A casting process digital management method based on robot action data

Abstract

This application provides a digital management method for casting processes based on robot motion data, applied in the field of digital management technology for casting. The method includes: dividing the casting process into multiple processes and corresponding process dependencies; collecting robot motion data and key process data for each process; constructing a first process node diagram; mapping the robot motion data and key process data to the first process node diagram to obtain a second process node diagram and calculating the deviation propagation value of the process nodes; generating a process deviation matrix based on the second process node diagram and identifying abnormal process nodes by combining the deviation propagation value; acquiring historical process node diagrams to perform alignment analysis on the deviations of abnormal process nodes and generating a robot correction strategy; controlling the intelligent robot to execute the correction strategy and recording the corrected dataset; and generating a digital management report based on the abnormal process nodes, virtual critical paths, and the recorded dataset for casting process status assessment and process adjustment.

Description

Technical Field

[0001] This application relates to the field of digital management technology for casting, and in particular to a digital management method for casting processes based on robot motion data. Background Technology

[0002] In machine-intelligent casting production workshops, robots are gradually taking on operations such as pouring, mold closing, mold opening, part removal, cooling and transfer, and grinding. Robot motion has become the main process expression of the casting process. As the number of robots involved in the process increases, the state information of the casting process has gradually changed from relying on equipment instructions or timing logic to being expressed by the motion behavior characteristics of robots at different process nodes, such as trajectory changes, posture changes, and torque changes. Therefore, in order to enable machine-intelligent casting process management to obtain continuous and complete process information, it is necessary to use robot motion data to construct a process expression of the process execution.

[0003] In existing technologies, the digital management method for casting processes based on robot motion data in intelligent machine casting production workshops typically treats processes such as pouring, mold closing, mold opening, part removal, and transfer as independent control objects. By collecting robot motion data or equipment operation data corresponding to each process, isolated digital models are established, and the status of a single process is monitored and judged according to its preset process parameter thresholds, thereby achieving segmented management of each process.

[0004] The above-mentioned solutions still have some problems in practical applications. Existing technologies usually do not perform unified modeling and correlation analysis on the process connection relationship, execution timing dependency, and the transmission and accumulation characteristics of process status in the process. When multiple processes are executed continuously and the production conditions change, local deviations may propagate along the process and be difficult to be identified as a whole. This makes it impossible for existing technologies to effectively characterize and analyze the overall status of the casting process from the process structure level, and it is also difficult to achieve digital management and process-level monitoring for multi-process processes. Summary of the Invention

[0005] This application provides a digital management method for casting process based on robot motion data. Its purpose is to achieve unified modeling, abnormal deviation analysis, and closed-loop optimization of robot motion for multiple processes and their dependencies in the casting process, thereby providing data-driven decision-making basis for the evaluation of the casting process's operating status and process management.

[0006] To achieve the above objectives, this application adopts the following technical solution: This application provides a digital management method for a casting process based on robot motion data. The method includes: dividing the casting process into multiple processes and corresponding process dependencies; collecting robot motion data and corresponding key process data for each process; and constructing a first process node graph based on the processes and corresponding process dependencies; mapping the robot motion data and the key process data to the first process node graph to obtain a second process node graph and calculating the deviation propagation value of the process nodes, wherein the second process node graph includes node attributes and edge weights; generating a process deviation matrix based on the second process node graph; identifying abnormal process nodes by combining the deviation propagation value; and determining the deviation propagation value of the abnormal process nodes and their corresponding process dependencies. Based on the corresponding edge weights, node deviation analysis is performed to form a virtual critical path, which represents the node sequence that has the greatest impact on casting quality. A historical process node graph is obtained, and based on the deviation propagation distribution recorded in the historical process node graph, alignment analysis is performed on the deviation degree of the abnormal process nodes in the virtual critical path to generate a corresponding robot correction strategy. The robot is controlled to execute the robot correction strategy, and the process node status, robot motion trajectory, and casting results are recorded as a data set during execution. A digital management report is generated based on the abnormal process nodes, the virtual critical path, and the data set, providing data for evaluating the casting process operation status and adjusting processes.

[0007] In some possible implementations, constructing the first process node graph based on the process steps and their corresponding process dependencies includes: defining each process step in the casting process as a process node; setting an action sub-node with an initial empty state within each process node, the action sub-node being used to store the robot's action data within the process; constructing a directed connection relationship for the preceding and succeeding process nodes of any process node based on the process dependencies, the directed connection relationship being used to record the dependency relationship and dependency type information between process nodes; establishing a hierarchical relationship between the process nodes and the action sub-nodes, and constructing the first process node graph in combination with the directed connection relationship, the first process node graph being used to structurally represent the dependencies between each process step in the casting process.

[0008] In some possible implementations, mapping the robot motion data and the process key data to the first process node graph to obtain a second process node graph and calculating the deviation propagation value of the process nodes includes: dividing the robot motion data into multiple motion execution segments according to preset motion decomposition rules; selecting motion sub-nodes of corresponding process nodes in the first process node graph and mapping the motion execution segments to the corresponding motion sub-nodes; extracting features from the motion execution segments to obtain motion feature data and assigning the motion feature data as node attributes of the motion sub-nodes; mapping the process key data to the corresponding process nodes in the first process node graph and determining the node attributes of the process nodes; performing hierarchical analysis based on the node attributes of the motion sub-nodes and the node attributes of the process nodes to calculate the deviation index between any two process nodes and using the deviation index as the edge weight of the second process node graph; outputting the second process node graph, which is used to characterize the internal motion differences of each process node and the dependencies between process nodes; and calculating the deviation propagation value based on the node attributes and edge weights of the process nodes in the second process node graph.

[0009] In some possible implementations, the step of performing hierarchical analysis based on the node attributes of the action sub-nodes and the node attributes of the process nodes to calculate the deviation index between any two process nodes includes: calculating an action deviation index for each lower-level action sub-node corresponding to a process node based on the node attributes of the action sub-node, wherein the action deviation index is used to quantify the degree of deviation in the execution of robot actions within the process node; combining the action deviation index with the node attributes of the corresponding process node to form a comprehensive deviation index, wherein the comprehensive deviation index is used to reflect action anomalies and process state deviations within the process node; and comparing and calculating the comprehensive deviation index between any two adjacent process nodes using the dependencies between process nodes in the second process node graph to obtain the deviation index between adjacent process nodes, wherein the deviation index is used to quantify dependency anomalies between process nodes.

[0010] In some possible implementations, the step of calculating the deviation propagation value based on the node attributes and edge weights of the process nodes in the second process node graph includes: for any process node in the second process node graph, reading the comprehensive deviation index and the edge weights with subsequent process nodes; multiplying the comprehensive deviation index by the corresponding edge weights to obtain the deviation increment transmitted to each subsequent process node; and accumulating the deviation increments according to the process dependency relationship to form the deviation propagation value of the process node.

