Production line anomaly diagnosis method, device, equipment and medium

Abstract

The present application relates to the technical field of industrial automation, and especially relates to a production line abnormality diagnosis method and device, computer equipment and medium, wherein, multi-source heterogeneous sensing data of a production line is acquired, and the multi-source heterogeneous sensing data is standardized to obtain multi-source homogeneous sensing data; the multi-source homogeneous sensing data is subjected to feature extraction processing to obtain production line features of the production line; a target diagnosis graph corresponding to the production line features is determined, the target diagnosis graph comprises a directed acyclic graph composed of multiple diagnosis nodes and directed edges, each diagnosis node corresponds to a diagnosis strategy, and the directed edges between the diagnosis nodes represent the execution order of the diagnosis strategies; the production line features are input to a starting diagnosis node of the target diagnosis graph, the diagnosis strategies corresponding to each diagnosis node are executed in turn, and a production line abnormality diagnosis result is output, and the present application can realize adaptive generation of a diagnosis path and enhance the adaptability to diversified production line scenes.

Description

Technical Field

[0001] This invention relates to the field of industrial automation technology, and in particular to a production line anomaly diagnosis method, a production line anomaly diagnosis device, a computer device, and a computer-readable storage medium. Background Technology

[0002] In modern intelligent manufacturing systems, production lines are highly integrated production systems whose operational efficiency and stability directly impact product quality and production costs. For example, a packaging production line is a combination of automated or semi-automated equipment used to complete the entire product packaging process, covering all stages from material input to finished product output. Its core functions include sorting, filling, sealing, labeling, and palletizing, and it is widely used in industries such as food, pharmaceuticals, chemicals, and electronics. However, in actual operation, production lines such as packaging lines often experience production interruptions due to equipment failures, material abnormalities, or environmental interference, leading to decreased efficiency. Related technologies, in order to diagnose potential anomalies in production lines, often focus on single technical paths, such as relying solely on sensor data monitoring or rule-based expert systems, which have limited generalization capabilities. Summary of the Invention

[0003] This invention provides a production line anomaly diagnosis method, a production line anomaly diagnosis device, a computer device, and a computer-readable storage medium, which can realize adaptive generation of diagnosis paths and enhance adaptability to diverse production line scenarios.

[0004] In a first aspect, the production line anomaly diagnosis method provided by the present invention includes:

[0005] Acquire multi-source heterogeneous sensing data from the production line and standardize the multi-source heterogeneous sensing data to obtain multi-source homogeneous sensing data;

[0006] Feature extraction processing is performed on multi-source isomorphic sensing data to obtain production line features of the production line;

[0007] Determine the target diagnostic map corresponding to the production line characteristics. The target diagnostic map includes a directed acyclic graph consisting of multiple diagnostic nodes and directed edges. Each diagnostic node corresponds to a diagnostic strategy, and the directed edges between diagnostic nodes represent the execution order of the diagnostic strategies.

[0008] Input the production line features into the starting diagnostic node of the target diagnostic map, execute the diagnostic strategies corresponding to each diagnostic node in sequence, and output the production line anomaly diagnostic results.

[0009] Secondly, the production line anomaly diagnosis device provided by the present invention includes:

[0010] The unified sensing module is used to acquire multi-source heterogeneous sensing data from the production line and to standardize the multi-source heterogeneous sensing data to obtain multi-source homogeneous sensing data.

[0011] The feature extraction module is used to extract features from multi-source isomorphic sensing data to obtain production line features.

[0012] The graph determination module is used to determine the target diagnostic graph corresponding to the production line characteristics. The target diagnostic graph includes a directed acyclic graph consisting of multiple diagnostic nodes and directed edges. Each diagnostic node corresponds to a diagnostic strategy, and the directed edges between each diagnostic node represent the execution order of the diagnostic strategy.

[0013] The collaborative reasoning module is used to input production line features into the starting diagnostic node of the target diagnostic map, execute the diagnostic strategies corresponding to each diagnostic node in sequence, and output the production line anomaly diagnostic results.

[0014] Optionally, in one embodiment, the diagnostic node includes a sensing node, an analysis node, and a fusion node. The collaborative reasoning module is used to input production line features into the sensing node of the target diagnostic map, perform anomaly detection on the production line features through the sensing node, and output anomaly type representation; input the anomaly type representation into the analysis node of the target diagnostic map, perform root cause analysis on the anomaly type representation through the analysis node, and output potential anomaly root causes; input the potential anomaly root causes output by different analysis nodes into the fusion node of the target diagnostic map, and fuse the potential anomaly root causes output by different analysis nodes through the fusion node to output a production line anomaly diagnostic result containing the target anomaly root cause.

[0015] Optionally, in one embodiment, the feature extraction module is used to extract features from multi-source isomorphic perception data in parallel by encapsulating multiple feature extraction microservices with different feature extraction algorithms, so as to obtain production line features of the production line.

[0016] Optionally, in one embodiment, the production line anomaly diagnosis device provided by the present invention further includes an optimization and adjustment module, used to obtain result verification information corresponding to the production line anomaly diagnosis result; and to optimize and adjust at least one of the feature microservices, diagnostic nodes, and directed edges according to the structural verification information.

