Equipment operation and maintenance security pre-control system multivariate heterogeneous data fusion method and system

Abstract

The invention discloses a multivariate heterogeneous data fusion method and system for an equipment operation and maintenance security pre-control system. The method comprises the following steps: receiving a multi-source heterogeneous modal data set; constructing a multivariate heterogeneous data fusion matrix, and calculating to obtain matrix variable elements; according to the matrix variable elements, constructing a multi-source heterogeneous modal data feature fusion calculation formula, and fusing multi-source heterogeneous modal data features; a dynamic metabolism model between the equipment health state and the external environment is constructed, and a digital metabolism rate is obtained; constructing a health prediction model, and analyzing the digital metabolic rate to obtain health degree prediction data; according to the equipment prediction health degree, the equipment prediction health degree is analyzed and judged, an early warning level is obtained, and a self-healing strategy is generated. The method is used for reflecting the operation and maintenance safety of the equipment according to the pre-equipment health degree by fusing the modal features, so that the pre-control of the operation and maintenance safety of the equipment is fed back more intelligently.

Description

Technical Field

[0001] This invention relates to the field of equipment operation and maintenance safety pre-control, specifically to a method and system for multi-source heterogeneous data fusion in an equipment operation and maintenance safety pre-control system. Background Technology

[0002] Equipment operation and maintenance safety pre-control system is a core component of modern industrial equipment intelligent management. Through real-time monitoring of equipment operating status, multi-source data fusion analysis, and intelligent early warning, it achieves early perception and proactive protection of equipment health status, effectively avoiding sudden failures, reducing operation and maintenance costs, and improving production continuity and safety. With the deepening of the Industrial Internet and intelligent manufacturing, this type of system has become a key technological support for enterprises to achieve predictive maintenance and intelligent operation and maintenance.

[0003] In existing technologies, equipment condition monitoring often relies on data from single or limited types of sensors, such as analyzing vibration or temperature data independently. While this can reflect the local state of the equipment to some extent, it is difficult to comprehensively capture the overall health status of the equipment and its evolution under the coupling effects of multiple physical fields. Some systems attempt to introduce multi-source heterogeneous modal data, such as vibration, temperature, acoustic, and environmental parameters, but effective fusion mechanisms are still lacking at the data level. Feature merging is often done using fixed weights or simple weighted averages, failing to reflect the dynamic correlation and causal influence between different multi-source heterogeneous modal data. Furthermore, existing early warning mechanisms are mostly based on static threshold judgments, which cannot adapt to the dynamic changes in equipment status under complex and variable operating conditions, leading to frequent false alarms and missed alarms, scattered early warning information, and difficulties in tracing the source.

[0004] The existing technology has the following main drawbacks: First, there is a lack of effective dynamic correlation modeling between multi-source heterogeneous modal data, making it difficult to identify fault propagation paths and coupling effects; second, the early warning mechanism is rigid, relying on fixed thresholds and unable to respond in real time to environmental changes and equipment status evolution; third, there is a lack of integrated indicators that can comprehensively analyze multi-source information and quantify the overall health status of equipment, resulting in limited system intelligence, and operation and maintenance decisions still heavily rely on human experience, with delayed response and inability to achieve true predictive maintenance and self-healing control. Summary of the Invention

[0005] This application provides a method and system for fusing multi-modal heterogeneous data in an equipment operation and maintenance safety pre-control system. It addresses the technical problems in existing technologies, such as fragmented early warning information, difficulty in tracing the source, rigid fixed threshold alarm mechanisms, inability to dynamically reflect the health status evolution of equipment under complex and changing operating conditions, and the lack of a comprehensive indicator that can integrate multimodal information and quantify the overall health status of equipment.

[0006] In view of the above problems, this application provides a method and system for multi-source heterogeneous data fusion in an equipment operation and maintenance safety pre-control system.

[0007] The first aspect of this application provides a method for fusing multi-source heterogeneous data in an equipment operation and maintenance safety pre-control system. The method is applied to a multi-source heterogeneous data fusion system for equipment operation and maintenance safety pre-control. The method includes: receiving a multi-source heterogeneous modal data set; constructing a multi-source heterogeneous data fusion matrix and calculating matrix variable elements; constructing a multi-source heterogeneous modal data feature fusion calculation formula based on the matrix variable elements, and fusing multi-source heterogeneous modal data features; constructing a dynamic metabolic model between equipment health status and the external environment to obtain a digital metabolic rate; constructing a health prediction model, analyzing the digital metabolic rate, and obtaining health prediction data; analyzing and judging the predicted health status of the equipment based on the predicted health status, obtaining an early warning level, and generating a self-healing strategy.