[0011] In some possible implementations, the step of generating a process deviation matrix based on the second process node diagram and identifying abnormal process nodes in conjunction with the deviation propagation value includes: initializing an empty process deviation matrix based on the number of process nodes in the second process node diagram; filling each process node in the second process node diagram and its corresponding deviation propagation value into the diagonal position of the process deviation matrix, and filling the deviation index between adjacent process nodes into the off-diagonal position of the process deviation matrix to form a structured process deviation matrix, wherein the rows and columns of the process deviation matrix correspond to process dependencies; for the process deviation matrix, accumulating and combining the diagonal value of each process node with the deviation index of the preceding process node in the corresponding column to generate a comprehensive anomaly index, the comprehensive anomaly index being used to reflect the process node's own deviation and dependency influence; comparing the comprehensive anomaly index with a preset index range, and marking the process nodes corresponding to the comprehensive anomaly index exceeding the preset index range as abnormal process nodes.

[0012] In some possible implementations, the step of performing node deviation analysis based on the deviation propagation value of the abnormal process node and the corresponding edge weight to form a virtual critical path includes: taking each marked abnormal process node as a starting node, sequentially visiting all subsequent process nodes along the process dependency relationship to form a continuous sequence of abnormal nodes; during the traversal, weighted and accumulated calculation of the deviation propagation value of each process node and the edge weight of the subsequent process nodes to evaluate the deviation impact of the abnormal process node on the subsequent process nodes; for all abnormal node sequences, based on the deviation impact, selecting the abnormal node sequence with the largest deviation impact as the virtual critical path, wherein the virtual critical path is used to identify the abnormal process nodes that have the greatest impact on the casting quality and their sequential relationship.

[0013] In some possible implementations, the step of acquiring a historical process node map and, based on the deviation propagation distribution recorded in the historical process node map, performing alignment analysis on the deviation degree of the abnormal process node in the virtual critical path to generate a corresponding robot correction strategy includes: acquiring a historical process node map, the historical process node map including a deviation propagation distribution, the deviation propagation distribution being obtained by recording the statistical distribution of deviation propagation values ​​of each process node in multiple historical casting processes; matching each abnormal process node in the virtual critical path with the corresponding node in the historical process node map; comparing the current deviation propagation value of each matched abnormal node with the deviation propagation distribution to determine the degree of deviation of the abnormal node from the historical baseline; generating motion parameter adjustment suggestions for the robot corresponding to the action sub-node for each abnormal process node according to a preset correspondence between the deviation degree and the adjustment range; and arranging the motion parameter adjustment suggestions of each abnormal process node into an ordered sequence of correction instructions according to the node order of the virtual critical path to form the robot correction strategy.

[0014] In some possible implementations, the controlled robot executes the robot correction strategy and records the process node status, robot motion trajectory, and casting results as a recording dataset during execution. This includes: according to the robot correction strategy, controlling the robot to adjust the motion parameters of each abnormal process node in sequence according to the virtual critical path; during the adjustment process, recording the status information of each process node, including the execution parameters of the action sub-nodes and the corresponding deviation propagation values; collecting and storing the motion trajectory data of the robot when executing the tasks of each process node; recording the casting quality inspection results of this casting process; and associating and integrating the process node status information, motion trajectory data, and casting quality inspection results to form a structured recording dataset, which is used to generate digital management reports and provide a basis for updating the process node diagram.

[0015] In some possible implementations, generating a digital management report based on the abnormal process nodes, the virtual critical path, and the recorded dataset includes: extracting the abnormal process nodes, deviation propagation values, and path sequences from the virtual critical path; extracting the state information, motion trajectory, and casting quality inspection results of the corresponding abnormal process nodes in the recorded dataset before and after executing the robot correction strategy; determining the impact contribution of each abnormal process node to the final casting quality based on the deviation propagation value, path sequence, and casting quality inspection results; determining the improvement effect of the robot correction strategy on the execution of each abnormal process node based on the state information and motion trajectory before and after execution; and organizing the impact contribution and improvement effect of the abnormal process nodes according to the path sequence to generate a structured digital management report. This digital management report is used to assess the casting process status and provide a reference for guiding process adjustments and robot motion corrections.

[0016] As can be seen from the above technical solution, this application has the following beneficial effects: 1. This application sets action sub-nodes within process nodes, mapping the robot's trajectory changes, posture changes, and torque changes in each process to the process node structure in a hierarchical manner. The deviation of each action sub-node is accumulated and analyzed between process nodes through deviation index and deviation propagation value, so that abnormal process nodes and their deviation impact on subsequent processes are quantified and form a virtual critical path. The robot's actions continuously track deviation changes in the hierarchical structure of the node diagram, enabling the centralized identification and localization of local anomalies that were originally scattered in each process. This establishes an analytical capability that can reflect process-level anomalies and the internal action state of nodes, solving the problem that the single-process threshold judgment in the prior art cannot fully characterize process deviations. 2. This application generates a robot correction strategy that matches the degree of deviation by aligning the current deviation of abnormal process nodes in the virtual critical path with the deviation propagation distribution in the historical process node diagram. When the robot performs action adjustments according to the path sequence, the correction action not only corrects the deviation of the current node, but also actively reduces the accumulation of deviations in subsequent processes through the inter-node dependencies. At the same time, it records the action parameters, execution status and casting quality changes. The structured record dataset quantitatively correlates action correction with deviation improvement, making each adjustment traceable and used for process optimization. This establishes a closed-loop digital management mechanism covering anomaly identification, action correction and effect verification, and realizes process-level control for quantitative evaluation and optimization. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a digital management method for the casting process based on robot motion data, as described in this application. Figure 2 This is a flowchart of a method for generating a robot correction strategy according to this application; Figure 3 This is a flowchart of a method for constructing a record dataset according to this application; Figure 4 This is a flowchart illustrating a method for generating digital management reports according to this application. Detailed Implementation

[0018] The terms "first," "second," and "third," etc., used in this application specification, claims, and drawings are used to distinguish different objects, not to limit a specific order.

[0019] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0020] Research has revealed that as machine intelligence is gradually applied to casting production workshops, robots have become the main executors of multiple processes, with process status increasingly reflected through robot actions. However, current technologies at the process management level primarily focus on data collection and status assessment for single processes or single devices, lacking a unified structural expression for multi-process processes. When multiple processes operate continuously and production conditions change, local state shifts introduced by changes in robot actions in each process are often recorded in a scattered manner. The transmission and accumulation of related states between processes lack effective correlation, making it difficult to identify the process node relationships that have a critical impact on casting quality as a whole. Therefore, current technologies are insufficient to meet the management needs of machine-intelligent casting processes for process structure perception and overall status analysis, and also struggle to achieve digital management and process-level monitoring for multi-process collaborative operation.