[0017] Optionally, in one embodiment, the diagram determination module is used to identify whether there is a diagnostic diagram that matches the production line features; if there is a matching diagnostic diagram, it is determined as the target diagnostic diagram; if there is no matching diagnostic diagram, a new diagnostic diagram is generated based on the production line features and determined as the target diagnostic diagram.

[0018] Optionally, in one embodiment, the production line anomaly diagnosis device provided by the present invention further includes a graph configuration module, which is used to display a visual diagnostic graph configuration interface, and receive input diagnostic graph drawing operations through the diagnostic graph configuration interface; and configure the type, number and directed edges of diagnostic nodes according to the drawing operations to generate a corresponding diagnostic graph.

[0019] Optionally, in one embodiment, the production line anomaly diagnosis device provided by the present invention further includes a report generation module, which is used to generate interpretable diagnostic basis corresponding to the production line anomaly diagnosis result based on the diagnostic path of the production line anomaly diagnosis result; and to generate a structured diagnosis report based on the production line anomaly diagnosis result and the interpretable diagnostic basis.

[0020] Thirdly, the computer device provided by the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the production line abnormality diagnosis method provided by the present invention.

[0021] Fourthly, the computer-readable storage medium provided by the present invention stores a computer program, which, when executed by a processor, implements the production line anomaly diagnosis method provided by the present invention.

[0022] This invention provides a production line anomaly diagnosis scheme. It acquires multi-source heterogeneous sensing data from the production line and standardizes this data to obtain multi-source homogeneous sensing data. Feature extraction is then performed on this homogeneous sensing data to obtain production line features. A target diagnostic graph corresponding to the production line features is determined. This target diagnostic graph comprises a directed acyclic graph (DAG) consisting of multiple diagnostic nodes and directed edges. Each diagnostic node corresponds to a diagnostic strategy, and the directed edges between nodes represent the execution order of the diagnostic strategies. The production line features are input to the starting diagnostic node of the target diagnostic graph, and the diagnostic strategies corresponding to each node are executed sequentially, outputting the production line anomaly diagnosis result. By constructing a collaborative mechanism between standardized data processing and feature extraction, the expression accuracy of production line features is improved. Combined with the topological characteristics of the DAG, the target diagnostic graph is dynamically matched to the production line features, enabling adaptive generation of diagnostic paths and enhancing adaptability to diverse production line scenarios. Attached Figure Description

[0023] 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.

[0024] Figure 1 This is a flowchart illustrating the production line anomaly diagnosis method provided in an embodiment of the present invention;

[0025] Figure 2 This is an example diagram of a composite diagnostic diagram involved in an embodiment of the present invention;

[0026] Figure 3 yes Figure 1 Detailed process diagram of S140;

[0027] Figure 4 yes Figure 1 Detailed process diagram of S140;

[0028] Figure 5 This is an example diagram of dynamically adjusting the diagnostic diagram in an embodiment of the present invention;

[0029] Figure 6 This is a schematic diagram of the production line abnormality diagnosis device provided in an embodiment of the present invention;

[0030] Figure 7 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0031] To make the technical problems solved, the technical solutions, and the beneficial effects 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 of the invention and are not intended to limit the invention.

[0032] It should be understood that, when used in this specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0033] It should also be understood that the term “and / or” as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0034] As used in this specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if [described condition or event] is detected" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once [described condition or event] is detected," or "in response to detection of [described condition or event]."

[0035] Furthermore, in the description of this invention and the appended claims, the terms "first," "second," "third," etc., are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance.

[0036] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of the invention include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0037] In related technologies, the focus for diagnosing potential anomalies on production lines often falls on a single technical path:

[0038] 1. Expert rule-based diagnostic approach: This approach relies on the experience of process experts to set static thresholds or rule chains. While this method can quickly identify known abnormal patterns, it is difficult to cope with dynamic changes in the production line and unknown fault types. Furthermore, the rule maintenance cost is high and the generalization ability is weak.

[0039] 2. Diagnostic methods based on a single artificial intelligence model: These typically employ deep learning (such as CNN, LSTM) or traditional machine learning models for anomaly detection or classification. While these methods possess some pattern recognition capabilities, the models are like "black boxes," resulting in a lack of interpretability in diagnostic conclusions. This leads to low trust levels in industrial settings and a heavy reliance on large amounts of labeled data, making it difficult to cope with changes in production line processes.

[0040] 3. Diagnostic methods based on isolated vision or sensors: Judging from only a single data dimension (such as image, vibration, temperature) makes it difficult to distinguish whether the root cause of the fault is from mechanical parts, electrical parameters, material batches or environmental interference, resulting in "easy alarm, difficult diagnosis".

[0041] As can be seen from the above, related technologies suffer from common problems such as low diagnostic accuracy and weak adaptability, making it difficult to meet the multiple demands for high precision and adaptability in complex industrial scenarios. Therefore, this invention provides a production line anomaly diagnosis method, a production line anomaly diagnosis device, a computer device, and a computer-readable storage medium. The production line anomaly diagnosis method can be executed by the production line anomaly diagnosis device or by a computer device integrating the production line anomaly diagnosis device. Specifically, it involves acquiring multi-source heterogeneous sensing data from the production line and standardizing the multi-source heterogeneous sensing data to obtain multi-source homogeneous sensing data; performing feature extraction processing on the multi-source homogeneous sensing data to obtain production line features; determining a target diagnostic graph corresponding to the production line features, the target diagnostic graph comprising a directed acyclic graph composed of multiple diagnostic nodes and directed edges, each diagnostic node corresponding to a diagnostic strategy, and the directed edges between diagnostic nodes representing the execution order of the diagnostic strategies; inputting the production line features to the starting diagnostic node of the target diagnostic graph, sequentially executing the diagnostic strategies corresponding to each diagnostic node, and outputting the production line anomaly diagnosis result.

