Abnormity analysis method and system applied to diborane production control system

Abstract

The invention provides an anomaly analysis method and system applied to a diborane production control system, and the method comprises the steps: firstly obtaining a multi-source production data set which is generated in the operation of the diborane production control system and comprises data with timestamp marks, such as equipment states, raw material conveying and reaction parameters, and then constructing a production association knowledge graph; the method comprises the following steps: carrying out physical connection between production equipment or parameters corresponding to production nodes and between equipment or logic association between parameters corresponding to production edges, then carrying out graph feature extraction on a production association knowledge graph, calling a pre-training graph neural network model to carry out graph structure analysis, and generating an anomaly detection result containing different anomaly type detection confidence degrees; and determining an abnormal event type and distribution characteristic information in the knowledge graph according to an anomaly detection result, and finally generating a production early warning instruction containing an event positioning identifier and sending the production early warning instruction to the production control terminal to trigger an anomaly response operation, thereby comprehensively and accurately analyzing the anomaly in diborane production, and improving the production safety and stability.

Description

Technical Field

[0001] The present application relates to the field of intelligent manufacturing technology, and more specifically, to an abnormality analysis method and system applied to a diborane production control system. Background Art

[0002] In the production of diborane, a key chemical raw material, its production process involves complex physical and chemical reactions and the coordinated operation of equipment. Traditional diborane production control systems rely primarily on manual experience and simple threshold alarm mechanisms to monitor production status. Manual experience is often limited by the operator's professional level and subjective judgment, making it difficult to fully and accurately grasp the various potential abnormalities in the production process. Simple threshold alarm mechanisms can only judge based on the fixed threshold of a single parameter and fail to account for the complex interrelationships between multiple parameters in the production process.

[0003] In actual production, diborane production control systems generate a large amount of multi-source production data, including equipment status data, raw material delivery data, and reaction parameter data. This data is closely correlated, but existing methods lack effective means to mine and utilize this correlated information. This results in a lack of timely detection of potential anomalies in the production process, which can lead to production accidents, affect product quality and production efficiency, and even pose safety risks. Summary of the Invention

[0004] In view of the above-mentioned problems, in combination with the first aspect of the present application, an embodiment of the present application provides an abnormality analysis method applied to a diborane production control system, the method comprising: Acquire a multi-source production data set generated during the operation of a diborane production control system, wherein the multi-source production data set includes equipment status data, raw material delivery data, and reaction parameter data with time stamps; Constructing a production-related knowledge graph based on the multi-source production data set, wherein the production-related knowledge graph includes a plurality of production nodes and production edges connecting the production nodes, wherein the production nodes correspond to diborane production equipment or production parameters, and the production edges correspond to physical connection relationships between production equipment or logical association relationships between production parameters; Performing graph feature extraction processing on the production-related knowledge graph to obtain a node feature set of the production node and an edge feature set of the production edge; Calling a pre-trained graph neural network model to perform graph structure analysis on the node feature set and the edge feature set to generate an anomaly detection result of the production-related knowledge graph, wherein the anomaly detection result includes detection confidence levels corresponding to different anomaly types; Determine the type of abnormal event present in the diborane production process and distribution characteristic information of the abnormal event in the production-related knowledge graph according to the abnormality detection result; A production warning instruction including an event location identifier is generated based on the abnormal event type and the distribution characteristic information, and the production warning instruction is sent to a diborane production control terminal to trigger an abnormal response operation.

[0005] On the other hand, an embodiment of the present application also provides an abnormality analysis system applied to a diborane production control system, including a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to run the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0006] Based on the above aspects, the embodiment of the present application obtains a multi-source production data set generated during the operation of the diborane production control system, constructs a production-related knowledge graph based on the multi-source production data set, and can intuitively display the association between production equipment and production parameters, which helps to deeply explore the potential information behind the data, perform graph feature extraction processing on the production-related knowledge graph, and call a pre-trained graph neural network model for graph structure analysis processing. It can fully utilize the advantages of graph neural networks in processing complex graph structure data, accurately identify abnormal situations in the production process, and generate abnormal detection results containing detection confidences corresponding to different abnormal types, thereby improving the accuracy and reliability of abnormal detection. The type of abnormal event and its distribution feature information in the production-related knowledge graph are determined according to the abnormal detection results, and the specific circumstances of the abnormal event can be fully understood. Based on the abnormal event type and distribution feature information, a production warning instruction containing an event location identifier is generated, and the production warning instruction is sent to the diborane production control terminal to trigger an abnormal response operation, thereby achieving timely warning and rapid response to abnormal events, effectively avoiding the occurrence of production accidents, improving the safety and stability of diborane production, and ensuring product quality and production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 It is a schematic diagram of the execution flow of the abnormality analysis method applied to the diborane production control system provided in an embodiment of the present application.

[0008] Figure 2 Schematic diagram of the hardware architecture of an abnormality analysis system applied to a diborane production control system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0009] The present application will be described in detail below with reference to the accompanying drawings. Figure 1This is a flow chart of an abnormality analysis method applied to a diborane production control system provided by an embodiment of the present application. The abnormality analysis method applied to a diborane production control system is introduced in detail below.

[0010] Step S110: Acquire a multi-source production data set generated during the operation of the diborane production control system, wherein the multi-source production data set includes equipment status data with timestamps, raw material delivery data, and reaction parameter data.

[0011] During the diborane production process, the production control system continuously generates a large amount of production-related data. In order to effectively analyze anomalies in the diborane production process, it is first necessary to obtain this multi-source production data set.

[0012] Equipment status data describes the operating status of diborane production equipment. Different production equipment has different status characteristics. For example, the status of a reactor may include temperature control status, pressure control status, and agitator operating status; the status of a pump may include speed and flow stability. This equipment status data is collected in real time by various sensors installed on the equipment. Each equipment status data point is assigned a timestamp that is accurate to the specific moment, which facilitates subsequent analysis of equipment status trends over time. For example, at a specific moment, the reactor's temperature sensor records temperature data within the reactor. This temperature data is also stamped with the precise time for subsequent traceability and analysis. Parameters reflecting the core operating status of the equipment, such as the reactor's temperature control status, agitator operating status, and pump speed stability, are key data for subsequent anomaly analysis. They play a vital role in constructing real-time operating status attributes for knowledge graph nodes, generating device node feature vectors, and locating the device nodes corresponding to anomaly events.

[0013] Raw material delivery data primarily covers the transportation of various raw materials during the production process. Raw material delivery is crucial to diborane production, as factors such as the delivery volume and speed directly impact the reaction. Raw material delivery data includes information such as raw material flow rate, pressure, and delivery time. This data is also collected by sensors installed on the raw material delivery pipelines, and each data point is timestamped. For example, flow sensors and pressure sensors are installed on the raw material delivery pipelines to monitor the raw material flow and pressure in real time and transmit time-stamped data to the production control system. This flow rate, pressure, and delivery rate data are used to analyze material transfer anomalies between equipment. They play a key role in determining the transmission rate attributes of the physical connections between production equipment, generating transmission rate signatures, and analyzing the spread of anomalies along these connections.

[0014] Reaction parameter data is key data reflecting the diborane production reaction process. These parameters include reaction temperature, reaction pressure, reaction time, and reactant concentration. The accuracy of reaction parameter data is crucial for controlling the reaction process and ensuring product quality. This data is collected by sensors installed inside or around the reaction equipment, and each data point is timestamped. For example, temperature and pressure sensors are installed in the reactor to monitor the temperature and pressure in real time and transmit the timestamped data to the production control system. Data such as reaction temperature, pressure, reactant concentration, and reaction time are important data for analyzing logical anomalies in the reaction process. They are used to construct real-time measurement value attributes for production parameter nodes, generate parameter node feature vectors, and extract logical associations between parameters based on reaction kinetic models.

[0015] Step S120: Construct a production-related knowledge graph based on the multi-source production data set, wherein the production-related knowledge graph includes multiple production nodes and production edges connecting the production nodes, wherein the production nodes correspond to diborane production equipment or production parameters, and the production edges correspond to the physical connection relationship between production equipment or the logical association relationship between production parameters.

[0016] After acquiring a multi-source production data set, the next step is to construct a production-related knowledge graph based on this data set. A production-related knowledge graph is a graph structure used to represent the relationships between various production factors in the diborane production process. By constructing this knowledge graph, we can more intuitively understand the relationships between various production equipment and production parameters in the diborane production process.

[0017] The production equipment nodes here specifically include reactors (such as "reactor 001"), raw material delivery pumps (such as "delivery pump 002"), heaters, coolers, gas compressors, purification towers, storage tanks, etc. Each node contains equipment type attributes (such as "reactor") and real-time operation status attributes (such as "normal operation"); production parameter nodes specifically include reaction temperature, reaction pressure, reaction time, reactant concentration (such as diborane concentration, hydrogen concentration), raw material flow (such as sodium borohydride solution flow, hydrochloric acid flow), raw material purity, agitator speed, etc. Each node contains parameter type attributes (such as "reaction temperature") and real-time Measurement value attributes (such as "300K"); production edges are specifically divided into three categories: the first is the physical connection edge between devices, such as the pipeline connection between the reactor and the delivery pump, whose attributes include connection type (such as "pipeline connection") and transmission rate (such as "50L / min"); the second is the logical association edge between parameters, such as the association between reaction temperature and reaction rate, whose attributes include influence direction (such as "temperature increases - rate increases") and association strength (such as "0.9"); the third is the device-parameter association edge, such as the measurement association between the reactor and reaction temperature, whose attributes include association type (such as "measurement association") and acquisition frequency (such as "1 time / second").

