Mechanical processing system collaborative optimization and whole-process tracing method and storage medium

Abstract

The application discloses a kind of mechanical processing system collaborative optimization and whole-process tracking method and storage medium.The method comprises the following steps: obtaining multi-source heterogeneous data in mechanical processing system, the multi-source heterogeneous data is synchronized in space-time by space-time synchronization technology, and unified multi-source data set is obtained;The label information related to processing process that user enters at least one modal form is semantically annotated in space-time dimension, is associated with unified multi-source data set, and the structured associated unit is obtained;The structured associated unit is semantically understood and structured implicit knowledge extraction, and constructs mechanical processing knowledge experience base;The decision demand that user initiates at least one modal form is obtained, and response suggestion for decision demand is generated and output based on mechanical processing knowledge experience base;The actual effect of response suggestion in actual processing is obtained, and mechanical processing knowledge experience base is updated and iterated according to actual effect.Therefore, mechanical processing system collaborative optimization and whole-process tracking are realized.

Description

Technical Field

[0001] This invention relates to the field of machining technology, specifically to a method and storage medium for collaborative optimization and full-process traceability of machining systems. Background Technology

[0002] Machining is a core and fundamental process in high-end equipment manufacturing and precision parts production. Monitoring the status of the machining system, optimizing the process, and tracing the process are key to ensuring machining accuracy, operational stability, and product quality.

[0003] Currently, optimization and traceability solutions in the machining field mainly rely on fixed threshold alarms, time-series signal anomaly detection, and independent information management systems for basic control. These traditional solutions have low integration and intelligence levels, making them unsuitable for precision intelligent machining scenarios with multi-parameter coupling and dynamically changing operating conditions. They suffer from weak multi-source data fusion capabilities, lack of multi-modal semantic association, poor reusability of domain knowledge, and the absence of closed-loop iterative optimization mechanisms, failing to meet the high-precision, traceable, and sustainable iterative production demands of modern machining. Summary of the Invention

[0004] The purpose of this invention is to provide a collaborative optimization and full-process traceability method and storage medium for machining systems, aiming to solve the problems of weak multi-source data fusion capability, lack of multimodal semantic association, poor reusability of domain knowledge, and lack of closed-loop iterative optimization mechanism in existing machining optimization and traceability technologies.

[0005] To address the above technical problems, the first aspect of this invention provides a method for collaborative optimization and full-process traceability of a machining system, comprising: Acquire and integrate multi-source heterogeneous data from a machining system, and perform spatiotemporal synchronization of the multi-source heterogeneous data using spatiotemporal synchronization technology to obtain a unified multi-source dataset. The user-entered label information related to the processing process in at least one modality is semantically annotated in the spatiotemporal dimension and associated with the unified multi-source dataset to obtain structured association units; Semantic understanding and structured implicit knowledge extraction are performed on the structured associated units to construct a mechanical processing knowledge and experience base; The system acquires decision-making requirements initiated by users in at least one modality, and generates and outputs response suggestions for the decision-making requirements based on the machining knowledge and experience base. Obtain the actual effectiveness of the response suggestions in actual processing, and update and iterate the machining knowledge and experience base based on the actual effectiveness.

[0006] In some specific embodiments, the acquisition of multi-source heterogeneous data from the machining system, and the spatiotemporal synchronization of the multi-source heterogeneous data using spatiotemporal synchronization technology to obtain a unified multi-source dataset, includes: Acquire at least two types of data from different sources from the machining system; By using spatiotemporal synchronization technology, the data from different sources are aligned to a unified processing timeline and a unified coordinate domain, forming a unified time domain dataset and a spatial domain dataset. The unified multi-source dataset includes a unified time domain dataset and a unified spatial domain dataset.

[0007] In some specific embodiments, the step of performing spatiotemporal semantic annotation on the tag information related to the processing process entered by the user in at least one modality, and associating it with the unified multi-source dataset to obtain structured association units, includes: Acquire label information related to the processing procedure entered by the user in at least one modal format; Automatically label each tag with its corresponding timestamp and / or spatial coordinates on the processing timeline to obtain multimodal tags with spatiotemporal information; The multimodal labels with spatiotemporal information are associated with the unified multi-source dataset to obtain structured association units.

[0008] In some specific embodiments, the step of performing semantic understanding and structured implicit knowledge extraction on the structured associated units to construct a machining knowledge and experience base includes: The semantic understanding model is invoked to perform semantic understanding on the structured related units, and structured knowledge units that can be stored, retrieved and reused by computers are extracted; The extracted structured knowledge units are organized into a mechanical processing knowledge and experience base.

[0009] In some specific embodiments, the decision-making requirement is a consultation request. The process of obtaining the decision-making requirement initiated by the user in at least one modal form, and generating and outputting response suggestions for the decision-making requirement based on the machining knowledge and experience base, includes: Obtain consultation requests initiated by users in at least one modality; The semantic understanding model is invoked to perform semantic understanding and intent recognition on the consultation request, extracting the mechanical processing system elements, question types, and granularity levels of interest involved, and obtaining the parsed consultation request; Based on the parsed consultation request, query conditions are constructed, and historical knowledge related to the consultation request is retrieved from the mechanical processing knowledge and experience base; Based on the retrieved historical knowledge, a response suggestion for the consultation request is generated and output through an intelligent decision-making algorithm.

[0010] In some specific embodiments, the method further includes: while outputting response suggestions for the consultation request, also outputting the source of the basis for the response suggestions, and automatically recording the consultation request and its response suggestions as input for subsequent tracing and iteration.

[0011] In some specific embodiments, the decision requirement is a traceability query. The process of obtaining the decision requirement initiated by the user in at least one modal form, and generating and outputting response suggestions for the decision requirement based on the machining knowledge and experience base, includes: Obtain traceability queries initiated by users in at least one modality; The semantic understanding model is invoked to perform semantic understanding and intent recognition on the traceability query, extracting the mechanical processing system elements, problem types, and granularity levels of interest involved, and obtaining the parsed traceability query; Based on the parsed traceability query, query conditions are constructed, query results related to the traceability query are retrieved from the machining knowledge and experience base, and response suggestions including the query results are output.