[0008] A second aspect of this application provides a system comprising: a first receiving unit for receiving a multi-source heterogeneous modal data set; a first calculation module for constructing a multi-source heterogeneous data fusion matrix and calculating matrix variable elements; a second calculation module for constructing a multi-source heterogeneous modal data feature fusion calculation formula based on the matrix variable elements and fusing multi-source heterogeneous modal data features; a third calculation module for constructing a dynamic metabolic model between device health status and the external environment and obtaining a digital metabolic rate; a fourth calculation module for constructing a health prediction model, analyzing the digital metabolic rate, and obtaining health prediction data; and a first analysis and judgment unit for analyzing and judging the device's predicted health status, obtaining an early warning level, and generating a self-healing strategy.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application embodiment receives a multi-source heterogeneous modal data set; constructs a multi-source heterogeneous data fusion matrix and calculates the matrix variable elements; based on the matrix variable elements, constructs a multi-source heterogeneous modal data feature fusion calculation formula to fuse the multi-source heterogeneous modal data features; constructs a dynamic metabolic model between the equipment health status and the external environment to obtain a digital metabolic rate; constructs a health prediction model, analyzes the digital metabolic rate, and obtains health prediction data; based on the predicted equipment health, analyzes and judges the predicted equipment health, obtains an early warning level, and generates a self-healing strategy. This application is used to preset the equipment health status to reflect the equipment operation and maintenance safety by fusing the features of various modalities, thereby providing more intelligent feedback and pre-control of equipment operation and maintenance safety.

[0010] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0012] Figure 1 This is a flowchart of the method in this application. Detailed Implementation

[0013] This application provides a remote intelligent unified control system and device based on the Internet of Things (IoT) to address the technical problems of existing IoT control technologies, such as reliance on manual subjective operation and poor adjustment effects.

[0014] To address the aforementioned technical problems, the overall approach of the technical solution provided in this application is as follows: This application embodiment fuses various modal features and presets the device health level to reflect the device operation and maintenance safety, thereby providing more intelligent feedback and pre-control of device operation and maintenance safety.

[0015] After introducing the basic principles of this application, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0016] Example 1 like Figure 1 As shown, this application provides a method for fusing multi-source heterogeneous data in an equipment operation and maintenance safety pre-control system. The method is characterized by being applied to a multi-source heterogeneous data fusion system for equipment operation and maintenance safety pre-control, and includes: S100: Receives a set of multi-source heterogeneous modal data; In this embodiment, the multi-source heterogeneous modal data set refers to data packets collected from different physical sources, such as vibration, temperature, and sound, reflecting the overall state of the equipment. The system needs to be capable of processing multiple interface protocols and data formats, and align all data under a unified microsecond-level timestamp to ensure spatiotemporal consistency in subsequent analysis. This process is like performing a comprehensive "physical examination" on the equipment, collecting various "vital signs" to provide basic data support for subsequent diagnosis.

[0017] Step S100 in the method provided in this application embodiment includes: S110: Receive equipment vibration data as the first multi-source heterogeneous mode data; S120: Receive device surface temperature distribution data as second multi-source heterogeneous mode data; S130: Receive equipment operating noise data as third-source heterogeneous modal data; S140: Receive ambient temperature and humidity as the fourth multi-source heterogeneous modal data; S150: Receive dust concentration as the fifth multi-source heterogeneous mode data; S160: The first multi-source heterogeneous modal data, the second multi-source heterogeneous modal data, the third multi-source heterogeneous modal data, the fourth multi-source heterogeneous modal data, and the fifth multi-source heterogeneous modal data are taken as a multi-source heterogeneous modal data set.

