Method and system for milling surface quality regulation of thin-walled parts based on dynamic knowledge graph

Abstract

This invention discloses a method and system for controlling the surface quality of thin-walled parts milling based on a dynamic knowledge graph, belonging to the field of intelligent monitoring technology for thin-walled parts. The method includes: transforming the milling mechanism into structured knowledge, defining entity categories and semantic relationships, and constructing a dynamic knowledge graph representing the cutting mechanism; acquiring single-channel vibration signals, which, after adaptive filtering and feature extraction, are input into a pre-trained lightweight monitoring model to output real-time surface roughness estimation data; using the surface roughness estimation data as real-time state input and synchronously updating it to the dynamic knowledge graph; and establishing a fault cause confidence scoring model by combining real-time monitoring data and prior knowledge. This invention outputs surface roughness data in real-time based on an adaptive filtering strategy and a lightweight surface roughness perception model, and integrates real-time monitoring data with domain expert knowledge using a dynamic knowledge graph to form a closed-loop control scheme for precise surface quality control of thin-walled parts milling.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring technology for thin-walled parts, and in particular to a method and system for controlling the surface quality of milled thin-walled parts based on dynamic knowledge graphs. Background Technology

[0002] Thin-walled components are core elements in the lightweight design of high-end aerospace equipment, widely used in critical structures such as aircraft fuselage frames, integral bladed disks of engines, and rocket fuel tanks. Due to their weak rigidity, thin-walled components are prone to chatter and tool deformation during milling, resulting in nonlinear fluctuations in surface roughness. This can easily lead to stress concentration and reduce component fatigue life, while also disrupting the aerodynamic morphology and laminar boundary layer of the aircraft surface, increasing drag. Therefore, achieving precise monitoring, root cause analysis, and stable control of machined surface quality is crucial for ensuring the reliable operation of high-end equipment.

[0003] Existing surface quality assessment and control technologies for thin-walled parts are mainly divided into two categories: online quality monitoring and offline knowledge modeling. Some online monitoring methods can only identify quality anomalies such as roughness through characteristic signals, but cannot reveal the fundamental physical causes such as tool wear, spindle speed instability, or sudden changes in workpiece stiffness. Moreover, existing monitoring systems often isolate real-time sensing data from expert experience and cutting mechanism knowledge, making it impossible to make decisions and diagnose based on the machining status. Although knowledge graph-related technologies can assist in monitoring analysis and early warning, they cannot quantitatively assess fault nodes, making it difficult to achieve quantitative diagnosis and priority differentiation of fault root causes. Existing monitoring methods are mostly based on passive early warning, unable to provide quantitative control basis based on causal relationships, and difficult to meet the autonomous control needs in complex machining scenarios. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the purpose of this invention is to provide a method and system for controlling the surface quality of thin-walled parts milling based on dynamic knowledge graphs. The system outputs surface roughness data in real time based on an adaptive filtering strategy and a lightweight surface roughness perception model. It integrates real-time monitoring data with domain expert knowledge using dynamic knowledge graphs to form a closed-loop control scheme for precise control of the surface quality of thin-walled parts milling.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solution: In a first aspect, embodiments of the present invention provide a method for controlling the surface quality of milled thin-walled parts based on dynamic knowledge graphs, including: The milling mechanism is transformed into structured knowledge, entity categories and semantic relationships are defined, and a dynamic knowledge graph representing the cutting mechanism is constructed. The system collects single-channel vibration signals, performs adaptive filtering and feature extraction, and then inputs them into a pre-trained lightweight monitoring model to output surface roughness estimation data in real time. The surface roughness estimation data is used as real-time status input and synchronously updated to the dynamic knowledge graph. By combining real-time monitoring data with prior knowledge, a confidence scoring model for fault causes is established, forming a cognitive reasoning and parameter control scheme for the surface quality of thin-walled parts milling.