[0012] In some specific embodiments, obtaining the actual effectiveness of the response suggestions in actual processing and updating the machining knowledge and experience base based on the actual effectiveness includes: Obtain the actual effectiveness of the response suggestions in actual processing, evaluate the actual effectiveness, and generate evaluation results; The machining knowledge and experience base is updated and iterated based on the evaluation results.

[0013] A second aspect of the present invention provides an electronic device, including a processor and a memory, wherein the memory stores a computer program, and the processor, when executing the program, implements the collaborative optimization and full-process traceability method for machining systems provided in the first aspect.

[0014] A third aspect of the present invention provides a computer storage medium storing a computer program, the computer program including program instructions, which, when executed by a processor, cause the processor to execute the collaborative optimization and full-process traceability method for machining systems provided in the first aspect above.

[0015] Compared with related technologies, this invention provides an embodiment of a method, equipment, and storage medium for collaborative optimization and full-process traceability of machining systems. This method constructs a core process of "multi-source heterogeneous data integration - multimodal semantic annotation - structured knowledge and experience base construction - general intelligent consultation - full lifecycle traceability and closed-loop iteration." It provides an intelligent management and control solution for machining systems capable of integrating multi-source heterogeneous information, supporting multimodal semantic association, achieving structured knowledge accumulation, providing general intelligent consultation, and possessing full lifecycle closed-loop iteration capabilities. This enables general intelligent consultation, decision support, and collaborative optimization of all elements in the machining system, including machine tools, cutting tools, fixtures, workpieces, process parameters, quality results, and environmental conditions, truly promoting the development of machining towards intelligence and knowledge-driven approaches. This solves the problems of existing machining optimization and traceability technologies, such as weak multi-source data fusion capabilities, lack of multimodal semantic association, poor reusability of domain knowledge, and lack of closed-loop iterative optimization mechanisms. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating a collaborative optimization and full-process traceability method for a machining system provided by the present invention; Figure 2 This is a schematic diagram of the structure of an electronic device provided by the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without inventive effort are within the scope of protection of this invention.

[0019] It should be noted that, unless otherwise specified, the various features in the embodiments of this invention can be combined with each other, all of which are within the protection scope of this invention. Furthermore, although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than the module division in the device or the order in the flowchart. Moreover, the terms "first," "second," and "third" used in this invention do not limit the data or execution order, but only distinguish identical or similar items with essentially the same function and effect.

[0020] Machining is a core foundational process in high-end equipment manufacturing and precision component production. Monitoring the status of the machining system, optimizing processes, and ensuring traceability are crucial for guaranteeing machining accuracy, operational stability, and product quality. Currently, optimization and traceability solutions in the machining field primarily rely on fixed threshold alarms, timing signal anomaly detection, and independent information management systems to achieve basic control.

[0021] Traditional solutions of this type have low integration and intelligence levels, making them difficult to adapt to precision intelligent machining scenarios with multi-parameter coupling and dynamic changes in working conditions. They suffer from numerous technical defects and cannot meet the high-precision, traceable, and sustainable iterative production needs of modern machining. The specific defects are as follows: First, the fragmentation of multi-source heterogeneous data and insufficient data fusion capabilities are problems. Machining systems generate massive amounts of multi-dimensional data during operation, encompassing equipment motion status data, physical response data during processing, equipment and material properties, and processing simulation data—all from diverse sources and exhibiting heterogeneous information. However, traditional monitoring and management solutions employ independent data acquisition and processing systems, lacking a unified data fusion architecture and collaborative processing mechanism. This prevents effective integration and correlation analysis of the aforementioned heterogeneous data, resulting in scattered data storage and independent applications at each stage, creating a severe data silo problem.

[0022] Second, it lacks multimodal semantic association capabilities. Traditional solutions rely solely on sensor time-series numerical data for anomaly detection, achieving only numerical-level anomaly identification and failing to achieve spatiotemporal correlation and fusion of multimodal information. In actual manufacturing scenarios, processing anomalies, equipment malfunction characteristics, and product quality defects can often be intuitively reflected through non-numerical modal information such as text descriptions, on-site images, and operating audio. Examples include textual descriptions of tool wear and vibration anomalies recorded by engineers, real-life images of workpiece surface defects, and audio recordings of abnormal equipment operation. Therefore, traditional solutions cannot perform spatiotemporal correlation matching between quantitative numerical data and multimodal information such as textual descriptions of working conditions, workpiece defect images, and equipment operating audio, resulting in a lack of contextual semantic interpretation and poor interpretability of the detection and analysis results. Furthermore, this approach cannot effectively integrate engineers' on-site practical experience and struggles to leverage prior industry knowledge to assist in intelligent analysis and anomaly assessment.

[0023] Third, tacit knowledge in the machining industry cannot be structured, preserved, and reused. The machining industry is highly experience-dependent; tacit knowledge such as machining process debugging experience, equipment fault diagnosis cases, daily equipment maintenance records, and abnormal operating condition handling plans are core resources supporting the efficient and stable operation of machining systems. Currently, this type of tacit knowledge mostly relies on the long-term accumulation and retention of senior engineers, or is stored and archived only in unstructured forms such as paper documents and scattered electronic files. The industry lacks a systematic semantic understanding, intelligent knowledge extraction, and standardized preservation mechanism specifically for the machining field. This results in the inability to mine, organize, and integrate massive amounts of tacit machining knowledge into a structured, standardized, searchable, and callable general knowledge base. Knowledge reuse is low, failing to provide systematic knowledge support for machining status monitoring, fault diagnosis, process optimization, and intelligent decision-making, severely restricting the iterative upgrading of the intelligent decision-making capabilities of machining systems.

[0024] Fourth, the decision-making functions are limited and the system lacks versatility. Existing intelligent analysis and decision-making systems for machining are all developed for specific purposes, with limited functional modules that can only adapt to a single specific business task. For example, they can only achieve specific functions such as optimizing a single process parameter or predicting the lifespan of a single type of tool. Such fragmented technical solutions cannot adapt to the diversified and all-scenario decision-making needs of modern machining. They are unable to cover all-dimensional business scenarios such as comprehensive assessment of the overall health status of machine tools, tracing and diagnosing machining quality anomalies, intelligent analysis of equipment failures, and targeted maintenance strategy recommendations. They cannot build a universal intelligent consultation and decision-making system for the entire machining system. The system's adaptability, scalability, and versatility are extremely poor, making it difficult to meet the needs of large-scale, standardized, and intelligent machining production.