[0042] Please refer to Figure 1 This is a flowchart illustrating a production line anomaly diagnosis method disclosed in an embodiment of the present invention, as shown below. Figure 1 As shown, the process of this production line anomaly diagnosis method can be as follows:

[0043] In S110, multi-source heterogeneous sensing data from the production line is acquired, and the multi-source heterogeneous sensing data is standardized to obtain multi-source homogeneous sensing data.

[0044] The following section uses computer equipment as the execution subject to describe in detail the production line anomaly diagnosis method provided by this invention.

[0045] To achieve seamless integration and plug-and-play functionality of multimodal data, this invention constructs a unified data representation and access standard, seamlessly integrating heterogeneous data from multiple sources such as physical signals, visual information, process parameters, and production work orders.

[0046] The computer equipment first acquires multi-source heterogeneous sensing data from various sensors, vision devices, controllers, and MES / SCADA systems, covering information such as vibration, temperature, current, images, and production records. Then, the data from different dimensions is normalized, time-series aligned, and format unified, transforming it into multi-source homogeneous sensing data with a consistent structure, laying the data foundation for subsequent feature extraction and fusion analysis.

[0047] For example, computer devices can standardize multi-source heterogeneous sensing data according to the following data format:

[0048] {

[0049] "atom_id": "Unique identifier", used to uniquely mark each piece of perceived data to ensure the accuracy of data tracing and correlation analysis;

[0050] "source_id": "Data source physical identifier", used to identify the physical device or system from which the data originates, enabling precise tracking and management of the data source;

[0051] "timestamp": "nanosecond-level precision timestamp", accurately records the data acquisition time, ensuring the synchronization and consistency of multi-source data in the time dimension;

[0052] "data_type": "Enumerated types, such as: vibration_raw, rgb_image, production_speed", used to explicitly perceive the type attribute of data;

[0053] "payload": "raw or lightly preprocessed data body", containing specific data content, which can be a numerical array, image matrix or structured text;

[0054] "metadata": {

[0055] "confidence": "Data quality confidence level (0-1)", used to assess the reliability and completeness of data collection;

[0056] "status": "Data acquisition device status code", used to monitor the operating status of data acquisition devices in real time;

[0057] "coordinates": "3D coordinates of the data source in the production line", used to locate the physical location of the data source in the production line space;

[0058] }

[0059] }

[0060] For example, a standardized vibration sensor data set can be represented as: {"atom_id": "vib_001", "source_id": "sensor_vib_1001", "timestamp": 1705219200000000000, "data_type":"vibration_raw", "payload": [0.12, -0.34, 0.56, ...], "metadata": {"confidence": 0.98, "status": 200, "coordinates": [12.3, 45.6, 7.8]}}; Similarly, a standardized visual inspection image data set can be represented as: {"atom_id": "img_001", "source_id": "camera_vis_2001", "timestamp": 1705219200000000001, "data_type": "rgb_image","payload": "base64_encoded_string", "metadata": {"confidence": 0.95, "status": 200, "coordinates": [8.9, 33.2, 6.1]}}, etc.

[0061] The above approach provides standardized input for production line anomaly diagnosis by unifying the expression format of multi-source sensing data.

[0062] In S120, feature extraction processing is performed on the multi-source isomorphic sensing data to obtain the production line features of the production line.

[0063] As described above, after standardizing the acquired multi-source heterogeneous sensing data from the production line into a unified data structure, the computer equipment performs feature extraction based on the data type. For example, taking a packaging production line as an example, for the acceleration signal of the filling valve spindle collected by the vibration sensor, the computer equipment can perform a fast Fourier transform on the time-domain signal to extract its frequency-domain spectral features, obtain the vibration amplitude at different frequencies, and extract key features such as the amplitude of the fundamental frequency and its harmonic components; for the temperature distribution image of the sealing area acquired by the infrared thermal imager, spatial thermal features such as the average temperature, the location of the highest temperature point, and the rate of change of the thermal gradient can be extracted; for the image of the sealing area acquired by the RGB camera, visual features such as the gray-scale mean, variance, and edge gradient intensity of the sealing area can be extracted.

[0064] Furthermore, the extracted production line features can inherit the data structure of the source perception data, that is, maintain the same metadata association such as atomic identifiers, timestamps and spatial coordinates, and add feature name fields, feature value fields and extraction algorithm fields to form a structured production line feature dataset, so that features of different modalities are comparable and fusionable under a unified spatiotemporal benchmark.

[0065] Through the above multi-dimensional feature extraction, the raw perception data is transformed into quantifiable production line operation status indicators, providing high-dimensional and resolvable feature inputs for subsequent anomaly diagnosis.

[0066] In S130, a target diagnostic graph corresponding to the production line characteristics is determined. The target diagnostic graph includes a directed acyclic graph consisting of multiple diagnostic nodes and directed edges. Each diagnostic node corresponds to a diagnostic strategy, and the directed edges between each diagnostic node represent the execution order of the diagnostic strategies.