[0018] Step S121: performing data alignment processing on the multi-source production data set to obtain an aligned production data sequence.

[0019] Because the production data in a multi-source production data set comes from different data sources and may be collected at different frequencies and times, it is necessary to align these production data. The purpose of data alignment is to make the production data from different data sources consistent in time to facilitate subsequent analysis and processing.

[0020] First, the timestamps of the production data in the multi-source production data set need to be unified. Since production data from different data sources may have different timestamp formats, these timestamps need to be converted to a unified format. For example, the timestamps of production data from different data sources should be converted to a standard time format, such as year-month-day hour:minute:second.

[0021] Then, the data is sorted according to the unified timestamps. Sorting the data in chronological order makes the data appear as a continuous sequence in time.

[0022] Next, production data needs to be interpolated or sampled to ensure temporal alignment of data from different data sources. If different data sources have different data collection frequencies, data at certain time points may be missing. Interpolation can be used to supplement this missing data. For example, linear interpolation can be used to estimate the data at the missing time points based on data from adjacent time points. If the data collection frequency of some data sources is too high, while that of other data sources is low, sampling can be used to downsample the data collected at the high frequency to align it with the data collected at the low frequency.

[0023] Through the above data alignment processing steps, an aligned production data sequence is finally obtained. The data in the aligned production data sequence is consistent in time and can be easily analyzed and processed later.

[0024] Step S122: extracting diborane production equipment identification information from the aligned production data sequence, and mapping the data set corresponding to each diborane production equipment identification information into a production equipment node, wherein the production equipment node includes an equipment type attribute and a real-time operation status attribute.

[0025] After obtaining the aligned production data sequence, it is necessary to extract the diborane production equipment identification information from the aligned production data sequence. This diborane production equipment identification information uniquely identifies each production equipment, such as the equipment number or name. This identification information allows accurate identification of each production equipment and mapping its corresponding data set to a production equipment node.

[0026] For each diborane production equipment identification information, the corresponding data set can be integrated and processed to form a production equipment node. The production equipment node contains two important attributes: equipment type attribute and real-time operation status attribute.

[0027] The equipment type attribute is used to describe the type of production equipment, such as reactors, transfer pumps, and heaters. Different types of production equipment have different functions and roles in the diborane production process, so the equipment type attribute is very important for subsequent analysis and processing.

[0028] Real-time operating status attributes reflect the current operating status of production equipment. These attributes can be determined based on equipment status data, such as whether the equipment is operating normally and whether there are any fault warnings. Using these attributes, you can promptly understand the operating status of production equipment and take appropriate measures.

[0029] For example, for a reactor, its equipment identification information is "reactor 001." Relevant data about the reactor, including equipment status data, is extracted from the aligned production data sequence. This data is then integrated into a production equipment node, whose equipment type attribute is "reactor." The real-time operating status attribute is determined based on the equipment status data, such as "normal operation" or "fault warning."

[0030] Step S123: extracting diborane production parameter identification information from the aligned production data sequence, and mapping the data set corresponding to each diborane production parameter identification information into a production parameter node, wherein the production parameter node includes a parameter type attribute and a real-time measurement value attribute.

[0031] In addition to the production equipment nodes, it is also necessary to extract diborane production parameter identification information from the aligned production data sequence and map its corresponding data set to a production parameter node. This diborane production parameter identification information uniquely identifies each production parameter, such as its name and number. This identification information allows accurate identification of each production parameter and mapping its corresponding data set to a production parameter node.

[0032] For each diborane production parameter identification information, the corresponding data set can be integrated and processed to form a production parameter node. The production parameter node contains two important attributes: parameter type attribute and real-time measurement value attribute.

[0033] The parameter type attribute is used to describe the type of production parameters, such as reaction temperature, reaction pressure, and reactant concentration. Different types of production parameters have different roles and meanings in the diborane production process, so the parameter type attribute is very important for subsequent analysis and processing.

[0034] Real-time measurement attributes reflect the current measured value of a production parameter. These attributes can be determined based on reaction parameter data, such as the specific values ​​of reaction temperature and reaction pressure. Using these attributes, you can monitor changes in production parameters and adjust the production process accordingly.

[0035] For example, for the production parameter reaction temperature, its parameter identification information is "Reaction Temperature 001." Relevant data for this reaction temperature, including reaction parameter data, is extracted from the aligned production data sequence. This data is then integrated into a production parameter node, whose parameter type attribute is "Reaction Temperature." The real-time measurement value attribute is determined based on the reaction parameter data, such as the specific temperature value.

[0036] Step S124: extracting the physical connection relationship between diborane production equipment based on the diborane production process flow chart, and establishing production edges between equipment nodes with direct material transmission or energy exchange, wherein the production edges include connection type attributes and transmission rate attributes.

[0037] In this embodiment, it is necessary to supplement the association edges between production equipment and production parameters. For example, the reactor and reaction temperature and reaction pressure are "measurement associations", and the edge attributes include parameter acquisition frequency; the delivery pump and raw material flow are "control associations", and the edge attributes include flow regulation accuracy, etc. Such edges are used to establish a direct association between equipment and parameters to ensure the integrity of the knowledge graph network.

[0038] After determining the production equipment nodes and production parameter nodes, it is necessary to extract the physical connection relationships between production equipment based on the diborane production process flow chart. The diborane production process flow chart details the physical connection methods and material transmission paths between each production equipment during the production process.

[0039] By analyzing the diborane production process flow chart, we can identify which production equipment has direct material transfer or energy exchange relationships. For equipment nodes with direct material transfer or energy exchange, we need to establish production edges between them. Production edges represent the physical connection between production equipment and contain two important attributes: connection type and transfer rate.

[0040] The connection type attribute is used to describe the connection method between production equipment, such as pipe connections and cable connections. Different connection types have different characteristics and functions in the production process, so the connection type attribute is very important for subsequent analysis and processing.

[0041] The transfer rate attribute reflects the rate of material transfer or energy exchange between production equipment. This attribute can be determined based on material delivery data and equipment operating data, such as material flow rate and energy transfer power. This attribute can be used to understand the material transfer and energy exchange between production equipment, enabling optimization of the production process.

[0042] For example, in a diborane production process flow chart, the reactor and the delivery pump are connected by a pipeline, and there is a raw material transmission relationship. In this case, a production edge can be established between the production equipment node corresponding to the reactor and the production equipment node corresponding to the delivery pump. The connection type attribute of this production edge is "Pipeline Connection", and the transmission rate attribute is determined by the raw material transmission data, such as the raw material delivery flow rate.

[0043] Step S125: extracting the logical association relationship between diborane production parameters based on the diborane production reaction kinetics model, and establishing production edges between parameter nodes with causal influence or coordinated changes, wherein the production edges include an influence direction attribute and an association strength attribute.

[0044] The correlation strength attribute is calculated by the Pearson correlation coefficient, and the formula is: , where xi and yi are the historical measured values ​​of the two parameters, 、 For example, the correlation coefficient between reaction temperature and reaction rate is calculated to be 0.9, indicating a strong positive correlation.

[0045] In addition to the physical connections between production equipment, it is also necessary to extract the logical relationships between production parameters based on the diborane production reaction kinetics model. The diborane production reaction kinetics model describes the interactions and changes between various reaction parameters during the production process.

[0046] By analyzing the diborane production reaction kinetics model, we can identify production parameters that have causal or synergistic relationships. For parameter nodes with causal or synergistic relationships, we need to establish production edges between them. Production edges represent the logical associations between production parameters and contain two important attributes: the direction of influence and the strength of the association.

[0047] The influence direction attribute describes the direction of causal influence between production parameters. For example, a change in one parameter leads to an increase or decrease in another parameter. The influence direction attribute can be determined based on the reaction kinetics model and actual production data. For example, the influence direction can be determined by analyzing the trend of parameter changes during the reaction.

[0048] The strength of association attribute reflects the degree of association between production parameters. This can be determined by calculating the correlation coefficient, for example, by analyzing the data of two parameters at multiple time points and calculating the correlation coefficient between them. The magnitude of the correlation coefficient reflects the strength of the association between the two parameters.

[0049] For example, in the diborane production reaction kinetics model, there's a causal relationship between reaction temperature and reaction rate: an increase in reaction temperature leads to an increase in reaction rate. In this case, a production edge can be established between the production parameter node corresponding to the reaction temperature and the production parameter node corresponding to the reaction rate. The influence direction attribute of this production edge is "an increase in reaction temperature leads to an increase in reaction rate," and the strength of the association attribute is determined by calculating the correlation coefficient.

[0050] Step S126: Construct an initial production-related knowledge graph based on the production equipment nodes, production parameter nodes and corresponding production edges, perform connectivity verification on the initial production-related knowledge graph, delete isolated nodes and corresponding edges that are not connected to the main production process, and generate a final production-related knowledge graph.

[0051] After determining the production equipment nodes, production parameter nodes, and the production edges between them, these nodes and edges are combined to construct an initial production-related knowledge graph. This initial production-related knowledge graph is a graph structure consisting of multiple production nodes and the production edges connecting them. It preliminarily reflects the relationships between the various production factors in the diborane production process.

