Intelligent operation and maintenance and evaluation method and system for cut tobacco making equipment, storage medium and equipment

Abstract

The invention discloses an intelligent operation and maintenance and evaluation method and system of cut tobacco making equipment, a storage medium and the equipment, and belongs to the technical field of intelligent operation and maintenance of industrial equipment, and the method comprises the steps: collecting multi-source heterogeneous data related to the cut tobacco making equipment; the multi-source heterogeneous data is preprocessed; based on the preprocessed multi-source heterogeneous data, executing the following two evaluation paths to obtain a numerical health score and a semantic health score; calculating an uncertainty measure between the numerical health score and the semantic health score; fusing the numerical health score and the semantic health score based on the uncertainty measure to obtain a fused health score; based on the fused health score, combining a safety index and a failure index, and calculating the comprehensive efficiency of the tobacco shred making equipment through a nonlinear comprehensive efficiency evaluation function; and automatically executing a corresponding operation and maintenance decision according to the risk level of the comprehensive efficiency. According to the invention, precise, self-adaptive and interpretable intelligent operation, maintenance and evaluation of the state of the cut tobacco making equipment are realized.

Description

Technical Field

[0001] This invention relates to the field of intelligent operation and maintenance technology for industrial equipment, and in particular to an intelligent operation and maintenance and evaluation method, system, storage medium and equipment for silk-making equipment. Background Technology

[0002] In the tobacco processing industry, the operating status of core equipment such as shredders, dryers, and rehumidifiers directly determines the physical properties of the tobacco and the sensory quality of the final cigarettes. Currently, equipment maintenance in tobacco processing workshops mainly relies on Industrial Internet of Things (IIoT) technology. This involves deploying sensors for vibration, temperature, and current to collect time-series data, which is then monitored using Supervisory Control and Data Acquisition (SCADA) systems or Equipment Management Systems (EAM).

[0003] The mainstream technical solutions for existing equipment operation and maintenance generally fall into two categories: one is traditional algorithms based on physical models, such as using Statistical Process Control (SPC) and Fast Fourier Transform (FFT) for frequency domain analysis, or constructing a fixed health evaluation system based on the Analytic Hierarchy Process (AHP); the other is data-driven shallow machine learning, such as using Random Forest, XGBoost, or K-Means clustering algorithms to classify and regress historical fault data. These technologies have reached a certain level of maturity in processing structured numerical data and are widely used in equipment threshold alarms and simple trend prediction.

[0004] Although existing technologies have enabled digital monitoring of equipment, the following technical bottlenecks remain insurmountable when dealing with the complex failure mechanisms and highly dynamic production environments of yarn-making equipment: 1. Numerical and semantic heterogeneous data cannot be effectively integrated, resulting in low knowledge utilization: Existing technologies primarily process structured numerical data from sensors, while a large amount of equipment maintenance knowledge (such as equipment maintenance manuals, fault case libraries, expert experience records, and shift handover logs) exists in unstructured text form. Traditional algorithms cannot understand the semantics of this text, resulting in a disconnect between sensor data and maintenance knowledge during fault diagnosis. For example, when a sensor captures an abnormal waveform, the system cannot automatically associate it with the corresponding description in the maintenance manual such as "incidental vibration caused by improper operation," still requiring manual document review, leading to low diagnostic efficiency and a high risk of misdiagnosis.

[0005] 2. Traditional evaluation models have static and rigid weights, lacking environmental adaptability: Existing health assessment systems (such as AHP or fixed-weighted averages) typically maintain fixed indicator weights (e.g., vibration 30%, temperature 20%) once established. However, the operating conditions of tobacco processing equipment are dynamic (e.g., different processing parameters for different grades of tobacco, or different stages of the equipment's aging cycle). Under specific failure modes (e.g., sudden overheating), the importance of temperature indicators should increase dramatically, but fixed-weighted models cannot respond dynamically to this. Consequently, when critical failure characteristics occur, the overall score is diluted by other normal indicators, thus masking the true risk.

[0006] 3. Lack of interpretability and reasoning ability, making it difficult to achieve closed-loop decision-making: Existing fault diagnosis models based on deep learning or ensemble learning (such as XGBoost) are generally considered "black boxes," only able to output fault classification probabilities (e.g., 85% probability of bearing failure) but unable to provide the logical reasoning chain (i.e., why it was determined to be a fault). Furthermore, existing systems cannot automatically generate specific maintenance strategies tailored to the current process context based on diagnostic results. When faced with alarm information, maintenance personnel still need to formulate maintenance plans based on personal experience, resulting in varying handling outcomes for the same fault and failing to achieve a standardized intelligent closed loop from "problem discovery" to "problem resolution."

