Multi-dimensional data fusion monitoring method and system for complex industrial scene

Abstract

The invention discloses a complex industrial scene-oriented multi-dimensional data fusion monitoring method and system, and belongs to the technical field of industrial automation and data analysis, and the method comprises the following steps: carrying out credibility evaluation and self-repairing processing on multi-source heterogeneous data of industrial equipment; performing cross-modal time sequence alignment on the repaired multi-source heterogeneous data, and enhancing the time sequence alignment precision; dynamically adjusting the fusion weight of the multi-source heterogeneous data based on the real-time working condition of the equipment; taking the output value after the weight fusion as an edge weight to guide the fusion of the multi-source heterogeneous data; and performing abnormal propagation path simulation analysis in combination with a digital twinborn model. According to the method, through multi-dimensional data fusion, dynamic adjustment and application of the digital twinborn model, efficient industrial fault monitoring and prediction are realized, and high reliability and safety of equipment are ensured.

Description

Technical Field

[0001] The present invention relates to the field of industrial automation and data analysis technology, and in particular to a multi-dimensional data fusion monitoring method and system for complex industrial scenarios. Background Art

[0002] In recent years, with the continuous advancement of industrial automation and digital transformation, multidimensional data monitoring technology in complex industrial scenarios has achieved significant development. In particular, driven by smart manufacturing, the Industrial Internet of Things, and big data technologies, the monitoring and operation and maintenance capabilities of industrial equipment have been greatly enhanced. Traditional single-source monitoring methods are no longer able to meet the high real-time, accuracy, and reliability requirements of modern industrial environments. Consequently, researchers have begun exploring methods for fusing multi-source heterogeneous data to achieve comprehensive monitoring of equipment status across multiple dimensions. By comprehensively analyzing information such as operating conditions, the environment, and operations, they improve the accuracy of monitoring equipment operating conditions and fault warning capabilities. For example, monitoring technology based on multidimensional data fusion has been applied to real-time monitoring and early warning of various equipment parameters, such as temperature, pressure, and current. Despite this, current multidimensional data fusion monitoring technology still faces numerous challenges, particularly in areas such as data credibility assessment, data self-repair, and time-series alignment, where significant room for improvement remains.

[0003] In existing technologies, the processing of multi-source heterogeneous data often lacks a systematic quality assessment system, resulting in insufficient data accuracy and reliability. For example, when packet loss, delays, or abnormal fluctuations occur during data acquisition, traditional methods fail to effectively identify and repair these problems, resulting in distorted monitoring results and an inability to provide a reliable basis for real-time decision-making. In addition, existing multidimensional data fusion methods often fail to dynamically adjust fusion weights in a timely manner and are unable to adapt to instantaneous changes in equipment operating status, resulting in a significant reduction in the response speed and accuracy of monitoring information during emergencies or changes in operating conditions. To a certain extent, relatively static data processing mechanisms also fail to fully consider the characteristics of different data sources and fail to effectively leverage the advantages of each data source. As a result, the early warning capabilities for potential hidden dangers are insufficient, which may result in delayed responses to equipment failures. Therefore, it is necessary to construct a multidimensional data fusion monitoring method that comprehensively considers data credibility assessment and self-repair processing. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem addressed by this invention is how to address the challenges of multi-source, heterogeneous data from different devices and sensors (such as device operating parameters, machine vision data, and environmental perception data), which often suffer from low reliability, data loss, and inconsistent timing. Traditional data processing methods are unable to effectively address these challenges, resulting in reduced accuracy and reliability of monitoring systems, which in turn affects production efficiency and equipment safety.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a multi-dimensional data fusion monitoring method for complex industrial scenes, which includes the following steps:

[0007] Conduct credibility assessment and self-repair processing on multi-source heterogeneous data of industrial equipment; perform cross-modal time series alignment on the repaired multi-source heterogeneous data and enhance the accuracy of time series alignment; dynamically adjust the fusion weight of multi-source heterogeneous data based on the real-time working conditions of the equipment; use the output value after fusion weight as the edge weight to guide the fusion of multi-source heterogeneous data; and combine the digital twin model to perform simulation analysis of abnormal propagation paths.

