Standard board production online intelligent detection and analysis method and system based on industrial Internet of Things

Abstract

The embodiment of the invention provides an online intelligent detection and analysis method and system for standard board production based on industrial Internet of Things, and the method comprises the steps: firstly obtaining first standard board production monitoring data, and determining a plurality of node diagram construction knowledge modes to construct a state node diagram; then, injecting identification attribute information into the first standard board production monitoring data to generate a plurality of second standard board production monitoring data corresponding to different knowledge modes, and determining a corresponding feature extraction path according to the injected mode identification information, so as to perform node graph construction after performing production parameter trajectory vector extraction; and on the basis of a state node diagram construction result, an abnormal state link in the production monitoring process of the target standard plate is estimated, so that the online intelligent detection and analysis level of standard plate production is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of home industrial Internet of Things, and in particular to an online intelligent detection and analysis method and system for standard board production based on industrial Internet of Things. Background Art

[0002] In modern industrial production, especially in the field of home furnishing industry manufacturing, the production process of standard boards involves the interaction of multiple complex processes and equipment. How to efficiently and accurately monitor this process and promptly detect and handle abnormal conditions in production is crucial to improving production efficiency and product quality. Traditionally, standard board production monitoring mainly relies on manual inspections and offline data analysis, which is not only inefficient, but also difficult to fully and real-time reflect the production status.

[0003] With the rapid development of industrial Internet of Things technology, more and more production equipment is connected to the Internet of Things platform, realizing the real-time collection and transmission of production data. This makes it possible to conduct online intelligent detection and analysis. However, the production monitoring data recorded in the industrial Internet of Things platform is often massive, heterogeneous, and high-dimensional. How to extract valuable information from these complex data is still a challenge. For example, related technologies rely too much on preset rules and models, making it difficult to adapt to dynamic changes in the production process and to fully tap the potential value of data. Summary of the invention

[0004] In order to at least overcome the above-mentioned deficiencies in the prior art, an object of the embodiments of the present application is to provide an online intelligent detection and analysis method and system for standard board production based on the industrial Internet of Things.

[0005] According to one aspect of the present application, there is provided an online intelligent detection and analysis method for standard board production based on the industrial Internet of Things, the method comprising: obtaining first standard board production monitoring data recording the target standard board production monitoring process in the industrial Internet of Things platform, and determining multiple node graph construction knowledge patterns for constructing a state node graph for the first standard board production monitoring data; for the multiple node graph construction knowledge patterns, injecting identification attribute information into the first standard board production monitoring data respectively to generate multiple second standard board production monitoring data; the second standard board production monitoring data in the multiple second standard board production monitoring data respectively corresponds to the node graph construction knowledge patterns in the multiple node graph construction knowledge patterns, and the identification attribute information in each of the second standard board production monitoring data includes the node graph construction knowledge pattern reflecting the corresponding pattern identification information; according to the pattern identification information included in the plurality of second standard plate production monitoring data, determine a plurality of feature extraction paths associated with the plurality of second standard plate production monitoring data; for each of the plurality of second standard plate production monitoring data, extract a production parameter trajectory vector according to the associated feature extraction path, and generate a set of production parameter trajectory vectors corresponding to the plurality of node graph construction knowledge patterns; perform node graph construction on the sets of production parameter trajectory vectors corresponding to the plurality of node graph construction knowledge patterns, and generate state node graph construction results corresponding to the plurality of node graph construction knowledge patterns, and estimate the abnormal state link of the target standard plate production monitoring process based on the state node graph construction results corresponding to the plurality of node graph construction knowledge patterns.

[0006] In the technical solutions provided in some embodiments of the present application, the embodiments of the present application first determine multiple node graphs to construct knowledge patterns, and generate multiple second standard board production monitoring data corresponding to different knowledge patterns by injecting identification attribute information into the first standard board production monitoring data, thereby realizing refined classification and processing of data. According to the pattern identification information in the second standard board production monitoring data, the feature extraction path associated with the data can be accurately matched. This accurate matching ensures the pertinence and effectiveness of feature extraction, and provides a high-quality data foundation for the subsequent state node graph construction. For each second standard board production monitoring data, the present invention efficiently extracts production parameter trajectory vectors according to the associated feature extraction path, generates a set of production parameter trajectory vectors corresponding to the knowledge pattern constructed by multiple node graphs, and can fully reflect the parameter changes and state transitions in the production process. By constructing a node graph for the production parameter trajectory vector set, multiple state node graph construction results are generated, which can intuitively display the state transition and logical relationship in the production process. Based on the state node graph, the abnormal state link in the target standard board production monitoring process can be accurately estimated, and potential problems in production can be discovered in time. Therefore, the level of online intelligent detection and analysis of standard board production has been significantly improved through the construction of multi-mode state node graphs, precise matching of feature extraction paths, efficient extraction of production parameter trajectory vectors, intelligent construction of state node graphs and accurate estimation of abnormal state links. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required to be used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be extracted in combination with these drawings without creative work.

[0008] Figure 1 A schematic diagram of a process for an online intelligent detection and analysis method for standard board production based on the industrial Internet of Things provided in an embodiment of the present application; Figure 2 This is a schematic block diagram of the architecture of an online intelligent detection and analysis system for standard board production based on the Industrial Internet of Things provided in an embodiment of the present application. DETAILED DESCRIPTION

[0009] The following description is intended to enable one of ordinary skill in the art to implement and combine the present application, and the description is provided in the context of a specific application scenario and its requirements. It will be apparent to one of ordinary skill in the art that various changes may be made to the disclosed embodiments, and that the general principles defined in the present application may be applied to other embodiments and application scenarios without departing from the principles and scope of the present application. Therefore, the present application is not limited to the described embodiments, but should be given the broadest scope consistent with the claims.

[0010] Figure 1 This is a flow chart of an online intelligent detection and analysis method for standard board production based on the industrial Internet of Things provided by an embodiment of the present application. The online intelligent detection and analysis method for standard board production based on the industrial Internet of Things is introduced in detail below.

[0011] Step S110, obtaining first standard board production monitoring data recording the target standard board production monitoring process in the industrial Internet of Things platform, and determining multiple node graph construction knowledge patterns for constructing a state node graph for the first standard board production monitoring data.

[0012] In detail, the industrial Internet of Things platform is a network that integrates various industrial equipment and systems. It realizes the interconnection between equipment through the Internet of Things technology, collects, analyzes and processes a large amount of data in the industrial production process to improve production efficiency and optimize resource allocation. The target standard board production monitoring process refers to the monitoring of the production process of standard boards (such as boards in furniture manufacturing) in the home furnishing industry manufacturing, for example, it can include the input of raw materials, the processing of each process, the quality inspection of finished products and other links.

[0013] The first standard board production monitoring data is obtained from the industrial Internet of Things platform, recording the data of the target standard board production monitoring process, which may include various types of data such as raw material information, equipment operation status, production parameters, quality inspection results, etc. The node graph construction knowledge model is used to guide how to construct the knowledge model or rules of the state node graph according to the production monitoring data, for example, it may include the single process node graph construction knowledge model, the whole batch node graph construction knowledge model, the associated equipment interaction node graph construction knowledge model, etc.

[0014] For example, in this embodiment, in the home furnishing industry manufacturing, for example, standard boards for furniture production. The server obtains the first standard board production monitoring data of the target standard board production monitoring process from the industrial Internet of Things platform. These data contain information from the input of raw materials to the initial forming of the standard board. For example, the humidity and texture direction data of the raw material wood, the rotation speed and temperature of the processing equipment, and the operation time of each process, etc.

[0015] Determine multiple node graph construction knowledge patterns for constructing state node graphs for the first standard board production monitoring data. For example, based on the characteristics of home furnishing manufacturing, the single process node graph construction knowledge pattern can be used to analyze the monitoring data of a single process such as wood cutting process to construct a state node graph. The whole batch node graph construction knowledge pattern can be used to analyze the overall monitoring data of a whole batch of standard boards, such as the state node graph of the overall process monitoring data of a batch of 100 standard boards from the beginning of production to the completion. The associated equipment interaction node graph construction knowledge pattern is suitable for analyzing the monitoring data related to the interaction between different devices to construct a state node graph, such as the monitoring data related to the transfer of standard boards between cutting equipment and grinding equipment. The multi-source data node graph construction knowledge pattern can be used to process the standard board production monitoring data from multiple data sources to construct a state node graph, such as the standard board production monitoring data integrated from multiple source data such as production equipment sensors, environmental monitoring sensors, and manual quality inspection records. The customized node graph construction knowledge pattern can construct a state node graph for specific needs, such as specifically analyzing the surface treatment process of standard boards and only focusing on the monitoring data of the usage and treatment time of specific chemicals to construct a state node graph. These different node graph construction knowledge patterns are helpful to comprehensively analyze the standard board production monitoring process from different perspectives.

[0016] Step S120: For the plurality of node graph construction knowledge patterns, identification attribute information is respectively injected into the first standard board production monitoring data to generate a plurality of second standard board production monitoring data. The second standard board production monitoring data in the plurality of second standard board production monitoring data respectively correspond to the node graph construction knowledge patterns in the plurality of node graph construction knowledge patterns, and the identification attribute information in each of the second standard board production monitoring data includes pattern identification information reflecting the corresponding node graph construction knowledge pattern.

[0017] In detail, the second standard board production monitoring data is data generated after injecting identification attribute information into the first standard board production monitoring data. Each second standard board production monitoring data corresponds to a specific node graph construction knowledge pattern and contains pattern identification information reflecting the pattern.

[0018] The identification attribute information is attribute information used to identify and distinguish different node graph construction knowledge patterns, and may include, for example, an initial data segment start identifier, an initial data segment end identifier, an end identifier, and pattern identification information reflecting the node graph construction knowledge pattern.

[0019] For example, in this embodiment, for the above-determined knowledge model building of the plurality of node graphs, the server begins to inject identification attribute information into the first standard board production monitoring data to generate a plurality of second standard board production monitoring data.

[0020] Take the construction of knowledge model of single process node graph as an example, if it is monitoring data for wood cutting process. Assuming that the amount of monitoring data about wood cutting process in the first standard board production monitoring data is not greater than the set quantity, the server injects pattern identification information reflecting the construction of knowledge model of single process node graph into the front end of these data, such as "wood cutting process pattern identification". Then, after this identification information and before the initial data segment of wood cutting process monitoring data, the initial data segment start identification is injected, such as "cutting process data segment start identification". The initial data segment end identification is injected into the back end of the wood cutting process monitoring data, such as "cutting process data segment end identification", and the termination identification "cutting process data termination identification" is injected after the injected initial data segment end identification, so as to generate the second standard board production monitoring data corresponding to the knowledge model constructed by single process node graph.

[0021] Let's look at the knowledge model constructed by the whole batch node graph. If the amount of whole batch production monitoring data contained in the first standard board production monitoring data is large (greater than the set number). The server injects the pattern identification information "whole batch production pattern identification" that reflects the knowledge model constructed by the whole batch node graph at the front end of the whole batch production monitoring data. After this identification information and before the initial data segment of the whole batch production monitoring data, the initial data segment start identification "whole batch initial data segment start identification" is injected. At the back end of the production monitoring data of the initial single process (such as the process corresponding to the start of production of the first batch of standard boards) in the whole batch production monitoring data, the initial data segment end identification "whole batch initial data segment end identification" is injected. Then determine the remaining single-process production monitoring data in the whole batch production monitoring data except the initial single-process production monitoring data, and inject a data segment segmentation identification, such as "whole batch data segment segmentation identification" between every two associated remaining single-process production monitoring data, thereby generating the second standard board production monitoring data corresponding to the knowledge model constructed by the whole batch node graph.

[0022] For the construction of knowledge patterns for the associated equipment interaction node graph, such as monitoring data involving the interaction between cutting equipment and polishing equipment, the server obtains the identification attribute information corresponding to the pattern, including specific pattern identification information "cutting-polishing equipment interaction pattern identification", initial data segment start identification, initial data segment end identification, etc., and injects them into the corresponding first standard board production monitoring data part according to the corresponding rules to generate the corresponding second standard board production monitoring data.

[0023] The multi-source data node graph construction knowledge model and the customized node graph construction knowledge model also inject the corresponding identification attribute information into the first standard board production monitoring data according to their respective requirements and data characteristics, thereby generating the corresponding second standard board production monitoring data. The identification attribute information in each second standard board production monitoring data contains the pattern identification information reflecting the corresponding node graph construction knowledge model, which helps the subsequent steps to accurately process the data.

[0024] Step S130 : determining a plurality of feature extraction paths associated with the plurality of second standard board production monitoring data according to the pattern identification information included in the plurality of second standard board production monitoring data.