[0007] 4. A single algorithm struggles to balance real-time performance with complex logical reasoning: Traditional numerical algorithms are fast but lack logical reasoning capabilities, while general large language models (LLMs), although possessing reasoning capabilities, have high computational overhead and are prone to illusions, making them unsuitable for direct application in demanding industrial real-time control. This results in insufficient capabilities when handling complex and rare systemic failures.

[0008] Therefore, there is an urgent need for an intelligent operation and maintenance and evaluation method for silk-making equipment that can integrate multi-source heterogeneous data, adaptive dynamic evaluation, and has interpretability while taking into account real-time performance and inference depth. Summary of the Invention

[0009] The purpose of this invention is to overcome the problems existing in the prior art and provide a method, system, storage medium and device for intelligent operation and maintenance and evaluation of silk-making equipment. By constructing a multi-agent collaborative architecture, it adaptively integrates traditional numerical algorithms (XGBoost, K-Means, AHP) based on sensor data with large model technology based on retrieval enhancement generation (RAG) of unstructured data, so as to achieve accurate evaluation and closed-loop management of the health, safety and failure status of equipment in the silk-making workshop.

[0010] The objective of this invention is achieved through the following technical solution: Firstly, a method for intelligent operation, maintenance, and evaluation of silk-making equipment is provided, including the following steps: S1. Collect multi-source heterogeneous data related to the silk-making equipment, wherein the multi-source heterogeneous data includes at least structured data collected in real time by sensors and unstructured text data related to operation and maintenance; S2. Preprocess the multi-source heterogeneous data; S3. Based on the preprocessed multi-source heterogeneous data, execute the following two evaluation paths: Numerical assessment path: Based on traditional numerical algorithms, calculations are performed on the preprocessed structured data to obtain a numerical health score; Semantic evaluation path: Based on a multi-agent collaborative architecture and retrieval enhancement generation technology, semantic understanding and reasoning are performed on the unstructured text data and structured data features to obtain a semantic health score; S4. Calculate the uncertainty measure between the numerical health score and the semantic health score; and fuse the numerical health score and the semantic health score based on the uncertainty measure to obtain the fused health score; S5. Based on the fused health score, combined with the safety index and failure index, the comprehensive performance of the yarn-making equipment is calculated using a nonlinear comprehensive performance evaluation function; S6. Based on the risk level of the overall performance, automatically execute the corresponding operation and maintenance decisions.

[0011] In some embodiments, the preprocessing includes: The structured data is subjected to outlier removal and standardization. The unstructured text data is transformed into a triplet knowledge graph, and the text paragraphs are transformed into vectors using a text embedding model.

[0012] In some embodiments, the numerical health score is calculated using the following formula: in, Indicates a numerical health score. This indicates the overall deviation of the silk-making equipment.

[0013] In some embodiments, the numerical evaluation path further includes: The K-Means algorithm is used to cluster historical data into normal operating condition clusters and abnormal operating condition clusters; Calculate the Euclidean distance from the real-time data point to the center of each operating condition cluster. If the Euclidean distance exceeds the threshold and the moving average trend is significantly negative, it is determined to be a potential failure.

[0014] In some embodiments, calculating the uncertainty measure between the numerical health score and the semantic health score includes: Cross-entropy is used to measure the degree of discrepancy between numerical health scores and semantic health scores.

[0015] In some embodiments, the fusion health score is calculated using the following formula: Among them, the This indicates the weight of the numerical health score. Indicates semantic health score, , This represents the weight adjustment parameter. It represents a measure of uncertainty.

[0016] In some embodiments, the overall performance is calculated by the following formula: Where H represents the health index and S represents the safety index. Indicates the failure index. Indicates the steepness coefficient. , , All of these represent weighting coefficients.

[0017] Secondly, a smart operation and maintenance and evaluation system for silk-making equipment is provided, including: The data perception layer is used to collect multi-source heterogeneous data related to the silk-making equipment. The multi-source heterogeneous data includes at least structured data collected by sensors in real time and unstructured text data related to operation and maintenance. The algorithm processing layer is used to preprocess the multi-source heterogeneous data; and based on the preprocessed multi-source heterogeneous data, executes the following two evaluation paths: Numerical assessment path: Based on traditional numerical algorithms, calculations are performed on the preprocessed structured data to obtain a numerical health score; Semantic evaluation path: Based on a multi-agent collaborative architecture and retrieval enhancement generation technology, semantic understanding and reasoning are performed on the unstructured text data and structured data features to obtain a semantic health score; The agent collaboration layer is used to calculate the uncertainty measure between the numerical health score and the semantic health score; and to fuse the numerical health score and the semantic health score based on the uncertainty measure to obtain a fused health score. The integrated decision-making layer calculates the overall performance of the yarn-making equipment based on the integrated health score, combined with the safety index and failure index, through a nonlinear comprehensive performance evaluation function; and automatically executes the corresponding operation and maintenance decisions according to the risk level of the comprehensive performance.