[0025] Fifth, there is a lack of a closed-loop iterative optimization mechanism throughout the entire lifecycle. Current machining monitoring and management systems mostly operate in a unidirectional open-loop mode, only capable of collecting, displaying, and providing basic early warnings for machining process data. They cannot achieve closed-loop control across the entire machining scenario and process. Specifically, existing technologies cannot review, evaluate, extract knowledge, or summarize patterns from the entire lifecycle of data, including machining execution processes, intelligent consultation records, equipment maintenance operations, and anomaly handling results. This makes it difficult to dynamically update and continuously optimize the knowledge base. This open-loop operation mode prevents the system from forming an autonomous iterative closed loop of "scenario use - result feedback - knowledge extraction - system update." The system's monitoring accuracy, decision-making ability, and adaptability cannot be continuously optimized with production operations. After long-term operation, its intelligence level is difficult to improve, and it cannot adapt to complex and ever-changing machining conditions.

[0026] In summary, existing machining optimization and traceability technologies generally suffer from problems such as weak multi-source data fusion capabilities, lack of multimodal semantic association, poor reusability of domain knowledge, and lack of closed-loop iterative optimization mechanisms, making it difficult to meet the production needs of modern machining for high precision, traceability, and sustainable iteration.

[0027] Therefore, it is necessary to propose an intelligent control scheme for machining systems that can achieve multi-source data fusion, multi-modal knowledge association, structured knowledge reuse, intelligent decision-making in all scenarios, and closed-loop iteration capability throughout the entire life cycle.

[0028] Please see Figure 1 This invention provides a method for collaborative optimization and full-process traceability of a machining system, comprising: S1. Acquire multi-source heterogeneous data from the machining system, and perform spatiotemporal synchronization of the multi-source heterogeneous data using spatiotemporal synchronization technology to obtain a unified multi-source dataset. S2. Perform spatiotemporal semantic annotation on the tag information related to the processing process entered by the user in at least one modality, and associate it with the unified multi-source dataset to obtain structured association units; S3. Perform semantic understanding and structured implicit knowledge extraction on the structured associated units to construct a mechanical processing knowledge and experience base; S4. Obtain decision requirements initiated by users in at least one modality, and generate and output response suggestions for the decision requirements based on the machining knowledge and experience base; S5. Obtain the actual effectiveness of the response suggestions in actual processing, and update the machining knowledge and experience base according to the actual effectiveness.

[0029] This embodiment provides a method for collaborative optimization and full-process traceability of machining systems. By constructing a core process of "multi-source heterogeneous data integration - multimodal semantic annotation - structured knowledge and experience base construction - general intelligent consultation - full lifecycle traceability and closed-loop iteration," it can provide an intelligent management and control solution for machining systems that can integrate multi-source heterogeneous information, support multimodal semantic association, achieve structured knowledge accumulation, provide general intelligent consultation, and possess full lifecycle closed-loop iteration capabilities. This enables general intelligent consultation, decision support, and collaborative optimization of all elements in the machining system, including machine tools, cutting tools, fixtures, workpieces, process parameters, quality results, and environmental conditions, truly promoting the development of machining towards intelligence and knowledge-driven approaches. This solves the problems of existing machining optimization and traceability technologies, such as weak multi-source data fusion capabilities, lack of multimodal semantic association, poor reusability of domain knowledge, and lack of closed-loop iterative optimization mechanisms.

[0030] In one embodiment, in step S1, multi-source heterogeneous data from a machining system is acquired, and the multi-source heterogeneous data is spatiotemporally synchronized using spatiotemporal synchronization technology to obtain a unified multi-source dataset, specifically including: S11. Obtain at least two types of data from different sources from the machining system.

[0031] In this invention, the machining system is a whole encompassing all participating elements in the machining process, including but not limited to machine tools, cutting tools, fixtures, workpieces, coolant, process parameters, quality requirements, and environmental conditions.

[0032] Multi-source heterogeneous data originates from datasets with different channels, structures, and sampling characteristics, including but not limited to: (1) Data acquired based on CNC system communication and sensor acquisition technology: motion and status data, such as machine tool axis coordinates, spindle speed, feed rate, tool position, etc.; physical response data, such as vibration, power, current, temperature, acoustic emission, cutting force and other sensor signals; (2) Data obtained from enterprise information management channels: equipment and material data, such as tool model, workpiece material, fixture status, coolant parameters, etc.; processing requirements and quality data, such as quality requirements, quality inspection results, etc. (3) Data obtained based on machining simulation software: kinematic simulation data, such as coordinates, spindle speed, feed rate, etc. generated by machine tool machining motion simulation; physical simulation data, such as cutting force, cutting heat, cutting vibration, etc. generated by machine tool machining cutting simulation.

[0033] S12. Through spatiotemporal synchronization technology, the data from different sources are aligned to a unified processing timeline and a unified coordinate domain to form a unified time domain dataset and a spatial domain dataset. The unified multi-source dataset includes a unified time domain dataset and a unified spatial domain dataset.

[0034] In this invention, spatiotemporal synchronization technology is a technical process that aligns data from different sources and with different sampling frequencies on the time axis and registers them on the spatial coordinate system to form a unified time domain dataset and a spatial domain dataset, thereby forming a unified digital description of the processing system.

[0035] In step S12, using spatiotemporal synchronization technology, the data from different sources are aligned to a unified processing timeline and a unified coordinate domain, forming a unified time domain dataset and a spatial domain dataset. The specific process includes: S121. Time dimension synchronization: Align to a unified reference time axis (processing timeline), that is, based on high-precision hardware timestamps, unify multi-source heterogeneous data onto the reference time axis.