[0006] As a further implementation, the construction of a dynamic knowledge graph characterizing the cutting mechanism includes: Define the entity categories and semantic relationships of milling processes, and complete the mechanistic structure and digital modeling in the form of triples; Pre-defined nodes and relationships are established in the graph database. Unstructured prior knowledge is transformed into attributed entities and Cypher instructions are generated using the Py2neo library. By combining real-time roughness values ​​with thresholds to perform associated retrieval, a dynamic knowledge graph can be constructed and updated.

[0007] As a further implementation, a single-channel x-direction vibration signal is acquired, and interference is removed by adaptive filtering and vibration features are extracted. A lightweight monitoring model containing two unidirectional GRU layers, two fully connected layers and a time attention mechanism is constructed to establish a real-time calculation relationship between vibration features and surface roughness.

[0008] As a further implementation, the fault cause confidence scoring model is used to quantify and score the causes of anomalies, and output cutting parameter optimization suggestions based on the scoring results.

[0009] As a further implementation, the fault cause confidence scoring model is as follows:

[0010] in, A confidence score indicating the potential cause of the failure. This represents the base probability of each cause node in the knowledge graph. This indicates the number of symptoms that match the processing fault. Indicates the weighting factor. Indicates the adjustment factor. This represents the real-time surface roughness monitoring value obtained through the cognitive service layer. This indicates the preset acceptable surface roughness threshold.

[0011] As a further implementation, the surface roughness estimation data is used as real-time status input and synchronously updated to the dynamic knowledge graph, including: Monitor the status of knowledge graph nodes. When the real-time monitoring results reach or exceed a preset threshold, they are considered abnormal nodes in the knowledge graph. By comprehensively assessing the prior probability, symptom matching degree, and numerical exceedance level of abnormal nodes, the abnormal cognitive reasoning process of the dynamic graph is triggered.

[0012] As a further implementation, a pathfinding algorithm based on a graph database is used to search for the source cause node by starting with the node whose surface roughness exceeds the standard and tracing back along the influence or cause relationship; and by combining prior probability, symptom matching degree and numerical exceedance level, the confidence score of each potential fault cause is calculated.

[0013] Secondly, embodiments of the present invention also provide a surface quality control system for thin-walled parts milling based on a dynamic knowledge graph, comprising: The dynamic knowledge graph construction module is used to transform milling machining mechanisms into structured knowledge, define entity categories and semantic relationships, and construct a dynamic knowledge graph representing the cutting mechanism. The surface roughness output module is used to collect single-channel vibration signals, which are then processed by adaptive filtering and feature extraction before being input into a pre-trained lightweight monitoring model to output surface roughness estimation data in real time. The dynamic knowledge graph update module is used to take the surface roughness estimation data as real-time status input and update it synchronously to the dynamic knowledge graph. The control scheme formation module is used to combine real-time monitoring data and prior knowledge to establish a fault cause confidence scoring model, and to form a cognitive reasoning and parameter control scheme for the milled surface quality of thin-walled parts.

[0014] Thirdly, embodiments of the present invention also provide an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to perform the described method for controlling the surface quality of thin-walled parts milling based on dynamic knowledge graphs.

[0015] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned method for controlling the surface quality of thin-walled parts milling based on a dynamic knowledge graph.

[0016] The beneficial effects of this invention are as follows: This invention requires only a single-channel vibration signal to monitor the surface roughness Ra value, reducing system complexity and cost. A confidence scoring model is constructed to evaluate machining faults. This model combines real-time monitoring data with prior knowledge to build a dynamic knowledge graph representing the cutting mechanism of thin-walled parts. Through this model, the knowledge graph outputs a list of fault causes sorted by confidence level, realizing a shift from a visual knowledge base to quantitative reasoning. By integrating real-time roughness monitoring and the knowledge graph, not only is real-time monitoring of the surface quality of machined workpieces achieved, but a dynamic cognitive chain from phenomenon monitoring to root cause tracing is also constructed. Attached Figure Description