[0018] Thirdly, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program, characterized in that, when the computer program is executed by a processor, it implements the intelligent operation and maintenance and evaluation method for the silk-making equipment described in the first aspect.

[0019] Fourthly, an electronic device is provided, including a memory and a processor, wherein the memory stores computer instructions that can be executed on the processor, characterized in that the processor executes the intelligent operation and maintenance and evaluation method for the silk-making equipment described in the first aspect when executing the computer instructions.

[0020] It should be further noted that the technical features corresponding to the above-mentioned options and embodiments can be combined or substituted with each other to form new technical solutions without conflict.

[0021] Compared with the prior art, the beneficial effects of the present invention are: 1. It breaks through the barriers of heterogeneous data fusion, significantly improving the comprehensiveness and accuracy of fault diagnosis. This invention introduces Retrieval Enhanced Generation (RAG) technology and vectorized knowledge graphs, constructing a dual-engine architecture of "traditional numerical algorithms + large language models." It overcomes the limitation of existing technologies that can only utilize structured sensor data (such as vibration and temperature) and cannot effectively utilize massive amounts of unstructured text data (such as maintenance logs, operation manuals, and expert experience). By semantically associating real-time sensing features with historical fault cases, this invention significantly reduces misjudgments caused by a single data source, greatly improving the diagnostic accuracy for complex and sporadic faults, and enabling the discovery of implicit association rules that traditional algorithms cannot identify.

[0022] 2. It achieves adaptive dynamic adjustment of the evaluation model, solving the problem of poor robustness of a single model. This invention designs an adaptive weight adjustment mechanism based on an uncertainty metric (cross-entropy), which dynamically adjusts the fusion weights according to the degree of discrepancy between the scores of traditional algorithms and large models. This solves the problem of false negatives or missed positives caused by fixed and rigid weights in traditional AHP or machine learning models when facing novel, unseen faults or drastic fluctuations in operating conditions. Under normal fault conditions, this invention utilizes traditional algorithms to ensure response speed and accuracy; under complex fault conditions, it automatically increases the weights of the large model to leverage its inference capabilities. This mechanism significantly improves the system's adaptability and robustness in the variable production environment of the yarn-making workshop.

[0023] 3. It endows the intelligent operation and maintenance system with interpretability, reducing reliance on highly skilled personnel. This invention employs a multi-agent collaborative architecture, specifically setting up a reasoning agent and an interpretation / generation agent, capable of outputting natural language diagnostic reports containing logical reasoning chains. This solves the problem that existing deep learning models can only output alarm codes or probabilities, failing to inform maintenance personnel of the cause of the fault and the reasoning process. It enables ordinary maintenance workers to possess expert-level diagnostic capabilities, effectively alleviating the pressure of a shortage of highly skilled maintenance personnel in enterprises and shortening fault diagnosis time.

[0024] 4. The introduction of a nonlinear performance evaluation function significantly improves the sensitivity to early equipment degradation. This invention abandons the traditional linear weighted summation in comprehensive performance evaluation, instead employing an improved Sigmoid nonlinear transformation function to fuse scores across the three dimensions of health, safety, and failure. This solves the problem of insufficient sensitivity in linear evaluation models when equipment is at a critical point (i.e., the sub-healthy stage transitioning from normal to failure), leading to missed optimal maintenance windows. This nonlinear function exhibits a higher slope when indicators deviate from normal values ​​(the 0.5 critical zone), significantly amplifying signals of minor performance degradation. This means the system can detect potential equipment problems earlier than traditional methods, truly achieving predictive maintenance and thus extending equipment lifespan.