[0018] In this application example, the system's perception of equipment status is the starting point. Its core task is to synchronously collect and aggregate multi-dimensional data from different physical sensors to form a raw dataset that can be used for analysis. Through the collaborative work of a multi-source heterogeneous sensor network, the system can comprehensively capture various characteristics of the equipment's operating status, providing a sufficient data foundation for subsequent in-depth analysis and intelligent decision-making. This step solves the limitations of traditional single-source data source analysis, ensuring that subsequent analysis is based on a comprehensive and multi-faceted data foundation. Specifically, an accelerometer installed on the compressor rotor bearing housing collects vibration acceleration data in three directions at a sampling frequency of 25,600 times per second, serving as the first multi-source heterogeneous modal data. An infrared thermal imager aligned with the compressor housing bearing area collects surface temperature distribution image data at a rate of 25 frames per second, serving as the second multi-source heterogeneous modal data. Acoustic sensors placed near the equipment collect noise data during equipment operation at a sampling rate of 44,100 times per second, serving as the third multi-source heterogeneous modal data. Temperature and humidity sensors installed in the machine room collect ambient temperature and humidity data, serving as the fourth multi-source heterogeneous modal data. By using air quality sensors, dust concentration data in the environment is collected as the fifth multi-source heterogeneous modal data. For example, in the daily monitoring of a centrifugal compressor, the system simultaneously receives vibration acceleration signals, infrared thermal images of the bearing area, audio streams of operating noise, and temperature, humidity, and dust concentration readings of the machine room environment, forming a complete and synchronous multi-source heterogeneous modal data set.

[0019] S200: Construct a multivariate heterogeneous data fusion matrix and calculate the matrix variable elements; In this embodiment, the matrix serves as the system's brain, with its elements representing the probability that an anomaly in the j-th mode leads to an anomaly in the i-th mode. It is a self-optimizing model based on historical data, ensuring that the more experience the system accumulates, the more accurate its fault tracing becomes. This process simulates the diagnostic thinking of human experts, continuously observing and summarizing the correlations between fault phenomena to form increasingly accurate fault diagnosis experience. The establishment of the causal matrix enables the system to understand fault propagation mechanisms, providing crucial support for accurate diagnosis and prediction.

[0020] Step S200 in the method provided in this application embodiment includes: S210: Construct a multi-source heterogeneous data fusion matrix, where the matrix variable elements represent the degree of influence of multi-source heterogeneous modal data anomalies on the anomalies of associated multi-source heterogeneous modal data; S220: Real-time calculation to obtain the mean value of multi-source heterogeneous modal data; S230: Preset threshold for monitoring abnormal fluctuations in multi-source heterogeneous modal data, and learning parameters; S240: When abnormal fluctuations are detected in multi-source heterogeneous modal data, output abnormal modal fluctuation data; when abnormal fluctuations are detected in correlated multi-source heterogeneous modal data, output correlated modal abnormal fluctuation data. S250: Construct a dynamically updated matrix variable element model, inputting a multi-source heterogeneous modal data set, modal anomaly fluctuation data, correlated modal anomaly fluctuation data, and learning parameters. The model is as follows: in For the elements of the new matrix variables, These are the elements of the original matrix variables. For learning parameters, To correlate modal anomaly fluctuation data, This is modal anomaly fluctuation data.

[0021] In this embodiment, online learning quantifies the causal relationships between anomalies in different multi-source heterogeneous modal data, forming a dynamically evolving fault knowledge graph. By establishing quantitative correlations between modalities, the system can gain a deep understanding of the intrinsic connections between various fault phenomena, providing crucial causal support for subsequent fault diagnosis and prediction. This step solves the problems of isolated fault mode analysis and inability to identify fault propagation paths in traditional methods. Specifically, a five-row, five-column multivariate heterogeneous data fusion matrix is ​​initialized, where rows and columns correspond to five different multi-source heterogeneous modal data. The initial value of all elements is set to 0.1. Matrix initialization establishes the basic framework for inter-modal correlation. The arithmetic mean and standard deviation of the multi-source heterogeneous modal data are calculated in real time within a sliding time window. Statistical feature calculation provides a basic reference for anomaly detection. The anomaly fluctuation monitoring threshold is set to 3 times the standard deviation, and a parameter controlling the learning speed is preset with a value of 0.1. The threshold setting ensures that only significant anomalies are responded to, avoiding false alarms. When the real-time value of a certain multi-source heterogeneous modal data continuously exceeds its threshold range, it is determined that the mode has experienced abnormal fluctuations, and the absolute value of its fluctuation amplitude is recorded. Anomaly detection can promptly capture changes in equipment status. Within a preset time window after the vibration anomaly occurs, other modes are continuously monitored. If abnormal fluctuations are also found in the infrared temperature data, the absolute value of its fluctuation amplitude is recorded. Time-series correlation analysis can reveal fault propagation patterns. The dynamically updated model is invoked to calculate the new values ​​of the influence factors of the vibration mode on the temperature mode and update the matrix. Through continuous learning, the matrix can accurately characterize the causal strength between modes. The online learning mechanism enables the system to have self-optimization capabilities. For example, the system detects that the vibration amplitude of the compressor first increases abnormally, and then its bearing temperature also rises significantly. Through model calculation, the values ​​of the elements representing the influence of vibration on temperature in the matrix are dynamically adjusted upwards, indicating that the system autonomously learns and reinforces the knowledge that abnormal vibration is an important cause of the bearing temperature increase.