[0017] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0018] Figure 1 This is a block diagram illustrating the principle of the control method according to one or more embodiments of the present invention; Figure 2 This is a schematic diagram of the monitoring and diagnostic results according to one or more embodiments of the present invention; Figure 3 This is a knowledge graph diagram of the present invention according to one or more embodiments; Figure 4 This is a flowchart of the control method according to one or more embodiments of the present invention. Detailed Implementation

[0019] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0020] Example 1: This embodiment provides a method for controlling the surface quality of thin-walled parts milled based on dynamic knowledge graphs, such as... Figure 1 As shown, it includes: The milling mechanism is transformed into structured knowledge, entity categories and semantic relationships are defined, and a dynamic knowledge graph representing the cutting mechanism is constructed. The system collects single-channel vibration signals, performs adaptive filtering and feature extraction, and then inputs them into a pre-trained lightweight monitoring model to output surface roughness estimation data in real time. The surface roughness estimation data is used as real-time status input and synchronously updated to the dynamic knowledge graph. By combining real-time monitoring data with prior knowledge, a confidence scoring model for fault causes is established, forming a cognitive reasoning and parameter control scheme for the surface quality of thin-walled parts milling.

[0021] The specific regulation process is as follows: Figure 4 As shown, it includes the following steps: Step S1: By reviewing metal cutting principles and process manuals, define entity categories and semantic relationships in the milling machining field, including: machining equipment, tool geometry parameters, process parameters, workpiece characteristics, and machining status indicators. Structure the mechanism by establishing triples (entity, relation, entity), for example: Tool wear manifests as noise fluctuations; tool wear manifests as increased cutting power; cutting vibration manifests as abnormal amplitude exceeding limits; cutting heat accumulation manifests as increased surface roughness; severe tool wear manifests as increased surface roughness; excessive spindle speed leads to tool wear; overfeed leads to increased surface roughness; cutting heat leads to tool chipping; insufficient fixture rigidity leads to cutting vibration; cutting vibration leads to abnormal surface morphology; etc.

[0022] The aforementioned unstructured prior knowledge is transformed into a structured dataset, enabling digital modeling of the processing mechanism.

[0023] Step S2: Before data transformation, it is necessary to define the nodes and the relationships between them in the Neo4j graph database. For example, a specific triplet in Step S1, unstructured machining knowledge such as cutting principles and process manuals are parsed into entities with attributes through Py2neo. The machining condition parameters are transformed into node attributes in the database, and the attributed entities are transformed into graph database operations. Through the Cypher commands generated by Py2neo, complex association retrieval commands can be executed accurately based on the real-time roughness values ​​generated during the machining process and the preset threshold, triggering the dynamic graph generation logic.

[0024] Step S3: Install a single-channel accelerometer at the machine tool workpiece fixture to acquire high-frequency vibration signals. Both model training and actual monitoring verification use single-channel vibration signals, i.e., vibration data in the x-direction. An adaptive filtering strategy is employed to eliminate machine tool background noise and unstable dynamic interference during no-load operation. Time-domain and frequency-domain analysis methods are used to extract feature vectors reflecting tool wear and stress vibration of thin-walled parts. These feature vectors are then input into a pre-trained lightweight GRU model, which outputs the calculated surface roughness value at the current moment in real time. .

[0025] Step S4: Construct a surface roughness recognition network; After obtaining the input feature matrix through feature adaptive filtering, a nonlinear real-time mapping relationship model between the input feature matrix and the target feature vector can be constructed based on gated recurrent units (GRUs) and a fully connected network. Since feature adaptive filtering eliminates the adverse effects of the unstable dynamic characteristics of the cutting system on the monitoring accuracy, accurate identification of the surface roughness of milled titanium alloy thin-walled parts can be achieved using only a 4-layer network structure; The structure of the constructed roughness recognition network includes 2 unidirectional GRU layers and 2 fully connected layers. The number of units in each of the 2 GRU layers is 128, the number of units in the first fully connected layer is 64, and the number of units in the last fully connected layer is 1.