[0008] As a preferred solution of the multi-dimensional data fusion monitoring method for complex industrial scenarios described in the present invention, the reliability assessment and self-repair processing include collecting multi-source heterogeneous data of industrial equipment, including equipment operating parameter data, machine vision data, environmental perception data, maintenance log data, and personnel operation data, and performing preliminary cleaning and normalization processing on the collected data;

[0009] Build a multi-dimensional data quality assessment system that includes signal integrity testing, timing consistency testing, and physical constraint compliance verification. Specifically:

[0010] Calculating the packet loss rate and the checksum matching degree through the signal integrity detection to generate an integrity score. When the integrity score exceeds a preset device standard threshold, it is marked as a signal anomaly.

[0011] Use time series consistency analysis to detect the variance of the time intervals between adjacent data points. When the variance exceeds the set device sampling period threshold, it is marked as a time series anomaly.

[0012] Compare sensor readings with the physical limits of the equipment through physical constraint compliance, identify out-of-limit data, generate constraint conflict reports, and mark them as constraint anomalies;

[0013] When any of the following anomalies occur: signal anomaly, timing anomaly, or constraint anomaly, the current data will be marked as abnormal data;

[0014] Treating each piece of data as a node, a local trusted consensus mechanism is established between nodes based on abnormal data. When any node detects abnormal data, it broadcasts a verification request containing the hash value, integrity score, and timestamp of the original data to adjacent nodes.

[0015] The node receiving the request retrieves data in the same time window in the local database. If more than 60% of the nodes return matching constraint conflict reports, the data repair process is triggered;

[0016] For data lost within a set time, the historical average value of the same operating parameters under normal operating conditions of the equipment is used for interpolation and filling;

[0017] For abnormal data that persists beyond the set time, the device thermodynamic model is called to generate a predicted value, which is then weighted and fused with adjacent sensor readings to output a repair value for the abnormal data.

[0018] As a preferred solution of the multi-dimensional data fusion monitoring method for complex industrial scenarios described in the present invention, the cross-modal time series alignment includes adding a repaired mark to the repaired repair value, and giving priority to the original timestamp of the data with the repaired mark during the time series alignment process;

[0019] The repaired data is time-based, and the signal chain delay of the collected sensor is calculated based on the total length of sampling time, repair time, and transmission time. The delay is added to the original timestamp of each data to obtain an accurate timestamp;

[0020] Build a time alignment buffer, extract the compensation timestamps for the repaired and marked multi-source data, align the windows within the set time, and mark all sensor data in the window as the same time series batch;

[0021] At the same time, the time alignment accuracy is reversely calculated based on the kinematic formula.

[0022] As a preferred solution of the multi-dimensional data fusion monitoring method for complex industrial scenarios described in the present invention, wherein: the fusion weight includes real-time judgment of the equipment operating status through embedded working condition recognition of a two-level judgment mechanism;

[0023] Primary determination: The primary determination unit calculates the current fluctuation rate every 2 seconds. When the current fluctuation rate is less than 0.05 for 5 consecutive times, it is marked as a steady-state working condition and a steady-state confirmation signal is sent to the secondary determination unit. Otherwise, an emergency stop signal is sent to the stimulation determination unit.

[0024] Secondary determination: the secondary determination unit monitors the confirmation signal of the primary determination in real time, and immediately seizes control when no steady-state signal is received or an emergency stop signal is received, forcibly switches to transient mode and marks it as a transient operating condition.

[0025] As a preferred solution of the multi-dimensional data fusion monitoring method for complex industrial scenarios described in the present invention, wherein: the fusion weight also includes generating a dynamic fusion weight allocation strategy based on time alignment accuracy combined with fuzzy logic algorithm;

[0026] Under steady-state conditions, the sensors are assigned fusion weights of 0.9, 0.8, 0.7...0.1 to the data collected by the current sensors in the order of priority of accuracy ranking;

[0027] Under transient conditions, the sensors are assigned fusion weights of 0.8, 0.7, 0.6...0.1 to the data collected by the current sensors in the order of priority of response speed.

[0028] When it is detected that the data integrity score of any sensor is lower than 0.5, no weight will be assigned to the current sensor.