[0025] 5. It has achieved closed-loop management from monitoring to decision-making, directly improving production efficiency. This invention integrates a direct linkage system between the decision-making layer and the decision-making intelligence agent—a work order system and a spare parts inventory system—to automatically generate early warning work orders, spare parts requisition forms, or shutdown instructions based on risk levels. It solves the problem of existing technologies that "emphasize alarms but neglect handling," requiring manual intervention to check inventory and fill out work orders after an alarm, resulting in delayed response and cumbersome processes. It achieves a fully automated closed-loop operation and maintenance process. By reducing intermediate manual steps, it significantly shortens the mean time to repair (MTTR). It is expected to reduce equipment failure downtime by at least 5%, significantly improving the overall operational efficiency (OEE) and economic benefits of the yarn-making workshop. Attached Figure Description

[0026] Figure 1 This is a flowchart of an intelligent operation, maintenance, and evaluation method for silk-making equipment according to the present invention; Figure 2 This is a graph of the nonlinear comprehensive performance evaluation function of the present invention; Figure 3 This is a schematic diagram of the data flow for evaluating the effectiveness of this invention. Detailed Implementation

[0027] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] It should be noted that the defects in the solutions in the prior art are all the results of the inventors' practice and careful research. Therefore, the discovery process of the above problems and the solutions proposed by the embodiments of this application in the following text should be the inventors' contributions to this application in the process of invention and creation, and should not be understood as technical content known to those skilled in the art.

[0029] In view of the technical problems pointed out in the background art, the present invention provides the following embodiments: In one exemplary embodiment, a method for intelligent operation, maintenance, and evaluation of a silk-making equipment is provided, such as... Figure 1 As shown, it includes the following steps: S1. Collect multi-source heterogeneous data related to the silk-making equipment, wherein the multi-source heterogeneous data includes at least structured data collected in real time by sensors and unstructured text data related to operation and maintenance; S2. Preprocess the multi-source heterogeneous data; S3. Based on the preprocessed multi-source heterogeneous data, execute the following two evaluation paths: Numerical assessment path: Based on traditional numerical algorithms, calculations are performed on the preprocessed structured data to obtain a numerical health score; Semantic evaluation path: Based on a multi-agent collaborative architecture and retrieval enhancement generation technology, semantic understanding and reasoning are performed on the unstructured text data and structured data features to obtain a semantic health score; S4. Calculate the uncertainty measure between the numerical health score and the semantic health score; and fuse the numerical health score and the semantic health score based on the uncertainty measure to obtain the fused health score; S5. Based on the fused health score, combined with the safety index and failure index, the comprehensive performance of the yarn-making equipment is calculated using a nonlinear comprehensive performance evaluation function; S6. Based on the risk level of the overall performance, automatically execute the corresponding operation and maintenance decisions.

[0030] Specifically, in step S1, the operating parameters of the device are collected in real time. And construct an unstructured knowledge base: Transform text data such as repair manuals and fault cases into triplet knowledge graphs. ,in For physical objects (such as bearings). For relationships (e.g., "caused by wear and tear").

[0031] Transforming text paragraphs into vectors using text embedding models Store it in a vector database.

[0032] In step S2, the structured data utilizes 3 Outliers are removed in principle, and standardized using the Z-score formula: Eliminate the influence of dimensions.

[0033] For example, the numerical evaluation path is executed by a traditional algorithm engine, focusing on real-time performance and numerical accuracy, specifically including: Health score calculation: Calculate the overall deviation of each monitoring parameter. Combined with the weights determined by AHP Calculate the overall deviation of the equipment. Convert the deviation into a percentage score. .

[0034] Failure Feature Extraction: The K-Means algorithm is used to cluster historical data into normal operating condition clusters and abnormal operating condition clusters.

[0035] Calculate the Euclidean distance from real-time data points to the centers of each cluster. If the distance exceeds a threshold and the moving average trend is... If the value is significantly negative, it is considered a potential failure.

[0036] The semantic evaluation path is based on multi-agent collaborative semantic evaluation. This step is executed by the agent collaboration layer and focuses on logical reasoning and the utilization of unstructured information. Specifically, it includes: Information retrieval agent (A2) operation: Generate query vectors when an abnormal signal or periodic evaluation request is received. .

[0037] Retrieve the top results with the highest similarity from the vector database. Context document The cosine similarity formula is used: .

[0038] The reasoning agent (A3) works as follows: Combining real-time data features and retrieved context Input the large language model for causal reasoning.

[0039] Generate semantic-based health scores using the RAG mechanism And risk probability prediction.

[0040] Explaining the workings of the generative agent (A4): Generate diagnostic reports in natural language, such as: "Based on the vibration spectrum and historical cases (Case ID: 2025-001), the current spindle exhibits early wear characteristics." Furthermore, in step S4, adaptive fusion decision-making is performed to address the traditional weight rigidity problem, specifically including: 1. Calculate the uncertainty measure : Using cross-entropy to measure traditional scoring With large model scoring Degree of disagreement between them: The greater the disagreement, the higher the uncertainty.

[0041] 2. Dynamic weight adjustment: Define the adaptive fusion formula: .