[0021] To address the aforementioned problems, this application provides a digital management method for casting processes based on robot motion data. The method includes: dividing the casting process into multiple processes and corresponding process dependencies; collecting robot motion data and corresponding key process data for each process; and constructing a first process node graph based on the processes and their dependencies; mapping the robot motion data and the key process data onto the first process node graph to obtain a second process node graph and calculating the deviation propagation value of the process nodes; the second process node graph includes node attributes and edge weights; generating a process deviation matrix based on the second process node graph; identifying abnormal process nodes by combining the deviation propagation values; and determining the deviation propagation value of the abnormal process nodes. Based on the corresponding edge weights, node deviation analysis is performed to form a virtual critical path, which represents the node sequence that has the greatest impact on casting quality. A historical process node graph is obtained, and based on the deviation propagation distribution recorded in the historical process node graph, alignment analysis is performed on the deviation degree of the abnormal process nodes in the virtual critical path to generate a corresponding robot correction strategy. The robot is controlled to execute the robot correction strategy, and the process node status, robot motion trajectory, and casting results are recorded as a data set during execution. A digital management report is generated based on the abnormal process nodes, the virtual critical path, and the data set, providing data for evaluating the casting process operation status and adjusting processes.

[0022] Example 1 like Figure 1 As shown, this application provides a digital management method for the casting process based on robot motion data, and the specific implementation steps are as follows: S1, the casting process is divided into multiple processes and corresponding process dependencies, robot motion data and corresponding key process data for each process are collected, and a first process node diagram is constructed based on the process and corresponding process dependencies.

[0023] Specifically, S101 analyzes the casting production process, identifying the main steps in the entire process, such as pouring, mold closing, mold opening, part removal, cooling and transfer, and grinding, as independent processes. Each process generates a unique process node identifier, which is used to represent the process-level process structure in the node diagram.

[0024] S102, under each process node, an action sub-node is set as a data storage unit. The initial state is empty. It is used to store the actual action data of the robot in the process, including joint angle sequence, end effector trajectory, pose change, execution speed and applied torque, etc. At the same time, key process data such as pouring rate, cooling time, fixture temperature and material state are collected, and the action data and key process data are associated with the corresponding action sub-node and process node respectively to ensure data traceability and synchronization.

[0025] S103. Based on the process execution sequence and process constraints, establish directed connections between process nodes to express the dependencies and dependency types between processes, such as sequential dependencies, condition-triggered dependencies, or parallel execution dependencies. At the same time, assign an influence weight to each directed connection. This weight can be determined by statistically analyzing the degree of process deviation propagation based on historical data or by the experience value of process engineers.

[0026] S104. Establish a hierarchical relationship between the process nodes and their action child nodes, and construct a complete first process node diagram by combining the directed connections between process nodes. This node diagram is used to structurally represent the logical dependencies of each process in the casting process and the ownership of action data within the process. At the same time, the node diagram structure can be imported into the computing system to realize automatic access and analysis of node attributes and connection relationships.

[0027] It should be noted that the action sub-nodes remain empty during the first process node diagram construction stage, and their action data can be dynamically filled in during the casting process. For example, in the pouring process, the corresponding action sub-nodes are filled in by collecting data from the robot nozzle position sensor trajectory and pouring rate sensor. In the cooling and transfer process, the corresponding action sub-nodes are filled in by collecting trajectory and temperature change data from the fixture motion encoder and workpiece temperature sensor. In the grinding process, the corresponding action sub-nodes are filled in by collecting attitude changes, applied torque, and running speed data from the end effector attitude sensor and torque sensor. This ensures that the process node diagram has the ability to fully express the process structure and can dynamically reflect the execution of actions within the process, providing an implementable data foundation for deviation analysis and digital management.

[0028] It should be noted that the following is a brief introduction to the relevant terms used in this embodiment: Casting process: refers to a series of procedures performed in the casting workshop according to the process sequence, including pouring, mold closing, mold opening, part removal, cooling and transfer, grinding, etc., to form a complete casting production process.

[0029] Process: refers to a single production link in the casting process. Each process has independent execution goals and operational requirements, such as the pouring process and the mold closing process.

[0030] Process dependency: refers to the logical constraints between processes in terms of execution order or conditions, including the dependency between preceding and subsequent processes, which is used to guide the construction of directed connections in the process node diagram.

[0031] Robot motion data refers to the motion information generated by the robot during each process, including joint angle sequence, end effector trajectory, pose change, execution speed and applied torque, etc., which are used to analyze the execution deviation of the motion within the process.

[0032] Key process data: refers to key parameters that reflect the execution status and quality of the process, such as pouring rate, cooling time, fixture temperature, material condition, etc., which are used to associate action sub-nodes and perform deviation calculations.

[0033] The first process node diagram is a process node diagram, which is a casting process model represented by a graph structure. It includes process nodes, action sub-nodes, and directed connections between processes. It is used to structurally represent the logical dependencies of processes and the ownership of action data within processes.

[0034] Process node: The node unit in the process node diagram. Each node corresponds to a casting process and is used to record process-level information, associated action sub-nodes, and logical position in the process.

[0035] Action sub-node: The lower-level node located under the process node, used to store robot action data and related information with key process data, and dynamically populated during process execution.

[0036] Directed connection: Directed edges between process nodes are used to represent the dependencies and types between processes, such as sequential dependencies, conditional triggering dependencies, or parallel dependencies. Influence weights can be attached for deviation propagation analysis.

[0037] Impact weight: A value assigned to the directed connection relationship of process nodes, used to quantify the potential deviation transmission impact of the preceding process on the subsequent process, which can be set through historical data statistics or engineering experience.

[0038] Hierarchical relationship: The hierarchical organization between process nodes and action sub-nodes, used to ensure that the ownership relationship between process nodes and internal action data is clear, facilitating subsequent data access and analysis.

[0039] By executing step S1, this application achieves a structured representation of the hierarchical organization of process nodes and action sub-nodes, as well as the directed dependencies between processes. This allows the execution sequence of each process, changes in robot actions, and process status to be clearly presented in the node diagram. As action sub-nodes gradually fill in robot action data during process execution, deviations within the process naturally emerge and are transmitted along the directed connections between nodes. Ultimately, a quantifiable deviation distribution is formed in the first process node diagram, enabling intuitive identification of process anomalies, a complete reflection of the overall process status, and providing a reliable and traceable data foundation for deviation analysis, virtual critical path identification, and digital management.

[0040] S2, map the robot motion data and the process key data to the first process node diagram to obtain the second process node diagram and calculate the deviation propagation value of the process node.

[0041] Specifically, in S201, the robot motion data collected in the action sub-nodes under each process node is divided into motion completion segments. Each complete motion execution is considered as a motion execution segment. For example, in the casting process, the process from starting to pour from the nozzle to completing a certain amount of pouring is considered as a motion execution segment. In the part picking process, the process of grasping and placing is considered as a motion execution segment. In the grinding process, the process of completing a preset trajectory grinding is considered as a motion execution segment. Each motion execution segment contains a complete joint angle sequence, end effector trajectory, pose change, execution speed, and applied torque data.

[0042] S202, the divided action execution segments are mapped to the action sub-nodes under the corresponding process nodes. Each action sub-node can store multiple action execution segments to record the actual execution of complete actions within the process. At the same time, the collected key process data is mapped to the corresponding process nodes to reflect the process status and key process parameters.