[0022] S300: Based on the matrix variable elements, construct a multi-source heterogeneous modal data feature fusion calculation formula to fuse multi-source heterogeneous modal data features; In this embodiment, intelligent weighted fusion based on causal relationships is employed. This ensures that the final feature vector used for health assessment is not only comprehensive but also highlights key features most likely to cause current problems, laying the foundation for accurate prediction. This is similar to a doctor focusing on test results related to primary symptoms during diagnosis, rather than treating all test results equally. Intelligent feature fusion improves the accuracy of condition assessment and the sensitivity of fault identification.

[0023] Step S300 in the method provided in this application embodiment includes: S310: Based on the multivariate heterogeneous data fusion matrix where the elements of the new matrix variables are located; S320: Receives feature vectors of each modality, with preset vector-specific operators; S330: Construct a multi-source heterogeneous modal data feature fusion calculation formula, inputting the multivariate heterogeneous data fusion matrix, each modal feature vector, and vector-specific operators to fuse the multi-source heterogeneous modal data features. The calculation formula is as follows: *[F_v,F_i,F_a,F_n]*T in To enhance the post-fused feature vector, The matrix variable elements are: F_v, which is the vibration feature vector; F_i, which is the infrared feature vector; F_a, which is the acoustic feature vector; F_n, which is the environmental feature vector; and T, which is the vector-specific operator.

[0024] In this embodiment, the causal matrix obtained in the previous step is used as weights to deeply fuse the features extracted from the original data of each modality, generating a fused feature vector that enhances key information. Through deep fusion at the feature level, the system can extract the feature representation that best reflects the true state of the equipment, providing high-quality input data for subsequent health assessment and prediction. This step solves the problem of fixed weights and inability to reflect fault characteristics in traditional feature fusion methods. Specifically, a fast Fourier transform is performed on the vibration time-series data to extract its spectral features, forming a 256-dimensional vibration feature vector. Frequency domain analysis can reveal the periodicity and impact characteristics of the vibration signal. A pre-trained convolutional neural network model is used to process infrared thermal images, extracting their deep texture features and outputting a 1024-dimensional infrared feature vector. Deep learning can automatically learn the correlation pattern between temperature distribution and faults. Mel-frequency cepstral coefficient analysis is performed on the acoustic data to extract its auditory features, forming a 128-dimensional acoustic feature vector. Acoustic features can reflect the internal working state of the equipment. Temperature, humidity, and dust concentration data are normalized to form a 5-dimensional environmental feature vector. Environmental parameter normalization ensures the comparability of data across different dimensions. The aforementioned feature vectors are then multiplied with the causal influence matrix. The result is a new fused feature vector, where modal features with strong causal influence are given higher weights. Weighted fusion highlights key fault characteristics. For example, for this compressor, the system calculates its vibration spectrum features, infrared image features, noise features, etc., together with the updated causal matrix. In the generated fused feature vector, the feature dimensions related to vibration-temperature coupled fault modes are significantly enhanced.

[0025] S400: Construct a dynamic metabolic model between the device's health status and the external environment to obtain a digital metabolic rate; In this embodiment, the digital metabolic rate is an indicator that simulates biological metabolism. It indicates that when the external environment deteriorates and the equipment itself is in a sub-optimal state, its health declines at a non-linear rate. This concept effectively describes the equipment's resilience under harsh operating conditions; the worse the environmental conditions and the poorer the equipment's condition, the faster its health deteriorates. The introduction of the digital metabolic rate enables the prediction model to dynamically respond to environmental changes, improving prediction accuracy.