[0052] However, the initial production-related knowledge graph may contain some isolated nodes and corresponding edges that are not connected to the main production process. These isolated nodes and edges may be caused by data collection errors, equipment failures, or other reasons. They are not practical for analyzing abnormalities in the diborane production process. Therefore, it is necessary to perform a connectivity check on the initial production-related knowledge graph and delete these isolated nodes and corresponding edges.

[0053] Step S1261: extracting a core node set corresponding to the main production process from the initial production-related knowledge graph, wherein the core node set includes a raw material input node, a reactor node, and a product output node.

[0054] First, we need to extract the core node set corresponding to the main production process from the initial production-related knowledge graph. The main production process is the main process in the diborane production process, which includes key links such as raw material input, reaction, and product output. The core node set includes three important nodes: the raw material input node, the reactor node, and the product output node.

[0055] The raw material input node represents the raw material input link in the production process and is connected to the raw material transportation equipment and raw material-related production parameter nodes. The reactor node represents the reaction link in the production process. As the core equipment for diborane production, it is closely related to other production equipment nodes and reaction parameter nodes. The product output node represents the product output link in the production process and is connected to the product transportation equipment and product-related production parameter nodes.

[0056] By extracting the core node set, the key nodes of the main production process can be identified.

[0057] Step S1262: Use a breadth-first search algorithm to traverse the initial production-related knowledge graph starting from the core node set, and mark all reachable nodes and corresponding edges that have path connections with the core nodes.

[0058] After determining the core node set, the initial production-related knowledge graph is traversed using the breadth-first search algorithm starting from the core node set. The breadth-first search algorithm is an algorithm used to traverse a graph structure. It starts from the starting node and traverses the nodes in the graph layer by layer until all reachable nodes are traversed.

[0059] Starting from each core node in the core node set, a breadth-first search algorithm is used to traverse the network. During the traversal, all reachable nodes and corresponding edges that have paths connecting to the core node can be marked. A reachable node is a node that can be reached from the core node via production edges.

[0060] By traversing the breadth-first search algorithm, we can determine which nodes and edges are connected to the main production process and which nodes and edges are isolated.

[0061] Step S1263: Identify unmarked nodes and corresponding edges in the initial production-related knowledge graph, where the unmarked nodes and corresponding edges are isolated nodes and edges that are not connected to the main production process.

[0062] After completing the breadth-first search algorithm, we can identify unlabeled nodes and corresponding edges in the initial production-related knowledge graph. These unlabeled nodes and corresponding edges are isolated nodes and edges that are not connected to the main production process. They are not practical for analyzing abnormalities in the diborane production process, so they need to be deleted.

[0063] Step S1264: Delete the isolated nodes and corresponding edges from the initial production-related knowledge graph to generate a production-related knowledge graph after connectivity verification.

[0064] After identifying isolated nodes and their corresponding edges, we delete them from the initial production-related knowledge graph. This deletion results in a connectivity-verified production-related knowledge graph. All nodes and edges in this production-related knowledge graph are connected to the main production process, making it more suitable for subsequent anomaly analysis.

[0065] Step S1265: Counting the number of nodes and edges of the production-related knowledge graph after the connectivity check to ensure that the number of remaining nodes and edges meets the basic structural requirements of the diborane production process, and generating a final production-related knowledge graph.

[0066] After obtaining the connectivity-verified production-related knowledge graph, we need to count the number of nodes and edges. The diborane production process has certain basic structural requirements, such as a certain number of production equipment nodes and production parameter nodes, as well as the connection relationships between them.

[0067] By counting the number of nodes and edges, we can ensure that the production-related knowledge graph, after connectivity verification, meets the basic structural requirements of the diborane production process. If the number of nodes or edges does not meet the requirements, further inspection and adjustment of the graph structure may be necessary, such as adding missing nodes or edges or deleting redundant nodes or edges.

[0068] After counting the number of nodes and edges, a final production-related knowledge graph that meets the basic structural requirements of the diborane production process is generated. This final production-related knowledge graph accurately reflects the relationships between the various production factors in the diborane production process.

[0069] Step S130: performing graph feature extraction processing on the production-related knowledge graph to obtain a node feature set of the production nodes and an edge feature set of the production edges.

[0070] After obtaining the final production-related knowledge graph, graph feature extraction is performed to obtain node feature sets for production nodes and edge feature sets for production edges. Graph feature extraction converts node and edge information in the production-related knowledge graph into feature vectors that can be used for subsequent analysis, thereby improving anomaly analysis.

[0071] Step S131: performing feature coding processing on the real-time operating status attributes of the production equipment node to generate a feature vector of the equipment node, wherein the feature coding processing includes state category hot coding and state change rate normalization.

[0072] The state change rate is calculated as follows: State Change Rate = (S(t2) - S(t1)) / (t2 - t1), where S(t) is the state of the device at time t ("Normal" is 1, "Warning" is 0.5, and "Fault" is 0), and t2 - t1 is the time interval. For example, if the reactor state is "Normal" (1) at 10:00 and changes to "Warning" (0.5) at 10:05, the state change rate is (0.5 - 1) / 5 minutes = -0.1 / minute.

[0073] First, feature encoding is performed on the real-time operating status attributes of production equipment nodes. Real-time operating status attributes describe the current operating status of production equipment, such as whether the equipment is operating normally or has any fault warnings. Feature encoding is required to convert this information into a feature vector that can be used for analysis.

[0074] The feature encoding process consists of two steps: state category hot encoding and state change rate normalization.

[0075] Category hot encoding converts the operational status of production equipment into a binary vector. For example, the operational status of production equipment may be classified as normal operation, fault warning, or shutdown. Each status category can be encoded as a binary vector with only one position set to 1 and all other positions set to 0. Category hot encoding converts the operational status of production equipment into digital features.

[0076] State change rate normalization is the process of normalizing the rate of change of the operating state of production equipment. The operating state of production equipment may change over time, and the state change rate reflects the rate of change. To eliminate differences in the state change rate between different equipment, the state change rate needs to be normalized. Normalization converts the state change rate into a numerical value between 0 and 1, making the state change rates of different equipment comparable. Normalization typically uses linear normalization to map the state change rate to the range of 0 to 1. For example, the maximum and minimum state change rates of all production equipment are first calculated, and then the state change rate of each equipment is converted to a numerical value between 0 and 1. Through the two steps of state category hot encoding and state change rate normalization, a device node feature vector is ultimately generated. This device node feature vector contains the category information and state change rate information of the production equipment's operating state, and can comprehensively reflect the real-time operating status of the production equipment.

[0077] Step S132: performing feature coding processing on the real-time measurement value attributes of the production parameter node to generate a parameter node feature vector, wherein the feature coding processing includes parameter value range binning and parameter change trend symbolization.

[0078] Next, feature encoding is performed on the real-time measurement attributes of the production parameter nodes. Real-time measurement attributes describe the current value of production parameters, such as reaction temperature and reaction pressure. To convert this information into feature vectors suitable for analysis, feature encoding is required. This involves two steps: binning the parameter range and symbolizing parameter change trends.

[0079] Parameter range binning is the process of dividing the range of a production parameter into several intervals. Different production parameters have different ranges of values. By binning their ranges, continuous values ​​can be converted into discrete categories. For example, the range of the reaction temperature parameter may fall within a large interval. This interval can be divided into several smaller intervals, each corresponding to a category. For each real-time measurement of a production parameter, the interval within which it falls is determined and then encoded into the category corresponding to that interval. This allows the specific numerical information of the production parameter to be converted into more easily processed categorical information.

[0080] Parameter trend symbolization is the process of symbolizing the changing trends of production parameters. The values ​​of production parameters may change over time, and the trend reflects the direction of this change, such as an increase, decrease, or no change. Symbolization is used to easily and intuitively represent these trends. For example, a positive sign indicates an increase in the parameter value, a negative sign indicates a decrease, and zero indicates a constant value. By symbolizing parameter trends, we can convert production parameter trend information into digital features.

[0081] After two steps of feature encoding, namely, parameter range binning and parameter change trend symbolization, a parameter node feature vector is generated. This parameter node feature vector contains the value range information and change trend information of the production parameter, which can effectively reflect the real-time status of the production parameter.

[0082] Step S133: merging the device node feature vector and the parameter node feature vector to obtain a node feature set of the production node.

[0083] After generating the device node feature vector and parameter node feature vector, they need to be merged to obtain the node feature set of the production node. In order to make the merged feature vector reflect the information of the production node more comprehensively and accurately, some preprocessing operations need to be performed on the device node feature vector and parameter node feature vector.

[0084] Step S1331: performing dimension expansion processing on the device node feature vector to generate an expanded device node feature vector, wherein the dimension expansion processing includes adding a device type code dimension.

[0085] The device node feature vector is dimensionally expanded. Device type is an important attribute of production equipment. Different types of equipment have different functions and characteristics during the production process. In order to reflect the device type information in the feature vector, it is necessary to add a device type encoding dimension. Device type encoding is the process of digitally representing the type of device. For example, a one-hot encoding method can be used to encode each device type as a binary vector where only one position is 1 and the rest are 0. By adding the device type encoding dimension, the dimension of the device node feature vector is expanded, generating an expanded device node feature vector. This expanded device node feature vector not only contains the real-time operating status information of the production equipment, but also contains information about the device type.