[0042] Weighting coefficient Dynamically adjust based on uncertainties: .

[0043] Explanation of principle: When the conclusions of the traditional algorithm are consistent with those of the large model ( When the difference between the two is large, maintain the benchmark weight; when the difference is small, ( (Large), through parameters This method reduces the weights of traditional numerical algorithms, relying more on large models with reasoning capabilities to handle complex or unseen fault modes. For the first time, it semantically integrates maintenance textual knowledge from the yarn-making workshop with sensor numerical data, solving the data silo problem. Simultaneously, it utilizes dynamic weights... The introduction of this feature enables the system to maintain high accuracy under both common faults (with obvious numerical characteristics) and difficult faults (requiring empirical reasoning).

[0044] Furthermore, in step S5, a comprehensive performance evaluation is performed, utilizing a nonlinear Sigmoid function to fuse health scores ( Safety Index ) and failure index ( ): Where H represents the health index and S represents the safety index. Indicates the failure index. Indicates the steepness coefficient. , , All represent weighting coefficients. This formula amplifies the sensitivity of the critical state (around 0.5) through nonlinear transformation, making it easier to detect equipment degradation in its early stages.

[0045] Furthermore, the decision is executed in step S6, specifically including: Low-risk status: Only logs are recorded, without disrupting production.

[0046] Medium risk status: The decision-making intelligent agent (A5) automatically generates an early warning work order, pushes it to the maintenance personnel's PDA handheld terminal, and includes repair guidance for similar faults retrieved from the database.

[0047] High-risk status: An alarm is triggered, and the system sends a stop or speed reduction command to the PLC via the OPC UA protocol. At the same time, a spare parts requisition form is automatically generated based on inventory data.

[0048] In another exemplary embodiment, an intelligent operation and maintenance and evaluation system for silk-making equipment is provided, including: The data perception layer is used to collect multi-source heterogeneous data related to the silk-making equipment. The multi-source heterogeneous data includes at least structured data collected by sensors in real time and unstructured text data related to operation and maintenance. The algorithm processing layer is used to preprocess the multi-source heterogeneous data; and based on the preprocessed multi-source heterogeneous data, executes the following two evaluation paths: Numerical assessment path: Based on traditional numerical algorithms, calculations are performed on the preprocessed structured data to obtain a numerical health score; Semantic evaluation path: Based on a multi-agent collaborative architecture and retrieval enhancement generation technology, semantic understanding and reasoning are performed on the unstructured text data and structured data features to obtain a semantic health score; The agent collaboration layer is used to calculate the uncertainty measure between the numerical health score and the semantic health score; and to fuse the numerical health score and the semantic health score based on the uncertainty measure to obtain a fused health score. The integrated decision-making layer calculates the overall performance of the yarn-making equipment based on the integrated health score, combined with the safety index and failure index, through a nonlinear comprehensive performance evaluation function; and automatically executes the corresponding operation and maintenance decisions according to the risk level of the comprehensive performance.

[0049] The specific design of each floor is as follows: Data perception layer: Hardware configuration: Includes IO-Link smart vibration sensors, temperature sensors, motor current transformers installed in key parts of filament-making equipment (such as filament cutters, rehumidifiers, and filament dryers), as well as PoE switches and industrial-grade wireless APs for data transmission.

[0050] Data type: Structured data: Time series data such as vibration frequency / amplitude, bearing temperature, motor speed, current load, and energy consumption.

[0051] Unstructured data: historical maintenance logs, equipment operation manuals, fault case databases, technical specification documents, and shift handover records.

[0052] Edge computing nodes: Edge servers deployed on the factory floor are responsible for the initial cleaning and normalization of data, as well as feature extraction of high-frequency data.

[0053] Algorithm processing layer: It includes two parallel computing engines: Traditional algorithm engine: Includes data preprocessing module (Z-score normalization), feature engineering module (Pearson correlation coefficient screening), state clustering module (K-Means), trend prediction module (moving average method) and weight calculation module (AHP analytic hierarchy process).

[0054] Semantic Cognition Engine: Includes a knowledge graph construction module (TransE / RotatE embedding), a vector database (storing text vectors), a RAG retrieval module, and a large language model (LLM) inference interface.

[0055] Multi-Agent Collaboration Layer: A set of intelligent agents with clearly defined functions is constructed based on a large model, including: a data processing agent (A1), an information retrieval agent (A2), a reasoning agent (A3), an interpretation and generation agent (A4), and a decision-making agent (A5). These agents interact with each other through a message passing mechanism.