[0043] S203. Extract motion feature data for each motion execution segment, including average motion speed, execution trajectory deviation, end effector pose change amplitude and average applied torque, and assign the extracted feature data as node attributes of the motion sub-node; extract key node attributes for process key data, such as pouring rate, cooling time, fixture temperature and material state, and assign them to the corresponding process node attributes.

[0044] S204 involves step-by-step analysis of data in action sub-nodes and process nodes. For each action sub-node, the degree of deviation from the target is calculated, such as comparing the difference between the robot's actual execution trajectory and the preset trajectory, speed changes, and applied torque deviations to obtain the action deviation value for each action segment. The deviation values ​​of all action segments within a process node are combined with key data collected for that process, such as pouring rate, fixture temperature, or cooling time, to generate a comprehensive deviation index within the process. This index reflects abnormal process actions and process status deviations. The comprehensive deviation index of adjacent process nodes is used to assess the potential impact of a process deviation on downstream process execution, forming quantitative information on inter-process deviations.

[0045] S205 In the second process node diagram, the comprehensive deviation index of each process node is transmitted and accumulated according to its dependency relationship with subsequent process nodes. The comprehensive deviation of the current process is allocated to the subsequent process according to the connection weight, and the cumulative deviation value of each subsequent process is accumulated to form the deviation propagation value of each process node. This value intuitively reflects the accumulation and diffusion of local motion deviation in the whole process, providing a basis for identifying key abnormal processes and optimizing robot motion in machine intelligence.

[0046] It should be noted that by analyzing motion deviations and cumulative deviations layer by layer, the deviation of motions within each process can be visually represented in the graph, while also reflecting the transmission effect of deviations along the process flow, providing actionable data support for adjusting the process sequence or optimizing robot execution strategies.

[0047] It should be noted that by mapping motion data to motion sub-nodes according to the completion of the motion, and combining the key data of the process to generate node attributes and edge weights, the internal motion deviation and inter-process dependency relationship can be reflected in the second process node graph. The deviation propagation value reflects the cumulative effect of the deviation along the process, thus providing an implementable data basis for identifying abnormal processes, generating virtual critical paths and robot correction strategies. At the same time, it ensures that the deviation analysis is closely related to the actual execution of the process motion, avoiding the problem of fragmented motion information caused by dividing based on fixed time intervals.

[0048] It should be noted that the following is a brief introduction to the relevant terms used in this embodiment: Action execution segment: refers to a complete action performed by the robot in a single process, such as a pouring, a gripping and placing, or a trajectory grinding, used to ensure the integrity and continuity of action data.

[0049] Motion characteristic data: Quantitative indicators extracted from the motion execution segment, including average speed, trajectory deviation, end effector pose change amplitude, and average applied torque, used to analyze motion execution deviation.

[0050] Node attributes: Feature data assigned to action sub-nodes or process nodes, used to reflect the execution status of the action or the status of the process.

[0051] Comprehensive deviation index: A quantitative index obtained by combining action deviation values ​​with key process data, used to reflect internal action abnormalities and process status deviations.

[0052] Deviation index: By analyzing the comprehensive deviation index, the degree of difference or abnormality between any two process nodes is calculated, which is used to quantify the dependency anomalies between processes.

[0053] Deviation propagation value: Based on the dependencies and edge weights between process nodes, the result of accumulating the comprehensive deviation of process nodes along the process flow is used to reflect the transmission and cumulative impact of local motion deviations on the entire process.

[0054] Edge weight: A numerical value assigned to the directed connection between process nodes, used to quantify the potential impact of deviation transmission from the preceding process to the following process.

[0055] The second process node diagram is a structured graph that maps action execution segments to action sub-nodes based on the first process node diagram, and combines key process data to generate node attributes and inter-process edge weights. It is used to simultaneously represent internal process action deviations and inter-process dependencies.

[0056] This application, through the execution of step S2, maps robot motion data to motion sub-nodes based on motion completion, and assigns process node attributes by combining process key data. Motion deviation values ​​are calculated from the motion deviations of the motion sub-nodes, and then combined with process key data to obtain a comprehensive deviation index. The comprehensive deviation index is accumulated and propagated using inter-process node dependencies and edge weights, with deviations accumulating layer by layer along the process flow. This allows the second process node diagram to simultaneously reflect the degree of motion deviation within a process and anomalies in inter-process dependencies, deriving the impact of anomalies in each local motion on downstream processes. This enables intuitive identification of critical abnormal processes, providing reliable data support for virtual critical path generation and robot correction strategies, and achieving quantitative evaluation and precise control of the casting process state using machine intelligence.

[0057] S3. Generate a process deviation matrix based on the second process node graph, identify abnormal process nodes by combining the deviation propagation value, and perform node deviation analysis based on the deviation propagation value of the abnormal process node and the corresponding edge weight to form a virtual critical path.

[0058] Specifically, in step S301, a process deviation matrix is ​​created based on the number of process nodes in the second process node diagram, and all elements in the matrix are initialized to zero. Each row and each column corresponds to a process node in the second process node diagram, used to record internal process deviations and inter-process deviation propagation. In practice, the total number of nodes N in the second process node diagram can be read first, and then an N×N matrix can be created. Floating-point type storage space is reserved for each matrix element to ensure that subsequent deviation values ​​can be directly written into the matrix, forming structured storage.

[0059] S302: Fill the deviation propagation value of each process node into the diagonal position of the matrix to represent the deviation of the process itself. Between dependent process node pairs, calculate the off-diagonal elements based on the action deviation of the action child nodes and key process data. For example, multiply the comprehensive deviation of the preceding process by the connection weight to obtain the propagated deviation value, and write it into the corresponding matrix element position. In practice, all process nodes and their dependencies can be traversed, action child node data can be read, action feature data such as trajectory deviation and average applied torque can be extracted, and then the propagated deviation can be calculated by combining the edge weights, ensuring that the matrix reflects both the deviation within the process and the deviation propagation between processes.

[0060] S303: For each process node in the matrix, its diagonal deviation value is summed with the off-diagonal deviations of the preceding process nodes in the corresponding column to form a comprehensive anomaly index. In practice, each process node's column is traversed, and all off-diagonal elements in that column are summed to the diagonal deviation value to obtain the comprehensive anomaly index for the current process node. This index quantifies the cumulative impact of the current process node's own deviation and the deviations transmitted from preceding processes. For example, deviations in the casting action are transmitted to the mold-opening process through edge weights, thus comprehensively reflecting the overall impact of process anomalies on the process flow.

[0061] S304 compares the comprehensive anomaly index with a preset anomaly threshold, and marks process nodes that exceed the threshold as abnormal process nodes. In implementation, a threshold range can be set for each process node, the comprehensive anomaly index of the node is read, a conditional judgment operation is performed, and when the index is greater than the threshold, an anomaly mark field is added to the node for subsequent virtual critical path analysis and priority handling of abnormal processes.