[0086] Step S1332: performing dimension expansion processing on the parameter node feature vector to generate an extended parameter node feature vector, wherein the dimension expansion processing includes adding a parameter type encoding dimension.

[0087] Similarly, the parameter node feature vector is dimensionally expanded. Parameter type is a key attribute of production parameters, and different types of parameters have different meanings and roles in the production process. To reflect parameter type information in the feature vector, a parameter type encoding dimension needs to be added. Parameter type encoding uses a method similar to device type encoding, encoding each parameter type as a binary vector. By adding the parameter type encoding dimension, the dimension of the parameter node feature vector is expanded, generating an extended parameter node feature vector. This extended parameter node feature vector not only contains the real-time measurement values ​​and change trend information of the production parameters, but also includes parameter type information, which can more accurately reflect the characteristics of the production parameters.

[0088] Step S1333: concatenate the extended device node feature vector and the extended parameter node feature vector to generate a merged node feature vector.

[0089] After obtaining the extended device node feature vector and the extended parameter node feature vector, they are concatenated. Concatenation involves connecting the two vectors in a predetermined order. This concatenation integrates the feature information of the device node and the parameter node into a single vector, generating a merged node feature vector. This merged node feature vector incorporates multiple aspects of production equipment and production parameters, providing a more comprehensive description of the production node's characteristics.

[0090] Step S1334: Arrange the merged node feature vectors in the order of the timestamps of the production nodes to generate a node feature set of the production nodes.

[0091] Finally, the merged node feature vectors are sorted in the order of the production node's timestamp. Since the production process continues over time, the characteristics of production nodes also change over time. Sorting the merged node feature vectors in timestamp order allows the node feature set to reflect how production node characteristics change over time. This sorted node feature set is therefore more useful for subsequent anomaly analysis, as anomalies may be related to temporal changes in production node characteristics. This ultimately generates the node feature set for the production node.

[0092] Step S134: performing feature extraction processing on the transmission rate attribute of the production edge to generate transmission rate features, wherein the feature extraction processing includes calculating the historical mean value of the rate and statistics of the rate fluctuation range.

[0093] For the transmission rate attribute on the production edge, feature extraction is required to generate a transmission rate feature. The transmission rate attribute reflects the speed of material transfer or energy exchange between production equipment. Feature extraction can help better analyze anomalies in the production process. Feature extraction involves calculating the historical mean of the rate and analyzing the rate fluctuation range.

[0094] Calculating the historical average rate is the process of calculating the average transmission rate of the production side over a period of time. The transmission rate of the production side may fluctuate over time. By calculating the historical average rate, a relatively stable reference value can be obtained. First, all data on the production side transmission rate over a historical period is collected. Then, this data is added and divided by the number of data points to obtain the historical average rate. This historical average rate reflects the overall level of the production side transmission rate and provides a benchmark for determining whether the current transmission rate is abnormal.

[0095] Rate fluctuation range statistics are used to measure the degree of transmission rate fluctuation at the production side over a period of time. The transmission rate at the production side may fluctuate within a certain range, and the fluctuation range reflects the stability of the transmission rate. To calculate the rate fluctuation range, we first find the maximum and minimum transmission rate values ​​within the historical time period and then calculate the difference between them. This difference is the rate fluctuation range. The rate fluctuation range reflects the magnitude of the transmission rate change at the production side and is important for determining whether the transmission rate is experiencing abnormal fluctuations.

[0096] Through the two steps of calculating the historical average of the rate and summarizing the rate fluctuation range, a transmission rate feature is finally generated. This transmission rate feature contains the average level and fluctuation range of the production-side transmission rate, which can fully reflect the production-side transmission rate situation.

[0097] Step S135: performing feature extraction processing on the association strength attribute of the production edge to generate an association strength feature, wherein the feature extraction processing includes calculating the correlation coefficient of historical data and statistics of the impact lag time.

[0098] The historical data correlation coefficient here differs from the correlation strength attribute in step S125. The correlation strength in step S125 is a "theoretical correlation coefficient" based on a reaction kinetic model, such as the relationship between temperature and rate derived through a formula. The historical data correlation coefficient in this step is a "measured correlation coefficient" calculated based on actual production data, reflecting the dynamic correlation deviations in actual production. The combination of the two comprehensively describes the correlation strength. The impact lag time statistics are analyzed using a sliding window method. Specifically, the historical data of nodes A and B are sorted by time series. A window size is set (e.g., 10 minutes), and the difference between the change time tA of A and the corresponding change time tB of B is calculated as |tB - tA|. The value with the highest frequency of occurrence of this difference is counted as the impact lag time. For example, if after A (temperature) increases, B (rate) increases most frequently 2 minutes later, the lag time is 2 minutes.

[0099] We perform feature extraction on the correlation strength attributes of production edges to generate correlation strength features. This attribute reflects the degree of correlation between production parameters or equipment. Extracting features from this attribute helps analyze inherent connections and anomalies within the production process. Feature extraction involves calculating the correlation coefficient of historical data and calculating the impact lag time.

[0100] Calculating the historical data correlation coefficient involves calculating the correlation coefficient between the relevant parameters of two nodes connected by a production edge in historical data. The correlation coefficient measures the strength of the linear relationship between two variables. By calculating the correlation coefficient, we can understand the closeness of the connection between the two nodes connected by a production edge. First, we collect all the relevant parameters of the two nodes connected by the production edge within a historical time period, and then calculate the correlation coefficient between these two sets of data. The correlation coefficient ranges from -1 to 1. The closer the absolute value is to 1, the stronger the connection between the two nodes; the closer the absolute value is to 0, the weaker the connection between the two nodes.

[0101] Impact lag statistics are the process of calculating the impact delay between two nodes connected by a production edge. During the production process, changes at one node may not immediately affect another node, but rather with a certain time delay. By calculating impact lag statistics, we can understand the impact transmission mechanism between two nodes connected by a production edge. To calculate impact lag, we need to analyze the time difference between changes at one node and the corresponding changes at another node in historical data and identify the most common time difference as the impact lag. Impact lag reflects the dynamic relationship between the two nodes connected by a production edge.

[0102] By calculating the historical data correlation coefficient and analyzing the impact lag time, we ultimately generate a correlation strength feature. This feature contains information about the closeness of the association between the two nodes connected by the production edge and the impact lag time, comprehensively reflecting the strength of the production edge's association.

[0103] Step S136: Merge the transmission rate feature and the association strength feature to obtain an edge feature set of the production edge.

[0104] When merging, it is necessary to include the connection type and influence direction features of the production edge. The connection type (such as pipeline connection - 1, cable connection - 2) is converted into a feature through one-hot encoding, and the influence direction (positive - 1, negative - -1) is converted into a feature through sign encoding. The final edge feature set includes "transmission rate + connection type + association strength + influence direction + historical correlation coefficient + lag time".

[0105] After generating the transmission rate and association strength features, they are merged to form the edge feature set for production edges. This merging process combines the transmission rate and association strength features. Because these two features reflect the characteristics of production edges from different perspectives, merging them provides a more comprehensive description of production edge information. When merging, attention must be paid to feature dimensionality matching to ensure that the merged edge feature set accurately reflects the various characteristics of production edges. This merging method ultimately generates the edge feature set for production edges.

[0106] Step S140: Call the pre-trained graph neural network model to perform graph structure analysis on the node feature set and the edge feature set to generate anomaly detection results for the production-related knowledge graph, where the anomaly detection results include detection confidence levels corresponding to different anomaly types.

[0107] After obtaining the node feature sets of production nodes and the edge feature sets of production edges, a pre-trained graph neural network model is used to perform graph structure analysis on these feature sets to generate anomaly detection results for the production-related knowledge graph. A graph neural network model is an artificial intelligence model specifically designed to process graph-structured data. It can automatically learn the complex relationships between nodes and edges in the graph to perform anomaly detection.

[0108] Step S141: Input the node feature set and the edge feature set into the input layer of the graph neural network model, and the input layer maps the node feature vector and the edge feature vector into a node embedding vector and an edge embedding vector, respectively.

[0109] First, the node feature set and edge feature set are fed into the input layer of the graph neural network model. The input layer is the first layer of the graph neural network model. Its primary function is to map the input node feature vectors and edge feature vectors into node embedding vectors and edge embedding vectors, respectively. Node embedding vectors and edge embedding vectors are low-dimensional vector representations that better preserve node and edge feature information and facilitate processing in subsequent layers. The input layer transforms the node feature vectors and edge feature vectors through a series of linear transformations and nonlinear activation functions, allowing them to be processed in a more suitable form in subsequent layers of the graph neural network model.

[0110] Step S142: Perform neighborhood information aggregation processing on the node embedding vector and the edge embedding vector through the graph convolution layer of the graph neural network model to generate an aggregated node vector containing local structural information, wherein the neighborhood information aggregation processing includes the weighted summation of the node's own features and the features of adjacent nodes.

[0111] Next, the graph convolutional layer of the graph neural network model performs neighborhood information aggregation on the node and edge embedding vectors. The graph convolutional layer is one of the core layers of the graph neural network model, capable of capturing local structural information between nodes in the graph. Neighborhood information aggregation is the process of weighted summing the features of a node with those of its neighboring nodes.