[0036] The goal of time dimension synchronization is to ensure that all data is based on the same hardware time reference and accurately records the moment of its occurrence. This can be achieved by setting a unified high-precision hardware clock reference for all acquisition devices (such as CNC systems and external sensors) or by using a shared triggering mechanism (such as a single external signal triggering all devices to start acquisition simultaneously).

[0037] S122. Spatial Dimension Registration: Alignment to a unified coordinate domain, that is, mapping the acquired coordinates to the simulation coordinates to form a dataset of "processing position - acquired or simulated signal" based on the spatial domain.

[0038] The goal of spatial registration is to ensure that all measurement data are based on the same coordinate measurement datum (usually the actual coordinate location where the data was acquired), thus clearly defining the physical location where each measurement occurred. The process includes: Coordinate System 1 and Transformation: Select a reference coordinate system (usually the currently used machining coordinate system), and through coordinate system transformation, convert both the coordinate data calculated in the virtual simulation and the coordinate data actually collected to this unified reference coordinate system.

[0039] Spatial correlation of virtual and real data: Using virtual and real coordinate data as a medium, non-coordinate data from virtual simulation and actual acquisition are correlated with the actual acquired tool tip coordinate data. For example, simulated cutting force, vibration, and other signals at a specific moment are mapped to the actual spatial coordinate point where the tool is located at that moment through methods such as spatial grid mapping and nearest-distance coordinate correlation.

[0040] Therefore, spatiotemporal synchronization involves aligning heterogeneous data from different sources and with different sampling frequencies onto a unified reference time axis (processing timeline) in terms of time, using a unified timestamp; and in terms of space, mapping all virtual or real-sampled data onto a unified coordinate reference through coordinate system transformation and spatial correlation, forming a spatial domain-based dataset of "processing location - acquired or simulated signal". The ultimate goal of spatiotemporal synchronization is to create a high-quality dataset that is precisely aligned in both time and space, i.e., to form a unified time-domain dataset and a spatial-domain dataset, providing a complete data foundation for subsequent analysis and decision-making.

[0041] In this embodiment, by acquiring and effectively integrating at least two types of multi-source heterogeneous data from different sources with varying structures and sampling characteristics from the machining system, and then using spatiotemporal synchronization technology, data with different sampling frequencies, sources, and structures are aligned to a unified machining timeline and a unified coordinate domain, forming a time-domain-based dataset and a space-domain-based dataset. This results in a unified digital description of the machining system, enabling the fusion of multi-source heterogeneous data and comprehensive perception of the machining system. This overcomes the problem of data silos and forms a unified digital description of all elements, including machine tools, cutting tools, fixtures, workpieces, processes, and quality, providing a complete data foundation for subsequent analysis and decision-making.

[0042] In one embodiment, in step S2, the user-entered label information related to the processing procedure in at least one modality is semantically annotated in the spatiotemporal dimension and associated with the unified multi-source dataset to obtain structured association units, including: S21. Obtain label information related to the processing process entered by the user in at least one modal form.

[0043] The system provides an interactive interface that allows users to input label information related to the processing in one or more modalities, such as text, images, and audio.

[0044] The label information can cover any dimension of the machining system, such as machine tool status, tool information, process parameters, quality results, etc. At the same time, the granularity level of the label information can be selected according to traceability requirements, such as a certain production batch, a certain specific workpiece, a certain process or several processes on a single workpiece, a certain local area on a single workpiece, etc.

[0045] S22. Automatically label each tag with its corresponding timestamp and / or spatial coordinates on the processing timeline, thereby binding the tag to the actual processed product in the spatiotemporal dimension and obtaining a multimodal tag with spatiotemporal information.

[0046] Multimodal semantic annotation allows users to input tag information related to the processing process in one or more modal forms such as text, images, and audio. The system automatically records the timestamp and / or spatial coordinates, thus binding the tags to the processed products in the spatiotemporal dimension.

[0047] Specifically, the system automatically records the timestamp and / or spatial coordinates on the corresponding processing timeline for each tag, thereby achieving precise binding between the tag and the actual processed product in the spatiotemporal dimension, forming a semantic anchor point of "spatiotemporal-element-event".

[0048] In this invention, the "spatiotemporal-element-event" semantic anchor is an associated structure consisting of a timestamp or spatial coordinate, machining system elements (e.g., machine tools, cutting tools), and event descriptions (e.g., anomalies, defects), used to accurately identify specific states or events in the machining process.

[0049] S23. Associate the multimodal labels with spatiotemporal information with the unified multi-source dataset to obtain structured association units.

[0050] Specifically, the system combines the unified multi-source dataset obtained in step S1, the multimodal labels obtained in step S22, and their spatiotemporal information into a structured "system state-event" association unit.

[0051] The associated unit is a structured data unit that combines multi-source heterogeneous data, multimodal labels and their spatiotemporal information. It can simultaneously contain multimodal raw data or their feature representations, such as numerical data, text, images, and audio, to retain rich semantic information for use in the subsequent construction of a mechanical processing knowledge and experience base.

[0052] In this embodiment, by performing spatiotemporal semantic annotation on the tag information related to the processing process entered by the user in at least one modality such as text, image, and audio, and associating it with the unified multi-source dataset, structured association units are obtained. This enables multimodal semantic association, binding processing phenomenon tags with spatiotemporal numerical data, allowing the machining system to understand and utilize human experience and knowledge. This enhances the interpretability and experience utilization of processing status analysis, and fills the gap in the difficulty of incorporating semantic information into purely numerical-driven methods.

[0053] In one embodiment, step S3, which involves semantic understanding and structured implicit knowledge extraction of the structured associated units to construct a machining knowledge and experience base, specifically includes: S31. Call the semantic understanding model to perform semantic understanding on the structured associated units, and extract structured knowledge units that can be stored, retrieved and reused by the computer.

[0054] Specifically, the semantic understanding model is invoked to perform semantic understanding on the associated units generated in step S2, including but not limited to: natural language parsing of text labels, visual content recognition of images (such as wear, cracks, and defects), and speech-to-text conversion and acoustic event detection of audio.

[0055] After semantic understanding is completed, multimodal information is fused to identify the mechanical processing system elements involved (such as machine tools, cutting tools, fixtures, workpieces, etc.), state descriptions (such as wear, loosening, overheating, failure, etc.), causal relationships (such as "B is caused by A"), and granular level information.