[0026] Step S400 in the method provided in this application embodiment includes: S410: Current health status of the receiving device, overall environmental stress; S420: Preset device environmental tolerance threshold, digital metabolic coefficient; S430: Construct a digital metabolic rate calculation formula, inputting the current health status of the device, comprehensive environmental stress, the device's environmental tolerance threshold, and the digital metabolic coefficient, to calculate the digital metabolic rate. The formula is as follows: in For digital metabolic rate, Digital metabolic coefficient To take into account environmental stress, This refers to the equipment's environmental tolerance threshold. This indicates the current health status of the device.

[0027] In this embodiment, the concept of digital metabolic rate is innovatively introduced to quantify the rate of deterioration of equipment health under specific environmental stress. By establishing a dynamic relationship model between equipment health and environmental factors, the system can more accurately assess the evolution trend of equipment under specific operating conditions. This step solves the problem that traditional prediction models ignore the influence of environmental factors. Specifically, an upper limit of environmental stress for the long-term normal operation of the compressor model is preset from the equipment specifications or historical operating data. The threshold setting provides a benchmark for environmental stress assessment. Based on real-time collected temperature, humidity, and dust data, the current comprehensive environmental stress value is calculated through weighted fusion. Environmental stress quantification reflects the degree of influence of external conditions on the equipment. Based on the fused feature vector, an assessment model calculates the current health index of the equipment, with a value between 0 and 1, where 1 represents perfect health. The health index comprehensively reflects the current state of the equipment. The above parameters are substituted into the digital metabolic rate calculation formula to obtain the digital metabolic rate. The metabolic rate calculation quantifies the rate at which environmental stress erodes the health of the equipment. For example, the current ambient temperature in the computer room is high, causing the comprehensive environmental stress to exceed the threshold. At the same time, the assessment shows that the health of the compressor bearing has dropped to 0.75. The system calculated a high digital metabolic rate, indicating that the health condition of the bearing is deteriorating rapidly under the current harsh environment.

[0028] S500: Build a health prediction model, analyze digital metabolic rate, and obtain health prediction data; In this embodiment, the model is a forward-looking predictor. Its core idea is that future health equals current health minus the product of the damage rate caused by environmental stress and time. The model concisely quantifies the cumulative damage effect of environmental stress on equipment lifespan. This method considers not only the current state of the equipment but also the influence of environmental factors, making the prediction results more accurate and practical. Forward-looking prediction enables maintenance personnel to take preventative measures in advance, achieving true preventative maintenance.

[0029] Step S500 in the method provided in this application embodiment includes: S510: Preset prediction time step, attenuation coefficient; S520: Construct a health prediction model by inputting the device's current health status, decay coefficient, prediction time step, and digital metabolic rate into the health prediction model to calculate the predicted health status of the device. The model is as follows: in Predict the health status of the equipment. The current health status of the device. The attenuation coefficient is... For digital metabolic rate, To predict the time step.

[0030] In this embodiment, the future health trend of the equipment is quantitatively predicted based on the current state and metabolic rate. By establishing a forward-looking health prediction model, the system can anticipate changes in the equipment's state in advance, providing decision support for preventative maintenance. This step solves the problem of large lag in traditional early warning systems. Specifically, a prediction time step and an environment-related decay coefficient are preset. Parameter settings ensure that the prediction model adapts to specific application scenarios. The current health level of the equipment, decay coefficient, prediction time step, and digital metabolic rate are substituted into the health prediction model to calculate the predicted health level of the equipment at a specific future time. The health prediction quantifies the future state change trend. For example, the system calculates the current health level of the compressor bearing to be 0.75. Combined with the calculated high digital metabolic rate, the model predicts that its health level will decrease to 0.58 after 12 hours.

[0031] S600: Based on the predicted health status of the equipment, analyze and judge the predicted health status of the equipment, obtain the warning level, and generate a self-healing strategy.

[0032] In this embodiment, the analysis results from the previous steps are transformed into specific operational and maintenance actions. Intelligent grading ensures the accuracy of early warnings, avoiding over- or under-warning. The generation of self-healing strategies is based on a deep understanding of the causes of failures, ensuring that the measures taken are targeted. This closed-loop processing method greatly improves operational and maintenance efficiency and reduces the need for manual intervention. The intelligent operational and maintenance closed loop achieves complete automation from state perception to decision execution, significantly improving equipment reliability and operational and maintenance efficiency.