[0112] Step S1421: define a neighborhood range for each production node, where the neighborhood range includes first-order neighbor nodes and corresponding edges directly connected to the production node.

[0113] First, define a neighborhood for each production node. A neighborhood is the set of nodes and edges directly connected to the production node. Here, we primarily consider first-order neighbors—those directly connected to the production node. Each production node's neighborhood includes its first-order neighbors and the edges connecting them. By defining a neighborhood, we can clearly define the local environment of each node in the graph.

[0114] Step S1422: Perform a weighted summation process on the node embedding vector of each production node and the node embedding vectors of all neighboring nodes in its neighborhood to obtain a weighted summation result, where the weight of the weighted summation process is calculated by the edge embedding vector of the corresponding edge.

[0115] Then, a weighted summation is performed on the node embedding vector of each production node and the node embedding vectors of all neighboring nodes in its neighborhood. The weight of the weighted summation is calculated from the edge embedding vector of the corresponding edge. The edge embedding vector contains the feature information of the edge, and the weight used for the weighted summation can be obtained from the edge embedding vector. For each production node, its own node embedding vector is multiplied by a weight, and the node embedding vector of each neighboring node in its neighborhood is also multiplied by the corresponding weight. These weighted vectors are then added together to obtain the weighted summation result. This weighted summation result combines the characteristics of the production node itself and the characteristics of the neighboring nodes in its neighborhood, and can better reflect the local structural information of the node in the graph.

[0116] Step S1423: concatenate the weighted summation result with the original node embedding vector of the production node to generate an intermediate node vector containing the production node's own features and neighborhood features.

[0117] Next, the weighted summation result is concatenated with the original node embedding vector of the production node. Concatenation involves connecting two vectors in a set order. This concatenation integrates the production node's own characteristics and those of its neighborhood into a single vector, generating an intermediate node vector. This intermediate node vector incorporates both the original characteristics of the production node itself and the characteristics of its neighboring nodes within its neighborhood, providing a more comprehensive representation of the local characteristics of the production node in the graph.

[0118] Step S1424: performing nonlinear activation processing on the intermediate node vector to generate an activated intermediate node vector, wherein the nonlinear activation processing uses a ReLU activation function.

[0119] To enhance the nonlinear expressiveness of the graph neural network model, nonlinear activation processing is performed on the intermediate node vectors. The ReLU activation function is used here. The ReLU activation function is a commonly used nonlinear activation function that converts negative values ​​in the intermediate node vectors to 0 while leaving positive values ​​unchanged. This nonlinear activation process introduces nonlinear factors, enabling the graph neural network model to learn more complex feature relationships, improving the model's expressiveness and performance. After processing with the ReLU activation function, an activated intermediate node vector is generated.

[0120] Step S1425: performing layer normalization processing on the activated intermediate node vector to generate an aggregated node vector containing local structural information.

[0121] Finally, the activated intermediate node vectors are subjected to layer normalization. Layer normalization is a technique used to normalize the output of layers in a neural network. It can make the feature vectors of each node have a similar distribution, which helps to improve the training stability and convergence speed of the model. Through layer normalization, the activated intermediate node vectors are normalized so that the values ​​of their various dimensions are comparable. After layer normalization, an aggregated node vector containing local structural information is finally generated. This aggregated node vector combines the characteristics of the production node itself and the characteristics of its neighboring nodes in the neighborhood, and has undergone nonlinear activation and layer normalization processing, which can effectively reflect the local structural information of the production node in the graph.

[0122] Step S143: Perform global correlation analysis on the aggregated node vector through the attention mechanism layer of the graph neural network model to generate an attention node vector containing global dependencies. The global correlation analysis includes calculating the attention weights between any two nodes and performing feature fusion based on the weights.

[0123] The attention mechanism layer of the graph neural network model performs global correlation analysis on aggregated node vectors. This layer captures the global dependencies between nodes in the graph, enabling better analysis of anomalies in the production-related knowledge graph. This global correlation analysis involves calculating the attention weights between any two nodes and then performing feature fusion based on these weights.

[0124] Calculating the attention weight between any two nodes is the process of determining the degree of connection between them. The attention weight reflects the degree of attention one node pays to another; a larger weight indicates a stronger connection between the two nodes. The attention mechanism layer calculates the attention weight between any two aggregated node vectors. This calculation is based on the feature information of the aggregated node vectors, and different node vectors generate different attention weights.

[0125] Weighted feature fusion combines node features according to their attention weights. For each aggregated node vector, the aggregated node vectors of other nodes are multiplied by the corresponding attention weights. These weighted vectors are then summed to produce a new vector. This new vector integrates the feature information of all nodes and is adjusted based on the attention weights to better reflect the global dependencies between nodes. This approach ultimately generates an attention node vector that incorporates global dependencies. This attention node vector comprehensively reflects the global relationships between nodes in the production-related knowledge graph.

[0126] Step S144: Perform abnormality probability calculation processing on the attention node vector through the classification layer of the graph neural network model to generate an abnormality detection result including detection confidence of different abnormality types. The abnormality probability calculation processing includes inputting the attention node vector into a fully connected network and outputting the probability distribution through a softmax function.

[0127] The classification layer of the graph neural network model calculates anomaly probabilities on the attention node vectors to generate anomaly detection results with different confidence levels for different anomaly types. The classification layer is the last layer of the graph neural network model. Its primary function is to determine the presence and type of anomalies in the production-related knowledge graph based on the input attention node vectors.

[0128] The anomaly probability calculation process involves two steps: inputting the attention node vector into a fully connected network and outputting a probability distribution through a softmax function. A fully connected network is a common neural network structure that takes the attention node vector as input and processes it through a series of linear transformations and nonlinear activation functions to produce an intermediate output vector. This intermediate output vector contains preliminary information about different anomaly types.

[0129] The intermediate output vector is then fed into the softmax function. The softmax function is a commonly used probability conversion function that converts each element of the intermediate output vector into a probability value between 0 and 1, with the sum of all probability values ​​being 1. These probability values ​​represent the detection confidence levels for different anomaly types, each representing the likelihood of the corresponding anomaly type appearing in the production-related knowledge graph. This approach ultimately generates anomaly detection results that include the detection confidence levels for different anomaly types.

[0130] Step S150: determining the type of abnormal event existing in the diborane production process and the distribution characteristic information of the abnormal event in the production-related knowledge graph according to the abnormality detection result.

[0131] After obtaining the anomaly detection results of the production-related knowledge graph, it is necessary to determine the types of abnormal events existing in the diborane production process and the distribution characteristic information of abnormal events in the production-related knowledge graph based on the anomaly detection results, which will help to further understand the scope and extent of the impact of abnormal situations on the production process.

[0132] Step S151: parsing the detection confidence in the anomaly detection result, and extracting the anomaly type whose detection confidence exceeds a preset threshold as the target anomaly type.

[0133] First, analyze the detection confidence in the anomaly detection results. Detection confidence indicates the likelihood of different anomaly types occurring in the production-related knowledge graph. To determine the actual anomaly type, a preset threshold is required. The preset threshold is a pre-set probability value. When the detection confidence of a particular anomaly type exceeds the preset threshold, it is considered to have actually occurred. By traversing all detection confidences in the anomaly detection results, the anomaly types with detection confidence exceeding the preset threshold are extracted and designated as target anomaly types. Target anomaly types are the actual anomalies that occurred during the diborane production process, and subsequent analysis and processing will focus on these anomaly types.

[0134] Step S152: Locate the production node or production edge corresponding to the target abnormality type, extract the timestamp information of the production node or production edge, and determine the start time point and end time point of the abnormal event.

[0135] After determining the target anomaly type, we need to locate the production nodes or production edges corresponding to these anomaly types. Each production node and production edge in the production association knowledge graph is associated with specific production equipment or production parameters. Abnormal events often manifest themselves at certain production nodes or production edges. By analyzing the association between the target anomaly type and the production nodes and production edges, we can locate the corresponding production nodes or production edges.

[0136] Next, extract the timestamp information for these production nodes or edges. Timestamp information records the time each data point was collected. By analyzing timestamp information, we can determine the start and end time points of the abnormal event. The start time point is when the abnormal event began, and the end time point is when the abnormal event ended. By determining these two time points, we can understand the duration of the abnormal event.

[0137] Step S153: performing a change trend analysis on the historical feature vectors of the production nodes or production edges to determine the diffusion direction of abnormal events in the production-related knowledge graph. The change trend analysis includes calculating the difference value of the feature vector in the time dimension.

[0138] To determine the direction of abnormal events' spread within the production-related knowledge graph, we need to analyze the changing trends of the historical feature vectors of production nodes or production edges. Historical feature vectors record the characteristic information of production nodes or production edges at different points in time. By analyzing the changing trends of these feature vectors, we can understand the propagation path of abnormal events.

[0139] Trend analysis involves calculating the difference in feature vectors over time. For each production node or edge's historical feature vectors, the difference between feature vectors at adjacent time points is calculated. This difference reflects the degree of temporal variation in the feature vector. By analyzing the magnitude and direction of these differences, the direction of the abnormal event's spread can be determined. If the difference in the feature vector of a production node or edge increases significantly after a certain point in time, this indicates that the abnormal event may have spread from that node or edge to other nodes or edges. By performing trend analysis on the historical feature vectors of all relevant production nodes and edges, the direction of the abnormal event's spread within the production-related knowledge graph can be determined.