[0056] Furthermore, the semantic understanding model is invoked to perform in-depth mining of the semantically understood information, extracting structured knowledge units that can be stored, retrieved, and reused by computers. For example, rules or patterns can be formed in the form of "when a machining system is in state combination X, it usually triggers event Y" or "at the granularity level L, the label Z of dimension D corresponds to the quality result Q".

[0057] In this invention, the semantic understanding model is an intelligent model used to parse, fuse, and deeply mine one or more modal information such as text, images, and audio. It can identify processing system elements, state descriptions, causal relationships, and granular levels.

[0058] Structured knowledge units are forms of knowledge extracted from multimodal semantic information that can be stored, retrieved, and reused by computers. Examples include rules or patterns such as "when a machining system is in state combination X, it usually triggers event Y" or "at the granularity level L, the label Z of dimension D corresponds to the quality result Q".

[0059] Transforming structured relational units into reusable structured knowledge units is a process of refining data into knowledge. Its core lies in invoking a semantic understanding model, based on a pre-defined knowledge representation model, to accurately extract entities, attributes, and their relationships from multimodal data. The specific process is as follows: Step 1: Call the semantic understanding model to perform semantic parsing on the structured related units.

[0060] The purpose of this step is to transform raw multimodal data (such as text, images, and audio) into preliminary semantic information that machines can understand.

[0061] Text parsing: For text tags or descriptions in associated units, the semantic understanding model performs entity recognition and relation extraction. Specifically, the semantic understanding model performs professional natural language parsing tasks, segmenting, extracting entities, disambiguating semantics, and understanding intent from text content such as machining process descriptions, quality inspection notes, equipment alarm texts, and manually labeled defect tags. It accurately extracts core entities such as machine tools, machining processes, workpiece defects, and equipment malfunctions, as well as textual semantic information such as state descriptions, degree representations, and operation instructions, discarding invalid and redundant text and outputting standardized textual semantic features. In the field of machining, this typically means identifying key entities such as cutting tools, workpieces, cutting parameters, machining features, and machine tool status. For example, for a text tag "High-speed steel cutting tools were used in the precision milling process," the semantic understanding model will identify key information such as "precision milling" (process), "high-speed steel" (material), and "cutting tool" (entity).

[0062] Image Analysis: For image data, the semantic understanding model performs visual content recognition, including target detection and classification of part images (such as identifying "scratches," "cracks," or "wear"), or analysis of engineering drawings. Specifically, the semantic understanding model performs industrial visual semantic recognition, performing feature detection and content understanding on workpiece surface images, tool edge images, and machining area images in related units. It accurately identifies various machining defects and state features such as tool wear, workpiece cracks, surface scratches, dimensional deviations, and machining burrs, while locating the spatial location, morphological features, and severity of defects, outputting visual recognition results with spatial and attribute semantics.

[0063] Audio parsing: For audio data, the semantic understanding model performs speech-to-text conversion or acoustic event detection, transforming equipment operating sounds into text or structured event descriptions. Specifically, the semantic understanding model performs acoustic semantic parsing, detecting acoustic events and identifying abnormal voiceprints in cutting vibration audio and equipment operating acoustic signals. It then translates valid acoustic events into standardized text descriptions, accurately distinguishing between normal cutting sounds, spindle noises, tool chipping noises, and abnormal equipment friction, extracting the semantic information of the equipment's operating status corresponding to the audio. For example, "Abnormal vibration occurred during the cutting process."

[0064] The second step is to use a pre-defined knowledge representation model to extract knowledge from the initial semantic information and construct structured knowledge units that can be stored, retrieved, and reused by computers.

[0065] In this invention, a pre-defined knowledge representation model predefines concepts, attributes, and explicit relationships within the field of machining. This model informs the semantic understanding model what knowledge to extract and how to organize it. The semantic understanding model operates based on this pre-defined knowledge representation model. For example, if the pre-defined knowledge representation model defines a "process" as including attributes such as "tools used," "cutting parameters," and "machining time," the semantic understanding model will selectively extract this information from text and images. The pre-defined knowledge representation model organizes knowledge around the core concept of "completing a specific machining task," enabling the extracted information to directly serve specific business scenarios such as process planning and fault diagnosis.

[0066] Guided by a pre-defined knowledge representation model, a semantic understanding model is invoked to extract knowledge from preliminary semantic information, constructing structured knowledge units that can be stored, retrieved, and reused by computers. The representation forms of these structured knowledge units include, but are not limited to, triples, rule sets, decision trees, and case libraries.

[0067] Taking the structured knowledge unit representation as a triple as an example, the specific process of constructing a structured knowledge unit is as follows: Triple extraction: The semantic understanding model extracts a massive number of triples from the semantic parsing results.

[0068] Entity recognition: Identifying the subject and object within a knowledge unit. For example, tool A, workpiece B, and finish milling process.

[0069] Relationship extraction: Identifying the semantic relationships between subjects and objects. For example, tool A is "used" in the finish milling process, and the finish milling process "produces" workpiece B. The semantic understanding model is based on spatiotemporal binding relationships, associating process numerical states, image defect features, acoustic anomaly signals, and text annotation information under the same processing time period and processing coordinates. It deeply mines the element associations, state couplings, causal logic, and hierarchical granular relationships within the processing system. On the one hand, it distinguishes data granularity levels, clarifying different levels of semantics such as overall processing conditions, single-process states, and single-workpiece defects; on the other hand, it identifies the linkage relationships of multi-dimensional elements, mining implicit association rules between process parameters, equipment status, processing quality, and equipment failures, breaking the limitations of isolated single-modal semantics.

[0070] Generating Rules or Patterns: In addition to specific fact triples, semantic understanding models can also generate higher-order rules or patterns through induction. For example, by analyzing a large number of images of "tool wear," time-series data of "increased cutting force," and text records of "excessive surface roughness," the semantic understanding model can summarize a rule: "When the tool flank wear (VB) exceeds 0.3mm (from the image) and the cutting force continues to increase by more than 15% (from the time-series data), the surface roughness is likely to exceed the standard (from the text)." This is similar to the knowledge form mentioned above, such as "When a machining system is in state combination X, it usually triggers event Y."