[0062] S305: Starting with each abnormal process node, subsequent process nodes are visited along the process dependencies to form an abnormal node sequence. During the traversal, the deviation propagation value of subsequent process nodes is weighted and accumulated with the edge weights. For example, the deviation of the current process node is multiplied by the edge weight and accumulated to the cumulative deviation field of the next node, which is used to quantify the impact of abnormal process nodes on downstream processes. In specific implementations, a queue or recursive traversal method can be used to perform cumulative calculations for each dependency chain and record each abnormal node sequence and its total cumulative deviation value.

[0063] S306. Compare the cumulative deviation values ​​of all abnormal node sequences and select the sequence with the largest cumulative deviation as the virtual critical path. In implementation, the cumulative deviation of each abnormal node sequence is stored in a list, the sequence corresponding to the maximum value is found through iteration, and this sequence is output as the virtual critical path. The order of process nodes in this path represents the execution order of abnormal processes that have the most significant impact on casting quality. This can be used to guide robot motion optimization or adjust the process sequence, such as adjusting the execution priority of the grinding process or modifying the pouring speed.

[0064] It should be noted that through this node-by-node analysis and cumulative calculation method, the motion deviation of each process, the key data of the process, and the inter-process dependencies are all mapped to a matrix and gradually quantified, realizing an executable process from motion data to virtual critical path, while ensuring that the analysis results closely correspond to the actual robot motion execution.

[0065] It should be noted that the following is a brief introduction to the relevant terms used in this embodiment: Process deviation matrix: refers to a two-dimensional matrix constructed with the process nodes of the second process node diagram as row and column indices. The diagonal elements store the comprehensive deviation index of the corresponding process node, and the off-diagonal elements store the deviation transmission value between adjacent process nodes. It is used to structurally reflect the internal deviation of the process and the inter-process dependency effect.

[0066] Comprehensive anomaly index: This refers to an index calculated by adding the diagonal deviation value of a certain process node in the process deviation matrix to the off-diagonal deviation transmitted from the preceding process. It is used to quantify the deviation of the process node itself and the cumulative deviation affected by the dependent processes.

[0067] Abnormal process nodes: These are process nodes whose comprehensive abnormality indicators exceed preset thresholds. They are used to identify important processes in the process that may lead to quality problems or the spread of action deviations.

[0068] Abnormal node sequence: refers to a continuous sequence formed by starting from an abnormal process node and visiting its subsequent process nodes along the process dependency relationship. Each node records the cumulative deviation value, which is used to analyze the transmission path of the abnormal impact in the process.

[0069] Cumulative deviation: refers to the cumulative value formed by each process node in the abnormal node sequence after receiving the deviation transmitted from the previous process and adding its own comprehensive deviation. It is used to quantify the degree of impact of the abnormal process on the subsequent processes.

[0070] Virtual critical path: refers to the sequence with the largest cumulative deviation among all abnormal node sequences. It is used to identify the abnormal process that has the greatest impact on casting quality and its execution order, providing guidance for optimizing the process and adjusting robot actions.

[0071] Weighted accumulation: refers to the operation of multiplying the deviation propagation value of the process node according to the weight of the connecting edge and accumulating it to the subsequent node during the traversal of the abnormal node sequence. It is used to calculate the diffusion of the abnormal impact along the process dependency chain.

[0072] This application, through the execution of step S3, quantifies and structurally stores the internal motion deviations and critical process states of each process during the node-by-node construction of the process deviation matrix. By accumulating the diagonal and off-diagonal elements of the matrix to calculate a comprehensive anomaly index, it can simultaneously reflect the process's own deviation and its cumulative deviation influenced by preceding processes. Abnormal process nodes are marked by threshold judgment, and the abnormal nodes are weighted and cumulatively traversed along the process dependencies to form an abnormal node sequence. Then, the path with the largest cumulative deviation from all abnormal node sequences is selected as the virtual critical path, thereby achieving accurate identification of critical abnormal processes affecting casting quality and their sequential relationships. This process ensures the correspondence between internal process deviations, inter-process dependencies, and deviation propagation effects, allowing the virtual critical path to be directly mapped to the actual process, providing operable data basis for machine intelligence to optimize the robot's motion execution sequence and adjust process operations.

[0073] S4. Obtain the historical process node map, and based on the deviation propagation distribution recorded in the historical process node map, perform alignment analysis on the deviation degree of the abnormal process node in the virtual critical path, and generate the corresponding robot correction strategy.

[0074] Specifically, such as Figure 2 As shown in S401, obtain the historical process node diagram and deviation propagation distribution.

[0075] Specifically, the process node diagrams for each casting process are extracted from the historical casting process database, and the deviation propagation value of each process node in the historical process is recorded. In practice, the mean, maximum, minimum, and distribution trends of the deviations of nodes across multiple historical processes can be statistically analyzed to form a historical deviation distribution table. In this way, when analyzing the current process, the real-time deviation of each process node can be compared with the historical benchmark to determine the magnitude of anomalies.

[0076] S402 matches abnormal process nodes on the virtual critical path with historical nodes.

[0077] Specifically, for each abnormal process node in the virtual critical path, based on the process number, process type, and node position in the path, the corresponding node is located in the historical process node graph, and the historical deviation distribution information of that node is extracted. In implementation, a mapping table can be established to correspond each abnormal node to its historical nodes, ensuring that subsequent deviation alignment analysis can accurately calculate the deviation of each abnormal node relative to its history.

[0078] S403, calculate the degree of deviation.

[0079] Specifically, the current deviation propagation value of each abnormal process node is compared with the historical deviation distribution. In practice, it can be logically determined whether the current deviation falls within the historical average or historical fluctuation range, or exceeds the historical maximum value. The degree of deviation is quantified by a deviation level, such as slight deviation, moderate deviation, or severe deviation. This degree of deviation directly reflects the severity of the current process anomaly relative to the historical benchmark, providing a basis for generating adjustment recommendations.

[0080] S404, Generate motion parameter adjustment suggestions.

[0081] Specifically, based on the degree of deviation of each abnormal node, specific motion parameter adjustment suggestions are generated for the robot corresponding to the motion sub-node. In implementation, the deviation level can be mapped to adjustment strategies. For example, if the deviation is severe, the end effector torque or execution speed is increased; if the deviation is moderate, the trajectory is fine-tuned or the speed is slightly adjusted; and if the deviation is slight, the motion parameters remain unchanged. Each suggestion includes the motion sub-node identifier, adjustment type, and adjustment range, ensuring that each suggestion can be directly used to correct the robot's motion execution.

[0082] S405, arranges an ordered sequence of correction instructions.

[0083] Specifically, the abnormal process nodes in the virtual critical path are arranged according to the path order, and the action adjustment suggestions generated by each node are sequentially formed into an ordered sequence of correction instructions. In implementation, each instruction includes a node identifier, action sub-nodes, adjustment parameters, and execution sequence number, ensuring that the robot executes the adjustment operations sequentially according to the path order. This minimizes the deviation impact of abnormal processes on downstream processes, thereby achieving deviation control of the entire process.