[0140] Step S154: constructing an abnormal event propagation model based on the starting time point, the ending time point and the diffusion direction, wherein the abnormal event propagation model includes a propagation rate parameter in the time dimension and a propagation range parameter in the space dimension.

[0141] After determining the start and end time points and diffusion direction of an abnormal event, we construct an abnormal event propagation model based on this information. This model describes the propagation patterns of abnormal events within the production-related knowledge graph and includes parameters for propagation rate in the time dimension and propagation range in the spatial dimension.

[0142] The propagation rate parameter in the time dimension reflects the speed at which an abnormal event propagates over time. This parameter is calculated by calculating the duration of the abnormal event from its start to end time points and the distance it propagates during this time. The propagation rate parameter can help predict the spread of abnormal events in the future.

[0143] The spatial propagation range parameter reflects the spatial propagation range of an abnormal event within the production-related knowledge graph. This parameter can be determined by analyzing the set of production nodes and production edges involved in the direction of the abnormal event's spread. This parameter helps understand the extent of an abnormal event's impact on the production process.

[0144] By combining information about the start and end time points and the diffusion direction, we construct an abnormal event propagation model that includes parameters for the propagation rate in the time dimension and the propagation range in the spatial dimension. This abnormal event propagation model can more comprehensively describe the propagation patterns of abnormal events in the production-related knowledge graph.

[0145] Step S155: Generate distribution feature information of the abnormal event in the production-related knowledge graph through the abnormal event propagation model, where the distribution feature information includes a node set and an edge set affected by the abnormality and a corresponding time coverage interval.

[0146] The abnormal event propagation model is used to generate the distribution characteristics of abnormal events in the production-related knowledge graph. The distribution characteristics are a detailed description of the impact range and time of the abnormal event in the production-related knowledge graph, which includes the set of nodes and edges affected by the abnormality and the corresponding time coverage interval.

[0147] The set of nodes affected by the anomaly is the set of production nodes affected during the propagation of the anomaly event. These affected production nodes can be determined using the spatial dimension propagation range parameter of the anomaly propagation model. The set of edges affected by the anomaly is the set of production edges connecting the affected production nodes. These edges can also be determined using the anomaly propagation model.

[0148] The corresponding time coverage interval refers to the time range in which the abnormal event affects each affected production node and production edge. The time coverage interval of each affected production node and production edge can be determined through the time dimension propagation rate parameter and the information of the start time point and the end time point of the abnormal event propagation model.

[0149] By comprehensively considering the propagation patterns of abnormal events and the structure of the production-related knowledge graph, the abnormal event propagation model generates distribution feature information for the set of nodes and edges affected by the abnormality, as well as the corresponding time coverage interval. This distribution feature information clearly demonstrates the distribution of abnormal events in the production-related knowledge graph.

[0150] Step S160: generating a production warning instruction including an event location identifier based on the abnormal event type and the distribution characteristic information, and sending the production warning instruction to the diborane production control terminal to trigger an abnormal response operation.

[0151] After determining the abnormal event type and its distribution characteristics within the production-related knowledge graph, a production warning instruction containing an event location identifier is generated based on this information. This instruction is then sent to the diborane production control terminal to trigger an abnormal response. The production warning instruction notifies the production control terminal of the abnormal event and directs it to take appropriate measures. The event location identifier accurately indicates the location of the abnormal event, facilitating its processing.

[0152] Step S161: querying a preset exception type response rule library to extract a response priority identifier and an emergency handling strategy code associated with the target exception type.

[0153] In order to formulate reasonable production early warning instructions, it is necessary to query the preset exception type response rule base. The exception type response rule base is a pre-established database that stores response rules corresponding to different exception types. Each exception type is associated with a response priority identifier and an emergency handling strategy code. The response priority identifier indicates the processing priority of the exception type. Different exception types may have different priorities, and exceptions with high priorities need to be handled first. The emergency handling strategy code is a code used to identify the emergency handling strategy for the exception type. Each emergency handling strategy code corresponds to a processing method. By querying the exception type response rule base, the response priority identifier and emergency handling strategy code associated with the target exception type are extracted. The above information will serve as an important part of the production early warning instruction to guide the production control terminal to take corresponding measures.

[0154] Step S162: extracting the node set and edge set affected by the abnormality from the distribution feature information, and obtaining the device identifier or parameter identifier corresponding to the node set, and the connection identifier or association identifier corresponding to the edge set.

[0155] The node set and edge set affected by the anomaly are extracted from the distribution feature information. The node set and edge set affected by the anomaly record the scope of the impact of the abnormal event in the production association knowledge graph. For the node set affected by the anomaly, each production node is associated with a specific production device or production parameter. The corresponding device identifier or parameter identifier can be obtained through the node set. The device identifier is information used to uniquely identify the production device, and the parameter identifier is information used to uniquely identify the production parameter. For the edge set affected by the anomaly, each production edge is associated with a specific device connection or parameter association. The corresponding connection identifier or association identifier can be obtained through the edge set. The connection identifier is information used to uniquely identify the connection relationship between devices, and the association identifier is information used to uniquely identify the association relationship between parameters. By extracting this identification information, the location where the abnormal event occurred can be accurately located.

[0156] Step S163: converting the equipment identifier, parameter identifier, connection identifier and association identifier into positioning coordinate information recognizable by the production control system, wherein the positioning coordinate information includes the physical position coordinates of the equipment and the logical position coordinates of the parameters.

[0157] In order for the production control system to accurately identify the location where an abnormal event occurred, it is necessary to convert the device identifier, parameter identifier, connection identifier, and association identifier into positioning coordinate information that can be recognized by the production control system. The positioning coordinate information includes the device physical location coordinates and the parameter logical location coordinates. The device physical location coordinates are the physical location information of the production equipment in the actual production environment. The corresponding device physical location coordinates can be queried through the device identifier. The parameter logical location coordinates are the logical location information of the production parameters in the production control system. The corresponding parameter logical location coordinates can be queried through the parameter identifier. Similarly, the connection identifier and association identifier can also be converted into corresponding location coordinate information. By converting these identification information into positioning coordinate information, the production control system can accurately locate the location where the abnormal event occurred.

[0158] Step S164: Based on the propagation rate parameter and propagation range parameter of the abnormal event propagation model, predict the extended node set and extended edge set of the abnormal event in the future time period, and generate extended positioning coordinate information including the predicted impact range.

[0159] The propagation rate and range parameters of the abnormal event propagation model are used to predict the set of expanded nodes and edges that will be affected by the abnormal event in a future time period. The propagation rate parameter reflects the speed of the abnormal event's propagation over time, while the propagation range parameter reflects the spatial extent of the abnormal event's propagation. By combining these two parameters, we can predict the set of production nodes and production edges that the abnormal event may affect within a certain future time period. These sets of nodes and edges are then used as the set of expanded nodes and edges, respectively.

[0160] Then, for the expanded node set and the expanded edge set, extended location coordinate information containing the predicted impact range is generated. This extended location coordinate information, similar to the previous location coordinate information, includes the physical location coordinates of the device and the logical location coordinates of the parameters, except that this coordinate information is generated based on the predicted impact range of the anomaly. This extended location coordinate information helps production control terminals understand the potential impact range of an anomaly in advance, allowing them to take appropriate preventive measures.

[0161] Step S165: Aggregate the response priority identifier, emergency handling strategy code, positioning coordinate information and extended positioning coordinate information to generate a production warning instruction containing an event positioning identifier chain with timestamp alignment, where each identifier node in the event positioning identifier chain corresponds one-to-one to a node or edge in the production-related knowledge graph.

[0162] The response priority identifier, emergency response strategy code, location coordinate information, and extended location coordinate information are aggregated to generate a production warning instruction. To ensure that the production warning instruction accurately indicates the location and handling method of the abnormal event, it is necessary to generate an event location identifier chain with timestamp alignment. The event location identifier chain consists of multiple identifier nodes, each of which corresponds one-to-one with a node or edge in the production-related knowledge graph. Timestamp alignment ensures that each identifier node in the event location identifier chain is temporally consistent with the corresponding production node or edge. The response priority identifier, emergency response strategy code, location coordinate information, and extended location coordinate information are integrated into the event location identifier chain to generate a production warning instruction containing this information. The production warning instruction includes information such as the abnormal event handling priority, emergency response strategy, location of the abnormal event, and predicted expansion range.

[0163] Step S166: parsing the event location identifier chain in the production warning instruction, and extracting the device identifier, parameter identifier, connection identifier or association identifier corresponding to each identifier node.

[0164] After a production warning instruction is generated, it needs to be parsed so that the production control terminal can understand and execute it. The event location identification chain in the production warning instruction is parsed to extract the device identification, parameter identification, connection identification, or association identification corresponding to each identification node. Each identification node in the event location identification chain contains information about a specific production node or production edge. By parsing these identification nodes, the corresponding device identification, parameter identification, connection identification, or association identification can be obtained. This identification information is key to the processing performed by the production control terminal. With this identification information, the production control terminal can accurately locate the location of the abnormal event and take appropriate measures.

[0165] Step S167: querying the device control table according to the device identification to obtain a corresponding device adjustment instruction template, wherein the device adjustment instruction template includes device start / stop control parameters and operating parameter adjustment ranges.