[0071] S32. Organize the extracted structured knowledge units into a mechanical processing knowledge and experience base.

[0072] To facilitate computer storage, retrieval, and reuse, all extracted structured knowledge units are organized into a machining knowledge and experience base. This machining knowledge and experience base can be retrieved, reasoned about, and applied in subsequent processes. It supports incremental updates and can continuously absorb new multimodal tag information.

[0073] A machining knowledge and experience base is a searchable and reusable knowledge repository formed through semantic understanding and structured extraction. Its implementation forms include, but are not limited to, knowledge graphs, vector databases, document databases, rule bases / expert system knowledge bases, etc. For example, in the case of using a knowledge graph as the experience base implementation, the experience base consists of interconnected nodes (representing entities) and edges (representing relationships). Nodes in a multimodal knowledge graph not only contain textual information but also directly relate to original images (such as photos of parts), audio (such as equipment voiceprints), or other non-textual data, enabling computers to efficiently perform complex semantic queries and reasoning. For example, it is easy to retrieve "all carbide cutting tools used in finish milling processes." The machining knowledge and experience base supports incremental updates, providing knowledge support for intelligent consultation and decision-making.

[0074] In this embodiment, by performing semantic understanding and structured tacit knowledge extraction on the structured associated units, unstructured multimodal user feedback is automatically transformed into searchable and reusable structured knowledge units, constructing a mechanical processing knowledge and experience base. This reduces reliance on personal experience, improves the efficiency of knowledge inheritance and reuse, realizes the structured accumulation of tacit knowledge, enables enterprise-level reuse of knowledge, and provides knowledge support for intelligent consultation and decision-making.

[0075] In one embodiment, in step S4, a decision requirement initiated by a user in at least one modal form is obtained, and a response suggestion for the decision requirement is generated and output based on the machining knowledge and experience base.

[0076] This embodiment addresses diverse decision-making needs in machining systems (such as machine tool health assessment, tool performance evaluation, parameter optimization, quality anomaly diagnosis, and maintenance recommendations), providing general intelligent consultation and decision support capabilities based on historical knowledge, rather than being limited to specific tasks. It caters to any decision-making need within a machining system, offering users intelligent consultation and decision support services based on historical knowledge. These decision-making needs include, but are not limited to: machine tool health status assessment and maintenance recommendations, tool performance evaluation and selection recommendations, machining parameter optimization, quality anomaly cause diagnosis and improvement measures, process path planning, and fixture selection.

[0077] The decision-making needs include consultation requests and traceability queries. Consultation requests are intelligent consultation requests initiated by users in one or more modalities, such as text, images, and audio, and are tailored to any decision-making need. The system provides general intelligent consultation responses based on historical knowledge.

[0078] Traceability query is full lifecycle traceability, which is the ability to record, query and analyze the knowledge, processes and results generated in all scenarios such as processing execution, intelligent consulting, equipment maintenance and quality improvement.

[0079] In a specific embodiment, if the decision-making requirement is a consultation request, then in step S4, the decision-making requirement initiated by the user in at least one modal form is obtained, and response suggestions for the decision-making requirement are generated and output based on the machining knowledge and experience base, including: SA41. Obtain consultation requests initiated by users in at least one modality.

[0080] The system receives consultation requests from users in one or more modalities, such as natural language text, images, and audio. For example, a user can upload a photo of tool wear and ask "Can this tool still be used?", or enter the text "The current spindle vibration is too large, what are the possible causes?", or upload a surface image of an abnormal workpiece and ask "What are the possible factors that cause this defect?"

[0081] SA42. Call the semantic understanding model to perform semantic understanding and intent recognition on the consultation request, extract the mechanical processing system elements, problem type and the level of granularity of concern involved, and obtain the parsed consultation request.

[0082] The system calls the semantic understanding model to perform semantic understanding and intent recognition on the consultation request, extracting the mechanical processing system elements involved (such as machine tools, cutting tools, workpieces, process parameters, etc.), problem types (such as evaluation, diagnosis, optimization, prediction, etc.), and the granularity level of interest (such as single piece, batch, region, process, etc.) to obtain the parsed consultation request.

[0083] SA43. Based on the parsed consultation request, construct query conditions and retrieve historical knowledge related to the consultation request from the machining knowledge and experience base.

[0084] Based on the parsed consultation request, the system automatically constructs query conditions and retrieves historical knowledge related to the consultation request from the machining knowledge and experience base. This historical knowledge includes: successful cases under similar working conditions, failure modes and avoidance rules, and correlation patterns between elements. The system supports cross-element and cross-granularity knowledge association reasoning. For example, it can retrieve "when the tool wear degree is X, which combinations of process parameters in history led to surface quality problems".

[0085] SA44. Based on the retrieved historical knowledge, generate and output response suggestions for the consultation request through an intelligent decision-making algorithm.

[0086] Specifically, generating response suggestions based on retrieved historical knowledge is a reasoning process that transforms data into decisions. Its core is invoking intelligent decision-making algorithms to generate interpretable and actionable outputs. These intelligent decision-making algorithms can take various forms, including but not limited to rule engines and intelligent reasoning, knowledge graph reasoning, and causal reasoning. This invention is not limited to specific combinations of algorithm types.

[0087] An implementation case using a combination of three specific algorithms—rule engine, knowledge graph reasoning, and causal reasoning—illustrates the process, which includes the following stages: The first stage is knowledge fusion and intelligent reasoning, which involves combining retrieved historical knowledge with system rules to conduct in-depth reasoning and obtain the reasoning results.

[0088] The system rules include a rule engine, inference strategies, and causal reasoning. The rule engine handles deterministic logic, such as checking if recommended parameters are within safe limits. Inference strategies include knowledge graph reasoning, which utilizes relationships between entities in a multimodal knowledge graph for multi-hop reasoning; for example, tracing the root cause of defects through the "process → material → equipment" chain. Causal reasoning traces the root cause of problems by analyzing causal relationships in data. For example, it uses techniques such as causal discovery and counterfactual reasoning to locate the source of a problem from multi-source data.