[0084] It should be noted that the following is a brief introduction to the relevant terms used in this embodiment: Historical process node diagram: refers to a structured diagram that records each process node and its deviation propagation value in multiple historical casting processes, used to provide a deviation benchmark for the current abnormal process.

[0085] Deviation propagation distribution: By statistically analyzing the deviation propagation values ​​of historical process nodes in multiple processes, quantitative distribution information is formed, including the deviation mean, fluctuation range, and trend, which is used to assess the degree to which the current process deviates from the historical benchmark.

[0086] Deviation degree: refers to the position of the deviation propagation value of the current abnormal process node relative to the historical deviation distribution. It is quantified by level or classification to reflect the severity of the abnormal action.

[0087] Motion parameter adjustment suggestions: Specific adjustment measures are proposed for the robot's execution of the corresponding motion sub-node, including the adjustment type, magnitude and target motion sub-node, to correct deviations and reduce the impact on subsequent processes.

[0088] Correction instruction sequence: A set of motion parameter adjustment instructions arranged in the order of the virtual critical path. Each instruction contains an action sub-node identifier, adjustment type and magnitude, and execution sequence number, which is used by the robot to perform correction operations.

[0089] This application performs alignment analysis on the deviation propagation distribution in the abnormal process nodes of the virtual critical path and the historical process node diagram by executing step S4. The degree of deviation is determined by comparing the current deviation with the historical benchmark, and the degree of deviation is mapped to the parameter adjustment suggestions of the robot corresponding to the action sub-node. These adjustments are executed sequentially according to the order of the virtual critical path, so that the action correction of each abnormal process node directly reduces its cumulative impact on the deviation of the downstream process, thereby realizing the controllable accumulation of deviation and the optimization of process quality in the entire machine intelligent casting process.

[0090] S5, control the robot to execute the robot correction strategy, and record the process node status, robot motion trajectory and casting results as a recording dataset during the execution process.

[0091] Specifically, such as Figure 3 As shown, S501, perform the adjustment operation.

[0092] Specifically, following the sequence of the virtual critical path, robot motion parameters are adjusted for each sub-node of an abnormal process. For example, the nozzle speed and output are adjusted in the casting process, the end effector trajectory and applied torque are fine-tuned in the grinding process, and the gripper closing speed is adjusted in the part-picking process. In practice, the upper-level controller calls the robot control interface to send the adjusted motion parameters to the robot driver. After receiving the instructions, the robot modifies its motion trajectory and execution torque in real time, so that motion deviations are suppressed in a timely manner, while ensuring the continuity and integrity of the motion, thereby reducing the impact of local deviations on downstream processes.

[0093] S502, collect process node status information.

[0094] Specifically, during the adjustment of each action sub-node, information such as joint encoder, end effector pose, torque sensor, and execution speed is read through the robot controller's data port. The process node number and time stamp are then added to the data acquisition module, and the data is written to a real-time database or local file system. In this way, the deviation correction effect of each process node can be accurately recorded, providing traceable status information for subsequent data analysis.

[0095] S503 collects robot motion trajectory data.

[0096] Specifically, during motion adjustment, the angle changes of each robot joint, the trajectory of the end effector, and the execution speed are collected and stored in structured objects. For example, a data object is created for each motion sub-node, with internal fields including joint angle, end effector pose, and execution speed. This structured trajectory data can be used to analyze the effect of adjustment strategies on motion deviation correction and the potential impact on subsequent processes, achieving a closed-loop correlation between motion and effect.

[0097] S504 records the results of casting quality inspection.

[0098] Specifically, after each process is completed, casting quality information, including the location, type, and dimensional deviation of surface defects, is collected using visual inspection systems, laser scanning, or contact measurement, and a structured measurement file is generated. Each quality record is associated with the process node number, action sub-node number, and time stamp to ensure that the action correction effect of each process corresponds to the final casting quality, and to verify the effectiveness of the action adjustment.

[0099] S505 generates a structured record dataset.

[0100] Specifically, the status information, motion trajectory, and quality data collected in S502 to S504 are integrated in the order of process nodes to establish a process record data model. Each instance contains motion parameters, trajectory data, deviation correction effect, and quality result, and a complete dataset is generated through serialization.

[0101] It should be noted that the following is a brief introduction to the relevant terms used in this embodiment: Process node status: This refers to the robot motion execution data collected during the execution of each process node, including joint angles, end effector pose, applied torque, execution speed, etc., and corresponds to the process node number and time stamp, which is used to quantify motion execution deviation and adjustment effect.

[0102] Robot motion trajectory: This refers to the robot motion trajectory data, which includes the changes in joint angles and end effector path information when the robot executes each action sub-node. It is stored in structured objects and used to analyze the effect of motion adjustment strategies on motion execution and the potential impact on downstream processes, so as to realize a closed-loop correlation between actions and effects.

[0103] Casting results: These are the casting quality inspection results, which refer to the surface and dimensional deviation information of the casting obtained through visual inspection, laser scanning, or contact measurement after the completion of the process. This includes the location, type, and numerical deviation of defects, and corresponds to the process node and action sub-node number. It is used to verify the actual effect of action adjustment on quality improvement.

[0104] Record dataset: This refers to the integration of process node status information, robot motion trajectory, and casting quality inspection results according to the process sequence to form a unified data model. Each data instance contains motion parameters, trajectory, deviation correction effect, and quality result, which is used to generate digital management reports and update process node diagrams, realizing a traceable closed loop from motion adjustment to quality result.

[0105] This application, through step S5, adjusts the motion parameters of each abnormal process node in the virtual critical path sequence, and sends adjustment instructions to the robot driver via the machine intelligence control interface. This allows the robot to modify its motion trajectory and apply torque in real time, thereby suppressing the spread of local motion deviations to downstream processes while ensuring the continuity and integrity of the motion. During the adjustment process, status information such as joint encoder, end effector pose, torque, and execution speed are collected through data ports and strictly corresponded to the process node number and action sub-node number. At the same time, the robot's motion trajectory and casting quality measurement results are recorded, achieving a closed-loop correspondence between motion correction effect, status change, and final quality. The collected status information, motion trajectory, and quality data are integrated into a structured record dataset according to the process sequence, so that each data can be traced back to the specific motion adjustment operation. This provides a reliable basis for the generation of digital management reports, process optimization, and subsequent updates of process node diagrams, thus forming a complete data closed loop from motion adjustment to effect verification, ensuring the effectiveness and traceability of deviation correction measures.

[0106] S6. A digital management report is generated based on the abnormal process nodes, virtual critical paths, and the record dataset. The digital management report is used to provide data basis for evaluating the operating status of the casting process and adjusting the process.

[0107] Specifically, such as Figure 4 As shown in S601, extract information about the virtual critical path.

[0108] In practical implementation, the node number, process type, and deviation propagation value of each abnormal process node are sequentially read from the virtual critical path list. For example, the node number for the casting process is G12, the deviation propagation value is 0.35, and it is the first node in the path order. A data object is created for each node, and the node number, process type, deviation propagation value, and path order are saved to a table or database to ensure that subsequent analysis can directly access each node in sequence and obtain deviation information.