[0067] In this embodiment of the invention, heterogeneous diagnostic strategies are encapsulated in a diagnostic graph, and their logical dependencies are explicitly modeled, thereby decoupling multimodal features from diagnostic logic. Each diagnostic node in the diagnostic graph encapsulates a diagnostic strategy for a specific anomaly pattern, such as temperature anomaly detection based on threshold rules, vibration fault identification based on spectral analysis, or visual defect classification based on deep learning. The execution order and conditional jumps between diagnostic nodes are defined by directed edges, enabling modular orchestration of complex diagnostic processes. The diagnostic graph supports dynamic loading and reconstruction, and can be flexibly configured according to production line type, process stage, or equipment model, ensuring the scalability and maintainability of the diagnostic logic.

[0068] Correspondingly, after extracting the production line characteristics, the computer equipment determines the diagnostic map corresponding to the production line characteristics, which is denoted as the target diagnostic map.

[0069] Optionally, the computer equipment can use pre-defined mapping rules to input the data type, feature dimensions, and process-related information of the production line features, and then match a suitable diagnostic map from existing diagnostic maps as the target diagnostic map. In specific implementation, the mapping rules can be pre-constructed based on the correlation between the production line features and the input requirements of the diagnostic map. For example, in a packaging production line, if the production line features include three types of features: vibration spectrum, thermal imaging, and visual features, then a composite diagnostic map that integrates multimodal diagnostic strategies is matched and invoked. This map includes vibration diagnostic nodes, thermal feature diagnostic nodes, visual diagnostic nodes, and a fusion node that execute in parallel. The visual diagnostic node identifies whether there are defects such as wrinkles, offsets, or incomplete seals in the sealing area by analyzing the grayscale mean and edge gradient intensity. If defects are found, the vibration diagnostic node and the thermal feature diagnostic node are triggered to perform collaborative analysis. The vibration diagnostic node diagnoses whether there is imbalance or looseness in the main shaft bearing based on the fundamental frequency and harmonic amplitude. The thermal feature diagnostic node diagnoses whether the heating element has local overheating or uneven heat dissipation by combining the average temperature and thermal gradient change rate of the sealing area. The fusion node integrates the diagnostic results of the vibration diagnostic node and the thermal feature diagnostic node and outputs the root cause of the sealing abnormality.

[0070] In addition, computer equipment can dynamically construct target diagnostic maps in real time based on production line characteristics through a pre-trained diagnostic map generation model. This diagnostic map generation model can be jointly optimized based on graph neural networks and reinforcement learning, and is configured to automatically infer the optimal combination of diagnostic nodes and connection topology based on the distribution characteristics of production line features.

[0071] Based on the above, those skilled in the art can choose a suitable diagnostic map determination method according to actual needs. It can be the mapping rule matching or dynamic generation method explicitly listed in the above embodiments of the present invention, or other reasoning matching based on experience knowledge base or adaptive selection mechanism combined with online learning, etc. The embodiments of the present invention do not impose specific limitations, as long as accurate matching or generation of diagnostic maps for specific production line characteristics can be achieved.

[0072] In S140, the production line features are input to the starting diagnostic node of the target diagnostic map, the diagnostic strategies corresponding to each diagnostic node are executed in sequence, and the production line abnormality diagnostic results are output.

[0073] In this embodiment of the invention, after determining the target diagnostic map corresponding to the production line features, the computer device inputs the extracted production line features into the starting diagnostic node of the target diagnostic map, executes the diagnostic strategies configured by each diagnostic node in sequence according to the topology of the target diagnostic map, and finally outputs the production line anomaly diagnostic results including the anomaly type, root cause and confidence level.

[0074] For example, please refer to Figure 2The determined target diagnostic map is a composite diagnostic map that integrates multimodal diagnostic strategies. This target diagnostic map includes vibration diagnostic nodes, thermal feature diagnostic nodes, visual diagnostic nodes, and fusion nodes. The computer equipment first inputs the visual features from the production line features into the visual diagnostic node, performs grayscale mean and edge gradient intensity analysis of the sealing area to determine whether there are wrinkles, offsets, or incomplete seals. If a defect is detected, the vibration and thermal feature diagnostic nodes are triggered for collaborative diagnosis: the vibration diagnostic node, based on the fundamental frequency and harmonic amplitude, diagnoses whether the spindle bearing is unbalanced or loose, and outputs the vibration anomaly diagnosis result and confidence level; the thermal feature diagnostic node, combined with the average temperature and thermal gradient change rate of the sealing area, determines whether the heating element has local overheating or uneven heat dissipation, and outputs the thermal anomaly diagnosis result and confidence level; the fusion node integrates the vibration anomaly diagnosis result and the thermal anomaly diagnosis result, combines the confidence level weights to perform evidence fusion, infers the root cause of the sealing anomaly, and outputs the production line anomaly diagnosis result including the anomaly type, root cause, and overall confidence level.

[0075] Optionally, in one embodiment, the diagnostic node includes a sensing node, an analysis node, and a fusion node, please refer to... Figure 3 The production line characteristics are input into the starting diagnostic node of the target diagnostic map, and the diagnostic strategies corresponding to each diagnostic node are executed sequentially. The production line anomaly diagnostic results are output, including:

[0076] In S1410, production line features are input to the perception node of the target diagnostic map. The perception node performs anomaly detection on the production line features and outputs an anomaly type representation.