[0166] The device control table is queried based on the extracted device ID. The device control table is a pre-established database that stores the device adjustment instruction templates corresponding to each production device. These templates contain the device's start / stop control parameters and operating parameter adjustment ranges. The device start / stop control parameters control the device's start and stop, while the operating parameter adjustment ranges define the adjustable ranges of the device's operating parameters. By querying the device control table, the device adjustment instruction template corresponding to the device ID is retrieved. These templates serve as the basis for the production control terminal to adjust the device.

[0167] Step S168: querying a parameter correction table according to the parameter identifier to obtain a corresponding parameter correction instruction template, wherein the parameter correction instruction template includes a parameter target value and a correction rate limit.

[0168] The parameter correction table is queried based on the extracted parameter identifier. The parameter correction table is a pre-established database that stores parameter correction instruction templates corresponding to each production parameter. A parameter correction instruction template contains a target parameter value and a correction rate limit. The target parameter value is the value the parameter should reach under normal circumstances, while the correction rate limit specifies the maximum speed at which the parameter can be corrected. By querying the parameter correction table, the parameter correction instruction template corresponding to the parameter identifier is obtained. These parameter correction instruction templates serve as the basis for the production control terminal to correct the parameter.

[0169] Step S169: querying the association adjustment table according to the connection identifier or the association identifier to obtain a corresponding association adjustment instruction template, wherein the association adjustment instruction template includes a collaborative adjustment strategy of the associated devices and a synchronous correction rule of the associated parameters.

[0170] The association adjustment table is queried based on the extracted connection identifier or association identifier. The association adjustment table is a pre-established database that stores the association adjustment instruction template corresponding to each connection identifier or association identifier. The association adjustment instruction template contains the collaborative adjustment strategy of the associated equipment and the synchronous correction rule of the associated parameters. The collaborative adjustment strategy of the associated equipment specifies how to coordinate and adjust multiple associated equipment to ensure the stability of the production process. The synchronous correction rule of the associated parameters specifies how to synchronously correct multiple associated parameters to ensure the correlation between the parameters. By querying the association adjustment table, the association adjustment instruction template corresponding to the connection identifier or association identifier is obtained. These association adjustment instruction templates will serve as the basis for the production control terminal to adjust the associated equipment and associated parameters.

[0171] Step S1610: combining the device adjustment instruction template, the parameter modification instruction template, and the associated adjustment instruction template according to the time sequence of the event location identifier chain to generate a comprehensive adjustment instruction including multi-stage operations.

[0172] Equipment adjustment instruction templates, parameter correction instruction templates, and associated adjustment instruction templates are combined in the chronological order of the event location identifier chain. The event location identifier chain records the temporal development of abnormal events. By combining these instruction templates in this chronological order, a comprehensive adjustment instruction containing multiple stages of operations can be generated. The comprehensive adjustment instruction guides the production control terminal to adjust equipment, parameters, and associated relationships in stages based on the development of the abnormal event, ensuring that the abnormal event is handled promptly and effectively.

[0173] Step S1611: Send the comprehensive adjustment instruction to the production control terminal so that the production control terminal executes equipment adjustment operations, parameter correction operations and associated adjustment operations in sequence according to the time sequence of the comprehensive adjustment instruction until the detection confidence of the abnormal event is lower than the preset threshold.

[0174] Finally, the comprehensive adjustment instruction is sent to the diborane production control terminal. After receiving the comprehensive adjustment instruction, the production control terminal executes the equipment adjustment operation, parameter correction operation, and associated adjustment operation in the chronological order of the comprehensive adjustment instruction. The equipment adjustment operation controls the start and stop of the production equipment and adjusts the operating parameters according to the equipment adjustment instruction template. The parameter correction operation corrects the production parameters according to the parameter correction instruction template. The associated adjustment operation coordinates the adjustment of associated equipment and associated parameters according to the associated adjustment instruction template. Through these operations, the impact of abnormal events is gradually eliminated. During the execution of the operation, the detection confidence of the abnormal event is continuously monitored. When the detection confidence of the abnormal event falls below the preset threshold, it indicates that the abnormal event has been effectively controlled, and the execution of the comprehensive adjustment instruction can be stopped. Through the above process, the effective detection and handling of abnormal events in the diborane production process is achieved, ensuring the stability and safety of the production process.

[0175] Figure 2 FIG. 1 shows the hardware structure of an abnormality analysis system 100 for a diborane production control system for implementing the abnormality analysis method for a diborane production control system provided in an embodiment of the present application. Figure 2 As shown, the abnormality analysis system 100 applied to a diborane production control system may include a processor 110 , a machine-readable storage medium 120 , a bus 130 , and a communication unit 140 .

[0176] In one possible design, the anomaly analysis system 100 applied to the diborane production control system can be a single server or a server group. The server group can be centralized or distributed (for example, the anomaly analysis system 100 applied to the diborane production control system can be a distributed system). In some embodiments, the anomaly analysis system 100 applied to the diborane production control system can be local or remote. For example, the anomaly analysis system 100 applied to the diborane production control system can access information and / or data stored in the machine-readable storage medium 120 via a network. For another example, the anomaly analysis system 100 applied to the diborane production control system can be directly connected to the machine-readable storage medium 120 to access the stored information and / or data. In some embodiments, the anomaly analysis system 100 applied to the diborane production control system can be implemented on the anomaly analysis system applied to the diborane production control system. By way of example only, the anomaly analysis system applied to the diborane production control system can include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an internal cloud, a multi-layer cloud, or any combination thereof.

[0177] Machine-readable storage medium 120 can store data and / or instructions. In some embodiments, machine-readable storage medium 120 can store data obtained from an external terminal. In some embodiments, machine-readable storage medium 120 can store data and / or instructions used by abnormality analysis system 100 for diborane production control system to execute or perform the exemplary methods described herein.

[0178] During the specific implementation process, one or more processors 110 execute computer executable instructions stored in the machine-readable storage medium 120, so that the processor 110 can execute the abnormality analysis method applied to the diborane production control system in the above method embodiment. The processor 110, the machine-readable storage medium 120 and the communication unit 140 are connected through the bus 130, and the processor 110 can be used to control the sending and receiving actions of the communication unit 140.

[0179] The specific implementation process of the processor 110 can refer to the various method embodiments executed by the above-mentioned abnormality analysis system 100 applied to the diborane production control system. The implementation principles and technical effects are similar and will not be repeated here in this embodiment.

[0180] In addition, an embodiment of the present application further provides a readable storage medium having computer executable instructions set therein. When a processor runs the computer executable instructions, the above-mentioned abnormality analysis method applied to the diborane production control system is implemented.

[0181] It should be noted that, in order to simplify the description of the present disclosure and thus facilitate understanding of one or more embodiments of the invention, in the foregoing description of the embodiments of the present disclosure, multiple features may sometimes be combined into one embodiment, figure, or description thereof. Similarly, it should be noted that, in order to simplify the description of the present disclosure and thus facilitate understanding of one or more embodiments of the invention, in the foregoing description of the embodiments of the present disclosure, multiple features may sometimes be combined into one embodiment, figure, or description thereof.

Claims

1. An abnormality analysis method applied to a diborane production control system, characterized in that: The method comprises: Acquire a multi-source production data set generated during the operation of a diborane production control system, wherein the multi-source production data set includes equipment status data, raw material delivery data, and reaction parameter data with time stamps; Constructing a production-related knowledge graph based on the multi-source production data set, wherein the production-related knowledge graph includes a plurality of production nodes and production edges connecting the production nodes, wherein the production nodes correspond to diborane production equipment or production parameters, and the production edges correspond to physical connection relationships between production equipment or logical association relationships between production parameters; Performing graph feature extraction processing on the production-related knowledge graph to obtain a node feature set of the production node and an edge feature set of the production edge; Calling a pre-trained graph neural network model to perform graph structure analysis on the node feature set and the edge feature set to generate an anomaly detection result of the production-related knowledge graph, wherein the anomaly detection result includes detection confidence levels corresponding to different anomaly types; Determine the type of abnormal event present in the diborane production process and distribution characteristic information of the abnormal event in the production-related knowledge graph according to the abnormality detection result; A production warning instruction including an event location identifier is generated based on the abnormal event type and the distribution characteristic information, and the production warning instruction is sent to a diborane production control terminal to trigger an abnormal response operation.

2. The abnormality analysis method applied to the diborane production control system according to claim 1, characterized in that: The constructing of a production-related knowledge graph based on the multi-source production data set includes: Performing data alignment on the multi-source production data set to obtain an aligned production data sequence; Extracting diborane production equipment identification information from the aligned production data sequence, and mapping the data set corresponding to each diborane production equipment identification information into a production equipment node, wherein the production equipment node includes an equipment type attribute and a real-time operation status attribute; Extracting diborane production parameter identification information from the aligned production data sequence, mapping the data set corresponding to each diborane production parameter identification information into a production parameter node, wherein the production parameter node includes a parameter type attribute and a real-time measurement value attribute; Based on the diborane production process flow chart, the physical connection relationship between diborane production equipment is extracted. Production edges are established between equipment nodes that have direct material transmission or energy exchange. The production edges contain connection type attributes and transmission rate attributes. Based on the diborane production reaction kinetics model, the logical correlation between diborane production parameters is extracted, and production edges are established between parameter nodes with causal effects or synergistic changes. The production edges contain influence direction attributes and association strength attributes. An initial production-related knowledge graph is constructed based on the production equipment nodes, production parameter nodes and corresponding production edges. The connectivity of the initial production-related knowledge graph is verified, and isolated nodes and corresponding edges that are not connected to the main production process are deleted to generate a final production-related knowledge graph.