[0109] S602, Extract the data of the corresponding abnormal process node from the record dataset.

[0110] Specifically, for each abnormal process node, the corresponding process status information, motion trajectory, and casting quality results are retrieved from the record dataset. For example, in the casting process, before the adjustment, the end effector trajectory deviated from the target trajectory by 0.8 mm; after the adjustment, the deviation decreased to 0.2 mm, and the number of surface defects decreased from 5 to 1. This information is saved as node data objects, and the objects are labeled with status indicators before and after the adjustment to ensure quantitative comparison and analysis.

[0111] S603, Analyze the impact contribution of abnormal process nodes.

[0112] In practice, the impact of each node on the final casting quality is determined based on the node deviation propagation value and path sequence logic. For example, pouring processes that are earlier in the path and have high deviation values ​​contribute significantly to surface defects and dimensional deviations in the casting and can be marked as high-impact nodes. Grinding processes with smaller deviations and later in the path have a moderate impact on the final surface quality. The impact contribution value of each node is quantified and recorded in the node data object for subsequent evaluation and optimization of process priorities.

[0113] S604, Evaluate the improvement effect of the robot correction strategy.

[0114] Specifically, the robot's motion state and trajectory changes before and after implementing the correction strategy are compared for each abnormal process node. For example, after fine-tuning the trajectory of the end effector in the grinding process, the deviation decreased from 0.5 mm to 0.1 mm, the applied torque decreased from 7 Nm to 5 Nm, and the trajectory curve became smoother. Through this comparison, the actual effect of each motion parameter adjustment suggestion is determined, and the improvement values ​​are recorded in the node data object to ensure that the improvement effect is quantifiable and correlated with the actual robot motion.

[0115] S605 generates structured digital management reports.

[0116] Specifically, the report is compiled in chronological order of the path for each abnormal process node, including its node number, deviation propagation value, impact contribution, and improvement effect. For example, in tabular form: the first column is the node number, the second is the process type, the third is the deviation propagation value, the fourth is the impact contribution, and the fifth is the improvement effect. The generated report provides a clear view of the impact of each abnormal process node on casting quality and the effectiveness of the robot correction strategy, offering concrete data for casting process optimization, process adjustment, and robot motion correction.

[0117] It should be noted that the following is a brief introduction to the relevant terms used in this embodiment: Virtual critical path: includes abnormal process node number, process type, deviation propagation value and path sequence, used to identify the process that has the greatest impact on quality.

[0118] Abnormal process nodes: Record the execution status, robot motion trajectory and casting quality results corresponding to the virtual critical path nodes in the data set, which are used to quantitatively analyze motion deviations and correction effects.

[0119] Impact contribution: By logically judging the actual impact of node deviation and path sequence on the quality of the final casting, for example, the larger the deviation, the higher the contribution of the process at the beginning of the path.

[0120] Improvement effect: The quantitative results of the robot correction strategy on the improvement of motion execution accuracy and deviation, such as the reduction of trajectory deviation, end torque and surface defects.

[0121] Digital Management Report: A structured report organized according to the path sequence, including node number, deviation propagation value, impact contribution, and improvement effect, used for process status assessment and guidance for process adjustment and robot motion correction.

[0122] This application, through the execution of step S6, compares and analyzes the deviation propagation value of each abnormal process node in the virtual critical path with the state information, motion trajectory, and casting quality results of the corresponding node in the recorded dataset before and after the execution of the robot correction strategy. By extracting the deviation changes and path sequence of each abnormal node, the deviation propagation of the node is mapped to the actual impact on downstream processes. At the same time, the improvement effect of the robot correction strategy on motion deviation is evaluated by combining the changes in trajectory and state information before and after execution. Thus, the contribution and adjustment effect of each abnormal node to the final casting quality are reflected in the digital management report, realizing closed-loop linkage management of abnormal process identification, improvement evaluation, and process optimization.

[0123] The technical solution provided in this application breaks down the casting process in a machine-intelligent workshop into process nodes with clear dependencies. At the node level, both robot motion data and process key data are introduced. This allows even minute deviations generated by each robot action to be mapped as measurable and transferable process node deviations. Based on the dependencies between processes, these deviations are propagated and analyzed to form a virtual critical path reflecting the concentrated propagation direction of the deviations. This concentrates previously difficult-to-locate quality fluctuations into a small number of abnormal process nodes that have the greatest impact on casting quality. By introducing the deviation propagation distribution from historical process node diagrams, the degree of deviation of the current abnormal process in the virtual critical path is aligned and analyzed, generating robot correction strategies for the corresponding action sub-nodes. This approach provides clear historical benchmarks and adjustment bases for corrective actions. During the execution of corrective strategies, the status of process nodes, robot motion trajectories, and casting results are recorded simultaneously. This establishes a one-to-one correspondence between changes in motion parameters, deviation convergence processes, and the final casting quality results. Ultimately, a digital management report is generated based on abnormal process nodes, virtual critical paths, and recorded datasets. This enables the reconstruction of a complete causal link from robot motion execution and process deviation evolution to casting quality results. The operational status of the casting process, the sources of key anomalies, and their corrective effects can be continuously recorded, compared, and evaluated in the form of structured data. This provides quantifiable, traceable, and reusable management data for the digital management, process control, and robot motion adjustment of the casting process in the intelligent machine workshop.

[0124] The foregoing has shown and described the basic principles, main features, and advantages of this application. Those skilled in the art should understand that this application is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of this application. Various changes and modifications can be made to this application without departing from the spirit and scope thereof, and all such changes and modifications fall within the scope of this application as claimed. The scope of protection of this application is defined by the appended claims and their equivalents.

Claims

1. A digital management method for casting processes based on robot motion data, characterized in that, The method includes: The casting process is divided into multiple steps and corresponding steps dependencies. Robot motion data and corresponding key data of each step are collected, and a first step node diagram is constructed based on the steps and corresponding steps dependencies. The robot motion data and the process key data are mapped to the first process node graph to obtain the second process node graph and the deviation propagation value of the process node is calculated. The second process node graph includes node attributes and edge weights. A process deviation matrix is ​​generated based on the second process node graph. Abnormal process nodes are identified by combining the deviation propagation value. Node deviation analysis is performed based on the deviation propagation value of the abnormal process nodes and the corresponding edge weights to form a virtual critical path. The virtual critical path is used to represent the node sequence that has the greatest impact on the casting quality. Obtain historical process node diagrams, and based on the deviation propagation distribution recorded in the historical process node diagrams, perform alignment analysis on the degree of deviation of the abnormal process nodes in the virtual critical path, and generate corresponding robot correction strategies. The robot is controlled to execute the robot correction strategy, and the process node status, robot motion trajectory and casting results are recorded as a data set during the execution process; A digital management report is generated based on the abnormal process nodes, virtual critical paths, and the recorded dataset. This digital management report provides data support for evaluating the operational status of the casting process and adjusting processes.