[0029] As a preferred solution of the multi-dimensional data fusion monitoring method for complex industrial scenarios described in the present invention, the fusion includes fusing multi-source heterogeneous data according to the obtained fusion weights and calculating the comprehensive anomaly score of the multi-source features;

[0030] When the comprehensive anomaly score is greater than the set threshold, the matching degree of each fault type is calculated. k :

[0031]

[0032] in, For the input data, is the characteristic reference value of typical fault k, α m is the fusion weight;

[0033] Select Match k <0.3 of the fault types generate a fault hypothesis set {Fault k ∣1-Match k}, encapsulate the fault hypothesis parameters into a structured data packet, including the fault type code and the impact range radius.

[0034] As a preferred solution of the multi-dimensional data fusion monitoring method for complex industrial scenarios described in the present invention, the digital twin model includes accepting items with a confidence level higher than 0.7 in the fault hypothesis set as injection parameters;

[0035] The fault type code is matched to the corresponding component model in the digital twin, and the impact radius is converted into the spatial boundary conditions of the simulation environment. A virtual sensor array synchronized with the actual equipment is set up in the digital twin. The distribution of actual monitoring data and simulation data is compared every 20 seconds. When the trend deviation of any node data exceeds the safety margin, the model parameters are triggered to perform adaptive calibration.

[0036] Record the time series of changes in system state variables during the simulation process, identify the components that first exhibit abnormal characteristics and whose propagation rate exceeds the historical average of similar faults, define priority detection areas based on the radius of the impact range, and generate a fault tree with hierarchical relationships;

[0037] The system connects to the Internet and enters equipment maintenance knowledge into the equipment maintenance knowledge base, matches the terminal nodes of the fault tree with typical solutions in the system's equipment maintenance knowledge base, calculates the remaining life credible interval based on the historical replacement cycle of the current data, and outputs a decision instruction set including emergency shutdown recommendations and spare parts preparation lists.

[0038] Another object of the present invention is to provide a multi-dimensional data fusion monitoring system for complex industrial scenarios.

[0039] To solve the above technical problems, the present invention provides the following technical solutions: a multi-dimensional data fusion monitoring system for complex industrial scenarios, comprising: a data acquisition and preprocessing module, a credibility assessment and self-repair module, a cross-modal time series alignment module, a fusion weight dynamic adjustment module, and a digital twin and fault prediction module;

[0040] The data acquisition and preprocessing module is responsible for collecting multi-source heterogeneous data from industrial equipment, including equipment operating parameters, machine vision information, environmental perception data, maintenance logs, and personnel operation records; and performing preliminary cleaning and normalization on the collected data to ensure data comparability and consistency;

[0041] The credibility assessment and self-repair module implements multi-dimensional data quality assessment, builds a signal integrity detection, timing consistency detection and physical constraint compliance verification system; when data anomalies are found, the local trusted consensus mechanism is used to repair the data;

[0042] The cross-modal temporal alignment module performs temporal alignment on the repaired multi-source heterogeneous data to ensure temporal consistency of data from different sources;

[0043] The fusion weight dynamic adjustment module dynamically calculates and allocates fusion weights based on the real-time working conditions of the equipment and the integrity of the sensor data, ensuring that sensor data is selected for fusion during the data fusion process and optimizing the accuracy of the data output;

[0044] The digital twin and fault prediction module combines the fused data with the digital twin model to perform real-time simulation and anomaly analysis. When abnormal features are detected, it automatically triggers model parameter calibration and provides decision support based on fault tree analysis, including emergency shutdown recommendations and spare parts preparation lists.

[0045] The present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program and is characterized in that when the processor executes the computer program, the steps of the multidimensional data fusion monitoring method for complex industrial scenarios are implemented.

[0046] The present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of the multi-dimensional data fusion monitoring method for complex industrial scenarios are implemented.