[0077] In S1420, the anomaly type representation is input to the analysis node of the target diagnostic map. The analysis node performs root cause analysis on the anomaly type representation and outputs the potential anomaly root causes.

[0078] In S1430, the potential abnormal root causes output by different analysis nodes are input to the fusion node of the target diagnostic map. The fusion node merges the potential abnormal root causes output by different analysis nodes and outputs the production line abnormal diagnosis result containing the target abnormal root causes.

[0079] This invention provides a hierarchical diagnostic graph architecture, comprising three types of diagnostic nodes: sensing nodes, analysis nodes, and fusion nodes, forming a three-level pipeline of "detection-attribution-decision". The sensing node focuses on capturing local features, such as transient impacts in vibration signals or pixel distortions in images, outputting anomaly type representations to provide preliminary basis for subsequent analysis. The analysis node performs deep reasoning on the anomaly type representations, combining relevant features from the production line characteristics to conduct causal correlation analysis, uncovering potential root causes of the anomalies, such as equipment aging, parameter drift, or external interference, and outputting corresponding confidence levels. The fusion node integrates the potential root causes output by multiple analysis nodes, performing conflict resolution and confidence-weighted fusion to generate a final production line anomaly diagnostic result that includes the anomaly type, root cause, and comprehensive confidence level.

[0080] Correspondingly, the computing device inputs feature data from the production line characteristics that match the sensing nodes in the target diagnostic map to the sensing nodes. The sensing nodes then perform anomaly detection on the input feature data, identifying significant deviations in the production line's operating status and outputting corresponding anomaly type representations and confidence levels, such as excessive vibration, abnormal temperature, or positional deviation. In practical implementation, the sensing nodes can use preset thresholds, statistical process control (SPC), or lightweight machine learning models to perform real-time discrimination of the input features and output anomaly type representations.

[0081] After the sensing node outputs an anomaly type representation, the computer equipment inputs this representation to the analysis node connected to the sensing node. Furthermore, the computer equipment extracts the relevant features required by the analysis node from the production line characteristics and inputs them to the corresponding analysis node. Combining the anomaly type representation with the analysis node's input, causal reasoning and root cause analysis are performed, outputting potential anomaly root causes and their confidence levels. In specific implementations, the analysis node can be based on Bayesian networks, decision trees, deep neural networks, rule networks, or knowledge graph query engines. It should also be noted that the analysis node can receive anomaly type representations from one or more sensing nodes, as well as intermediate root cause results from other analysis nodes, enabling multi-path reasoning and collaborative diagnosis.

[0082] Finally, the computer equipment inputs the potential root causes of anomalies and their confidence levels from each analysis node to the fusion node. The fusion node then resolves conflicts and performs confidence-weighted fusion of the multi-source root causes, ultimately achieving root cause aggregation and outputting a production line anomaly diagnosis result that includes the target anomaly root cause, anomaly type, and overall confidence level. In practical implementation, confidence-weighted fusion can be achieved using methods such as DS evidence theory or Bayesian averaging.

[0083] Optionally, in one embodiment, feature extraction processing is performed on the multi-source isomorphic sensing data to obtain production line features of the production line, including:

[0084] By encapsulating multiple feature extraction microservices with different feature extraction algorithms, feature extraction is performed in parallel on multi-source isomorphic sensing data to obtain production line features.

[0085] To enhance the flexibility and scalability of feature extraction, each feature extraction process is encapsulated as an independent, containerized microservice in this embodiment of the invention. Each microservice receives perception data and outputs the extracted features through a standardized interface.

[0086] Correspondingly, when extracting features from multi-source isomorphic sensing data, the computer equipment calls multiple feature extraction microservices for parallel processing. Each microservice performs specific types of feature extraction on the sensing data according to its encapsulated algorithm, such as time-domain statistics, spectral analysis, or wavelet transform. The extraction results are uniformly output in a standardized format, forming a set of production line features containing multi-dimensional production line operation characteristics, which can be used by subsequent sensing and analysis nodes.

[0087] By adopting the above microservice architecture for feature extraction, not only is decoupling and reuse of feature extraction achieved, but it also supports dynamic expansion of new feature algorithms without system reconstruction. Each microservice can be upgraded, replaced, or invoked in parallel, adapting to diverse analysis needs in different production line scenarios. Furthermore, based on the elastic scheduling capabilities of containerized deployment, computer equipment can automatically adjust the number of microservice instances according to the data scale, ensuring real-time performance and stability under high concurrency.

[0088] Optionally, in one embodiment, after inputting production line features to the starting diagnostic node of the target diagnostic map, sequentially executing the diagnostic strategies corresponding to each diagnostic node, and outputting the production line anomaly diagnostic results, the method further includes:

[0089] Obtain the result verification information corresponding to the production line anomaly diagnosis results;

[0090] Based on the structural verification information, at least one of the feature microservices, diagnostic nodes, and directed edges is optimized and adjusted.

[0091] In this embodiment of the invention, the production line anomaly diagnosis result can be predictive, that is, to provide early warning of possible anomalies in the production line in the future.