[0025] In detail, the feature extraction path is a path or method for extracting feature information from the second standard board production monitoring data, and may include an interactive feature extraction path and a non-interactive feature extraction path, which are respectively used to process associated equipment interactive production monitoring data and non-associated equipment interactive production monitoring data.

[0026] For example, in this embodiment, the server determines a plurality of feature extraction paths associated with the plurality of second standard board production monitoring data according to the pattern identification information included in the plurality of second standard board production monitoring data.

[0027] For the second standard board production monitoring data containing the interactive task mode, such as the second standard board production monitoring data corresponding to the knowledge mode of the associated equipment interactive node graph construction (such as the monitoring data related to the interaction between the cutting equipment and the grinding equipment), the server determines the feature extraction path associated with it as the interactive feature extraction path for extracting features from the associated equipment interactive production monitoring data. Because in the manufacture of furniture standard boards, the interactive process of the cutting equipment and the grinding equipment has its unique features that need to be extracted, such as the time interval for the standard board after cutting to be transferred to the grinding equipment, the matching degree between the standard board size output by the cutting equipment and the preset receiving size of the grinding equipment, and other features.

[0028] For the second standard board production monitoring data in the non-interactive task mode, such as the second standard board production monitoring data corresponding to the knowledge mode of the whole batch node graph construction (the overall monitoring data of the whole batch of standard board production), the server determines that the feature extraction path associated with it is a non-interactive feature extraction path for feature extraction of non-associated device interactive production monitoring data. In this case, more attention may be paid to the overall time consumption and raw material utilization rate of the whole batch of standard board production process, rather than the interactive features between devices.

[0029] The second standard board production monitoring data corresponding to the knowledge model constructed by the single-process node graph, if it is the monitoring data of the wood cutting process, mainly focuses on the situation inside the single process and belongs to the non-interactive task mode, so the non-interactive feature extraction path is also adopted, mainly extracting features such as the wear speed of the cutting tool and the cutting accuracy. The second standard board production monitoring data corresponding to the knowledge model constructed by the multi-source data node graph may adopt part of the interactive feature extraction path (if it involves equipment interaction-related data sources) and part of the non-interactive feature extraction path (such as simple environmental monitoring data) according to the nature of different source data. The second standard board production monitoring data corresponding to the knowledge model constructed by the customized node graph also determines whether it is an interactive or non-interactive feature extraction path according to its customized content. For example, if the customization is about the special relationship monitoring data between two specific process equipment, the interactive feature extraction path is adopted, otherwise the non-interactive feature extraction path is adopted.

[0030] Step S140, for each of the plurality of second standard plate production monitoring data, extract the production parameter trajectory vector according to the associated feature extraction path, and generate a set of production parameter trajectory vectors corresponding to the plurality of node graph construction knowledge patterns.

[0031] In detail, the production parameter trajectory vector is generated by extracting features from the second standard board production monitoring data, and is a vector representing the change of production parameters over time, and is used to construct a state node graph to reflect state transitions and parameter changes in the production process.

[0032] For example, in this embodiment, for each of the multiple second standard plate production monitoring data, the server extracts the production parameter trajectory vector according to the associated feature extraction path, and generates multiple node graphs to construct production parameter trajectory vector sets corresponding to the knowledge patterns.

[0033] Taking the second standard board production monitoring data (monitoring data related to the interaction between cutting equipment and grinding equipment) corresponding to the knowledge model constructed by the associated equipment interaction node graph as an example, when the feature extraction path associated with the data is an interactive feature extraction path, for each single-process production monitoring data (such as the monitoring data corresponding to each cutting operation of the cutting equipment and the monitoring data corresponding to each grinding operation of the grinding equipment), the server determines the interactive link data of each single-process production monitoring data in the second standard board production monitoring data. For example, when the cutting equipment passes the standard board to the grinding equipment, the interactive link data such as the instantaneous stop time of the cutting equipment and the start-up preparation time of the grinding equipment are passed. Then, the determined interactive link data are characterized respectively, and the interactive link feature vectors corresponding to each single-process production monitoring data are generated.

[0034] Based on the interactive device labels of each single-process production monitoring data, such as cutting equipment and grinding equipment, the interactive devices associated with each single-process production monitoring data are determined, and the interactive devices associated with each single-process production monitoring data are characterized respectively, and the interactive device feature vectors corresponding to each single-process production monitoring data are generated. For example, for cutting equipment, its features may include the type of cutting tool, cutting power, etc.; for grinding equipment, its features may include the roughness of the grinding wheel, grinding speed, etc.

[0035] The state parameter feature representation is performed for each single-process production monitoring data, and a set of state parameter feature vectors corresponding to each single-process production monitoring data is generated. For example, the state parameters such as cutting depth and cutting angle of cutting equipment, and the state parameters such as grinding force and grinding time of grinding equipment are all represented as state parameter feature vectors.

[0036] Each single-process production monitoring data is represented by context features, and a set of context feature vectors corresponding to each single-process production monitoring data is generated. For example, the ambient temperature and humidity of the cutting equipment when it is operating may affect the cutting effect, and these environmental factors constitute the context feature vector; the noise level in the workshop when the grinding equipment is operating also belongs to the context feature vector.

[0037] Finally, based on the interaction link feature vector, interaction device feature vector, state parameter feature vector set and context feature vector set corresponding to each single-process production monitoring data, a production parameter trajectory vector set corresponding to the second standard board production monitoring data corresponding to the knowledge model constructed by the associated equipment interaction node graph is generated.

[0038] For the second standard board production monitoring data corresponding to the knowledge model constructed by the entire batch of node graphs, when the feature extraction path associated with it is a non-interactive feature extraction path, the server performs state parameter feature representation on each data node therein (such as each process link in the production process of each standard board is regarded as a data node) and generates a state parameter feature vector corresponding to each data node. For example, for the wood drying process of each standard board, the state parameters may include drying time, drying temperature, etc.

[0039] Contextual feature representation is performed on each data node to generate contextual feature vectors corresponding to each data node, such as the ventilation conditions of the warehouse during the wood drying process.

[0040] The state parameter feature vector and context feature vector corresponding to the same data node are fused to generate a fused feature vector corresponding to each data node. Based on the fused feature vector corresponding to each data node, a set of production parameter trajectory vectors corresponding to the second standard board production monitoring data corresponding to the knowledge model constructed by the entire batch node graph is generated.

[0041] The second standard board production monitoring data corresponding to the single-process node graph construction knowledge model (such as wood cutting process), the multi-source data node graph construction knowledge model, and the customized node graph construction knowledge model also perform similar production parameter trajectory vector extraction operations according to their respective associated feature extraction paths to generate their respective corresponding production parameter trajectory vector sets.

[0042] Step S150, node graphs are constructed for the production parameter trajectory vector sets corresponding to the multiple node graph construction knowledge patterns respectively, and state node graph construction results corresponding to the multiple node graph construction knowledge patterns are generated; based on the state node graph construction results corresponding to the multiple node graph construction knowledge patterns respectively, abnormal state links of the target standard board production monitoring process are estimated.

[0043] In detail, the state node graph construction result is constructed based on a set of production parameter trajectory vectors, reflecting a graphical representation of state transitions and logical relationships in the production monitoring process, including initial state nodes, derived state nodes, relationships between state nodes, etc.

[0044] The abnormal state link is a link in which an abnormal state occurs during the production monitoring process and causes the production process to deviate from the normal path. By estimating the abnormal state link, problems in the production process can be discovered in time and corresponding measures can be taken to solve them, so as to improve production efficiency and product quality.

[0045] For example, in this embodiment, the server constructs node graphs for the production parameter trajectory vector sets corresponding to the multiple node graph construction knowledge patterns respectively, generates state node graph construction results corresponding to the multiple node graph construction knowledge patterns respectively, and estimates the abnormal state link of the target standard board production monitoring process based on the state node graph construction results corresponding to the multiple node graph construction knowledge patterns respectively.

[0046] Taking the production parameter trajectory vector set (such as the wood cutting process) corresponding to the knowledge model constructed by the single process node graph as an example, for this production parameter trajectory vector set, the server generates an initial state node based on the traversed production parameter trajectory vector set and the initial state node identifier corresponding to the production parameter trajectory vector set. For example, if the initial state is the start of the cutting equipment, then the initial state node is represented as "the start of the cutting equipment".

[0047] Based on this initial state node and the traversed production parameter trajectory vector set, multiple stages of graph construction processing are performed. First, each production parameter trajectory vector in the production parameter trajectory vector set is classified and marked, and a preliminary association relationship is established between the initial state node and the first production parameter trajectory vector in the production parameter trajectory vector set. This preliminary association relationship is based on the starting logic of the production process. For example, the first production parameter trajectory vector may be the initial speed of the cutting equipment, and it is associated with the initial state node of the cutting equipment startup based on the production starting logic.

[0048] Then analyze the properties of the initial state node. If the initial state node represents the start-up state of production (such as the start-up of the cutting equipment), find the production parameter vector related to the start-up state in the production parameter trajectory vector set, establish an advanced association relationship between the production parameter vector and the initial state node, and thus preliminarily build a mapping relationship between the initial state node and some production parameter trajectory vectors based on production logic. For example, parameters such as the current size when the cutting equipment is started are associated with the start-up state of the cutting equipment.

[0049] Based on the preliminary association relationship, each associated production parameter trajectory vector is analyzed to find other production parameter trajectory vectors in the production parameter trajectory vector set that are logically associated with the associated production parameter trajectory vector, and the initial state node is gradually expanded based on the search results to continuously create representations of intermediate state nodes based on the initial state nodes to generate an intermediate state node structure. For example, after the cutting equipment is started, as the cutting proceeds, the wear of the cutting tool is logically associated with the startup state of the cutting equipment, and through this association, an intermediate state node such as "the tool begins to wear after the cutting equipment is started for a period of time" is expanded.

[0050] The intermediate state node structure is optimized for redundancy to obtain an optimized intermediate state node structure. The time information corresponding to each production parameter trajectory vector in the production parameter trajectory vector set is determined, and the association relationship in the intermediate state node in the optimized intermediate state node structure is time-sorted according to the time information, and the intermediate state node is expanded according to the advancement of the time sorting sequence to obtain a derived state node structure carrying the time sorting sequence information. For example, the degree of tool wear at different time points after the cutting equipment is started is different, and different derived state nodes are constructed in chronological order.

[0051] The initial state node and multiple derived state nodes are integrated to generate a state node graph construction result corresponding to the set of production parameter trajectory vectors corresponding to the knowledge model of the single process node graph construction.

[0052] Similar node graph construction operations are also performed on the production parameter trajectory vector sets corresponding to the whole batch node graph construction knowledge model, the associated equipment interaction node graph construction knowledge model, the multi-source data node graph construction knowledge model and the customized node graph construction knowledge model to generate the corresponding state node graph construction results.

[0053] Next, the server standardizes the state node graph corresponding to each state node graph construction result to unify the representation of state nodes in the state node graph and the description format of the relationship between state nodes. For example, the equipment status and process status in different modes are represented by unified symbols or formats.

[0054] Analyze each state node and the relationship between each state node in the standardized state node graph, find the state nodes corresponding to each link in the production process, and for each state node, determine the preceding and succeeding nodes of the state node in the production process, establish a complete node relationship mapping based on the production process sequence, and generate a state node graph structure that carries the production process mapping relationship. For example, in the production of standard furniture boards, the preceding node of the wood drying process may be the wood cutting process, and the succeeding node may be the board splicing process, so such a complete relationship mapping is established.

[0055] Based on the established node relationship mapping, the normal state link under normal production conditions is extracted from the state node graph, and the characteristic patterns in the normal state link are analyzed. The key nodes, relationships and characteristic patterns of the normal state link are combined to construct a characteristic model for describing the normal state link.

[0056] The actual state link in each state node graph is compared and analyzed with the characteristic model of the normal state link to preliminarily screen out candidate abnormal links in the state node graph. For example, if the time of the wood drying process in the actual state link is far longer than that in the normal state link, it may be preliminarily screened as a candidate abnormal link.

[0057] For the candidate abnormal links that are initially screened, refer to the historical production monitoring data for verification and generate a verified target abnormal link. For example, check the time data range of similar wood drying processes in the past. If the current abnormal situation is also judged as abnormal in the historical data, then it is determined as the target abnormal link.

[0058] For the verified target abnormal link, the root cause analysis is performed in the graph structure of the state node graph, the abnormal state node that first appeared abnormal state in the target abnormal link is traced back, and the abnormal cause of the abnormal state node is analyzed, and the association relationship information between the target abnormal link and other state links is evaluated. For example, if the target abnormal link is that the wood drying process takes too long, it may be traced back to the failure of the drying equipment, and whether this abnormal link will affect the subsequent board splicing process and other associations are analyzed.