[0047] Beneficial effects of the present invention: The present invention realizes efficient data processing and fault monitoring. First, credibility assessment and self-repair processing ensure data quality and improve the accuracy of subsequent analysis. Next, cross-modal timing alignment eliminates the inconsistency caused by time delays and enhances the accuracy of data in real-time monitoring. Dynamic adjustment of fusion weights flexibly configures data sources according to equipment operating conditions, thereby improving the responsiveness of fault detection. Subsequently, the fusion output value is set as the edge weight, making the anomaly scoring analysis more comprehensive and enhancing the ability to respond quickly to different fault types. Finally, the digital twin model is combined with simulation analysis of the abnormal propagation path to further enhance the ability to predict and respond to potential faults, ensuring the high reliability and safety of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0049] Figure 1 This is an overall flow chart of a multi-dimensional data fusion monitoring method for complex industrial scenarios provided by one embodiment of the present invention.

[0050] Figure 2 This is a system solution module diagram of a multi-dimensional data fusion monitoring system for complex industrial scenarios provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0051] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0052] Example 1, reference Figure 1 , which is the first embodiment of the present invention, provides a multi-dimensional data fusion monitoring method for complex industrial scenarios, including:

[0053] S1: Perform credibility assessment and self-repair processing on multi-source heterogeneous data of industrial equipment.

[0054] It should be noted that the multi-source heterogeneous data of industrial equipment is collected, including equipment operating parameter data, machine vision data, environmental perception data, maintenance log data, and personnel operation data;

[0055] Equipment operating parameter data includes but is not limited to continuous data: temperature, pressure, current, vibration amplitude; discrete data: valve open / close status (0 / 1), fault code (such as ISO 10816 standard code), etc.

[0056] Machine vision data collected by industrial cameras and infrared thermal imagers includes but is not limited to structured data: part size measurements (such as mm values ​​after pixel coordinate conversion); unstructured data: equipment surface defect images (RGB matrix), thermal imaging temperature field distribution maps, etc.

[0057] Environmental sensing data includes but is not limited to workshop environment: dust concentration, humidity; energy consumption: three-phase voltage fluctuation curve, compressed air flow, etc.

[0058] Maintenance log data includes but is not limited to structured: spare parts replacement records (timestamp + component code); semi-structured: maintenance work order text (including fault phenomena described in natural language), etc.

[0059] It should also be noted that a multi-dimensional data quality assessment system including signal integrity testing, timing consistency testing, and physical constraint compliance verification is constructed, specifically:

[0060] The signal integrity test calculates the packet loss rate and the check code matching degree to generate the integrity score S int =1-N loss / N total , where N loss is the number of lost packets, N total is the expected total number of packets; when the integrity score exceeds the preset device standard threshold, it is marked as a signal anomaly;

[0061] Use time series consistency analysis to detect the variance of the time intervals between adjacent data points. When the variance exceeds the set device sampling period threshold, it is marked as a time series anomaly.

[0062] Compare sensor readings with the physical limits of the equipment through physical constraint compliance, identify out-of-limit data, generate constraint conflict reports, and mark them as constraint anomalies;

[0063] When any of the following anomalies occur: signal anomaly, timing anomaly, or constraint anomaly, the current data will be marked as abnormal data;

[0064] Treating each piece of data as a node, a local trusted consensus mechanism is established between nodes based on abnormal data. When any node detects abnormal data, it broadcasts a verification request containing the hash value, integrity score, and timestamp of the original data to adjacent nodes.

[0065] The node receiving the request retrieves data in the same time window in the local database. If more than 60% of the nodes return matching constraint conflict reports, the data repair process is triggered;

[0066] For data lost within a set time, the historical average value of the same operating parameters under normal operating conditions of the equipment is used for interpolation and filling;

[0067] For abnormal data that persists beyond the set time, the device thermodynamic model is called to generate a predicted value, and the predicted value is weighted and fused with the adjacent sensor readings to output the repair value V of the abnormal data. repair :

[0068]

[0069] Among them, V model is the value predicted by the thermodynamic model, is the i-th sensor reading of n adjacent sensors of the same type, ρ is the weight assigned to different parameters, and n is the number of sensors.

[0070] By performing credibility assessment and self-repair on multi-source, heterogeneous data from industrial equipment, standardized data quality management is achieved. Specifically, a multidimensional data quality assessment system is established to ensure data integrity, temporal consistency, and compliance with physical constraints, effectively identifying and flagging abnormal data. This improves data reliability and lays a solid foundation for subsequent data processing and fusion. Ultimately, it achieves the beneficial effect of reducing data noise and errors, significantly improving the accuracy of subsequent analysis.