[0026] Meanwhile, a time attention mechanism is added between the gated recurrent unit layer and the fully connected layer. This mechanism can automatically learn the importance weights of different time steps, assign higher weights to moments with strong predictive capabilities, and suppress the impact of random interference. It can also comprehensively consider the differentiated impact of sensor data at different processing times on recognition accuracy.

[0027] The calculation process of the gated loop unit is expressed as follows: (1) (2) (3) (4) In the above formula, W , U , b ,( ∈{ h,r , z}),in W and U Let each represent a weight matrix of two different types. b For bias vectors, z i Indicates an update to the door. r i This indicates that the door is being reset. and The states are the current state and the previous state, respectively. Represents the candidate state, in the formula This indicates the calculation of the Hadamard product.

[0028] The formula for calculating a fully connected layer is: (5) In the formula, W Represents the network weight matrix. b This represents the network weight matrix and bias vector. X This represents the output vector of the previous layer.

[0029] Furthermore, this embodiment uses single-channel segmented data for model training and numerical comparison with actual monitoring, i.e., vibration data in the x-direction. The cutting data for each tool was divided into training and testing sets in an 8:2 ratio. To reduce the impact of feature amplitude variations on training efficiency, all feature vectors and target vectors were normalized to the range [0, 1] before being input into the model. The batch size was set to 16, the number of training epochs was set to 500, and the Adam optimizer was used for training with a learning rate of 0.001. To prevent overfitting, a dropout rate of 0.2 was applied between the unidirectional gate control recurrent unit layers.

[0030] like Figure 2 As shown, the surface roughness monitoring results and the measured surface roughness values ​​are compared. Figure 2 In the diagram, the horizontal axis represents the measured surface roughness value, and the vertical axis represents the corresponding detected roughness value. Ideally, the monitoring results should match the actual measured values.

[0031] Step S5: Using the Py2neo driving layer, process the real-time roughness values ​​obtained in step S3. Push it to the corresponding dynamic status node in the Neo4j database.

[0032] Furthermore, the status of knowledge graph nodes is monitored. When the real-time monitoring results reach or exceed a preset threshold, they are considered abnormal nodes in the knowledge graph. Then, the prior probability, symptom matching degree, and numerical exceedance level of the abnormal nodes are comprehensively evaluated to trigger the abnormal cognitive reasoning process of the dynamic knowledge graph, such as... Figure 3 As shown, only 15 nodes are displayed, effectively presenting a complex chain-like reasoning path such as excessive spindle speed → increased cutting temperature → tool wear → excessive surface roughness.

[0033] Once the calculated value exceeds the critical point, the system automatically locks the abnormal correlation subgraph, initiates the cognitive reasoning process, and marks the current processing deviation.

[0034] Step S6: Intelligent diagnosis is achieved by constructing a fault cause confidence scoring model based on knowledge graph; according to formula (6), the probability of various potential fault causes can be assessed.

[0035] The confidence scoring model for the cause of failure is as follows:

[0036] In the formula A confidence score indicating the cause of a potential failure. This represents the base probability of each cause node in the knowledge graph. This indicates the number of symptoms that match the processing fault. Indicates the weighting factor. This represents the adjustment factor, which is set to 0.5 in this embodiment to balance the contribution of the base probability and the real-time deviation. This represents the real-time surface roughness monitoring value obtained through the cognitive service layer. This indicates the preset acceptable surface roughness threshold. It reflects the real-time degree of roughness exceeding the standard; when the monitored value does not exceed the standard, this item is 0; when the exceedance is severe, this item increases linearly, thereby significantly improving the confidence score of the related causes.