3. The abnormality analysis method applied to the diborane production control system according to claim 2, characterized in that: The connectivity check process is performed on the initial production-related knowledge graph, and isolated nodes and corresponding edges that are not connected to the main production process are deleted to generate the final production-related knowledge graph, including: Extracting a core node set corresponding to the main production process from the initial production-related knowledge graph, wherein the core node set includes a raw material input node, a reactor node, and a product output node; Using a breadth-first search algorithm to traverse the initial production-related knowledge graph starting from the core node set, marking all reachable nodes and corresponding edges that have path connections with the core nodes; Identifying unlabeled nodes and corresponding edges in the initial production-related knowledge graph, wherein the unlabeled nodes and corresponding edges are isolated nodes and edges that are not connected to the main production process; Deleting the isolated nodes and corresponding edges from the initial production-related knowledge graph to generate a production-related knowledge graph after connectivity verification; The number of nodes and the number of edges of the production-related knowledge graph after the connectivity check are counted to ensure that the number of remaining nodes and the number of edges meet the basic structural requirements of the diborane production process, and generate a final production-related knowledge graph.

4. The abnormality analysis method applied to a diborane production control system according to claim 2, characterized in that: The performing graph feature extraction processing on the production-related knowledge graph to obtain a node feature set of the production node and an edge feature set of the production edge includes: Performing feature coding processing on the real-time operating status attributes of the production equipment node to generate a device node feature vector, wherein the feature coding processing includes state category hot coding and state change rate normalization; Performing feature coding processing on the real-time measurement value attributes of the production parameter node to generate a parameter node feature vector, wherein the feature coding processing includes parameter value range binning and parameter change trend symbolization; Merging the device node feature vector and the parameter node feature vector to obtain a node feature set of the production node; Performing feature extraction processing on the transmission rate attribute of the production edge to generate a transmission rate feature, wherein the feature extraction processing includes calculating a historical mean value of the rate and statistics on a rate fluctuation range; Performing feature extraction processing on the correlation strength attribute of the production edge to generate correlation strength features, wherein the feature extraction processing includes calculating the correlation coefficient of historical data and statistics of the impact lag time; The transmission rate feature and the association strength feature are combined to obtain an edge feature set of the production edge.

5. The abnormality analysis method applied to the diborane production control system according to claim 4, characterized in that: The merging of the device node feature vector and the parameter node feature vector to obtain the node feature set of the production node includes: Performing dimension expansion processing on the device node feature vector to generate an expanded device node feature vector, wherein the dimension expansion processing includes adding a device type code dimension; Performing dimension expansion processing on the parameter node feature vector to generate an extended parameter node feature vector, wherein the dimension expansion processing includes adding a parameter type encoding dimension; Concatenate the extended device node feature vector and the extended parameter node feature vector to generate a merged node feature vector; The merged node feature vectors are arranged in the order of the timestamps of the production nodes to generate a node feature set of the production nodes.

6. The abnormality analysis method applied to a diborane production control system according to claim 1, characterized in that: The calling of the pre-trained graph neural network model to perform graph structure analysis on the node feature set and the edge feature set to generate an anomaly detection result of the production-related knowledge graph includes: Inputting the node feature set and the edge feature set into the input layer of the graph neural network model, wherein the input layer maps the node feature vector and the edge feature vector into a node embedding vector and an edge embedding vector, respectively; Performing neighborhood information aggregation processing on the node embedding vector and the edge embedding vector through the graph convolution layer of the graph neural network model to generate an aggregated node vector containing local structural information, wherein the neighborhood information aggregation processing includes weighted summation of the node's own features and the features of adjacent nodes; Performing a global correlation analysis on the aggregated node vectors through the attention mechanism layer of the graph neural network model to generate an attention node vector containing a global dependency relationship, wherein the global correlation analysis includes calculating the attention weights between any two nodes and performing feature fusion based on the weights; The attention node vector is subjected to an abnormality probability calculation process through the classification layer of the graph neural network model to generate an abnormality detection result including detection confidence of different abnormality types. The abnormality probability calculation process includes inputting the attention node vector into a fully connected network and outputting a probability distribution through a softmax function.

7. The abnormality analysis method applied to a diborane production control system according to claim 6, characterized in that: The performing neighborhood information aggregation processing on the node embedding vector and the edge embedding vector by the graph convolution layer of the graph neural network model to generate an aggregated node vector containing local structural information includes: A neighborhood range is defined for each production node, where the neighborhood range includes first-order neighbor nodes and corresponding edges directly connected to the production node; Perform a weighted summation on the node embedding vector of each production node and the node embedding vectors of all neighboring nodes in its neighborhood to obtain a weighted summation result, where the weight of the weighted summation is calculated based on the edge embedding vector of the corresponding edge; Concatenating the weighted summation result with the original node embedding vector of the production node to generate an intermediate node vector containing the production node's own features and neighborhood features; Performing nonlinear activation processing on the intermediate node vector to generate an activated intermediate node vector, wherein the nonlinear activation processing uses a ReLU activation function; The activated intermediate node vectors are subjected to layer normalization processing to generate aggregated node vectors containing local structural information.

8. The abnormality analysis method applied to a diborane production control system according to claim 1, characterized in that: The determining, based on the abnormality detection result, the type of abnormal event present in the diborane production process and the distribution characteristic information of the abnormal event in the production-related knowledge graph includes: Analyze the detection confidence in the anomaly detection result, and extract the anomaly type whose detection confidence exceeds a preset threshold as the target anomaly type; Locate the production node or production edge corresponding to the target anomaly type, extract the timestamp information of the production node or production edge, and determine the start and end time points of the anomaly event; Performing a change trend analysis on the historical feature vectors of the production nodes or production edges to determine the diffusion direction of the abnormal event in the production-related knowledge graph, wherein the change trend analysis includes calculating the difference value of the feature vector in the time dimension; Constructing an abnormal event propagation model based on the starting time point, the ending time point, and the diffusion direction, wherein the abnormal event propagation model includes a propagation rate parameter in the time dimension and a propagation range parameter in the space dimension; The abnormal event propagation model is used to generate distribution feature information of the abnormal event in the production-related knowledge graph, and the distribution feature information includes a node set and an edge set affected by the abnormality and a corresponding time coverage interval.

9. The abnormality analysis method applied to a diborane production control system according to claim 8, characterized in that: The generating of a production warning instruction including an event location identifier based on the abnormal event type and the distribution characteristic information, and sending the production warning instruction to a diborane production control terminal to trigger an abnormal response operation, includes: Querying a preset exception type response rule library to extract a response priority identifier and an emergency handling strategy code associated with the target exception type; Extracting the node set and edge set affected by the abnormality from the distribution feature information, obtaining the device identifier or parameter identifier corresponding to the node set, and the connection identifier or association identifier corresponding to the edge set; Converting the device identifier, parameter identifier, connection identifier, and association identifier into positioning coordinate information recognizable by the production control system, wherein the positioning coordinate information includes the physical position coordinates of the device and the logical position coordinates of the parameters; Based on the propagation rate parameter and propagation range parameter of the abnormal event propagation model, predict the extended node set and extended edge set of the abnormal event in the future time period, and generate extended positioning coordinate information including the predicted impact range; Aggregate the response priority identifier, emergency handling strategy code, location coordinate information, and extended location coordinate information to generate a production warning instruction containing a timestamp-aligned event location identifier chain, where each identifier node in the event location identifier chain corresponds one-to-one to a node or edge in the production-related knowledge graph; Parsing the event location identification chain in the production early warning instruction, and extracting the device identification, parameter identification, connection identification or association identification corresponding to each identification node; Querying the device control table according to the device identification to obtain a corresponding device adjustment instruction template, wherein the device adjustment instruction template includes the device start and stop control parameters and the operating parameter adjustment range; Querying a parameter correction table according to the parameter identifier to obtain a corresponding parameter correction instruction template, wherein the parameter correction instruction template includes a parameter target value and a correction rate limit; Querying an association adjustment table according to the connection identifier or the association identifier to obtain a corresponding association adjustment instruction template, wherein the association adjustment instruction template includes a collaborative adjustment strategy of associated devices and a synchronization correction rule of associated parameters; Combining the device adjustment instruction template, the parameter correction instruction template, and the associated adjustment instruction template in the time sequence of the event location identifier chain to generate a comprehensive adjustment instruction including multi-stage operations; The comprehensive adjustment instruction is sent to the production control terminal so that the production control terminal performs equipment adjustment operations, parameter correction operations and associated adjustment operations in sequence according to the time sequence of the comprehensive adjustment instruction until the detection confidence of the abnormal event is lower than the preset threshold.

10. An abnormality analysis system applied to a diborane production control system, characterized in that: The abnormality analysis system applied to the diborane production control system includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to run the programs, instructions or codes in the memory to implement the abnormality analysis method applied to the diborane production control system according to any one of claims 1 to 9 above.

Images