[0071] S2: Perform cross-modal time series alignment on the repaired multi-source heterogeneous data and enhance the accuracy of time series alignment.

[0072] It should be noted that a repaired mark is added to the repaired value, and the original timestamp of the data with the repaired mark is preferentially used during the timing alignment process;

[0073] The repaired data is time-based, and the signal chain delay of the collected sensor is calculated based on the total length of sampling time, repair time, and transmission time. The delay is added to the original timestamp of each data to obtain an accurate timestamp;

[0074] Build a time alignment buffer, extract the compensation timestamps for the repaired and marked multi-source data, align the windows within the set time, and mark all sensor data in the window as the same time series batch;

[0075] Reversely calculate the time alignment accuracy based on the kinematic formula:

[0076] T recalc =T vision -(θ-θ ref ) / ω

[0077] Where ω is the average angular velocity of the device (the angular velocity of the device refers to the instantaneous angular velocity of the rotating parts of the device, such as the speed of the core moving parts such as the motor main shaft and gearbox output shaft), θ ref is the reference phase angle, T vision is the original timestamp, T recalc is the time alignment accuracy, and θ is the actual phase angle of the device detected by the laser rangefinder at the moment when the data is collected in the image of the visual data.

[0078] By performing cross-modal time alignment on the repaired, multi-source, heterogeneous data and enhancing its accuracy, data temporal consistency is achieved. This process prioritizes the original timestamps of the repaired, marked data, ensuring highly accurate time alignment. This eliminates data inconsistencies caused by time delays, thereby improving the quality of data fusion. Ultimately, this ensures that real-time monitoring data more accurately reflects equipment operating conditions, optimizing monitoring effectiveness.

[0079] S3: Dynamically adjust the fusion weight of multi-source heterogeneous data based on the real-time working conditions of the equipment.

[0080] It should be noted that the operating status of the equipment is determined in real time through the embedded working condition recognition of the two-level judgment mechanism;

[0081] Primary determination: The primary determination unit calculates the current fluctuation rate every 2 seconds. When the current fluctuation rate is less than 0.05 for 5 consecutive times, it is marked as a steady-state working condition and a steady-state confirmation signal is sent to the secondary determination unit. Otherwise, an emergency stop signal is sent to the stimulation determination unit.

[0082] Secondary determination: the secondary determination unit monitors the confirmation signal of the primary determination in real time, and immediately seizes control when no steady-state signal is received or an emergency stop signal is received, forcibly switches to transient mode and marks it as a transient operating condition.

[0083] It should also be noted that the dynamic fusion weight allocation strategy is generated based on the time alignment accuracy combined with the fuzzy logic algorithm;

[0084] Under steady-state conditions, the sensors are assigned fusion weights of 0.9, 0.8, 0.7...0.1 to the data collected by the current sensors in the order of priority of accuracy ranking;

[0085] Under transient conditions, the sensors are assigned fusion weights of 0.8, 0.7, 0.6...0.1 to the data collected by the current sensors in the order of priority of response speed.

[0086] When it is detected that the data integrity score of any sensor is lower than 0.5, no weight will be assigned to the current sensor;

[0087] A sliding time window algorithm is used to set a sliding time window of a specified length. When the operating condition mark changes more than three times within the window, the fast switching mode is enabled, that is, the fast switching between steady-state and transient operating conditions. During the mode switching transition period, a weighted smooth transition algorithm is used to avoid data jumps.

[0088] By dynamically adjusting the fusion weights of multi-source heterogeneous data based on the real-time operating conditions of the equipment, data processing strategies for both steady-state and transient conditions are clarified. This mechanism monitors equipment status in real time through multi-level determination, adjusting the weights of different sensor data to ensure that the most reliable data sources are always prioritized under different operating conditions. The goal is to improve the flexibility and adaptability of data fusion, thereby increasing the accuracy of monitoring data and ultimately enhancing fault detection and early warning capabilities.

[0089] S4: Use the output value after fusion weight as the edge weight to guide the fusion of multi-source heterogeneous data.