[0037] The failure cause confidence scoring model enables a shift from qualitative knowledge base to quantitative reasoning, generating a list of failure causes sorted by confidence level. The confidence score for a potential failure cause can then be derived using the confidence scoring formula.

[0038] in, It is a relative probability score used for ranking and comparison, not a strict probability value. Therefore, when multiple symptoms are matched and significantly elevated levels are detected, It may exceed 1.

[0039] The Neo4j-based pathfinding algorithm starts from nodes with excessive surface roughness and searches for related source nodes (such as tool wear, excessive spindle speed, etc.) along the influencing or causing relationships. It then integrates prior probabilities, symptom matching degree, and excess level to calculate a score for each cause. Based on the score priority, it outputs process parameter optimization suggestions, achieving closed-loop real-time monitoring of the machining quality of thin-walled parts.

[0040] Tables 1 and 2 present the monitoring results of the milling process and the recommended measures based on cognitive reasoning, respectively.

[0041] Table 1 shows the monitoring results of surface quality. Diagnostic results

[0042] Optimization suggestions

[0043] Cognitive Reasoning - Recommendations

[0044] Table 2: Results of Recommended Operations Scenario: Feed rate increased by 50%

[0045] Parameter optimization suggestions: To avoid the "poor surface roughness" fault.

[0046] Furthermore, this embodiment develops a user interface based on a B / S architecture. The constructed dynamic knowledge graph is stored in the backend Neo4j graph database, while the state recognition results are stored in the backend MySQL database. The user interface is implemented using HTML and CSS, with the front-end browser responsible for displaying the surface roughness recognition results. This interface integrates multiple module functions, including state information display, cutting diagnosis results, cognitively driven recommended operations, and parameter optimization suggestions.

[0047] To verify the effectiveness of the control method in this embodiment, the performance of the cognitive digital twin system was verified through specific processing experiments, as follows: This embodiment demonstrates a high-speed side milling experiment on a thin-walled Ti6Al4V titanium alloy part (150mm×100mm×5mm) with a large aspect ratio (>20) using a DMU 70V five-axis machining center. A 14mm diameter double-flute end mill was used as the cutting tool, and the machining parameters were set as follows: spindle speed 8000 r / min, cutting speed 351.68 m / min. The monitoring system uses an accelerometer (Dytran3663A1) in conjunction with an NI PXIe-4464 data acquisition card to acquire real-time x-axis vibration data of the workpiece at a sampling frequency of 5kHz, providing physical layer input for subsequent quality identification and inference.

[0048] In this embodiment, the entire lifecycle of multiple milling operations of the tool (numbered T1) is monitored. At the hardware level, in addition to using the single-channel accelerometer in step S3 to measure surface roughness... In addition to extraction, a microphone (acoustic sensor) and a Hall sensor are also configured as supplementary monitoring units to capture abnormal noise fluctuations and electromagnetic induction characteristics of the spindle motor during the processing, providing richer abnormal symptom inputs for the knowledge graph.

[0049] The preset roughness quality threshold for rough machined surfaces is: Real-time output using the GRU lightweight model Value. Taking the 29th cut with tool T1 as an example, when the monitoring system identifies the calculated value... Upon triggering the dynamic knowledge graph reasoning process based on Neo4j, a real-time alert is sent to the operating terminal. After the reasoning is triggered, the confidence scoring model in step S6 is executed. Using a preset fault diagnosis formula, combined with the currently extracted vibration characteristics, noise frequency distribution, and process parameters, the confidence level (quantitative probability) of each fault node is calculated in real time. This probability is not a fixed value but a dynamic value that changes with the machining conditions. Finally, based on the detected abnormal symptoms, the fault type is identified, and based on the knowledge graph path retrieval, an optimization scheme for cutting parameters is provided.

[0050] Table 3 Typical reasoning results for machining thin-walled parts

[0051] Based on actual continuous cutting experiments, as shown in Table 3, differentiated control measures were implemented by evaluating the degree to which the monitored values ​​deviated from the threshold: Normal state (30th cut): Noise fluctuations were detected, but predicted Not exceeding The system identifies the fluctuation as a fluctuation in operating conditions and does not output recommended operations to ensure continuous processing.