[0089] The second stage is response suggestion generation and output, where the system transforms the reasoning results into response suggestions that the user can understand.

[0090] Response suggestion generation: Directly outputs response suggestions for the current consultation request. The forms of response suggestions include, but are not limited to: quantitative evaluation results (such as health scores, risk levels), optimization plan recommendations (such as parameter adjustment suggestions, tool replacement timing), root cause analysis (such as "Based on historical data, this vibration anomaly is most likely caused by fixture loosening"), etc.

[0091] Explainability: A good system provides a basis for decision-making. For example, this parameter combination is recommended because it achieved a surface roughness of Ra0.8 in 87% of similar cases historically (material 45# steel, tool carbide).

[0092] Confidence assessment: The system can attach a normalized confidence score to the recommendations to help users assess the risks.

[0093] Multiple output formats for response suggestions: The final response suggestions can be presented in the form of text, charts, or even augmented reality (AR) visualizations.

[0094] SA45. While outputting response suggestions for the consultation request, it also outputs the source of the basis for the response suggestions and automatically records the consultation request and its response suggestions as input for subsequent tracking and iteration.

[0095] In this embodiment, by acquiring consultation requests initiated by users in at least one modal form, and generating and outputting response suggestions for the decision-making needs based on the machining knowledge and experience base, it is possible to provide intelligent responses based on historical knowledge for any decision-making needs such as machine tool health assessment, tool performance evaluation, parameter optimization, quality anomaly diagnosis, and maintenance suggestions. This breaks through the bottleneck of existing machining systems being limited to specific tasks, has high scene adaptability and scalability, enhances the general intelligent consultation capabilities of machining systems, and covers decision-making needs across all scenarios.

[0096] In another specific embodiment, where the decision requirement is a traceability query, then in step S4, the decision requirement initiated by the user in at least one modal form is obtained, and response suggestions for the decision requirement are generated and output based on the machining knowledge and experience base, including: SB41. Obtain traceability queries initiated by users in at least one modality.

[0097] The system receives traceability queries from users in one or more modal forms such as natural language text, images, and audio. For example: "What were the suggestions for spindle vibration in the last consultation?", "All consultation records related to tool wear in the past week", and "Show sensor data and user annotations for workpiece No. 5 in the third process".

[0098] SB42. Call the semantic understanding model to perform semantic understanding and intent recognition on the traceability query, extract the mechanical processing system elements, problem types and the granularity level of interest involved, and obtain the parsed traceability query.

[0099] SB43. Based on the parsed traceability query, construct query conditions, retrieve query results related to the traceability query from the machining knowledge and experience base, and output response suggestions including the query results.

[0100] The system uses semantic understanding and index retrieval to return query results related to the traceability query from the mechanical processing knowledge and experience base and historical records, and can also display the spatiotemporal tags, multimodal raw data and system suggestions at that time.

[0101] The specific implementation process of the traceability query embodiment of the present invention is similar to that of the consultation request embodiment. The specific implementation process of the traceability query embodiment can be referred to the specific implementation process of the consultation request embodiment, and will not be repeated here.

[0102] In this embodiment, by acquiring traceability queries initiated by users in at least one modality, and generating and outputting response suggestions for the traceability queries based on the machining knowledge and experience base, it is possible to achieve full lifecycle traceability of knowledge, processes and results generated in all scenarios such as machining execution, intelligent consultation, equipment maintenance, and quality improvement, thereby enhancing the traceability capability of the machining system.

[0103] In one embodiment, step S5, obtaining the actual effectiveness of the response suggestion in actual processing and updating the machining knowledge and experience base based on the actual effectiveness, includes: S51. Obtain the actual effectiveness of the response suggestions in actual processing, evaluate the actual effectiveness, and generate evaluation results.

[0104] The system acquires the actual processing results of response suggestions, the effectiveness of consultation adoption, or the actual results of maintenance actions, and evaluates the actual results based on evaluation criteria. Evaluation criteria include: subsequent user feedback (such as confirmed problem resolution, ratings, and correction suggestions), and objective result data collected by sensors (such as processing quality detection values ​​and vibration improvement amplitude).

[0105] The system automatically extracts new knowledge units or corrective information from the evaluation process, such as confirming the validity of a historical suggestion, discovering an anomaly pattern that was not previously recorded, or correcting an inaccurate association rule.

[0106] S52. Update and iterate the machining knowledge and experience base according to the evaluation results.

[0107] The new knowledge units or correction information extracted in step S51 are stored in the machining knowledge and experience base, triggering incremental updates and consistency maintenance of the machining knowledge and experience base.

[0108] This step enables the system to continuously evolve and self-optimize as it is used more, forming a complete closed-loop self-optimization mechanism of "use-feedback-refinement-update" without human intervention.

[0109] In this embodiment, by obtaining the actual effectiveness of the response suggestions in actual processing and updating the machining knowledge and experience base according to the actual effectiveness, it is possible to achieve result evaluation, knowledge extraction and continuous iteration of the machining knowledge and experience base for the entire life cycle of activities such as processing execution, consultation process and maintenance actions. This forms a full life cycle traceability and closed-loop self-optimization mechanism of "use-feedback-extraction-update", which enables the system to continuously evolve as the number of times it is used increases, thus getting rid of the limitations of traditional unidirectional open-loop data flow.

[0110] Based on the same inventive concept, please refer to Figure 2 , Figure 2This is a schematic diagram of the structure of an electronic device 500 provided in an embodiment of the present invention. In terms of hardware implementation, the electronic device 500 can be an AI computing platform. The electronic device 500 typically integrates heterogeneous computing units such as CPUs, GPUs, or NPUs to meet the parallel computing capabilities required for model inference.

[0111] Electronic device 500 includes one or more processors 51 and memory 52. ​​Memory 52 is connected to one or more processors 51, for example, via a bus.

[0112] Processor 51 is configured to support the electronic device in performing the corresponding functions in the methods described in the above method embodiments. Processor 51 may be a central processing unit (CPU), a network processor (NP), a hardware chip, or any combination thereof. The aforementioned hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The aforementioned PLD may be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.