[0059] Based on the abnormal cause and the association relationship information between the target abnormal link and other status links, the severity of each target abnormal link is determined. For example, if the drying equipment failure causes the wood drying process to take too long, if it will seriously affect the production progress and quality of the entire batch of standard boards, then the severity is high. If it can be remedied in time through some adjustment measures, the severity is relatively low. In this way, the abnormal status links in the production monitoring process of standard boards in home furnishing industry manufacturing can be comprehensively and accurately analyzed, which helps to find problems in time and take effective solutions to improve production efficiency and product quality.

[0060] Based on the above steps, the embodiment of the present application first determines multiple node graphs to construct knowledge patterns, and generates multiple second standard board production monitoring data corresponding to different knowledge patterns by injecting identification attribute information into the first standard board production monitoring data, thereby realizing refined classification and processing of data. According to the pattern identification information in the second standard board production monitoring data, the feature extraction path associated with the data can be accurately matched. This accurate matching ensures the pertinence and effectiveness of feature extraction, and provides a high-quality data foundation for the subsequent state node graph construction. For each second standard board production monitoring data, the present invention efficiently extracts production parameter trajectory vectors based on the associated feature extraction path, generates a set of production parameter trajectory vectors corresponding to the knowledge pattern constructed by multiple node graphs, and can fully reflect the parameter changes and state transitions in the production process. By constructing a node graph for the production parameter trajectory vector set, multiple state node graph construction results are generated, which can intuitively display the state transition and logical relationship in the production process. Based on the state node graph, the abnormal state link in the target standard board production monitoring process can be accurately estimated, and potential problems in production can be discovered in time. Therefore, the level of online intelligent detection and analysis of standard board production has been significantly improved through the construction of multi-mode state node graphs, precise matching of feature extraction paths, efficient extraction of production parameter trajectory vectors, intelligent construction of state node graphs and accurate estimation of abnormal state links.

[0061] In a possible implementation, step S110 includes: Step S111, generating a node graph construction knowledge pattern sequence based on a state node graph construction instruction for the first standard plate production monitoring data; Step S112, based on the hit instruction for the node graph construction knowledge pattern sequence, the multiple node graph construction knowledge patterns hit by the hit instruction are used as multiple node graph construction knowledge patterns for constructing the state node graph of the first standard board production monitoring data.

[0062] In this embodiment, for determining multiple node graph construction knowledge patterns for state node graph construction of the first standard board production monitoring data, when the server receives the state node graph construction instruction for the first standard board production monitoring data, it starts to generate a node graph construction knowledge pattern sequence. This sequence includes various patterns that may be applicable to the home standard board production monitoring data, such as single process node graph construction knowledge patterns for single processes such as cutting, polishing, and assembly in furniture production, whole batch node graph construction knowledge patterns considering the production of whole batches of standard boards, associated equipment interaction node graph construction knowledge patterns focusing on the interaction between different equipment such as cutting equipment and polishing equipment, multi-source data node graph construction knowledge patterns integrating multi-source data such as equipment sensors, manual quality inspection data, etc., and customized node graph construction knowledge patterns to meet special needs, etc. Then, when there is a hit instruction for this node graph construction knowledge pattern sequence, for example, when producing standard boards of furniture of a specific style, since the production process involves multiple complex processes and equipment interactions and has high requirements for quality stability, the hit instruction may select multiple patterns such as the associated equipment interaction node graph construction knowledge pattern, the whole batch node graph construction knowledge pattern, and the single process node graph construction knowledge pattern. These hit patterns become multiple node graph construction knowledge patterns for constructing the status node graph of the first standard board production monitoring data.

[0063] In a possible implementation, step S120 includes: Step S121, for each node graph construction knowledge pattern in the plurality of node graph construction knowledge patterns, obtaining identification attribute information corresponding to the traversed node graph construction knowledge pattern; Step S122, injecting identification attribute information corresponding to the knowledge pattern constructed by the traversed node graph into the first standard plate production monitoring data, and generating second standard plate production monitoring data corresponding to the knowledge pattern constructed by the traversed node graph.

[0064] In this embodiment, for injecting identification attribute information into the first standard board production monitoring data respectively in the knowledge mode of building multiple node graphs to generate multiple second standard board production monitoring data, the server will operate for each mode in the knowledge mode of building multiple node graphs. Taking the knowledge mode of building the node graph of associated equipment interaction as an example, the server will obtain the identification attribute information corresponding to this mode, which is specially set for the monitoring data identification related to the interaction of associated equipment, including the initial data segment start identifier indicating the start of the device interaction, the initial data segment end identifier indicating the end of the interactive part of the data, the end identifier indicating the end of the entire data segment, and the mode identification information that can clearly reflect that this is the knowledge mode of building the node graph of associated equipment interaction. Then, the server injects these obtained identification attribute information into the part of the first standard board production monitoring data related to the interaction of associated equipment, such as the data part when the cutting equipment interacts with the grinding equipment, so as to generate the second standard board production monitoring data corresponding to the knowledge mode of building the node graph of associated equipment interaction. Similarly, for the whole batch of node graph construction knowledge models, the server obtains the corresponding identification attribute information, such as the whole batch production start identifier, the segmentation identifier of different stages in each batch, etc., and injects these identification attribute information into the whole batch production related part of the first standard board production monitoring data to generate the corresponding second standard board production monitoring data. The single-process node graph construction knowledge model, the multi-source data node graph construction knowledge model, and the customized node graph construction knowledge model also obtain the corresponding identification attribute information and inject it into the corresponding part of the first standard board production monitoring data in a similar manner to generate the corresponding second standard board production monitoring data. Each second standard board production monitoring data generated in this way can clearly reflect the corresponding node graph construction knowledge model through the identification attribute information therein, which is convenient for subsequent processing and analysis.

[0065] In one possible implementation, the multiple node graph construction knowledge patterns include at least one of a single-process node graph construction knowledge pattern, a whole batch node graph construction knowledge pattern, an associated device interaction node graph construction knowledge pattern, a multi-source data node graph construction knowledge pattern, or a customized node graph construction knowledge pattern.

[0066] The node graph construction instance of the single-process node graph construction knowledge model is used to construct a state node graph for single-process production monitoring data. The node graph construction instance of the whole batch node graph construction knowledge model is used to construct a state node graph for the whole batch production monitoring data. The node graph construction instance of the associated device interaction node graph construction knowledge model is used to construct a state node graph for associated device interaction production monitoring data. The node graph construction instance of the multi-source data node graph construction knowledge model is used to construct a state node graph for standard board production monitoring data extracted from multi-source data. The node graph construction instance of the customized node graph construction knowledge model is used to construct a customized state node graph for standard board production monitoring data.

[0067] In this embodiment, in the production process of standard home boards, different node graph construction knowledge models play their own important roles. The single-process node graph construction knowledge model focuses on a single production process, taking the wood cutting process as an example. In furniture production, the cutting process has a key impact on the dimensional accuracy and quality of standard boards. The server collects various production monitoring data in the cutting process, such as the speed of the cutting tool, the cutting depth, the feeding speed of the wood, etc. These single-process production monitoring data are the basis for building a state node graph. The server starts to build a state node graph through a node graph construction instance of the single-process node graph construction knowledge model. First, it sets the starting state of the cutting process to an initial node, such as "cutting process start", which contains the basic parameters when the cutting equipment is started, such as the initial power of the equipment, the initial position of the tool, etc. As the cutting process progresses, different monitoring data form different status nodes, such as "cutting depth reaches 3 cm" and "feed speed is stable at 1 meter per minute", etc., and these nodes are associated with each other according to production logic. For example, changes in cutting depth will affect the adjustment of feed speed. In this way, a single-process status node diagram is constructed, which helps to analyze potential problems in a single cutting process, such as whether abnormal changes in cutting depth are related to tool wear or equipment failure.

[0068] The knowledge model constructed by the whole batch node graph is mainly aimed at the whole batch production monitoring data. Assume that the production quantity of a batch of home standard boards is 500 pieces. The server will obtain various data of this whole batch of standard boards from raw material input to finished product output, including the time nodes of different production stages, the overall raw material consumption, the quality sampling results of the whole batch of products, etc. Node graph construction example of knowledge model constructed by the whole batch node graph When constructing the state node graph, the initial node can be "whole batch production starts", which contains the basic information of the whole batch production, such as the batch of raw materials, the start time of the production plan, etc. As the production progresses, different state nodes will be generated, such as "the cutting process of 200 standard boards is completed" and "the whole batch of standard boards enters the assembly process". These nodes reflect the progress and state transition of the whole batch production. At the same time, by analyzing the relationship between different nodes, potential problems in the whole batch production process can be found. For example, if the unqualified rate of the quality sampling of the whole batch of products suddenly increases after a certain time node, the state node graph can be used to trace which production link has problems, whether it is the problem of raw materials or the operation error of a certain process that affects the quality of the whole batch of products.

[0069] The knowledge model constructed by the node graph of the interaction of associated equipment is used to process the production monitoring data of the interaction of associated equipment. In furniture production, there is a close interaction between the cutting equipment and the grinding equipment. The server collects relevant data when the cutting equipment transfers the standard board to the grinding equipment, such as the standard board size accuracy after cutting, the time interval of the transfer, the synchronization status of the cutting equipment and the grinding equipment, etc. The node graph construction example of the knowledge model constructed by the node graph of the interaction of associated equipment constructs the state node graph. The initial node may be "cutting-grinding equipment interaction starts", which contains the state information of the two equipment at the beginning of the interaction, such as the stop state of the cutting equipment and the ready-to-receive state of the grinding equipment. Then, as the interaction process proceeds, state nodes such as "the standard board is successfully transferred from the cutting equipment to the grinding equipment" and "the grinding equipment adjusts the grinding parameters according to the size of the standard board" will be generated. The relationship between these nodes can reflect the coordination of the equipment interaction process. For example, if a node of "grinding equipment frequently adjusts grinding parameters" appears, by analyzing the state node diagram, it can be traced whether the standard plate size accuracy output by the cutting equipment is unstable, or whether there is a problem with the detection and adjustment mechanism of the grinding equipment itself, so as to promptly resolve abnormal situations in the equipment interaction process and ensure smooth production.

[0070] The knowledge model constructed by the multi-source data node graph is targeted at the standard board production monitoring data extracted from the multi-source data. In home furnishing manufacturing, the data sources are diverse, including sensor data on production equipment, environmental monitoring data, manual quality inspection data, etc. For example, sensor data may provide equipment operating parameters, environmental monitoring data can reflect the impact of temperature and humidity in the production workshop on the quality of standard boards, and manual quality inspection data records the inspection results of workers on the appearance and structural quality of standard boards. When the server uses the node graph construction instance of the multi-source data node graph to construct the knowledge model, it will integrate the data from these different sources when constructing the state node graph. The initial node can integrate the initial information of multiple data sources, such as "production starts, equipment runs normally, workshop temperature is 25 degrees Celsius, humidity is 50%, and the first batch of standard boards pass the appearance inspection". As production progresses, the changes in different source data form different state nodes, such as "equipment temperature rises by 5 degrees, workshop humidity drops to 40%, and a certain standard board fails the structural inspection". By constructing such a state node diagram, the mutual influence between multi-source data can be comprehensively analyzed, such as whether changes in equipment operating parameters are related to changes in environmental conditions, and how these changes jointly affect the quality of standard boards, thereby achieving more comprehensive and accurate production monitoring.

[0071] The knowledge model of customized node graph construction is used to construct customized state node graphs for standard board production monitoring data. Assume that when producing standard boards for high-end customized furniture, special attention should be paid to the two indicators of wood grain direction and surface finish. The server will construct customized state node graphs specifically for the production monitoring data related to these two indicators. The initial node can be "customized production starts, pay attention to wood grain and surface finish", which contains the initial setting values ​​and requirements related to these two indicators. Then in the production process, the state node may be "the wood grain direction meets the design requirements, and the surface finish meets the preliminary standards" and "after the polishing process, the surface finish is further improved". Through the construction of customized state node graphs, precise monitoring under special production needs can be met to ensure the production of standard boards that meet high-end customization requirements. For example, it is possible to promptly discover the deviation of wood grain direction in a certain process, or the situation where the surface finish fails to meet the expected target, so as to adjust the production process in time.

[0072] In a possible implementation, the identification attribute information corresponding to the traversed node graph construction knowledge pattern includes an initial data segment start identifier, an initial data segment end identifier, an end identifier, and pattern identification information reflecting the traversed node graph construction knowledge pattern.