[0052] Slightly exceeding the limit (10th and 27th cuts): When the predicted value If the value exceeds the threshold but the increase is within 20%, the system will determine it as a [medium] level alarm and output process optimization suggestions (such as suggesting fine-tuning the feed rate or rotation speed).

[0053] Severe quality deviation (11th and 15th cuts): When the predicted value significantly exceeds the threshold and is accompanied by obvious vibration abnormalities, the system immediately triggers a [high] level warning, requiring the mandatory execution of recommended operations (such as stopping the machine for inspection or replacing the tool).

[0054] This bidirectional data flow enables precise control and closed-loop optimization of surface quality during the milling process of thin-walled parts.

[0055] This embodiment, by establishing a low-level logical framework and utilizing the declarative graph query characteristics of a graph database, not only achieves the structured reconstruction of unstructured mechanisms but also possesses the ability to trace back the causes of machining anomalies within seconds under large-scale multi-parameter correlations. Addressing the unstable dynamic characteristics easily exhibited during the milling of thin-walled parts, this embodiment integrates an adaptive filtering strategy with a lightweight GRU surface roughness monitoring model. This fusion mechanism, relying solely on a single-channel vibration signal, can filter strong interference while ensuring the identification and estimation of tool wear and workpiece state characteristics, offering advantages in engineering practicality and lightweight design. The confidence scoring model in this embodiment fully combines cutting prior probability, real-time symptom matching degree, and exceedance level. This multi-dimensional graph inference method based on graph structure achieves objective quantitative diagnosis and priority assessment, reducing false alarm and misjudgment rates.

[0056] This embodiment deeply integrates the underlying signal processing and graph database mapping to the upper-level decision engine. When an anomaly is triggered, the dynamic processing parameters can be transmitted to the cognitive reasoning network in real time and automatically generate process parameter optimization strategies, giving the thin-walled component processing system the ability to autonomously complete closed-loop decision-making from monitoring and tracing to intelligent control.

[0057] Example 2: This embodiment provides a surface quality control system for thin-walled part milling based on dynamic knowledge graphs, including: The dynamic knowledge graph construction module is used to transform milling machining mechanisms into structured knowledge, define entity categories and semantic relationships, and construct a dynamic knowledge graph representing the cutting mechanism. The surface roughness output module is used to collect single-channel vibration signals, which are then processed by adaptive filtering and feature extraction before being input into a pre-trained lightweight monitoring model to output surface roughness estimation data in real time. The dynamic knowledge graph update module is used to take the surface roughness estimation data as real-time status input and update it synchronously to the dynamic knowledge graph. The control scheme formation module is used to combine real-time monitoring data and prior knowledge to establish a fault cause confidence scoring model, and to form a cognitive reasoning and parameter control scheme for the milled surface quality of thin-walled parts.

[0058] Example 3: This embodiment provides an electronic device, including a memory and a processor. The memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to perform the thin-walled part milling surface quality control method based on dynamic knowledge graph described in Embodiment 1.

[0059] Example 4: This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method for controlling the surface quality of thin-walled parts milling based on dynamic knowledge graphs as described in Embodiment 1.

[0060] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for milling surface quality regulation of thin-walled parts based on a dynamic knowledge graph, characterized in that, include: The milling mechanism is transformed into structured knowledge, entity categories and semantic relationships are defined, and a dynamic knowledge graph representing the cutting mechanism is constructed. The system collects single-channel vibration signals, performs adaptive filtering and feature extraction, and then inputs them into a pre-trained lightweight monitoring model to output surface roughness estimation data in real time. The surface roughness estimation data is used as real-time status input and synchronously updated to the dynamic knowledge graph. By combining real-time monitoring data with prior knowledge, a confidence scoring model for fault causes is established, forming a cognitive reasoning and parameter control scheme for the surface quality of thin-walled parts milling.