[0092] For such predictive diagnostic results, the computer equipment can acquire result verification information corresponding to the production line anomaly diagnosis. This result verification information includes feedback data on whether the expected anomaly occurred during actual production. Based on this result verification information, the computer equipment performs parameter tuning or structural optimization on at least one of the feature microservices, diagnostic nodes, and directed edges. For example, when the predicted result is inconsistent with the actual feedback, the computer equipment adjusts the algorithm parameters of the relevant feature extraction microservices, retrains the diagnostic model of the diagnostic nodes, or even reconstructs the logical association rules of the directed edges, thereby improving diagnostic accuracy.

[0093] Alternatively, in one embodiment, please refer to Figure 4 Determine the target diagnostic map corresponding to the production line characteristics, including:

[0094] In S1310, identify whether a diagnostic map matching the production line characteristics currently exists;

[0095] In S1320, if a matching diagnostic map exists, the diagnostic map that matches the production line features is determined as the target diagnostic map.

[0096] In S1330, if no matching diagnostic map exists, a new diagnostic map is generated based on the production line characteristics and identified as the target diagnostic map.

[0097] In this embodiment of the invention, when determining a target diagnostic map corresponding to a production line feature, the computer device first identifies whether a diagnostic map matching the production line feature exists. For example, the computer device can, based on preset mapping rules, take the data type, feature dimension, and process association information of the production line feature as input, and search for a matching structural pattern in existing diagnostic maps.

[0098] If a matching diagnostic map is identified, the computer device directly identifies it as the target diagnostic map.

[0099] If no matching diagnostic map is identified, the computer device generates a new diagnostic map based on production line characteristics and identifies it as the target diagnostic map. For example, the computer device can dynamically construct the target diagnostic map in real time based on production line characteristics using a pre-trained diagnostic map generation model. This diagnostic map generation model can be jointly optimized based on graph neural networks and reinforcement learning and is configured to automatically infer the optimal combination of diagnostic nodes and connection topology based on the distribution characteristics of production line characteristics.

[0100] Furthermore, computer equipment can also dynamically adjust the diagram structure by locally expanding or pruning existing diagnostic diagrams based on production line characteristics, thereby obtaining a target diagnostic diagram that matches the production line characteristics and adapting to the dynamic evolution of the production line's operating status. For example, please refer to... Figure 5 , Figure 5The left side shows a typical initial structure of a production line diagnostic diagram, including sensing nodes P1 and P2, analysis node A1, and fusion node F1. Figure 5 The right side shows the target diagnostic map structure after dynamic adjustment based on production line characteristics, which includes the addition of a perception node P3 and an analysis node A2.

[0101] Optionally, in one embodiment, before acquiring the multi-source heterogeneous sensing data of the production line, the method further includes:

[0102] Display a visual diagnostic chart configuration interface, and receive input diagnostic chart drawing operations through the diagnostic chart configuration interface;

[0103] Based on the drawing operation configuration, the type, number, and directed edges of the diagnostic nodes are configured to generate the corresponding diagnostic graph.

[0104] To enhance the flexibility of diagnostic graph construction and improve user interaction, this embodiment of the invention provides a visual diagnostic graph configuration mechanism.

[0105] The computer equipment displays a visual diagnostic graph configuration interface. This interface integrates drag-and-drop editing components, allowing users to add and delete diagnostic nodes and directed edges by dragging and dropping, and configure the type and diagnostic strategies of each diagnostic node. Users can customize the layout and association logic of diagnostic nodes, such as perception nodes, analysis nodes, and fusion nodes, in this diagnostic graph configuration interface according to actual production line needs, thus flexibly constructing the diagnostic graph.

[0106] Correspondingly, the computer device receives the diagnostic diagram drawing operation input by the user through the diagnostic diagram configuration interface, parses the drawing operation to extract the type, number and directed edge connection relationship of the diagnostic nodes, and then generates the corresponding diagnostic diagram and stores it in the diagnostic diagram library for subsequent use.

[0107] Optionally, in one embodiment, after inputting production line features to the starting diagnostic node of the target diagnostic map, sequentially executing the diagnostic strategies corresponding to each diagnostic node, and outputting the production line anomaly diagnostic results, the method further includes:

[0108] Based on the diagnostic path of the production line anomaly diagnosis results, generate interpretable diagnostic basis corresponding to the production line anomaly diagnosis results;

[0109] A structured diagnostic report is generated based on the production line anomaly diagnosis results and interpretable diagnostic criteria.

[0110] In this embodiment of the invention, the computer device also acquires the complete diagnostic path of the production line abnormality diagnosis result, as well as the input and output data of each diagnostic node on the diagnostic path, and traces back the diagnostic logic and diagnostic basis of each node in the diagnostic path to generate interpretable diagnostic basis and clarify the key evidence and reasoning process for abnormality determination.

[0111] In addition, the computer equipment organizes the production line anomaly diagnosis results and the corresponding interpretable diagnostic basis in a structured manner, generating a structured diagnostic report that includes the production line anomaly diagnosis results (including anomaly type, confidence score, target anomaly root cause, etc.), interpretable diagnostic basis, maintenance recommendations (such as checking, replacing or calibrating relevant sensors, optimizing parameter configuration, etc.), and handling priorities.

[0112] For example, in the predictive diagnosis of "inaccurate liquid level" on a high-speed beverage bottling line, the generated structured diagnostic report may include the following:

[0113] Anomaly type: "Filling level deviation";

[0114] Root cause of the anomaly: Response delay of the front-end flow sensor;

[0115] Confidence level: 98.7%;

[0116] Diagnostic criteria: The measured flow rate deviated from the set value by more than the threshold within 12 consecutive sampling periods, and the timing analysis showed that the signal phase lag was 230 milliseconds.