[0073] Step S122 may include: when the number of single-process production monitoring data included in the first standard board production monitoring data is not greater than a set number, injecting pattern identification information reflecting the traversed node graph construction knowledge pattern into the front end of the first standard board production monitoring data.

[0074] The initial data segment start identifier is injected after the pattern identification information reflecting the knowledge pattern constructed by the traversed node graph and before the initial data segment of the first standard board production monitoring data.

[0075] The initial data segment completion identifier is injected into the back end of the first standard board production monitoring data, and the termination identifier is injected after the injected initial data segment completion identifier to generate the second standard board production monitoring data corresponding to the traversed node graph construction knowledge model.

[0076] In another possible implementation, the identification attribute information corresponding to the knowledge pattern constructed by the traversed node graph includes an initial data segment start identifier, an initial data segment end identifier, a data segment segmentation identifier, and pattern identification information reflecting the knowledge pattern constructed by the traversed node graph.

[0077] Step S122 may include: when the number of single-process production monitoring data included in the first standard board production monitoring data is greater than a set number, injecting pattern identification information reflecting the traversed node graph construction knowledge pattern into the front end of the first standard board production monitoring data.

[0078] The initial data segment start identifier is injected after the pattern identification information reflecting the knowledge pattern constructed by the traversed node graph and before the initial data segment of the first standard board production monitoring data.

[0079] The initial data segment completion identifier is injected into the back end of the initial single-process production monitoring data in the first standard board production monitoring data.

[0080] Determine the remaining single-process production monitoring data in the first standard board production monitoring data except the initial single-process production monitoring data, and inject the data segment segmentation identifier between every two associated remaining single-process production monitoring data to generate the second standard board production monitoring data corresponding to the knowledge model constructed by the traversed node graph.

[0081] In this embodiment, in the production process of furniture standard boards, when it comes to constructing knowledge models for different node graphs and injecting identification attribute information to generate corresponding second standard board production monitoring data, different injection methods need to be adopted according to the number of single-process production monitoring data.

[0082] Taking the construction of a knowledge model of a single process node graph as an example, assume that this single process is a wood cutting process. If the number of production monitoring data about the wood cutting process in the first standard board production monitoring data is not greater than the set number, the server will inject identification attribute information according to specific rules. First, the pattern identification information reflecting the construction of the knowledge model of the single process node graph is injected at the front end of the first standard board production monitoring data (here is the monitoring data related to the wood cutting process), such as "WOOD_CUTTING_SINGLE_PROCESS". This identification clarifies that this part of the data is the data of the node graph construction knowledge model related to the single process of wood cutting. Then, after this pattern identification information and before the initial data segment of the wood cutting process monitoring data, the initial data segment start identifier is injected, such as "START_CUTTING_SEGMENT", which indicates the beginning of a valid data segment in the wood cutting process. Then, the initial data segment end mark, such as "END_CUTTING_SEGMENT", is injected into the back end of the wood cutting process monitoring data to indicate the end of this initial data segment, and the termination mark "TERMINATE_CUTTING_DATA" is injected after the injected initial data segment end mark, so that the identification attribute information is completely added to this part of the data, and the second standard board production monitoring data corresponding to the single process node graph construction knowledge model is generated. This method helps to clearly define and identify this part of the data, and provides convenience for subsequent processing. For example, when performing data extraction or state node graph construction, the data segment related to the wood cutting process can be quickly located and its starting and ending positions can be clearly identified.

[0083] When the number of production monitoring data about the wood cutting process in the first standard board production monitoring data is greater than the set number, the operation of the server is different. Taking the single process node graph construction knowledge model as an example, the server injects the pattern identification information reflecting the single process node graph construction knowledge model at the front end of the wood cutting process monitoring data, such as "WOOD_CUTTING_SINGLE_PROCESS_LARGE_DATA", which corresponds to the single process wood cutting data with a large amount of data. Then, after this pattern identification information and before the initial data segment of the wood cutting process monitoring data, the initial data segment start identifier "START_CUTTING_SEGMENT_LARGE" is injected. At the back end of the initial single process production monitoring data (such as the monitoring data of the first batch of cutting wood) in the wood cutting process monitoring data, the initial data segment end identifier "END_FIRST_CUTTING_SEGMENT" is injected. After that, the server will determine the remaining single process production monitoring data in the wood cutting process monitoring data except the initial single process production monitoring data, and inject the data segment segmentation identifier, such as "DIVIDE_CUTTING_SEGMENT" between every two associated remaining single process production monitoring data. In this way, a large amount of wood cutting process monitoring data is reasonably divided and identified, and the second standard board production monitoring data corresponding to the knowledge model constructed by the single process node graph is generated. The advantage of this is that when a large amount of cutting process data is subsequently analyzed, the data at different stages can be quickly located based on these identifiers, which facilitates data mining and analysis, such as analyzing the differences or trends in the cutting process of different batches of wood.

[0084] The knowledge model for the whole batch node graph construction is similar. Assume that the whole batch is the monitoring data of the entire production process of producing 1,000 standard boards. If the number of single-process production monitoring data about a sub-process (such as the assembly process) in the whole batch production monitoring data is not greater than the set number, the server will inject the pattern identification information reflecting the knowledge model of the whole batch node graph construction, such as "WHOLE_BATCH_ASSEMBLY_PROCESS" at the front end of the monitoring data related to the assembly process. Then, after this identification and before the initial data segment of the assembly process monitoring data, the initial data segment start identification "START_ASSEMBLY_SEGMENT" is injected, and the initial data segment end identification "END_ASSEMBLY_SEGMENT" is injected at the back end, and finally the termination identification "TERMINATE_ASSEMBLY_DATA" is injected, thereby generating the assembly process-related part in the second standard board production monitoring data corresponding to the knowledge model of the whole batch node graph construction. This helps to clearly identify the data part of the assembly process in the whole batch production monitoring data, and facilitates the subsequent analysis of the assembly process from the perspective of the whole batch, such as analyzing the impact of the assembly process on the quality of the whole batch of products.

[0085] When the number of single-process production monitoring data about a sub-process (such as the polishing process) in the entire batch of production monitoring data is greater than the set number, the server injects the pattern identification information reflecting the knowledge model of the entire batch node graph construction at the front end of the monitoring data related to the polishing process, such as "WHOLE_BATCH_SANDING_PROCESS_LARGE_DATA". After this identification and before the initial data segment of the polishing process monitoring data, the initial data segment start identification "START_SANDING_SEGMENT_LARGE" is injected, and the initial data segment end identification "END_FIRST_SANDING_SEGMENT" is injected at the back end of the initial single-process production monitoring data (such as the first batch of polished standard board data). Then determine the remaining single-process production monitoring data, and inject the data segment division identification "DIVIDE_SANDING_SEGMENT" between every two associated remaining single-process production monitoring data, and generate the polishing process-related part in the second standard board production monitoring data corresponding to the knowledge model of the entire batch node graph construction. This allows for better data organization and analysis when dealing with large amounts of grinding process data, such as analyzing the contribution of different stages of grinding to the quality of a whole batch of standard plates or finding abnormal fluctuations in the grinding process.

[0086] In the knowledge model of the associated equipment interaction node graph construction, take the interaction between the cutting equipment and the grinding equipment as an example. If the number of standard board-related interaction monitoring data transmitted by the cutting equipment to the grinding equipment is not greater than the set number, the server injects the pattern identification information reflecting the knowledge model of the associated equipment interaction node graph construction into the front end of this part of the interaction monitoring data, such as "CUTTING_SANDING_INTERACTION". Then, after this identifier and before the initial data segment, the initial data segment start identifier "START_INTERACTION_SEGMENT" is injected, and the initial data segment end identifier "END_INTERACTION_SEGMENT" is injected into the back end, and then the termination identifier "TERMINATE_INTERACTION_DATA" is injected to generate the corresponding second standard board production monitoring data. This helps to accurately identify the scope and nature of this part of the interaction data when analyzing the equipment interaction process.

[0087] When the number of interactive monitoring data related to the standard board transmitted from the cutting device to the grinding device is greater than the set number, the server injects the pattern identification information reflecting the knowledge pattern of the associated device interaction node graph construction into the front end of this part of the interactive monitoring data, such as "CUTTING_SANDING_INTERACTION_LARGE_DATA". After this identification and before the initial data segment, the initial data segment start identification "START_INTERACTION_SEGMENT_LARGE" is injected, and the initial data segment end identification "END_FIRST_INTERACTION_SEGMENT" is injected into the back end of the initial interactive monitoring data (such as the interactive data of the standard board transmitted for the first time), and the remaining interactive monitoring data is determined and the data segment segmentation identification "DIVIDE_INTERACTION_SEGMENT" is injected between every two associated remaining interactive monitoring data to generate the corresponding second standard board production monitoring data. This can better manage and analyze a large amount of equipment interaction data, such as finding regularities or abnormalities in the equipment interaction process.

[0088] The knowledge model constructed by the node graph of multi-source data and the knowledge model constructed by the customized node graph also follow similar principles to inject identification attribute information. For the knowledge model constructed by the node graph of multi-source data, for example, when integrating equipment sensor data, environmental monitoring data and manual quality inspection data, if the number of single-process production monitoring data corresponding to a certain type of data (such as temperature sensor data in equipment sensor data) is not greater than the set number, the server will inject identification attribute information in the manner of not greater than the set number mentioned above. If the amount of data is greater than the set number, the identification attribute information is injected in the corresponding manner of greater than the set number, thereby generating the corresponding second standard board production monitoring data, which is convenient for effective organization and analysis of multi-source data. For the knowledge model constructed by the customized node graph, such as the data on the usage and processing time of a certain special chemical agent in the surface treatment process of the customized standard board, if the amount of this part of the data is not greater than the set number, the identification attribute information is injected according to the corresponding rules; if it is greater than the set number, it is also injected according to the corresponding rules to generate the second standard board production monitoring data corresponding to the customized requirements, which is helpful for accurate customized production monitoring and analysis.

[0089] In a possible implementation, step S130 includes: Step S131, for each second standard board production monitoring data in the plurality of second standard board production monitoring data, when the node graph construction knowledge mode reflected by the mode identification information included in the traversed second standard board production monitoring data belongs to the interactive task mode, determining that the feature extraction path associated with the traversed second standard board production monitoring data is an interactive feature extraction path for extracting features from the interactive production monitoring data of the associated device; Step S132, when the node graph construction knowledge pattern reflected by the pattern identification information included in the traversed second standard board production monitoring data belongs to the non-interactive task mode, determine that the feature extraction path associated with the traversed second standard board production monitoring data is a non-interactive feature extraction path used to extract features from non-associated equipment interactive production monitoring data.

[0090] In this embodiment, in the production process of home standard boards, a plurality of second standard board production monitoring data have been generated, and now it is necessary to determine the feature extraction path associated therewith. For each second standard board production monitoring data, the server checks the type of node graph construction knowledge pattern reflected by the pattern identification information contained therein.

[0091] Taking the second standard board production monitoring data corresponding to the knowledge model constructed by the associated device interaction node graph of the interaction between the cutting device and the grinding device as an example, since this mode belongs to the interactive task mode, the server determines that the feature extraction path associated with the second standard board production monitoring data is an interactive feature extraction path for feature extraction of the associated device interactive production monitoring data. In furniture production, this means paying attention to various interactive related features in the interaction process between the cutting device and the grinding device. For example, the transmission speed and accuracy of the standard board when the cutting device passes it to the grinding device, which involves whether the discharge speed of the cutting device matches the feed speed of the grinding device, and whether the position of the standard board is accurate during the transmission process. It also includes the signal transmission situation when the devices interact, such as whether the signal from the cutting device to the grinding device that the standard board is about to arrive is timely and accurate, and the response time of the grinding device after receiving the signal. In addition, the interactive feature extraction path will also pay attention to the impact of the cutting quality of the cutting device before passing the standard board on the subsequent grinding work of the grinding device, such as the flatness of the cutting surface will affect the difficulty and effect of grinding.

[0092] Let's look at the second standard board production monitoring data corresponding to the knowledge model constructed by the whole batch node graph. This mode belongs to the non-interactive task mode. The server determines that the feature extraction path associated with this second standard board production monitoring data is a non-interactive feature extraction path for feature extraction of non-associated interactive production monitoring data. For the whole batch of production monitoring data, the server will focus on some overall features, such as the total consumption of raw materials in the production process of the whole batch of standard boards, including the total amount of wood, the amount of auxiliary materials such as glue, etc. The total production time is also an important feature, which is the time spent from the start of production of the first batch of standard boards to the completion of production of the last batch, which can reflect the production efficiency. In addition, the overall distribution of the quality of the whole batch of products is also the content to be extracted by the non-interactive feature extraction path, such as the proportion of standard boards of different quality levels (such as high-quality, qualified, and defective) in the whole batch of products, and the changing trend of these proportions during the production process. Some macro-environmental factors in the production process, such as the comprehensive impact of the average temperature and humidity of the entire production workshop during the whole batch production on product quality, are also extracted through the non-interactive feature extraction path.