[0113] Memory 52 is used to store program code, etc. Memory 52 may include volatile memory (VM), such as random access memory (RAM); memory 52 may also include non-volatile memory (NVM), such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid-state drive (SSD); memory 52 may also include combinations of the above types of memory.

[0114] The memory 52 can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules, such as the program instructions / modules corresponding to the methods in the embodiments of the present invention. The processor 51 executes various functional applications and data processing of the method and apparatus by running the non-volatile software programs, instructions, and modules stored in the memory 52, that is, it realizes the functions of each module or unit of the method and apparatus provided in the above method embodiments.

[0115] The memory 52 may include a program storage area and a data storage area, wherein the program storage area may store the operating system and applications required for at least one function. The data storage area may store data created based on the use of the device. In some embodiments, the memory 52 may optionally include memory remotely located relative to the processor 51, and this remote memory may be connected to the device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0116] The one or more modules are stored in the memory 52. ​​When executed by the one or more processors 51, they perform the methods in any of the above method embodiments. For example, they perform the method steps described in the above method embodiments to realize the functions of the modules described in the above device embodiments.

[0117] This invention also provides a computer-readable storage medium storing a computer program, the computer program including program instructions that, when executed by a computer processor, cause the processor to perform the method described in the foregoing embodiments.

[0118] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0119] The above description discloses only preferred embodiments of the present invention and should not be construed as limiting the scope of the present invention. Therefore, equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.

Claims

1. A method for collaborative optimization and full-process traceability of a machining system, characterized in that, include: Multi-source heterogeneous data from a machining system is acquired, and the multi-source heterogeneous data is spatiotemporally synchronized using spatiotemporal synchronization technology to obtain a unified multi-source dataset. The user-entered label information related to the processing process in at least one modality is semantically annotated in the spatiotemporal dimension and associated with the unified multi-source dataset to obtain structured association units; Semantic understanding and structured implicit knowledge extraction are performed on the structured associated units to construct a mechanical processing knowledge and experience base; The system acquires decision-making requirements initiated by users in at least one modality, and generates and outputs response suggestions for the decision-making requirements based on the machining knowledge and experience base. Obtain the actual effectiveness of the response suggestions in actual processing, and update and iterate the machining knowledge and experience base based on the actual effectiveness.

2. The method according to claim 1, characterized in that, The process involves acquiring multi-source heterogeneous data from a machining system, and then using spatiotemporal synchronization technology to synchronize this data to obtain a unified multi-source dataset, including: Acquire at least two types of data from different sources from the machining system; By using spatiotemporal synchronization technology, the data from different sources are aligned to a unified processing timeline and a unified coordinate domain, forming a unified time domain dataset and a spatial domain dataset. The unified multi-source dataset includes a unified time domain dataset and a unified spatial domain dataset.

3. The method according to claim 1, characterized in that, The process involves performing spatiotemporal semantic annotation on the tag information related to the processing process entered by the user in at least one modality, and associating it with the unified multi-source dataset to obtain structured association units, including: Acquire label information related to the processing procedure entered by the user in at least one modal format; Automatically label each tag with its corresponding timestamp and / or spatial coordinates on the processing timeline to obtain multimodal tags with spatiotemporal information; The multimodal labels with spatiotemporal information are associated with the unified multi-source dataset to obtain structured association units.

4. The method according to claim 1, characterized in that, The step of performing semantic understanding and structured implicit knowledge extraction on the structured associated units to construct a machining knowledge and experience base includes: The semantic understanding model is invoked to perform semantic understanding on the structured related units, and structured knowledge units that can be stored, retrieved and reused by computers are extracted. The extracted structured knowledge units are organized into a mechanical processing knowledge and experience base.

5. The method according to claim 1, characterized in that, The decision-making requirement is a consultation request. The process of obtaining decision-making requirements initiated by users in at least one modal form, and generating and outputting response suggestions for the decision-making requirement based on the machining knowledge and experience base, includes: Obtain consultation requests initiated by users in at least one modality; The semantic understanding model is invoked to perform semantic understanding and intent recognition on the consultation request, extracting the mechanical processing system elements, question types, and granularity levels of interest involved, and obtaining the parsed consultation request; Based on the parsed consultation request, query conditions are constructed, and historical knowledge related to the consultation request is retrieved from the mechanical processing knowledge and experience base; Based on the retrieved historical knowledge, a response suggestion for the consultation request is generated and output through an intelligent decision-making algorithm.

6. The method according to claim 5, characterized in that, The method further includes: while outputting response suggestions for the consultation request, it also outputs the source of the basis for the response suggestions, and automatically records the consultation request and its response suggestions as input for subsequent tracing and iteration.

7. The method according to claim 1, characterized in that, The decision requirement is a traceable query. The process of obtaining decision requirements initiated by users in at least one modality, and generating and outputting response suggestions for the decision requirements based on the machining knowledge and experience base, includes: Obtain traceability queries initiated by users in at least one modality; The semantic understanding model is invoked to perform semantic understanding and intent recognition on the traceability query, extracting the mechanical processing system elements, problem types, and granularity levels of interest involved, and obtaining the parsed traceability query; Based on the parsed traceability query, query conditions are constructed, query results related to the traceability query are retrieved from the machining knowledge and experience base, and response suggestions including the query results are output.

8. The method according to claim 1, characterized in that, The step of obtaining the actual effectiveness of the response suggestions in actual processing and updating the machining knowledge and experience base based on the actual effectiveness includes: Obtain the actual effectiveness of the response suggestions in actual processing, evaluate the actual effectiveness, and generate evaluation results; The machining knowledge and experience base is updated and iterated based on the evaluation results.

9. An electronic device, characterized in that, It includes a processor and a memory, the memory storing a computer program, and the processor, when executing the program, implements the collaborative optimization and full-process traceability method for machining systems according to any one of claims 1 to 8.

10. A computer storage medium, characterized in that, The computer storage medium stores a computer program, which includes program instructions. When executed by a processor, the program instructions cause the processor to perform the collaborative optimization and full-process traceability method for machining systems as described in any one of claims 1 to 8.

Images