2. The thin-walled part milling surface quality regulation method based on a dynamic knowledge graph according to claim 1, characterized in that, The construction of the dynamic knowledge graph characterizing the cutting mechanism includes: Define the entity categories and semantic relationships of milling processes, and complete the mechanistic structure and digital modeling in the form of triples; Pre-defined nodes and relationships are established in the graph database. Unstructured prior knowledge is transformed into attributed entities and Cypher instructions are generated using the Py2neo library. By combining real-time roughness values ​​with thresholds to perform associated retrieval, a dynamic knowledge graph can be constructed and updated.

3. The thin-walled part milling surface quality regulation method based on a dynamic knowledge graph according to claim 1, characterized in that, Single-channel x-direction vibration signal is acquired, and interference is removed by adaptive filtering to extract vibration features; A lightweight monitoring model consisting of two unidirectional GRU layers, two fully connected layers, and a time attention mechanism is constructed to establish a real-time inference relationship between vibration characteristics and surface roughness.

4. The dynamic knowledge graph based milling surface quality regulation method for thin-walled parts according to claim 1, characterized in that, The fault cause confidence scoring model is used to quantify and score the causes of anomalies, and output cutting parameter optimization suggestions based on the scoring results.

5. The dynamic knowledge graph-based thin-walled part milling surface quality regulation method according to claim 1 or 4, characterized in that, The confidence scoring model for the cause of the failure is as follows: wherein, a confidence score representing potential failure causes, a base probability of each cause node in the knowledge graph, a number of symptoms matching the processing failure, a weight factor, an adjustment factor, a real-time surface roughness monitoring value obtained through the cognitive service layer, a pre-set surface roughness qualified threshold.

6. The dynamic knowledge graph based milling surface quality regulation method for thin-walled parts according to claim 1, wherein, The surface roughness estimation data is used as real-time status input and synchronously updated to the dynamic knowledge graph, including: Monitor the status of knowledge graph nodes. When the real-time monitoring results reach or exceed a preset threshold, they are considered abnormal nodes in the knowledge graph. By comprehensively assessing the prior probability, symptom matching degree, and numerical exceedance level of abnormal nodes, the abnormal cognitive reasoning process of the dynamic graph is triggered.

7. The method for controlling the surface quality of thin-walled parts milling based on dynamic knowledge graphs according to claim 6, characterized in that, The pathfinding algorithm based on graph databases starts with the node with excessive surface roughness and backtracks along the influence or cause relationship to search for the source cause node; and combines prior probability, symptom matching degree and numerical excess level to calculate the confidence score of each potential fault cause.

8. A surface quality control system for thin-walled parts milling based on dynamic knowledge graphs, characterized in that, include: The dynamic knowledge graph construction module is used to transform milling machining mechanisms into structured knowledge, define entity categories and semantic relationships, and construct a dynamic knowledge graph representing the cutting mechanism. The surface roughness output module is used to collect single-channel vibration signals, which are then processed by adaptive filtering and feature extraction before being input into a pre-trained lightweight monitoring model to output surface roughness estimation data in real time. The dynamic knowledge graph update module is used to take the surface roughness estimation data as real-time status input and update it synchronously to the dynamic knowledge graph. The control scheme formation module is used to combine real-time monitoring data and prior knowledge to establish a fault cause confidence scoring model, and to form a cognitive reasoning and parameter control scheme for the milled surface quality of thin-walled parts.

9. An electronic device, characterized in that, The device includes a memory and a processor, the memory being used to store a computer program, and the processor running the computer program to cause the electronic device to perform the method for controlling the surface quality of thin-walled parts milling based on dynamic knowledge graphs as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the method for controlling the surface quality of thin-walled parts milling based on a dynamic knowledge graph as described in any one of claims 1-7.

Images