[0093] If the knowledge model constructed by the single process node graph is for a relatively independent process such as the wood cutting process, it also belongs to the non-interactive task mode. The server also determines that the feature extraction path associated with the second standard board production monitoring data corresponding to this mode is a non-interactive feature extraction path. In this case, the server will focus on some characteristics within the wood cutting process, such as the wear rate of the cutting tool, which can be determined by analyzing the changes in the blade thickness of the cutting tool or the decrease in cutting accuracy in different time periods. The energy consumption of the cutting equipment is also an important feature, such as the difference in the amount of electricity consumed by the equipment when cutting wood of different hardness. In addition, quality control indicators in the cutting process, such as the dimensional accuracy distribution of the standard board after cutting, how many standard boards are within the allowable error range, and how many exceed the error range, etc., are all features extracted through non-interactive feature extraction paths.

[0094] For the second standard board production monitoring data corresponding to the knowledge model constructed by the multi-source data node graph, if it contains data related to equipment interaction, and the pattern identification information corresponding to this part of the data indicates that it is an interactive task mode, then the interactive feature extraction path is used for this part of the data. For example, there is data about the interaction between cutting equipment and quality inspection equipment in the multi-source data. The interactive feature extraction path will focus on interactive features such as the speed when the cutting equipment sends the standard board to the inspection equipment, the speed and accuracy of the inspection equipment to recognize the standard board. For other parts of the multi-source data, such as simple environmental monitoring data or raw material batch quality data in different time periods, if the corresponding pattern identification information indicates that it is a non-interactive task mode, the non-interactive feature extraction path is used. For example, the long-term impact of temperature and humidity changes in the production workshop in different seasons on the quality of the standard board is extracted from the environmental monitoring data, and the impact of wood provided by different suppliers on product quality is analyzed from the raw material batch quality data.

[0095] The second standard board production monitoring data corresponding to the knowledge model constructed by the customized node graph also determines the feature extraction path according to the task mode reflected by its mode identification information. If customization focuses on the interaction between specific processes, such as the interaction between the painting equipment and the drying equipment in the furniture surface treatment process, and the mode identification information indicates that it is an interactive task mode, then the interactive feature extraction path is adopted, focusing on the interactive features such as the influence of the thickness and uniformity of the paint on the drying time and effect, and the feedback of the temperature and wind speed settings of the drying equipment on the painting effect. If customization focuses on special indicators within a single process, such as customization focuses on the drying time and final wood humidity in the wood drying process, and the mode identification information indicates that it is a non-interactive task mode, the non-interactive feature extraction path is adopted, focusing on the changing law of the drying time under different power settings of the drying equipment, and the relationship between the final wood humidity and the drying time. By accurately determining the feature extraction path according to different node graph construction knowledge models, the server can effectively extract valuable features from the second standard board production monitoring data, and provide accurate data support for subsequent operations such as production parameter trajectory vector extraction and state node graph construction.

[0096] In a possible implementation, step S140 includes: for each second standard board production monitoring data among the multiple second standard board production monitoring data, when the feature extraction path associated with the traversed second standard board production monitoring data is an interactive feature extraction path, for each single-process production monitoring data in the traversed second standard board production monitoring data, respectively determine the interactive link data of each single-process production monitoring data in the traversed second standard board production monitoring data, respectively perform feature representation on each determined interactive link data, and generate an interactive link feature vector corresponding to each single-process production monitoring data in the traversed second standard board production monitoring data.

[0097] Based on the interactive device label corresponding to each single-process production monitoring data in the traversed second standard board production monitoring data, determine the interactive device associated with each single-process production monitoring data in the traversed second standard board production monitoring data, perform feature representation on the interactive device associated with each single-process production monitoring data, and generate the interactive device feature vector corresponding to each single-process production monitoring data in the traversed second standard board production monitoring data.

[0098] Each single-process production monitoring data in the traversed second standard plate production monitoring data is represented by state parameter characteristics, and a state parameter characteristic vector set corresponding to each single-process production monitoring data in the traversed second standard plate production monitoring data is generated.

[0099] Context feature representation is performed on each single-process production monitoring data in the traversed second standard board production monitoring data, and a context feature vector set corresponding to each single-process production monitoring data in the traversed second standard board production monitoring data is generated.

[0100] Based on the interaction link feature vector, interaction device feature vector, state parameter feature vector set and context feature vector set corresponding to each single-process production monitoring data in the traversed second standard board production monitoring data, a production parameter trajectory vector set corresponding to the traversed second standard board production monitoring data is generated.

[0101] Alternatively, for each second standard plate production monitoring data among the multiple second standard plate production monitoring data, when the feature extraction path associated with the traversed second standard plate production monitoring data is a non-interactive feature extraction path, state parameter feature representation is performed on each data node in the traversed second standard plate production monitoring data, and a state parameter feature vector corresponding to each data node in the traversed second standard plate production monitoring data is generated.

[0102] Context feature representation is performed on each data node in the traversed second standard board production monitoring data, and a context feature vector corresponding to each data node in the traversed second standard board production monitoring data is generated.

[0103] The state parameter feature vector and the context feature vector corresponding to the same data node are fused to generate a fused feature vector corresponding to each data node in the traversed second standard board production monitoring data.

[0104] Based on the fused feature vectors respectively corresponding to each data node in the traversed second standard plate production monitoring data, a set of production parameter trajectory vectors corresponding to the traversed second standard plate production monitoring data is generated.

[0105] In this embodiment, the second standard board production monitoring data corresponding to the knowledge model constructed by the associated equipment interaction node graph is taken as an example, because it is associated with the interactive feature extraction path. In furniture production, it is assumed that there is second standard board production monitoring data related to the interaction process of cutting equipment and polishing equipment.

[0106] For each single-process production monitoring data, such as the cutting operation process of the cutting equipment and the grinding operation process of the grinding equipment. First, determine the interactive link data of each single-process production monitoring data in the entire interactive process. For the cutting operation process of the cutting equipment, the interactive link data may include the transfer speed when the cutting equipment transfers the cut standard plate to the grinding equipment, the posture of the standard plate during the transfer (whether it is horizontal, tilted, etc.), and the stop state of the cutting equipment at the moment of transfer. For the grinding operation process of the grinding equipment, the interactive link data may include the initial detection data when the standard plate is received, such as the detected surface roughness of the standard plate, the deviation between the actual size of the standard plate and the expected size, and the startup parameters when starting grinding, such as the initial rotation speed of the grinding wheel, the initial setting value of the grinding pressure, etc. Then, the determined interactive link data are characterized respectively to generate the interactive link feature vector corresponding to each single-process production monitoring data. For example, the transmission speed of the cutting equipment is expressed in vector form, with the speed magnitude being the modulus of the vector and the transmission direction being the direction of the vector; for the initial detection data when the grinding equipment receives the standard plate, the surface roughness, dimensional deviation and other information are encoded into vector form according to certain rules, thereby obtaining the interaction link feature vector corresponding to each single-process production monitoring data.

[0107] Based on the interactive device labels corresponding to each single-process production monitoring data in the traversed second standard board production monitoring data, the interactive devices associated with each single-process production monitoring data are determined. For the cutting operation process, the associated interactive devices are the cutting device itself and the subsequent grinding device; for the grinding operation process, the associated devices are the grinding device and the previous cutting device. The interactive devices associated with each single-process production monitoring data are characterized respectively, and the interactive device feature vectors corresponding to each single-process production monitoring data are generated. For the cutting equipment, the feature representation may include the model of the cutting equipment, the specifications of the cutting tool (such as tool diameter, blade angle, etc.), the service life of the equipment, the power range of the equipment, and other information, which are converted into vector form. For the grinding equipment, it may include the model of the grinding wheel, the particle size of the grinding wheel, the maximum grinding pressure of the equipment, the speed range of the equipment, etc., which are also converted into vector form, so as to obtain the interactive device feature vector corresponding to each single-process production monitoring data.

[0108] Each single-process production monitoring data in the traversed second standard plate production monitoring data is represented by state parameter characteristics, and a state parameter feature vector set corresponding to each single-process production monitoring data is generated. For the cutting operation process of the cutting equipment, the state parameters may include the cutting depth, cutting angle, dynamic changes of the cutting speed during the cutting process, the real-time power of the equipment during cutting, etc. These state parameters are converted into vector form according to certain mathematical methods. For example, the cutting depth and cutting angle can be combined into a two-dimensional vector, and the dynamic changes of the cutting speed and the real-time power can also be combined into a multi-dimensional vector to form a state parameter feature vector set corresponding to the cutting operation process. For the grinding operation process of the grinding equipment, the state parameters include the change of the grinding wheel speed during grinding, the adjustment of the grinding pressure, the change of the surface roughness of the standard plate during grinding, etc., which are also converted into vector form to construct a state parameter feature vector set corresponding to the grinding operation process.

[0109] Each single-process production monitoring data in the traversed second standard board production monitoring data is represented by context features, and a context feature vector set corresponding to each single-process production monitoring data is generated. For the cutting operation process, the context features may include the ambient temperature and humidity of the workshop during cutting, because the temperature and humidity may affect the physical properties of the wood and thus affect the cutting effect; the noise level in the workshop may also affect the operator's operating accuracy. These environmental factors are quantified and converted into vector form to form a context feature vector set corresponding to the cutting operation process. For the grinding operation process of the grinding equipment, the context features may include the dust content of the surrounding air during grinding, because dust may affect the quality of grinding and the life of the equipment; the operating status of adjacent equipment (such as whether the nearby polishing equipment is running and the operating parameters) may also affect the grinding operation. These factors are quantified into vector form to obtain a context feature vector set corresponding to the grinding operation process.

[0110] Finally, based on the interaction link feature vector, interaction device feature vector, state parameter feature vector set and context feature vector set corresponding to each single-process production monitoring data in the traversed second standard board production monitoring data, a production parameter trajectory vector set corresponding to the traversed second standard board production monitoring data is generated. For example, the above-mentioned various types of vectors corresponding to the cutting operation process and the grinding operation process are combined according to certain rules to form a comprehensive vector set, which can comprehensively describe the production parameter trajectory in the interaction process between the cutting equipment and the grinding equipment.

[0111] Let’s look at the second standard board production monitoring data corresponding to the knowledge model constructed by the whole batch node graph. Since it is associated with a non-interactive feature extraction path, in furniture production, the production process of a whole batch of standard boards contains multiple data nodes, such as raw material procurement, production operations of each process, quality inspection, etc.

[0112] The state parameter characteristics of each data node in the traversed second standard board production monitoring data are represented respectively, and the state parameter characteristic vector corresponding to each data node is generated. For the data node of raw material procurement, the state parameters may include the type of wood, the moisture content of wood, the purchase volume of wood, the credit rating of the supplier, etc., and this information is converted into vector form. For the production operation data node of a certain process, such as the assembly process, the state parameters may include the number of assembly workers, the speed of assembly, the defective rate during the assembly process, etc., which are also converted into vector form to obtain the state parameter characteristic vector corresponding to each data node.

[0113] The context features of each data node in the traversed second standard board production monitoring data are represented respectively, and the context feature vector corresponding to each data node is generated. For the raw material procurement data node, the context features may include the market wood price fluctuations at the time of procurement, transportation distance, transportation method (because these factors may affect the cost and quality of raw materials), etc. These factors are quantified into vector form. For the assembly process data node, the context features may include the spatial layout of the assembly workshop (whether the reasonable layout may affect the assembly efficiency), the personnel flow of the workshop at that time (large personnel flow may affect the stability of assembly), etc. These factors are converted into vector form to obtain the context feature vector corresponding to each data node.

[0114] The state parameter feature vector and context feature vector corresponding to the same data node are fused to generate the fused feature vector corresponding to each data node in the traversed second standard board production monitoring data. For example, for the raw material procurement data node, the state parameter feature vector representing the wood type, moisture content, procurement volume, and supplier credit rating is fused with the context feature vector representing market price fluctuations, transportation distance, and transportation method according to a certain weighted summation or other fusion rules to obtain the fused feature vector corresponding to the data node. The same operation is performed for other data nodes.

[0115] Based on the fused feature vectors corresponding to each data node in the traversed second standard board production monitoring data, a set of production parameter trajectory vectors corresponding to the traversed second standard board production monitoring data is generated. This set can comprehensively describe the production parameter trajectory of the entire batch of standard boards in the production process by comprehensively considering the characteristics of each data node from the perspective of the entire batch.