[0090] Furthermore, the fusion includes fusing the multi-source heterogeneous data according to the obtained fusion weights, and simultaneously calculating the comprehensive anomaly score of the multi-source features;

[0091] When the comprehensive anomaly score is greater than the set threshold, the matching degree of each fault type is calculated. k :

[0092]

[0093] in, is the observed eigenvalue, is the characteristic reference value of typical fault k, α m is the feature importance coefficient;

[0094] Select Match k <0.3 Fault type generation hypothesis set {Fault k ∣1-Match k The fault hypothesis parameters are encapsulated into a structured data packet, which includes the fault type code (according to the IEC 62264 standard) and the impact range radius. The impact range radius is the product of the average historical propagation rate of similar faults and the current system response delay time.

[0095] By using the weighted output values ​​of fusion as edge weights to guide the fusion of multi-source heterogeneous data, we can integrate various features to perform anomaly scoring analysis, achieving accurate identification of potential faults. By effectively combining data from different sources and calculating a comprehensive anomaly score, we can better predict and identify equipment failures. Ultimately, by establishing feature matching and fault hypotheses, we improve the response speed and handling capabilities for different fault types.

[0096] S5: Combine the digital twin model to perform simulation analysis of abnormal propagation paths.

[0097] Specifically, the entries with screening confidence higher than 0.7 in the fault hypothesis set are accepted as injection parameters;

[0098] The fault type code is matched to the corresponding component model in the digital twin, and the impact radius is converted into the spatial boundary conditions of the simulation environment. A virtual sensor array synchronized with the actual equipment is set up in the digital twin. The distribution of actual monitoring data and simulation data is compared every 20 seconds. When the trend deviation of any node data exceeds the safety margin, the model parameters are triggered to perform adaptive calibration.

[0099] Record the time series of changes in system state variables during the simulation process, identify the components that first exhibit abnormal characteristics and whose propagation rate exceeds the historical average of similar faults, define priority detection areas based on the radius of the impact range, and generate a fault tree with hierarchical relationships;

[0100] The system connects to the Internet and enters equipment maintenance knowledge into the equipment maintenance knowledge base, matches the terminal nodes of the fault tree with typical solutions in the system's equipment maintenance knowledge base, calculates the remaining life credible interval based on the historical replacement cycle of the current data, and outputs a decision instruction set including emergency shutdown recommendations and spare parts preparation lists.

[0101] By combining digital twin models with simulation analysis of abnormal propagation paths, we can further predict and prevent the impact of faults. This approach injects high-confidence data into a virtual environment for real-time comparison and adaptive calibration. This enables real-time monitoring and analysis during actual operation, facilitating the identification of potential fault propagation paths. Ultimately, this improves the intelligence of equipment management and reduces potential losses.

[0102] Example 2 is the second embodiment of the present invention, which is different from the previous embodiment in that:

[0103] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0104] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0105] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0106] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0107] Example 3, reference Figure 2 , which is the third embodiment of the present invention, provides a multi-dimensional data fusion monitoring system for complex industrial scenarios, including: a data acquisition and preprocessing module, a credibility assessment and self-repair module, a cross-modal time series alignment module, a fusion weight dynamic adjustment module, and a digital twin and fault prediction module;

[0108] The data acquisition and preprocessing module is responsible for collecting multi-source heterogeneous data from industrial equipment, including equipment operating parameters, machine vision information, environmental perception data, maintenance logs, and personnel operation records; and performing preliminary cleaning and normalization on the collected data to ensure data comparability and consistency;

[0109] The credibility assessment and self-repair module implements multi-dimensional data quality assessment and builds a signal integrity detection, timing consistency detection, and physical constraint compliance verification system. When data anomalies are found, the local trusted consensus mechanism is used to repair the data.

[0110] The cross-modal timing alignment module performs timing alignment on repaired multi-source heterogeneous data to ensure temporal consistency of data from different sources.

