification of each node, where the rows represent modules and the columns represent relationship identification types.

[0106] A dependency influence weight distribution unit for constructing a dependency graph according to the data sources of the input data and data processing, calculating the influence degree of the data source on the functional block, and distributing the dependency influence weight.

[0107] A cross-weight matrix construction unit for constructing a cross-weight matrix according to the dependency influence weight of the data source on the functional block to identify the cross-influence relationship.

[0108] A secondary clustering unit for performing primary clustering based on the attribute matrix and secondary clustering according to the cross-weight matrix to obtain the clustering result.

[0109] Further, the hierarchical early warning information determination module 30 includes:

[0110] An input data connection establishment unit, configured to perform same data source normalization processing according to the clustering result, and establish an input data connection between the data source and the functional block.

[0111] An influence weight acquisition unit, configured to, based on the clustering result, start from the input end of the data source, perform path search according to the causal association library, determine a target warning functional block, and obtain the influence weight of the target warning functional block according to the path relevance between the search path and the path in the causal association library, where the target warning functional block includes a single device or a combination of devices.

[0112] A hierarchical warning information determination unit, configured to determine the hierarchical warning information based on the total amount of the influence weights of the target warning functional blocks in the search path.

[0113] It should be noted that the above sequence 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 the present specification. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0114] 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 principle of the present application shall be included in the protection scope of the present application.

[0115] 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 monitoring and early warning method for skid-mounted equipment, characterized in that: include: Conduct functional coordination and fault impact causal relationship analysis based on the collaborative functional blocks of skid-mounted equipment, and build a causal association library; According to the input data, data analysis, and result output logic framework, the causal association library is disassembled and clustered to obtain clustering results and corresponding relationship identifiers, wherein the relationship identifiers include input relationship identifiers, processing relationship identifiers, and output relationship identifiers, wherein the clustering process includes: clustering the attributes of functional similarity according to the causal relationship of the functional blocks, clustering the cross-influence relationships according to the data source dependency based on the attribute clustering results, identifying the cross-influence strength of the causal relationship, obtaining clustering results, and generating relationship identifiers of causal relationship nodes according to the clustering results; According to the clustering results and the corresponding relationship identifiers, a warning path search is performed based on the causal association library to determine the graded warning information; Perform early warning monitoring of the collaborative functional blocks of the skid-mounted equipment according to the relationship identifier and the hierarchical early warning information to obtain an early warning identification result; Build a causal association library, including: Obtaining the module connection relationship and the functional coordination relationship of the skid-mounted equipment, wherein the functional coordination relationship is the relationship between the modules that need to cooperate with each other to realize the equipment function during the operation of the skid-mounted equipment, and the module connection relationship is the physical connection relationship between the modules; According to the module connection relationship, a connection association is obtained; According to the functional synergy relationship, the success and abnormality of the synergy function are evaluated, and based on the influencing factors of the success and abnormality evaluation of the synergy function, a causal relationship analysis is performed to establish a causal relationship; Merging the connection association with the causal relationship to construct the causal association library; According to the module connection relationship, a connection association is obtained, including: Constructing an adjacency matrix according to the module connection relationship; Extract the type and weight of each module connection and configure the connection properties; Configuring connection weights according to the connection attributes; Converting the adjacency matrix into a weighted graph based on the connection weights to obtain the connection association, wherein the modules in the weighted graph are nodes, the connection edge weights represent the connection strength between the connection modules, and the connection association is used to represent the connection strength between the modules; The causal association library is disassembled and clustered according to the input data, data analysis, and result output logic framework to obtain clustering results and corresponding relationship identifiers, including: Decompose the causal relationship of each functional block according to the input data, data analysis, and result output logic framework, and determine the causal relationship of input data-data processing-result output of each functional block; Based on the causal relationship of input data-data processing-result output of each functional block, an attribute matrix is ​​constructed to record the relationship identification of each node, where the row represents the module and the column represents the relationship identification type; According to the input data and the data source of the data analysis, a dependency graph is constructed, the influence of the data source on the functional block is calculated, and the dependency influence weight is allocated; According to the dependency influence weights of the data sources on the functional blocks, a cross-weight matrix is ​​constructed to identify the cross-influence relationship; A clustering is performed based on the attribute matrix, and a secondary clustering is performed based on the cross weight matrix to obtain the clustering result.

2. The monitoring and early warning method for skid-mounted equipment according to claim 1 is characterized in that: Causal relationship analysis is performed based on the influencing factors of the success and abnormality evaluation of the synergistic function to establish causal relationships, including: Decomposing the influencing factors according to the evaluation results of the success and abnormality of the collaborative function to obtain a success factor set and an abnormal factor set; Based on the success factor set and the abnormal factor set, factor contribution is calculated by using mean square error; generating an initial causal diagram based on the factor contributions; Based on historical sample data, through time series causal analysis, the impact correlation of each factor on the synergistic function results is explored; An initial cause-effect graph is optimized according to the impact correlation to obtain the cause-effect relationship.

3. The monitoring and early warning method for skid-mounted equipment according to claim 2 is characterized in that: Based on the historical sample data, through time series causal analysis, the influence correlation of each factor on the synergistic function results is explored, including: Setting a dynamic time window and a factor variable screening gradient, wherein the dynamic time window is used to perform an impact correlation analysis of multiple time lengths, and the factor variable screening gradient is used to control the continuous increase in the number of factors; Variables are set according to the dynamic time window and factor variable screening gradient, and the dependency of each factor on the synergistic function result is identified through the time series analysis network model to obtain the influence correlation of each factor or factor combination on the synergistic function result.

4. The monitoring and early warning method for skid-mounted equipment according to claim 1 is characterized in that: According to the clustering results and the corresponding relationship identifiers, a warning path search is performed based on the causal association library to determine the graded warning information, including: Performing standardization processing on the same data source according to the clustering results, and establishing a connection between the data source and the input data of the functional block; Based on the clustering result, taking the input end of the data source as the starting point, performing path search according to the causal association library, determining the target early warning function block, and obtaining the influence weight of the target early warning function block according to the correlation between the search path and the path in the causal association library, wherein the target early warning function block includes a single device or a combination of devices; The graded warning information is determined based on the total amount of influence weights of the target warning function block in the search path.

5. A monitoring and early warning device for skid-mounted equipment, characterized in that: The device is used to implement the monitoring and early warning method for skid-mounted equipment according to any one of claims 1 to 4, and the device comprises: The causal association library construction module is used to perform functional coordination and fault impact causal relationship analysis based on the collaborative functional blocks of the skid-mounted equipment, and to build a causal association library; A relationship identification acquisition module is used to disassemble and cluster the causal association library according to the input data, data analysis, and result output logic framework to obtain clustering results and corresponding relationship identifications, wherein the relationship identifications include input relationship identifications, processing relationship identifications, and output relationship identifications, wherein the clustering processing includes: clustering the attributes of functional similarity according to the causal relationship of the functional blocks, clustering the cross-influence relationships according to the data source dependency based on the attribute clustering results, identifying the cross-influence strength of the causal relationship, obtaining clustering results, and generating relationship identifications of causal relationship nodes according to the clustering results; A hierarchical warning information determination module, configured to search for warning paths based on the causal association library according to the clustering results and corresponding relationship identifiers, and determine hierarchical warning information; The early warning identification result acquisition module is used to perform early warning monitoring of the collaborative functional blocks of the skid-mounted equipment according to the relationship identifier and the hierarchical early warning information to obtain the early warning identification result.

Images