[0117] Maintenance recommendations: Calibrate the sensor zero point and check the impedance of the signal transmission line;

[0118] Priority of handling: High.

[0119] As described above, the production line anomaly diagnosis scheme provided by this invention acquires multi-source heterogeneous sensing data from the production line and standardizes this data to obtain multi-source homogeneous sensing data. It then performs feature extraction on this data to obtain production line features. A target diagnostic graph corresponding to the production line features is determined. This target diagnostic graph comprises a directed acyclic graph (DAG) consisting of multiple diagnostic nodes and directed edges. Each diagnostic node corresponds to a diagnostic strategy, and the directed edges between nodes represent the execution order of the diagnostic strategies. The production line features are input to the starting diagnostic node of the target diagnostic graph, and the diagnostic strategies corresponding to each node are executed sequentially, outputting the production line anomaly diagnosis result. Thus, by constructing a collaborative mechanism between standardized data processing and feature extraction, the expression accuracy of production line features is improved. Combined with the topological characteristics of the DAG, the target diagnostic graph is dynamically matched to the production line features, achieving adaptive generation of diagnostic paths and enhancing adaptability to diverse production line scenarios.

[0120] To facilitate better implementation of the above production line anomaly diagnosis methods, this embodiment of the invention also provides a corresponding production line anomaly diagnosis device. The meanings of the terms used are the same as in the above production line anomaly diagnosis methods; for specific implementation details, please refer to the descriptions in the above method embodiments.

[0121] Please refer to Figure 6The production line anomaly diagnosis device may include a unified perception module 210, a feature extraction module 220, a graph determination module 230, and a collaborative reasoning module 240. Detailed descriptions of each functional module are as follows:

[0122] The unified sensing module 210 is used to acquire multi-source heterogeneous sensing data from the production line and to standardize the multi-source heterogeneous sensing data to obtain multi-source homogeneous sensing data.

[0123] Feature extraction module 220 is used to perform feature extraction processing on multi-source isomorphic sensing data to obtain production line features of the production line;

[0124] The graph determination module 230 is used to determine the target diagnostic graph corresponding to the production line characteristics. The target diagnostic graph includes a directed acyclic graph consisting of multiple diagnostic nodes and directed edges. Each diagnostic node corresponds to a diagnostic strategy, and the directed edges between each diagnostic node represent the execution order of the diagnostic strategy.

[0125] The collaborative reasoning module 240 is used to input production line features into the starting diagnostic node of the target diagnostic map, execute the diagnostic strategies corresponding to each diagnostic node in sequence, and output the production line abnormality diagnostic results.

[0126] Optionally, in one embodiment, the diagnostic node includes a sensing node, an analysis node, and a fusion node. The collaborative reasoning module 240 is used to input production line features into the sensing node of the target diagnostic map, perform anomaly detection on the production line features through the sensing node, and output anomaly type representation; input the anomaly type representation into the analysis node of the target diagnostic map, perform root cause analysis on the anomaly type representation through the analysis node, and output potential anomaly root causes; input the potential anomaly root causes output by different analysis nodes into the fusion node of the target diagnostic map, and fuse the potential anomaly root causes output by different analysis nodes through the fusion node to output a production line anomaly diagnostic result containing the target anomaly root cause.

[0127] Optionally, in one embodiment, the feature extraction module 220 is used to extract features from multi-source isomorphic perception data in parallel by encapsulating multiple feature extraction microservices with different feature extraction algorithms to obtain production line features of the production line.

[0128] Optionally, in one embodiment, the production line anomaly diagnosis device provided by the present invention further includes an optimization and adjustment module, used to obtain result verification information corresponding to the production line anomaly diagnosis result; and to optimize and adjust at least one of the feature microservices, diagnostic nodes, and directed edges according to the structural verification information.

[0129] Optionally, in one embodiment, the diagram determination module 230 is used to identify whether there is a diagnostic diagram that matches the production line features; if there is a matching diagnostic diagram, it is determined as the target diagnostic diagram; if there is no matching diagnostic diagram, a new diagnostic diagram is generated based on the production line features and determined as the target diagnostic diagram.

[0130] Optionally, in one embodiment, the production line anomaly diagnosis device provided by the present invention further includes a graph configuration module, which is used to display a visual diagnostic graph configuration interface, and receive input diagnostic graph drawing operations through the diagnostic graph configuration interface; and configure the type, number and directed edges of diagnostic nodes according to the drawing operations to generate a corresponding diagnostic graph.

[0131] Optionally, in one embodiment, the production line anomaly diagnosis device provided by the present invention further includes a report generation module, which is used to generate interpretable diagnostic basis corresponding to the production line anomaly diagnosis result based on the diagnostic path of the production line anomaly diagnosis result; and to generate a structured diagnosis report based on the production line anomaly diagnosis result and the interpretable diagnostic basis.

[0132] Specific limitations regarding the production line anomaly diagnostic device can be found in the limitations of the production line anomaly diagnostic method described above, and will not be repeated here. Each module in the aforementioned production line anomaly diagnostic device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0133] In one embodiment, a computer device is provided, the internal structure of which can be as follows: Figure 7 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface connects to external wireless clients, providing wireless network access services to the connected clients. When the computer program is executed by the processor, it implements the production line anomaly diagnosis method provided by this invention.