2. The method according to claim 1, characterized in that, The construction of the first process node diagram based on the process and the corresponding process dependencies includes: Each step in the casting process is defined as a process node; Within each process node, an action sub-node with an initial empty state is set. The action sub-node is used to store the robot's action data within the process. Based on the process dependency relationship, for any process node, a directed connection relationship is constructed with its predecessor and successor process nodes. The directed connection relationship is used to record the dependency relationship and dependency type information between process nodes. Establish a hierarchical relationship between the process node and the action sub-node, and construct a first process node graph by combining the directed connection relationship. The first process node graph is used to structurally represent the dependency relationship between each process in the casting process.

3. The method according to claim 2, characterized in that, The step of mapping the robot motion data and the key process data to the first process node diagram to obtain the second process node diagram and calculating the deviation propagation value of the process nodes includes: According to the preset motion decomposition rules, the robot motion data is divided into multiple motion execution segments; In the first process node diagram, select the action sub-node of the corresponding process node, and map the action execution segment to the corresponding action sub-node; Features are extracted from the action execution segment to obtain action feature data, and the action feature data is assigned as the node attribute of the action child node; Map the key data of the process to the corresponding process nodes in the first process node diagram, and determine the node attributes of the process nodes; Perform hierarchical analysis based on the node attributes of the action sub-nodes and the node attributes of the process nodes, calculate the deviation index between any two process nodes, and use the deviation index as the edge weight of the second process node graph. Output the second process node diagram, which is used to characterize the differences in actions within each process node and the dependencies between process nodes; The deviation propagation value is calculated based on the node attributes and edge weights of the process nodes in the second process node diagram.

4. The method according to claim 3, characterized in that, The step of performing hierarchical analysis based on the node attributes of the action sub-node and the node attributes of the process node, and calculating the deviation index between any two process nodes, includes: For each process node, the lower-level action sub-node is calculated based on the node attributes of the action sub-node. The action deviation index is used to quantify the degree of deviation in the execution of robot actions within the process node. The action deviation index is combined with the node attributes of the corresponding process node to form a comprehensive deviation index, which is used to reflect the internal action abnormalities and process status deviations of the process node. Using the dependencies between process nodes in the second process node diagram, the comprehensive deviation index of any two adjacent process nodes is compared and calculated to obtain the deviation index between adjacent process nodes. The deviation index is used to quantify the dependency anomalies between process nodes.

5. The method according to claim 4, characterized in that, The step of calculating the deviation propagation value based on the node attributes and edge weights of the process nodes in the second process node diagram includes: For any process node in the second process node graph, read the comprehensive deviation index and the edge weight with the subsequent process node; Multiply the comprehensive deviation index by the corresponding edge weight to obtain the deviation increment passed to each subsequent process node; The deviation increments are accumulated according to the process dependency relationship to form the deviation propagation value of the process node.

6. The method according to claim 5, characterized in that, The step of generating a process deviation matrix based on the second process node diagram and identifying abnormal process nodes by combining the deviation propagation value includes: Based on the number of process nodes in the second process node diagram, initialize an empty process deviation matrix; Each process node and its corresponding deviation propagation value in the second process node diagram are filled into the diagonal position of the process deviation matrix, and the deviation index between adjacent process nodes is filled into the off-diagonal position of the process deviation matrix to form a structured process deviation matrix, wherein the rows and columns of the process deviation matrix correspond to the process dependency relationship. For the process deviation matrix, the diagonal value of each process node is combined with the cumulative deviation index of the preceding process node in the corresponding column to generate a comprehensive anomaly index. The comprehensive anomaly index is used to reflect the process node's own deviation and dependent influence. The comprehensive anomaly index is compared with a preset index range, and the process nodes corresponding to comprehensive anomaly indices that exceed the preset index range are marked as abnormal process nodes.

7. The method according to claim 6, characterized in that, The step of performing node deviation analysis based on the deviation propagation value of the abnormal process node and the corresponding edge weight to form a virtual critical path includes: Each marked abnormal process node is used as the starting node, and all subsequent process nodes are visited sequentially along the process dependency relationship to form a continuous sequence of abnormal nodes. During the traversal, the deviation propagation value of each process node is weighted and accumulated with the edge weight of subsequent process nodes to evaluate the impact of abnormal process nodes on the deviation of subsequent process nodes. For all abnormal node sequences, based on the deviation impact, the abnormal node sequence with the largest deviation impact is selected as the virtual critical path. The virtual critical path is used to identify the abnormal process nodes and their sequential relationships that have the greatest impact on casting quality.

8. The method according to claim 7, characterized in that, The process of acquiring historical process node maps and, based on the deviation propagation distribution recorded in the historical process node maps, performing alignment analysis on the deviation degree of the abnormal process nodes in the virtual critical path, and generating corresponding robot correction strategies includes: Obtain a historical process node diagram, which includes a deviation propagation distribution. The deviation propagation distribution is obtained by statistically distributing the deviation propagation values ​​of each process node in multiple historical casting processes. Match each abnormal process node in the virtual critical path with the corresponding node in the historical process node graph; Compare the current deviation propagation value of each matched abnormal node with the deviation propagation distribution to determine the degree of deviation of the abnormal node from the historical benchmark; Based on the preset correspondence between the degree of deviation and the adjustment range, suggestions for adjusting the motion parameters of the robot corresponding to the action sub-node are generated for each abnormal process node; Based on the node order of the virtual critical path, the proposed adjustment of action parameters for each abnormal process node is arranged into an ordered sequence of correction instructions, forming the robot correction strategy.

9. The method according to claim 8, characterized in that, The controlled robot executes the robot correction strategy and records the process node status, robot motion trajectory, and casting results as a recording dataset during execution, including: Based on the robot correction strategy, the robot is controlled to adjust the motion parameters of each abnormal process node in the order of the virtual critical path; During the adjustment process, the status information of each process node is recorded, including the execution parameters of the action sub-node and the corresponding deviation propagation value. Collect and store the robot's motion trajectory data when performing tasks at each process node; Record the casting quality inspection results of this casting process; The process node status information, motion trajectory data, and casting quality inspection results are correlated and integrated to form a structured record dataset. This record dataset is used to generate digital management reports and provide a basis for updating the process node diagram.

10. The method according to claim 9, characterized in that, The process of generating a digital management report based on the abnormal process nodes, virtual critical paths, and the record dataset includes: Extract abnormal process nodes, deviation propagation values, and path sequence from the virtual critical path; Extract the state information, motion trajectory, and casting quality inspection results of the corresponding abnormal process nodes in the record dataset before and after the execution of the robot correction strategy; Based on the deviation propagation value, path sequence, and casting quality inspection results, determine the impact contribution of each abnormal process node on the final casting quality; Based on the state information and motion trajectory before and after execution, determine the improvement effect of the robot correction strategy on the execution of each abnormal process node; The impact and improvement effects of abnormal process nodes are organized in chronological order to generate a structured digital management report. This digital management report is used to assess the status of the casting process and to provide a reference for guiding process adjustments and robot motion corrections.

Images