[0056] Integration of decision-making levels: The final device performance score is calculated by combining the numerical score of the traditional algorithm with the semantic score of the agent through an adaptive weight adjustment algorithm.

[0057] For example, based on the above system, a system hardware architecture is provided, which mainly includes a physical device layer, an edge sensing layer, a data transmission layer, a computing power and storage layer, and an application interaction layer.

[0058] Physical equipment layer: This includes the core production equipment in the filament processing workshop, such as filament cutters, rehydration machines, and filament drying machines. These machines are equipped with actuators (such as motors and gearboxes).

[0059] Edge sensing layer: Vibration sensor: Model LY-XSA2700-A, installed in the bearing housing of a critical motor, used to collect acceleration signals, with a sampling frequency of not less than 10kHz, to capture the high-frequency characteristics of early mechanical wear.

[0060] Various types of transmitters, including temperature sensors (monitoring motor temperature rise), current transformers (monitoring load changes), etc.

[0061] Data transmission layer: The IO-Link protocol is used to connect the sensors to the field master station module (such as BalluffBNI008X), and then the data is aggregated to the workshop LAN via a PoE switch (supporting the IEEE 802.3at standard) through Category 6 network cable or fiber optic cable.

[0062] Computing and storage layer: Database server: Deploy a time-series database (InfluxDB) to store structured sensor data, and deploy a vector database (Milvus) to store unstructured knowledge vectors.

[0063] AI computing server: configured with high-performance GPUs (such as H100 or RTX 3080) to run traditional machine learning models (XGBoost / K-Means) and large language models (LLM).

[0064] Application interaction layer: including handheld PDA terminals (Android system, with maintenance APP installed) for maintenance personnel and large-screen visualization system in the central control room.

[0065] Furthermore, based on the above method and system, a specific example is provided, including the following steps: Step S1: Physical acquisition and standardization preprocessing of multi-source heterogeneous data The system collects the physical operating parameters of the equipment in real time. (Settings...) The sensor data vector collected at each moment is ,in This represents the amplitude of spindle vibration (mm / s). Represents the bearing temperature (°C).

[0066] To eliminate the influence of different physical dimensions, Z-score normalization is used: in, This is the average value of the physical quantity during the normal historical operation of the equipment. This represents the standard deviation. This step transforms the physical quantity into a dimensionless statistical characteristic. If it is, then it is marked as a physical anomaly.

[0067] Step S2: Construct a semantic vectorized knowledge base This step is offline preprocessing. The system will process text data such as the "Equipment Maintenance Manual" and historical fault work orders (e.g., "March 2024 Shredder Knife Gate Jam Repair Record"). Clean and divide into sections.

[0068] Text embedding models are used to map text blocks into high-dimensional vectors. : Storing vectors in a vector database creates a semantic index of physical fault phenomena (such as "high-frequency abnormal noise") and maintenance knowledge.

[0069] Step S3: Parallel assessment and diagnosis via dual pathways The system simultaneously launches two evaluation paths to process numerical features and semantic features respectively.

[0070] Path A: Numerical scoring based on traditional algorithms ( Analyze the current feature vector using the K-Means clustering algorithm Clustering centers under normal operating conditions Euclidean distance .

[0071] like Exceeding a preset threshold indicates that the device deviates from its normal physical state. A numerical health score is calculated based on moving average trend prediction. (Normalized to 0-100 points).

[0072] Physical meaning: This reflects the degree to which the equipment's operating parameters have statistically deviated from the "golden state".

[0073] Path B: Semantic health scoring based on multi-agent collaboration ( ) The data processing agent (A1) translates abnormal physical characteristics (such as "vibration value suddenly increases to 5.2 mm / s") into natural language descriptions.

[0074] The information retrieval agent (A2) generates a query vector based on this description and retrieves the Top-K most similar historical cases from the vector database.

[0075] The reasoning agent (A3) combines retrieved cases (RAG mechanism) with a large model to infer the risk level of the current state and outputs a semantic health score. .

[0076] Physical meaning: This reflects the similarity between current physical phenomena and known historical failure modes, as well as potential risks.

[0077] Step S4: Adaptive Weight Fusion Based on Uncertainty To address the instability of a single model, fusion weights are calculated.

[0078] Calculate the uncertainty (cross-entropy): Physical meaning: This represents the "degree of divergence" between the numerical model and the semantic model in their assessment of the device's status. If both models consider the device to be normal, then... Small; if the numerical value is normal but the semantic model issues a warning based on a weak signal, then big.

[0079] Dynamically adjust weights : in As the baseline weight (e.g., 0.6). This is the adjustment coefficient.