[0116] For the second standard board production monitoring data corresponding to the knowledge model constructed by the single process node graph (such as the wood drying process), if the associated non-interactive feature extraction path is used, for each data node in the wood drying process, such as the start-up of the drying equipment, the temperature and humidity control during the drying process, and the drying time.

[0117] The state parameter characteristics of each data node are represented. For example, the state parameters of the drying equipment startup data node may include the equipment power at startup, the initial temperature of the equipment, etc., which are converted into vector form. The state parameters of the temperature and humidity control data nodes during the drying process are the actual temperature and humidity values ​​and their fluctuations, which are also converted into vector form. The context characteristics of each data node are represented, such as the ventilation conditions of the workshop when the drying equipment is started, the heat dissipation conditions of the surrounding equipment, etc., which are converted into vector form. Then the state parameter feature vector and the context feature vector of the same data node are fused to obtain the fused feature vector corresponding to each data node. Finally, based on these fused feature vectors, a set of production parameter trajectory vectors of the second standard board production monitoring data corresponding to the wood drying process is generated.

[0118] The second standard board production monitoring data corresponding to the multi-source data node graph construction knowledge model and the customized node graph construction knowledge model also extracts production parameter trajectory vectors according to the associated feature extraction path (interactive feature extraction path or non-interactive feature extraction path) in a similar manner to generate the corresponding production parameter trajectory vector sets. Through such a comprehensive and detailed production parameter trajectory vector extraction process, the server can mine rich information from different second standard board production monitoring data, providing strong data support for subsequent node graph construction and abnormal state link estimation and other operations.

[0119] In a possible implementation, the method is performed by a state node graph construction model, and the training step of the state node graph construction model includes: Step A110, obtaining first template standard plate production monitoring data and a plurality of prior state node graphs corresponding to the first template standard plate production monitoring data; Step A120, for a plurality of preset node graph construction knowledge patterns, respectively inject identification attribute information into the first template standard plate production monitoring data, and generate a plurality of second template standard plate production monitoring data corresponding to the plurality of node graph construction knowledge patterns. The identification attribute information in each of the second template standard plate production monitoring data includes pattern identification information reflecting the corresponding node graph construction knowledge pattern; Step A130, determining a plurality of feature extraction paths associated with the plurality of second template standard plate production monitoring data according to the pattern identification information respectively included in the plurality of second template standard plate production monitoring data; Step A140, for each of the plurality of second template standard plate production monitoring data, respectively extracting a production parameter trajectory vector according to the associated feature extraction path, and generating a set of template production parameter trajectory vectors corresponding to the plurality of node graph construction knowledge patterns respectively; Step A150, performing node graph construction on the template production parameter trajectory vector sets corresponding to the plurality of node graph construction knowledge patterns respectively, to generate predicted state node graphs corresponding to the plurality of node graph construction knowledge patterns respectively; Step A160, training the state node graph construction model according to the predicted state node graphs and the multiple prior state node graphs corresponding to the multiple node graph construction knowledge patterns, and generating a trained state node graph construction model.

[0120] In this embodiment, the server first obtains the production monitoring data of the first template standard board and a plurality of prior state node graphs corresponding to these data. The production monitoring data of the first template standard board contains various information in the production process of the home standard board. For example, in furniture production, it covers the monitoring data of each process from the selection, cutting, grinding, assembly to the final surface treatment of wood raw materials. The prior state node graph is constructed based on past experience or expert knowledge, reflecting the ideal state node relationship of each link in the standard board production process under normal production conditions. For example, in the normal cutting process, the ideal relationship between cutting depth, cutting speed and wood type will be clearly indicated in the prior state node graph.

[0121] For the preset multiple node graph construction knowledge patterns, the server begins to inject identification attribute information into the first template standard board production monitoring data to generate multiple second template standard board production monitoring data. For example, the preset node graph construction knowledge pattern includes a single process node graph construction knowledge pattern, a whole batch node graph construction knowledge pattern, and an associated device interaction node graph construction knowledge pattern. Taking the single process node graph construction knowledge pattern as an example, if it is a wood cutting process, the server will inject the corresponding identification attribute information. The identification attribute information in each second template standard board production monitoring data contains the pattern identification information reflecting the corresponding node graph construction knowledge pattern, such as the single process pattern identification information "WOOD_CUTTING_SINGLE_PROCESS" for the wood cutting process, as well as the initial data segment start identification, end identification, etc. For the whole batch node graph construction knowledge pattern, if a batch is to produce 1,000 standard boards, the server will inject identification for each link data in the whole batch production process, and the pattern identification information may be "WHOLE_BATCH_PRODUCTION", and add corresponding start, end, etc. identifications, thereby generating the corresponding second template standard board production monitoring data. When constructing the knowledge model of the associated equipment interaction node graph, for example, the interaction between the cutting equipment and the grinding equipment, the pattern identification information such as "CUTTING_SANDING_INTERACTION" and other related identifications will be injected to clarify the scope and nature of the interaction-related data.

[0122] According to the pattern identification information respectively included in the multiple second template standard plate production monitoring data, the server determines multiple feature extraction paths associated with the multiple second template standard plate production monitoring data. For example, for the second template standard plate production monitoring data corresponding to the knowledge model constructed by the associated device interactive node graph, since its pattern identification information indicates that it is related to device interaction, the feature extraction path associated with it is determined to be an interactive feature extraction path used to extract features from the interactive production monitoring data of the associated device. In the process of interaction between the cutting device and the grinding device, the interactive feature extraction path will focus on the features related to the interactive link data such as the speed when the cutting device transfers the standard plate to the grinding device and the posture of the standard plate. For the second template standard plate production monitoring data corresponding to the knowledge model constructed by the entire batch node graph, the pattern identification information indicates that it is a non-interactive task mode, and the determined feature extraction path is a non-interactive feature extraction path used to extract features from the interactive production monitoring data of non-associated devices, and will focus on the total raw material consumption and total production time in the entire batch production process.

[0123] For each of the multiple second template standard plate production monitoring data, the server extracts the production parameter trajectory vector according to the associated feature extraction path, and generates a set of template production parameter trajectory vectors corresponding to the knowledge model constructed by multiple node graphs. Taking the second template standard plate production monitoring data corresponding to the knowledge model constructed by the associated device interaction node graph as an example, for each single-process production monitoring data, such as the cutting operation of the cutting equipment and the grinding operation of the grinding equipment. Determine the link data of the cutting operation in the interaction, such as the speed and stop state when the standard plate is transferred, perform feature representation to generate the interactive link feature vector; determine the associated cutting equipment and grinding equipment based on the interactive equipment label, perform feature representation on the equipment to generate the interactive equipment feature vector; perform state parameter feature representation on the cutting operation, such as cutting depth, angle, speed change, etc. to generate a state parameter feature vector set, and perform context feature representation, such as workshop temperature, humidity, etc. to generate a context feature vector set, and finally generate a template production parameter trajectory vector set corresponding to the second template standard plate production monitoring data based on these vectors. For the second template standard plate production monitoring data corresponding to the knowledge model of the entire batch node graph, state parameter feature representation and context feature representation are performed for each data node such as raw material procurement and process operations respectively. After fusion, a fused feature vector corresponding to each data node is generated, and then the corresponding template production parameter trajectory vector set is generated.

[0124] The server constructs node graphs for template production parameter trajectory vector sets corresponding to the knowledge patterns of multiple node graph construction, and generates predicted state node graphs corresponding to the knowledge patterns of multiple node graph construction. Taking the template production parameter trajectory vector set corresponding to the knowledge pattern of single-process node graph construction as an example, the initial state node is generated according to the vectors in the set and the corresponding initial state node identifier, such as the "start cutting" identifier of the cutting process. Then, a multi-stage graph construction process is performed, the vectors are classified and marked, and a preliminary association relationship is established between the initial state node and the first vector, such as the association between the start of cutting and the start parameter vector of the cutting equipment. If the initial state node represents the production start state, the vector related to the start is searched to further establish an advanced association relationship, such as the power vector at the start of cutting. Based on the preliminary association relationship, the associated vector is analyzed to find other logically associated vectors, and the initial state node is expanded to create an intermediate state node structure, such as the tool wear state node during the cutting process. After the redundancy optimization of the intermediate state node structure, the time information of the vector is determined, and the association relationship of the intermediate state nodes is sorted and expanded according to the time to obtain the derived state node structure. Finally, the predicted state node graph is generated by integrating the initial state node and the derived state node. Similar operations are performed on the template production parameter trajectory vector sets corresponding to other node graph construction knowledge patterns.

[0125] Finally, based on the predicted state node graphs and multiple prior state node graphs corresponding to the knowledge patterns of multiple node graphs, the state node graph construction model is trained to generate a trained state node graph construction model. For example, the predicted state node graph is compared with the prior state node graph. If there is a large deviation between the cutting depth displayed by the predicted state node graph and the ideal cutting depth in the prior state node graph during the cutting process, the model will adjust its own parameters according to this difference, so that the state node graph can be predicted more accurately in subsequent constructions. By comparing and adjusting the predicted results corresponding to the knowledge patterns of each node graph construction with the prior results, the model is continuously optimized, and finally a trained state node graph construction model is obtained. This model can more accurately construct the state node graph for the production monitoring data of home standard panels.

[0126] For example, in a possible implementation, step S150 includes: Step S151, for each production parameter trajectory vector set, an initial state node is generated according to the traversed production parameter trajectory vector set and the initial state node identifier corresponding to the traversed production parameter trajectory vector set. In the present embodiment, firstly, taking the wood cutting process in furniture production as an example, it is assumed that the corresponding production parameter trajectory vector set contains relevant vectors such as the start-up parameters of the cutting equipment, the initial state of the tool, and the wood feed speed. If the initial state node identifier is "cutting process start", the server generates an initial state node based on this identifier and the relevant production parameter trajectory vector set. This initial state node contains information such as the initial power of the cutting equipment, the initial position of the tool, the initial speed of the wood feed, etc., representing the state at the beginning of the wood cutting process; Step S152, performing a graph construction process in multiple stages based on the initial state node and the traversed production parameter trajectory vector set to generate multiple derived state nodes; Step S153, synthesizing the initial state node and the multiple derived state nodes, generating a state node graph construction result corresponding to the traversed production parameter trajectory vector set. For example, in the example of the wood cutting process, the initial state node "cutting process start" and the various nodes in the derived state node structure are combined to form a complete state node graph construction result. This state node graph clearly shows the state transition of the wood cutting process from start to each stage and the logical relationship between different states, including the mutual influence relationship of various parameters at different time points, such as the change and mutual correlation of parameters such as power, feed speed, tool speed, cutting depth, etc. in the cutting process over time.

[0127] For other production processes or node graph construction knowledge patterns corresponding to the production parameter trajectory vector set, such as furniture assembly process, whole batch production monitoring or related equipment interaction, similar steps are also followed. Taking the furniture assembly process as an example, the initial state node may be "assembly process starts", and the production parameter trajectory vector set includes vectors such as the number of workers, the initial state of assembly tools, and the initial placement of parts. According to the above graph construction processing steps, the vectors are first classified and marked and a preliminary association relationship is established, such as associating the number of workers vector with the initial state node based on the logic that the workers are in place when the assembly starts; then the initial state node attributes are analyzed, and the vectors related to the start are found to establish an advanced association relationship; then the initial state node is expanded based on the preliminary association relationship to create an intermediate state node, such as creating an intermediate state node such as "assembly process starts, the number of workers affects the assembly efficiency, and the tool status affects the assembly accuracy" according to the number of workers and the use of assembly tools; after performing redundancy optimization, the time information of each vector is determined, and the intermediate state node is expanded according to the time sorting to obtain the derived state node structure, and finally the initial state node and the derived state node are combined to obtain the state node graph construction result of the assembly process. The whole batch production monitoring starts from the macroscopic perspective of the whole batch production. The initial state node may be "whole batch production starts". The production parameter trajectory vector set includes vectors such as raw material batch information and the overall arrangement of each process. Similarly, these steps are followed to construct the state node graph construction result that reflects the state transition and logical relationship of the whole batch production process. In terms of the interaction of related equipment, taking the interaction between cutting equipment and grinding equipment as an example, the initial state node is "cutting-grinding equipment interaction starts". The production parameter trajectory vector set includes vectors such as signal transmission between equipment and standard board transmission status. Through the above steps, the state node graph construction result that can accurately reflect the state changes and logical relationships in the process of equipment interaction is constructed. In this way, for different production situations and node graph construction knowledge models, the state node graph construction result can be accurately constructed, providing an effective data basis for subsequent abnormal state link estimation and other operations.