[0134] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the production line anomaly diagnosis method described in the above embodiment.

[0135] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the production line anomaly diagnosis method described above.

[0136] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0137] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0138] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for diagnosing production line anomalies, characterized in that, include: Acquire multi-source heterogeneous sensing data from the production line and standardize the multi-source heterogeneous sensing data to obtain multi-source homogeneous sensing data. The production line features of the production line are obtained by performing feature extraction processing on the multi-source isomorphic sensing data. Based on preset mapping rules, the data type, feature dimension, and process association information of the production line features are used as input to search for whether there is a diagnostic map in the existing diagnostic maps that matches the production line features. If a matching diagnostic map exists, it is determined as the target diagnostic map. If no matching diagnostic map exists, the graph structure of the existing diagnostic map is adjusted according to the production line features, and it is determined as the target diagnostic map. The target diagnostic map includes a directed acyclic graph composed of multiple diagnostic nodes and directed edges. Each diagnostic node corresponds to a diagnostic strategy, and the directed edges between each diagnostic node represent the execution order of the diagnostic strategy. The diagnostic nodes include perception nodes, analysis nodes, and fusion nodes. The production line features are input to the perception nodes of the target diagnostic map, and anomaly detection is performed on the production line features through the perception nodes, and an anomaly type representation is output. The abnormality type representation is input into the analysis node of the target diagnostic map, and the root cause analysis is performed on the abnormality type representation through the analysis node to output the potential abnormal root causes. The potential root causes of anomalies output from different analysis nodes are input into the fusion node of the target diagnostic map. The potential root causes of anomalies output from different analysis nodes are fused through the fusion node, and the production line anomaly diagnosis result containing the target root causes of anomalies is output.

2. The production line anomaly diagnosis method according to claim 1, characterized in that, The step of performing feature extraction processing on the multi-source isomorphic sensing data to obtain the production line features of the production line includes: By encapsulating multiple feature extraction microservices with different feature extraction algorithms, feature extraction is performed in parallel on the multi-source isomorphic perception data to obtain the production line features of the production line.

3. The production line anomaly diagnosis method according to claim 2, characterized in that, After inputting the potential root causes of anomalies output from different analysis nodes into the fusion node of the target diagnostic map, and fusing the potential root causes of anomalies output from different analysis nodes through the fusion node to output a production line anomaly diagnostic result containing the target root causes of anomalies, the process further includes: Obtain result verification information corresponding to the production line anomaly diagnosis results; Based on the results, at least one of the feature extraction microservice, the diagnostic node, and the directed edge is optimized and adjusted.

4. The production line anomaly diagnosis method according to claim 1, characterized in that, Before acquiring the multi-source heterogeneous sensing data from the production line, the process also includes: Display a visual diagnostic chart configuration interface, and receive input diagnostic chart drawing operations through the diagnostic chart configuration interface; Based on the drawing operation configuration, the type, number, and directed edges of the diagnostic nodes are configured to generate the corresponding diagnostic graph.

5. The production line anomaly diagnosis method according to claim 1, characterized in that, After inputting the potential root causes of anomalies output from different analysis nodes into the fusion node of the target diagnostic map, and fusing the potential root causes of anomalies output from different analysis nodes through the fusion node to output a production line anomaly diagnostic result containing the target root causes of anomalies, the process further includes: Based on the diagnostic path of the production line anomaly diagnosis results, generate interpretable diagnostic basis corresponding to the production line anomaly diagnosis results; A structured diagnostic report is generated based on the production line anomaly diagnosis results and the interpretability diagnostic criteria.

6. A production line anomaly diagnosis device, characterized in that, include: A unified sensing module is used to acquire multi-source heterogeneous sensing data from the production line and to standardize the multi-source heterogeneous sensing data to obtain multi-source homogeneous sensing data. The feature extraction module is used to perform feature extraction processing on the multi-source isomorphic sensing data to obtain the production line features of the production line; The graph determination module is used to, based on preset mapping rules, take the data type, feature dimension, and process association information of the production line features as input, and search existing diagnostic graphs to see if there is a diagnostic graph that matches the production line features. If a matching diagnostic graph exists, it is determined as the target diagnostic graph. If no matching diagnostic graph exists, the graph structure of the existing diagnostic graph is adjusted according to the production line features, and it is determined as the target diagnostic graph. The target diagnostic graph includes a directed acyclic graph composed of multiple diagnostic nodes and directed edges. Each diagnostic node corresponds to a diagnostic strategy, and the directed edges between each diagnostic node represent the execution order of the diagnostic strategy. The diagnostic nodes include perception nodes, analysis nodes, and fusion nodes. The collaborative reasoning module is used to input the production line features into the perception nodes of the target diagnostic map, perform anomaly detection on the production line features through the perception nodes, and output anomaly type representation; input the anomaly type representation into the analysis nodes of the target diagnostic map, perform root cause analysis on the anomaly type representation through the analysis nodes, and output potential anomaly root causes; The potential root causes of anomalies output from different analysis nodes are input into the fusion node of the target diagnostic map. The potential root causes of anomalies output from different analysis nodes are fused through the fusion node, and the production line anomaly diagnosis result containing the target root causes of anomalies is output.

7. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the production line abnormality diagnosis method according to any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the production line abnormality diagnosis method according to any one of claims 1 to 5.

Images