[0080] Logical judgment: When the disagreement is large ( (High) The system automatically reduces the weight of traditional algorithms, relying more on the reasoning ability of large models to handle complex situations.

[0081] Calculate the final score: Step S5: Comprehensive performance calculation based on nonlinear functions To simulate the physical law of nonlinear degradation of equipment performance with wear, the Sigmoid variant function is used to calculate the overall performance index (PE).

[0082] Physical meaning of the parameters: (Health Index): by It is derived from this and represents the mechanical integrity of the equipment.

[0083] (Safety Index): Represents the degree to which the equipment's operating environment meets safety standards (e.g., no leakage current, suitable temperature).

[0084] (Failure Index): Represents the probability of a device experiencing a functional shutdown.

[0085] (Steepness coefficient): Take This is used to amplify the sensitivity near the critical state (0.5), causing the score to drop sharply when the device transitions from "sub-healthy" to "faulty," triggering an alarm. For example... Figure 2 As shown in the figure, the sigmoid function curve is displayed, with the horizontal axis representing the normalized index value and the vertical axis representing the performance score. This figure visually explains why the score change rate is greatest when the index approaches 0.5 (the critical point), demonstrating the ability to sensitively detect early equipment failures.

[0086] Step S6: Closed-loop execution and feedback The generated agent (A4) reports: "Overall score 78 (medium risk). Reason: Abnormal vibration of the shredder spindle was detected, and 3 similar cases of bearing wear were matched. Recommendation: Check the lubrication level and prepare to replace the 6205 bearing." The system automatically calls the API interface: Check the inventory of "6205 bearing" in the spare parts management system.

[0087] If there is sufficient inventory, a material requisition form will be automatically generated and pushed to the maintenance worker's PDA.

[0088] If the risk is extremely high (PE < 30), send a stop command to the PLC to prevent equipment damage. For example... Figure 3 As shown, this illustrates the data flow for performance evaluation.

[0089] In another exemplary embodiment, an electronic device is provided, including a memory and a processor. The memory stores computer instructions that can be executed on the processor. When the processor executes the computer instructions, it performs the intelligent operation and maintenance and evaluation method for silk-making equipment provided in the embodiments of the present invention.

[0090] The processor may be a single-core or multi-core central processing unit or a specific integrated circuit, or one or more integrated circuits configured to implement the present invention.

[0091] The embodiments of the subject matter and functional operation described in this specification can be implemented in: tangibly embodied computer software or firmware, computer hardware including the structures disclosed in this specification and their structural equivalents, or combinations thereof. Embodiments of the subject matter described in this specification can be implemented as one or more computer programs, i.e., one or more modules of computer program instructions encoded on a tangible, non-transitory program carrier for execution by a data processing device or for controlling the operation of a data processing device. Alternatively or additionally, the program instructions may be encoded on artificially generated propagation signals, such as machine-generated electrical, optical, or electromagnetic signals, which are generated to encode information and transmit it to a suitable receiving device for execution by the data processing device.

[0092] The processing and logic flow described in this specification can be executed by one or more programmable computers that execute one or more computer programs to perform corresponding functions by operating on input data and generating output. The processing and logic flow can also be executed by dedicated logic circuitry—such as FPGAs (Field-Programmable Gate Arrays) or ASICs (Application-Specific Integrated Circuits), and the device can also be implemented as dedicated logic circuitry.

[0093] Suitable processors for executing computer programs include, for example, general-purpose and / or special-purpose microprocessors, or any other type of central processing unit. Typically, the central processing unit receives instructions and data from read-only memory and / or random access memory. The basic components of a computer include a central processing unit for implementing or executing instructions and one or more memory devices for storing instructions and data. Typically, a computer will also include one or more mass storage devices for storing data, such as disks, magneto-optical disks, or optical disks, or the computer will be operatively coupled to such mass storage devices to receive data from or transfer data to them, or both. However, a computer is not required to have such devices. Furthermore, a computer can be embedded in another device, such as a mobile phone, a personal digital assistant (PDA), a mobile audio or video player, a game console, a global positioning system (GPS) receiver, or a portable storage device such as a universal serial bus (USB) flash drive, to name a few.

[0094] It should be understood that each block in a flowchart or block diagram can represent a module, segment, or portion of code, which contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those shown in the figures. For example, two consecutive blocks may actually be executed substantially in parallel, or they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0095] The above detailed embodiments are a description of the present invention. It should not be considered that the specific embodiments of the present invention are limited to these descriptions. For those skilled in the art, several simple deductions and substitutions can be made without departing from the concept of the present invention, and all of these should be considered to fall within the protection scope of the present invention.