[0128] For example, in a possible implementation, step S152 includes: Step S1521, classify and mark each production parameter trajectory vector in the production parameter trajectory vector set, and establish a preliminary association relationship between the initial state node and the first production parameter trajectory vector in the production parameter trajectory vector set, and the preliminary association relationship is based on the starting logic of the production process. For example, in the wood cutting process, each vector is classified and marked, such as marking the starting power vector of the cutting equipment as "equipment starting parameter-power", marking the initial position vector of the tool as "tool state-initial position", etc. Assuming that the first vector in the production parameter trajectory vector set is the starting power vector of the cutting equipment, since the cutting process first involves starting the equipment and setting the power, based on the starting logic of the production process, a preliminary association relationship is established between the initial state node "cutting process start" and this starting power vector; Step S1522, analyze the attributes of the initial state node, if the initial state node represents the start-up state of production, then search the production parameter vector related to the start-up state in the production parameter trajectory vector set, establish an advanced association relationship between the production parameter vector and the initial state node, thereby preliminarily constructing a mapping relationship based on production logic between the initial state node and part of the production parameter trajectory vector. For the wood cutting process, in addition to the start-up power vector, the production parameter vector related to the start-up state may also include the start-up current vector of the cutting equipment, the initial speed vector of the tool, etc. Establish an advanced association relationship between these vectors and the initial state node "cutting process start", so that a mapping relationship based on production logic is preliminarily constructed. For example, when the cutting process is started, the relationship between the power, current and speed of the equipment and the speed of the tool can be clearly defined; Step S1523, based on the preliminary association relationship, each associated production parameter trajectory vector is analyzed to find other production parameter trajectory vectors in the production parameter trajectory vector set that are logically associated with the associated production parameter trajectory vector, and based on the search result, the initial state node is gradually expanded to continuously create representations of intermediate state nodes based on the initial state node, and generate an intermediate state node structure, wherein the intermediate state node contains relevant information in the extended association relationship. For example, for the associated cutting equipment startup power vector, it is found that the setting of power will affect the feeding speed of wood, so the wood feeding speed vector is logically associated with the startup power vector. Based on this relationship, an intermediate state node is created based on the initial state node "cutting process startup", and this intermediate state node may be expressed as "cutting process startup, power setting affects feeding speed", which contains the association information between startup power and feeding speed. For other associated vectors, such as the initial rotation speed vector of the tool, it is found that it has a logical association with the cutting depth vector (the rotation speed affects the cutting depth), so another intermediate state node "cutting process starts, tool rotation speed affects cutting depth" is created. These intermediate state nodes together constitute the intermediate state node structure; Step S1524, perform redundancy optimization on the intermediate state node structure to obtain an optimized intermediate state node structure. For example, in the intermediate state node created earlier, there may be two nodes respectively representing "the cutting process is started, the power setting affects the feed speed" and "the cutting process is started, and the feed speed is affected by the power and the tool speed", and there are repeated parts about the relationship between power and feed speed. Through redundancy optimization, the repeated association relationships are removed to obtain an optimized intermediate state node structure. For example, the structure of "the cutting process is started, the power setting affects the feed speed, and the tool speed affects the cutting depth" is retained, so that the intermediate state node structure is more concise and clear, and accurately reflects the production logic relationship; Step S1525, determine the time information corresponding to each production parameter trajectory vector in the production parameter trajectory vector set, and according to the time information, time-sort the association relationship in the intermediate state nodes in the optimized intermediate state node structure, and according to the advancement of the time sorting sequence, expand the intermediate state nodes to obtain a derived state node structure carrying the time sorting sequence information.

[0129] In the wood cutting process, each vector has corresponding time information. For example, the time corresponding to the power vector of the cutting equipment is the moment t0 when the process starts. As time goes by, the tool starts to cut into the wood at moment t1, and the tool's cutting angle vector starts to work. According to these time information, the association relationship in the optimized intermediate state node structure is time-sorted. First, the power setting affects the feed speed (time period t0 - t1), and then the tool cuts into the wood at moment t1, and the tool speed affects the cutting depth. According to the advancement of this time sorting sequence, the intermediate state nodes are expanded to obtain a derived state node structure carrying time sorting sequence information such as "t0 - t1: cutting process starts, power setting affects feed speed; t1: tool cuts into wood, tool speed affects cutting depth".

[0130] For example, in a possible implementation, step S150 further includes: Step S154, standardize the state node graph corresponding to each state node graph construction result to unify the representation of the state nodes in the state node graph and the description format of the relationship between the state nodes. In this embodiment, in the production of home standard boards, for example, for the state node graph of the wood cutting process, different engineers or systems may use different representation methods when constructing the node graph. Some may describe the state node in detail with words, such as "the cutting device is in a high-speed cutting state, and the cutting depth is 3 cm", and some may be represented by codes or specific symbols. Through standardization, these different representation methods are unified. For example, it is uniformly stipulated that the state node is represented by a concise code and number, such as "CUT_01" represents a state of the cutting device, followed by a specific parameter value. For the relationship between the state nodes, the format is also unified. For example, if it is a causal relationship, it is uniformly represented by an arrow plus a specific symbol, such as "→(C)" represents a causal relationship, where "C" represents a specific causal relationship content, such as "power increase→(C) cutting speed increase", where "power increase" is the preceding node, "cutting speed increase" is the following node, and "power increase leads to cutting speed increase" is the specific description of the relationship. In this way, whether it is the state node diagram of the wood cutting process in the single process node diagram construction knowledge mode, or the state node diagram of the whole batch of standard boards in the whole batch node diagram construction knowledge mode, or the state node diagram of the cutting equipment and grinding equipment interaction in the associated equipment interaction node diagram construction knowledge mode, they can all be presented in a unified format, which is convenient for subsequent analysis operations; Step S155, analyze each state node and the relationship between each state node in the standardized state node diagram, wherein specifically search for the state nodes corresponding to each link in the production process, determine the preceding node and the succeeding node of each state node in the production process for each state node, establish a complete node relationship mapping based on the production process sequence, and generate a state node diagram structure carrying the production process mapping relationship. Next, taking the assembly process in furniture production as an example, in the standardized state node diagram, there are state nodes such as "parts sorting completed", "parts assembly started", "assembly tools ready", etc. For the state node "parts assembly started", it is found through analysis that its preceding node is "parts sorting completed", because only after the parts sorting is completed can the assembly start; the succeeding node is "preliminary assembly completed", because after the parts assembly starts, a series of operations will reach the state of preliminary assembly completed. In this way, each state node is analyzed, its preceding and succeeding node relationships are determined, and a complete node relationship mapping based on the production process sequence is established. For example, in the state node diagram of the production of a batch of standard boards, for the state node "the batch of standard boards enters the grinding process", the preceding node is "the batch of standard boards completes the cutting process", and the succeeding node is "the batch of standard boards begins the surface treatment process". Through such a comprehensive analysis, a state node diagram structure carrying the production process mapping relationship is generated. This state node diagram structure clearly shows the sequence of each link in the entire production process and the state conversion relationship; Step S156, based on the established node relationship mapping, extract the normal state link under normal production conditions from the state node graph, analyze the characteristic patterns in the normal state link, combine the key nodes, relationships and the characteristic patterns of the normal state link, construct a characteristic model for describing the normal state link, compare and analyze the actual state link in each state node graph with the characteristic model of the normal state link, and preliminarily screen out the candidate abnormal links in the state node graph. Then, in the wood cutting process, the normal state link may be "cutting equipment starts → (C) cutting tool operates normally → (C) wood feeding is stable → (C) cutting depth and speed meet the standards → (C) cutting is completed". Analyze the characteristic patterns in this normal state link, for example, the key node "cutting tool operates normally" may have a characteristic pattern that the rotation speed of the tool fluctuates within a certain range, and the fluctuation amplitude is small; the characteristic pattern of "wood feeding is stable" may be that the rate of change of the feeding speed is within a certain threshold. Combine these key nodes, relationships and characteristic patterns to construct a characteristic model for describing the normal state link. For the production of a batch of standard boards, the normal state link may involve the continuous normal operation of multiple processes, such as "Qualified raw material procurement → (C) Normal batch cutting process → (C) Normal batch polishing process → (C) Normal batch assembly process → (C) Qualified batch product quality inspection", and analyze the characteristic pattern of normal operation of each process. For example, the characteristic pattern of normal batch cutting process may be that the cutting size deviation of the whole batch of standard boards is within the allowable range. This information is constructed into a characteristic model of the normal state link of the production of a batch of standard boards. Later, in furniture production, if the actual state link in the state node diagram of a batch of standard boards is "Qualified raw material procurement → (C) Too large cutting size deviation of some standard boards in the whole batch cutting process → (C) Normal batch polishing process → (C) Normal batch assembly process → (C) Some unqualified batch product quality inspection", compared with the characteristic model of the normal state link, it is found that "Excessive cutting size deviation of some standard boards in the whole batch cutting process" does not meet the characteristics of the normal cutting process in the normal state link (the cutting size deviation is within the allowable range), then the link containing this node may be preliminarily screened as a candidate abnormal link. For the node graph of the interaction state of associated equipment, such as the actual state link of the interaction between the cutting equipment and the grinding equipment, if "when the cutting equipment transfers the standard board to the grinding equipment, the transfer speed is unstable and exceeds the normal fluctuation range → (C) the grinding equipment frequently adjusts parameters when starting grinding", compared with the normal interaction state link feature model, this candidate abnormal link may be preliminarily screened as a candidate abnormal link; Step S157, for the candidate abnormal links preliminarily screened out, refer to the historical production monitoring data for verification, and generate a verified target abnormal link. For the candidate abnormal links mentioned above in which the cutting size deviation is too large in the cutting process, the server will refer to the historical production monitoring data. If the historical data shows that under the same production conditions (such as the same raw materials, equipment settings, etc.), the cutting size deviation rarely exceeds a certain range under normal circumstances, and the current batch exceeds the range, then this candidate abnormal link is verified as a target abnormal link. For candidate abnormal links in which the transmission speed is unstable in the interaction between the cutting equipment and the grinding equipment and the grinding equipment frequently adjusts the parameters, check similar equipment interaction situations in the historical data. If it is found that the transmission speed was stable during normal interaction in the past and the grinding equipment rarely needed to adjust the parameters frequently, then this link is also verified as a target abnormal link; Step S158, for the verified target abnormal link, a root cause analysis is performed in the graph structure of the state node graph, tracing back to the abnormal state node that first appeared in the abnormal state in the target abnormal link, and analyzing the abnormal cause that caused the abnormal state node to be abnormal, and at the same time evaluating the association relationship information between the target abnormal link and other state links; Step S159: Determine the severity of each target abnormal link based on the abnormal cause and the association relationship information between the target abnormal link and other status links.

[0131] Finally, for the target abnormal link with excessive cutting size deviation in the cutting process, tracing back in the state node graph structure, it is found that the abnormal state node with the earliest abnormal state may be "cutting tool wear". The reason for the abnormality of this abnormal state node may be that the tool has been used for too long and has not been replaced in time, or the hardness of the cut wood exceeds expectations, resulting in accelerated tool wear. At the same time, the association relationship information between this target abnormal link and other state links is evaluated. For example, due to excessive cutting size deviation, the workload of the subsequent grinding process may increase, affecting the efficiency of the grinding process, and then may affect the production progress of the entire batch of products. Based on these abnormal reasons and association relationship information, if this problem is not solved in time, a large number of defective products will be produced in the entire batch of products, then the severity of this target abnormal link is high. For the target abnormal link in the interaction between the cutting equipment and the grinding equipment, the root cause analysis found that the earliest abnormal state node may be the failure of the transmission device of the cutting equipment, resulting in unstable transmission speed. The abnormal cause may be the damage of a component of the transmission device. This target abnormal link is closely related to the grinding process. Because the unstable transmission speed causes the grinding equipment to frequently adjust parameters, affecting the grinding quality and efficiency, if it continues, it will affect the quality of the entire batch of products, so the severity is also high. Through such comprehensive analysis, the server can accurately estimate the abnormal status links in the production monitoring process of the target standard board, so as to take effective measures for production optimization and problem solving.

[0132] Figure 2 The online intelligent detection and analysis system 100 for standard board production based on the industrial Internet of Things shown includes: a processor 1001 and a memory 1003. Among them, the processor 1001 and the memory 1003 are connected, such as through a bus 1002. Optionally, the online intelligent detection and analysis system 100 for standard board production based on the industrial Internet of Things may also include a transceiver 1004, which can be used for data interaction between the server and other servers, such as data transmission and / or data reception. It should be noted that in actual scheduling, the transceiver 1004 is not limited to one, and the structure of the online intelligent detection and analysis system 100 for standard board production based on the industrial Internet of Things does not constitute a limitation on the embodiments of the present application.