[0111] The fusion weight dynamic adjustment module dynamically calculates and allocates fusion weights based on the real-time working conditions of the equipment and the integrity of the sensor data, ensuring that sensor data is selected for fusion during the data fusion process and optimizing the accuracy of the data output;

[0112] The digital twin and fault prediction module combines fused data with the digital twin model to perform real-time simulation and anomaly analysis. When abnormal features are detected, it automatically triggers model parameter calibration and provides decision support based on fault tree analysis, including emergency shutdown recommendations and spare parts preparation lists.

[0113] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A multi-dimensional data fusion monitoring method for complex industrial scenarios, characterized by: include, Conduct credibility assessment and self-repair processing on multi-source heterogeneous data of industrial equipment; Perform cross-modal time series alignment on the repaired multi-source heterogeneous data and enhance the accuracy of time series alignment; Dynamically adjust the fusion weight of multi-source heterogeneous data based on the real-time working conditions of the equipment; The output value after fusion weight is used as edge weight to guide the fusion of multi-source heterogeneous data; Combine the digital twin model to perform simulation analysis of abnormal propagation paths.

2. The multi-dimensional data fusion monitoring method for complex industrial scenarios according to claim 1, characterized in that: The reliability assessment and self-repair processing includes collecting multi-source heterogeneous data of industrial equipment, including equipment operating parameter data, machine vision data, environmental perception data, maintenance log data, and personnel operation data, and performing preliminary cleaning and normalization processing on the collected data; Build a multi-dimensional data quality assessment system that includes signal integrity testing, timing consistency testing, and physical constraint compliance verification. Specifically: Calculating the packet loss rate and the checksum matching degree through the signal integrity detection to generate an integrity score. When the integrity score exceeds a preset device standard threshold, it is marked as a signal anomaly. Use time series consistency analysis to detect the variance of the time intervals between adjacent data points. When the variance exceeds the set device sampling period threshold, it is marked as a time series anomaly. Compare sensor readings with the physical limits of the equipment through physical constraint compliance, identify out-of-limit data, generate constraint conflict reports, and mark them as constraint anomalies; When any of the following anomalies occur: signal anomaly, timing anomaly, or constraint anomaly, the current data will be marked as abnormal data; Treating each piece of data as a node, a local trusted consensus mechanism is established between nodes based on abnormal data. When any node detects abnormal data, it broadcasts a verification request containing the hash value, integrity score, and timestamp of the original data to adjacent nodes. The node receiving the request retrieves data in the same time window in the local database. If more than 60% of the nodes return matching constraint conflict reports, the data repair process is triggered; For data lost within a set time, the historical average value of the same operating parameters under normal operating conditions of the equipment is used for interpolation and filling; For abnormal data that persists beyond the set time, the device thermodynamic model is called to generate a predicted value, which is then weighted and fused with adjacent sensor readings to output a repair value for the abnormal data.

3. The multi-dimensional data fusion monitoring method for complex industrial scenarios according to claim 2, characterized in that: The cross-modal time series alignment includes adding a repaired mark to the repaired repair value, and giving priority to the original timestamp of the data with the repaired mark during the time series alignment process; The repaired data is time-based, and the signal chain delay of the collected sensor is calculated based on the total length of sampling time, repair time, and transmission time. The delay is added to the original timestamp of each data to obtain an accurate timestamp; Build a time alignment buffer, extract the compensation timestamps for the repaired and marked multi-source data, align the windows within the set time, and mark all sensor data in the window as the same time series batch; At the same time, the time alignment accuracy is reversely calculated based on the kinematic formula.

4. The multi-dimensional data fusion monitoring method for complex industrial scenarios according to claim 3, characterized in that: The fusion weight includes real-time judgment of the equipment operating status through embedded working condition recognition of a two-level judgment mechanism; Primary determination: The primary determination unit calculates the current fluctuation rate every 2 seconds. When the current fluctuation rate is less than 0.05 for 5 consecutive times, it is marked as a steady-state working condition and a steady-state confirmation signal is sent to the secondary determination unit. Otherwise, an emergency stop signal is sent to the stimulation determination unit. Secondary determination: the secondary determination unit monitors the confirmation signal of the primary determination in real time, and immediately seizes control when no steady-state signal is received or an emergency stop signal is received, forcibly switches to transient mode and marks it as a transient operating condition.