Claims

1. A method for intelligent operation, maintenance, and evaluation of silk-making equipment, characterized in that, Includes the following steps: S1. Collect multi-source heterogeneous data related to the silk-making equipment, wherein the multi-source heterogeneous data includes at least structured data collected in real time by sensors and unstructured text data related to operation and maintenance; S2. Preprocess the multi-source heterogeneous data; S3. Based on the preprocessed multi-source heterogeneous data, execute the following two evaluation paths: Numerical assessment path: Based on traditional numerical algorithms, calculations are performed on the preprocessed structured data to obtain a numerical health score; Semantic evaluation path: Based on a multi-agent collaborative architecture and retrieval enhancement generation technology, semantic understanding and reasoning are performed on the unstructured text data and structured data features to obtain a semantic health score; S4. Calculate the uncertainty measure between the numerical health score and the semantic health score; and fuse the numerical health score and the semantic health score based on the uncertainty measure to obtain the fused health score; S5. Based on the fused health score, combined with the safety index and failure index, the comprehensive performance of the yarn-making equipment is calculated using a nonlinear comprehensive performance evaluation function; S6. Based on the risk level of the overall performance, automatically execute the corresponding operation and maintenance decisions.

2. The intelligent operation, maintenance, and evaluation method for silk-making equipment according to claim 1, characterized in that, The preprocessing includes: The structured data is subjected to outlier removal and standardization. The unstructured text data is transformed into a triplet knowledge graph, and the text paragraphs are transformed into vectors using a text embedding model.

3. The intelligent operation, maintenance, and evaluation method for silk-making equipment according to claim 1, characterized in that, The numerical health score is calculated using the following formula: in, Indicates a numerical health score. This indicates the overall deviation of the silk-making equipment.

4. The intelligent operation and maintenance and evaluation method for silk-making equipment according to claim 1, characterized in that, The numerical evaluation path also includes: The K-Means algorithm is used to cluster historical data into normal operating condition clusters and abnormal operating condition clusters; Calculate the Euclidean distance from the real-time data point to the center of each operating condition cluster. If the Euclidean distance exceeds the threshold and the moving average trend is significantly negative, it is determined to be a potential failure.

5. The intelligent operation and maintenance and evaluation method for silk-making equipment according to claim 1, characterized in that, The calculation of the uncertainty measure between the numerical health score and the semantic health score includes: Cross-entropy is used to measure the degree of discrepancy between numerical health scores and semantic health scores.

6. The intelligent operation and maintenance and evaluation method for silk-making equipment according to claim 5, characterized in that, The integrated health score is calculated using the following formula: Among them, the This indicates the weight of the numerical health score. Indicates semantic health score, , This represents the weight adjustment parameter. It represents a measure of uncertainty.

7. The intelligent operation and maintenance and evaluation method for silk-making equipment according to claim 1, characterized in that, The overall performance is calculated by the following formula: Where H represents the health index and S represents the safety index. Indicates the failure index. Indicates the steepness coefficient. , , All of these represent weighting coefficients.

8. An intelligent operation, maintenance, and evaluation system for silk-making equipment, characterized in that, include: The data perception layer is used to collect multi-source heterogeneous data related to the silk-making equipment. The multi-source heterogeneous data includes at least structured data collected by sensors in real time and unstructured text data related to operation and maintenance. The algorithm processing layer is used to preprocess the multi-source heterogeneous data; and based on the preprocessed multi-source heterogeneous data, executes the following two evaluation paths: Numerical assessment path: Based on traditional numerical algorithms, calculations are performed on the preprocessed structured data to obtain a numerical health score; Semantic evaluation path: Based on a multi-agent collaborative architecture and retrieval enhancement generation technology, semantic understanding and reasoning are performed on the unstructured text data and structured data features to obtain a semantic health score; The agent collaboration layer is used to calculate the uncertainty measure between the numerical health score and the semantic health score; and to fuse the numerical health score and the semantic health score based on the uncertainty measure to obtain a fused health score. The integrated decision-making layer calculates the overall efficiency of the yarn-making equipment based on the integrated health score, combined with the safety index and failure index, through a nonlinear comprehensive efficiency evaluation function. And based on the risk level of the overall performance, the system will automatically execute the corresponding operation and maintenance decisions.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the intelligent operation and maintenance and evaluation method for the silk-making equipment as described in any one of claims 1-7.

10. An electronic device comprising a memory and a processor, wherein the memory stores computer instructions executable by the processor, characterized in that, When the processor executes computer instructions, it performs the intelligent operation and maintenance and evaluation method for the silk-making equipment as described in any one of claims 1-7.

Images