[0133] Processor 1001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. Processor 1001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0134] The bus 1002 may include a path to transmit information between the above components. The bus 1002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus 1002 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 2 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0135] The memory 1003 may be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), magnetic disk storage medium, other magnetic storage devices, or any other medium that can be used to carry or store program code and can be read by a computer, without limitation herein.

[0136] The memory 1003 is used to store program codes for executing the embodiments of the present application, and the execution is controlled by the processor 1001. The processor 1001 is used to execute the program codes stored in the memory 1003 to implement the steps shown in the above method embodiments.

[0137] In addition, an embodiment of the present application also provides a readable storage medium, in which computer executable instructions are pre-installed. When the processor executes the computer executable instructions, the above-mentioned online intelligent detection and analysis method for standard board production based on industrial Internet of Things is implemented.

[0138] Similarly, it should be noted that in order to simplify the description disclosed in this application, thereby helping to understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present application, multiple features are sometimes combined into one embodiment, drawings, or descriptions thereof. Similarly, it should be noted that in order to simplify the description disclosed in this application, thereby helping to understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present application, multiple features are sometimes combined into one embodiment, drawings, or descriptions thereof.

Claims

1. An online intelligent detection and analysis method for standard board production based on industrial Internet of Things, characterized in that: The method comprises: obtaining first standard board production monitoring data recording the target standard board production monitoring process in an industrial Internet of Things platform, and determining a plurality of node graph construction knowledge patterns for constructing a state node graph for the first standard board production monitoring data; for the plurality of node graph construction knowledge patterns, injecting identification attribute information into the first standard board production monitoring data respectively to generate a plurality of second standard board production monitoring data; the second standard board production monitoring data in the plurality of second standard board production monitoring data respectively correspond to the node graph construction knowledge patterns in the plurality of node graph construction knowledge patterns, and the identification attribute information in each of the second standard board production monitoring data includes pattern identification information reflecting the corresponding node graph construction knowledge pattern; according to the plurality of second standard board production monitoring data The pattern identification information included in the production monitoring data is used to determine multiple feature extraction paths associated with the multiple second standard board production monitoring data; for each of the multiple second standard board production monitoring data, production parameter trajectory vector extraction is performed according to the associated feature extraction path to generate a set of production parameter trajectory vectors corresponding to the multiple node graph construction knowledge patterns; node graphs are constructed for the sets of production parameter trajectory vectors corresponding to the multiple node graph construction knowledge patterns to generate state node graph construction results corresponding to the multiple node graph construction knowledge patterns, and the abnormal state link of the target standard board production monitoring process is estimated based on the state node graph construction results corresponding to the multiple node graph construction knowledge patterns.

2. The online intelligent detection and analysis method for standard board production based on industrial Internet of Things according to claim 1 is characterized in that: The method of determining multiple node graph construction knowledge patterns for constructing a state node graph for the first standard board production monitoring data includes: generating a node graph construction knowledge pattern sequence based on a state node graph construction instruction for the first standard board production monitoring data; and based on a hit instruction for the node graph construction knowledge pattern sequence, using the multiple node graph construction knowledge patterns hit by the hit instruction as the multiple node graph construction knowledge patterns for constructing a state node graph for the first standard board production monitoring data.

3. The online intelligent detection and analysis method for standard board production based on industrial Internet of Things according to claim 1 is characterized in that: The method of constructing knowledge patterns for the multiple node graphs and injecting identification attribute information into the first standard plate production monitoring data respectively to generate multiple second standard plate production monitoring data includes: for each node graph construction knowledge pattern in the multiple node graph construction knowledge patterns, obtaining identification attribute information corresponding to the traversed node graph construction knowledge pattern; injecting the identification attribute information corresponding to the traversed node graph construction knowledge pattern into the first standard plate production monitoring data to generate second standard plate production monitoring data corresponding to the traversed node graph construction knowledge pattern.

4. The online intelligent detection and analysis method for standard board production based on industrial Internet of Things according to any one of claims 1 to 3, characterized in that: The multiple node graph construction knowledge patterns include at least one of a single-process node graph construction knowledge pattern, a whole batch node graph construction knowledge pattern, an associated device interaction node graph construction knowledge pattern, a multi-source data node graph construction knowledge pattern or a customized node graph construction knowledge pattern; the node graph construction instance of the single-process node graph construction knowledge pattern is used to construct a state node graph for single-process production monitoring data; the node graph construction instance of the whole batch node graph construction knowledge pattern is used to construct a state node graph for a whole batch of production monitoring data; the node graph construction instance of the associated device interaction node graph construction knowledge pattern is used to construct a state node graph for associated device interaction production monitoring data; the node graph construction instance of the multi-source data node graph construction knowledge pattern is used to construct a state node graph for standard board production monitoring data extracted from multi-source data; the node graph construction instance of the customized node graph construction knowledge pattern is used to construct a customized state node graph for standard board production monitoring data.

5. The online intelligent detection and analysis method for standard board production based on industrial Internet of Things according to claim 3 is characterized in that: The identification attribute information corresponding to the traversed node graph construction knowledge pattern includes an initial data segment start identifier, an initial data segment end identifier, an end identifier and a pattern identifier information reflecting the traversed node graph construction knowledge pattern; the identification attribute information corresponding to the traversed node graph construction knowledge pattern is injected into the first standard board production monitoring data to generate the second standard board production monitoring data corresponding to the traversed node graph construction knowledge pattern, including: when the number of single-process production monitoring data included in the first standard board production monitoring data is not greater than the set number, the pattern identifier information reflecting the traversed node graph construction knowledge pattern is injected into the front end of the first standard board production monitoring data; after the pattern identifier information reflecting the traversed node graph construction knowledge pattern and before the initial data segment of the first standard board production monitoring data, the initial data segment start identifier is injected; the initial data segment end identifier is injected into the back end of the first standard board production monitoring data, and the end identifier is injected after the injected initial data segment end identifier to generate the second standard board production monitoring data corresponding to the traversed node graph construction knowledge pattern.

6. The online intelligent detection and analysis method for standard board production based on industrial Internet of Things according to claim 3 is characterized in that: The identification attribute information corresponding to the traversed node graph construction knowledge pattern includes an initial data segment start identifier, an initial data segment end identifier, a data segment segmentation identifier, and a pattern identification information reflecting the traversed node graph construction knowledge pattern; the identification attribute information corresponding to the traversed node graph construction knowledge pattern is injected into the first standard board production monitoring data to generate the second standard board production monitoring data corresponding to the traversed node graph construction knowledge pattern, including: when the number of single-process production monitoring data included in the first standard board production monitoring data is greater than a set number, injecting the node graph construction reflecting the traversed knowledge pattern into the front end of the first standard board production monitoring data. The pattern identification information of the knowledge pattern; after the pattern identification information reflecting the traversed node graph to construct the knowledge pattern and before the initial data segment of the first standard board production monitoring data, the initial data segment start identifier is injected; at the back end of the initial single-process production monitoring data in the first standard board production monitoring data, the initial data segment end identifier is injected; the remaining single-process production monitoring data except the initial single-process production monitoring data in the first standard board production monitoring data is determined, and the data segment segmentation identifier is injected between every two associated remaining single-process production monitoring data to generate the second standard board production monitoring data corresponding to the knowledge pattern constructed by the traversed node graph.

7. The online intelligent detection and analysis method for standard board production based on industrial Internet of Things according to claim 1 is characterized in that: The determining of multiple feature extraction paths associated with the multiple second standard board production monitoring data includes: for each second standard board production monitoring data among the multiple second standard board production monitoring data, when the node graph construction knowledge pattern reflected by the pattern identification information included in the traversed second standard board production monitoring data belongs to the interactive task mode, determining that the feature extraction path associated with the traversed second standard board production monitoring data is an interactive feature extraction path used to extract features from the interactive production monitoring data of associated equipment; when the node graph construction knowledge pattern reflected by the pattern identification information included in the traversed second standard board production monitoring data belongs to the non-interactive task mode, determining that the feature extraction path associated with the traversed second standard board production monitoring data is a non-interactive feature extraction path used to extract features from the interactive production monitoring data of non-associated equipment.

8. The online intelligent detection and analysis method for standard board production based on industrial Internet of Things according to claim 7 is characterized in that: For each of the plurality of second standard board production monitoring data, production parameter trajectory vector extraction is performed respectively according to the associated feature extraction path, and a set of production parameter trajectory vectors corresponding to the plurality of node graph construction knowledge patterns are generated, including: for each of the plurality of second standard board production monitoring data, when the feature extraction path associated with the traversed second standard board production monitoring data is an interactive feature extraction path, for each single-process production monitoring data in the traversed second standard board production monitoring data, interactive link data of each single-process production monitoring data in the traversed second standard board production monitoring data is determined respectively, and the interactive link data of each single-process production monitoring data in the traversed second standard board production monitoring data is determined. The determined data of each interaction link are characterized respectively, and the interaction link feature vector corresponding to each single-process production monitoring data in the traversed second standard board production monitoring data is generated; based on the interaction device label corresponding to each single-process production monitoring data in the traversed second standard board production monitoring data, the interaction device associated with each single-process production monitoring data in the traversed second standard board production monitoring data is determined, and the interaction devices associated with each single-process production monitoring data are characterized respectively, and the interaction device feature vector corresponding to each single-process production monitoring data in the traversed second standard board production monitoring data is generated; for each single-process in the traversed second standard board production monitoring data The production monitoring data are represented by state parameter characteristics respectively, and a state parameter feature vector set corresponding to each single-process production monitoring data in the traversed second standard board production monitoring data is generated; each single-process production monitoring data in the traversed second standard board production monitoring data is represented by context characteristics respectively, and a context feature vector set corresponding to each single-process production monitoring data in the traversed second standard board production monitoring data is generated; based on the interaction link feature vector, interaction device feature vector, state parameter feature vector set and context feature vector set corresponding to each single-process production monitoring data in the traversed second standard board production monitoring data, a context feature vector set corresponding to the traversed second standard board production monitoring data is generated. A set of production parameter trajectory vectors corresponding to the data; or, for each second standard board production monitoring data in the multiple second standard board production monitoring data, when the feature extraction path associated with the traversed second standard board production monitoring data is a non-interactive feature extraction path, each data node in the traversed second standard board production monitoring data is respectively represented by state parameter features, and a state parameter feature vector corresponding to each data node in the traversed second standard board production monitoring data is generated; each data node in the traversed second standard board production monitoring data is respectively represented by context features, and a context feature vector corresponding to each data node in the traversed second standard board production monitoring data is generated;The state parameter feature vector and the context feature vector corresponding to the same data node are fused to generate a fused feature vector corresponding to each data node in the traversed second standard board production monitoring data; based on the fused feature vector corresponding to each data node in the traversed second standard board production monitoring data, a set of production parameter trajectory vectors corresponding to the traversed second standard board production monitoring data is generated. ; 9. The online intelligent detection and analysis method for standard board production based on industrial Internet of Things according to claim 1 is characterized in that: The method is executed by a state node graph construction model, and the training step of the state node graph construction model includes: obtaining first template standard plate production monitoring data and multiple prior state node graphs corresponding to the first template standard plate production monitoring data; constructing knowledge patterns for multiple preset node graphs, injecting identification attribute information into the first template standard plate production monitoring data respectively, and generating multiple second template standard plate production monitoring data corresponding to the multiple node graph construction knowledge patterns; the identification attribute information in each of the second template standard plate production monitoring data includes pattern identification information reflecting the corresponding node graph construction knowledge pattern; determining the multiple second template standard plate production monitoring data according to the pattern identification information respectively included in the multiple second template standard plate production monitoring data multiple feature extraction paths associated with production monitoring data; for each of the multiple second template standard plate production monitoring data, extracting production parameter trajectory vectors according to the associated feature extraction paths, generating sets of template production parameter trajectory vectors corresponding to the multiple node graph construction knowledge patterns; performing node graph construction on the sets of template production parameter trajectory vectors corresponding to the multiple node graph construction knowledge patterns, generating predicted state node graphs corresponding to the multiple node graph construction knowledge patterns; training the state node graph construction model according to the predicted state node graphs corresponding to the multiple node graph construction knowledge patterns and the multiple prior state node graphs, generating a trained state node graph construction model.

10. An online intelligent detection and analysis system for standard board production based on industrial Internet of Things, characterized in that: The online intelligent detection and analysis system for standard board production based on the industrial Internet of Things includes a processor and a machine-readable storage medium, wherein the machine-readable storage medium stores machine executable instructions, and the machine executable instructions are loaded and executed by the processor to implement the online intelligent detection and analysis method for standard board production based on the industrial Internet of Things as described in any one of claims 1 to 9.

Images