5. The multi-dimensional data fusion monitoring method for complex industrial scenarios according to claim 4, characterized in that: The fusion weight also includes generating a dynamic fusion weight allocation strategy based on time alignment accuracy combined with a fuzzy logic algorithm; Under steady-state conditions, the sensors are assigned fusion weights of 0.9, 0.8, 0.7...0.1 to the data collected by the current sensors in the order of priority of accuracy ranking; Under transient conditions, the sensors are assigned fusion weights of 0.8, 0.7, 0.6...0.1 to the data collected by the current sensors in the order of priority of response speed. When it is detected that the data integrity score of any sensor is lower than 0.5, no weight will be assigned to the current sensor.

6. The multi-dimensional data fusion monitoring method for complex industrial scenarios according to claim 4, characterized in that: The fusion includes fusing multi-source heterogeneous data according to the obtained fusion weights and calculating a comprehensive anomaly score of multi-source features; When the comprehensive anomaly score is greater than the set threshold, the matching degree of each fault type is calculated. k : in, For the input data, is the characteristic reference value of typical fault k, α m is the fusion weight; Select Match k <0.3 of the fault types generate a fault hypothesis set {Fault k ∣1-Match k }, encapsulate the fault hypothesis parameters into a structured data packet, including the fault type code and the impact range radius.

7. The multi-dimensional data fusion monitoring method for complex industrial scenarios according to claim 4, characterized in that: The digital twin model includes accepting items with a screening confidence level higher than 0.7 in the fault hypothesis set as injection parameters; The fault type code is matched to the corresponding component model in the digital twin, and the impact radius is converted into the spatial boundary conditions of the simulation environment. A virtual sensor array synchronized with the actual equipment is set up in the digital twin. The distribution of actual monitoring data and simulation data is compared every 20 seconds. When the trend deviation of any node data exceeds the safety margin, the model parameters are triggered to perform adaptive calibration. Record the time series of changes in system state variables during the simulation process, identify the components that first exhibit abnormal characteristics and whose propagation rate exceeds the historical average of similar faults, define priority detection areas based on the radius of the impact range, and generate a fault tree with hierarchical relationships; The system connects to the Internet and enters equipment maintenance knowledge into the equipment maintenance knowledge base, matches the terminal nodes of the fault tree with typical solutions in the system's equipment maintenance knowledge base, calculates the remaining life credible interval based on the historical replacement cycle of the current data, and outputs a decision instruction set including emergency shutdown recommendations and spare parts preparation lists.

8. A multi-dimensional data fusion monitoring system for complex industrial scenes, applying the multi-dimensional data fusion monitoring method for complex industrial scenes according to any one of claims 1 to 7, characterized in that: include: Data acquisition and preprocessing module, credibility assessment and self-repair module, cross-modal time series alignment module, fusion weight dynamic adjustment module, digital twin and fault prediction module; The data acquisition and preprocessing module is responsible for collecting multi-source heterogeneous data from industrial equipment, including equipment operating parameters, machine vision information, environmental perception data, maintenance logs, and personnel operation records; and performing preliminary cleaning and normalization on the collected data to ensure data comparability and consistency; The credibility assessment and self-repair module implements multi-dimensional data quality assessment, builds a signal integrity detection, timing consistency detection and physical constraint compliance verification system; when data anomalies are found, the local trusted consensus mechanism is used to repair the data; The cross-modal temporal alignment module performs temporal alignment on the repaired multi-source heterogeneous data to ensure temporal consistency of data from different sources; The fusion weight dynamic adjustment module dynamically calculates and allocates fusion weights based on the real-time working conditions of the equipment and the integrity of the sensor data, ensuring that sensor data is selected for fusion during the data fusion process and optimizing the accuracy of the data output; The digital twin and fault prediction module combines the fused data with the digital twin model to perform real-time simulation and anomaly analysis. When abnormal features are detected, it automatically triggers model parameter calibration and provides decision support based on fault tree analysis, including emergency shutdown recommendations and spare parts preparation lists.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the multi-dimensional data fusion monitoring method for complex industrial scenarios according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the multi-dimensional data fusion monitoring method for complex industrial scenarios according to any one of claims 1 to 7 are implemented.

Images