[0033] Step S600 in the method provided in this application embodiment includes: S610: Preset intelligent hierarchical early warning system, the system is divided into hierarchical levels including safety level, attention level, risk level and high level; S620: Preset self-healing strategy knowledge base; S630: Based on the intelligent hierarchical early warning system, analyze and judge the predicted health status of the equipment to obtain the early warning level; S640: Access the self-healing measurement library based on the warning level to obtain self-healing strategy instructions.

[0034] In this embodiment, the prediction results are transformed into executable operation and maintenance instructions, realizing a closed loop from perception and prediction to decision execution. By establishing an intelligent early warning and self-healing mechanism, the system can automatically respond to changes in equipment status, achieving intelligent and automated operation and maintenance. This step solves the problems of traditional operation and maintenance relying on manual experience and slow response speed. Specifically, the system has preset intelligent hierarchical early warning rules. For example: a health level greater than or equal to 0.8 is the safe level, 0.6 to 0.8 is the attention level, 0.4 to 0.6 is the risk level, 0.2 to 0.4 is the high-risk level, and less than 0.2 is the critical level. Hierarchical early warning ensures that the response intensity matches the risk level. The system has preset a self-healing strategy knowledge base, which stores the strategies that should be triggered for different early warning levels combined with different root causes. The strategy knowledge base accumulates expert experience and historical handling solutions. The system determines that the predicted health level is 0.58, which belongs to the risk level. The risk level determination triggers the corresponding handling process. The system queries the causal matrix and identifies that the most important causal chain at present is abnormal vibration leading to temperature rise. Accordingly, a medium-level early warning response strategy for this pattern is matched from the strategy library. Root cause analysis ensures that the response measures are highly targeted. For example, the system automatically executes the following strategies: triggers a yellow medium-level early warning on the central control room screen, indicating the risk of bearing overheating; sends a command to the control system to reduce the compressor load by 10% to reduce vibration sources; links the lubrication system to temporarily increase the oil supply rate by 15% for 2 hours; automatically generates a planned maintenance work order in the work order management system, recommending a condition check and maintenance of the centrifugal compressor bearing within 12 hours, and pushes the work order to the maintenance team.

[0035] Example 2 Based on the same inventive concept as the multi-heterogeneous data fusion method for an equipment operation and maintenance safety pre-control system described in the foregoing embodiments, this application provides a multi-heterogeneous data fusion system for an equipment operation and maintenance safety pre-control system. This system is equipped with the aforementioned method, wherein the system includes: The first receiving unit is used to receive a multi-source heterogeneous modal data set; The first calculation module is used to construct a multi-dimensional heterogeneous data fusion matrix and calculate the matrix variable elements. The second calculation module is used to construct a multi-source heterogeneous modal data feature fusion calculation formula based on the matrix variable elements, and to fuse the multi-source heterogeneous modal data features. The third calculation module is used to construct a dynamic metabolic model between the device's health status and the external environment to obtain the digital metabolic rate. The fourth calculation module is used to build a health prediction model, analyze digital metabolic rate, and obtain health prediction data. The first analysis and judgment unit analyzes and judges the predicted health status of the equipment based on the predicted health status, obtains the warning level, and generates a self-healing strategy.

[0036] Preferred options also include: The second receiving unit is used to receive equipment vibration data as the first multi-source heterogeneous mode data; The third receiving unit is used to receive the surface temperature distribution data of the device as the second multi-source heterogeneous mode data; The fourth receiving unit is used to receive equipment operating noise data as the third multi-source heterogeneous modal data; The fifth receiving unit is used to receive ambient temperature and humidity as the fourth multi-source heterogeneous modal data; The sixth receiving unit is used to receive dust concentration as the fifth multi-source heterogeneous mode data; The first processing output unit is used to take the first multi-source heterogeneous modal data, the second multi-source heterogeneous modal data, the third multi-source heterogeneous modal data, the fourth multi-source heterogeneous modal data, and the fifth multi-source heterogeneous modal data as a multi-source heterogeneous modal data set.

[0037] Preferred options also include: The first building unit is used to construct a multi-source heterogeneous data fusion matrix. The matrix variable elements represent the degree of influence of multi-source heterogeneous modal data anomalies on the anomalies of associated multi-source heterogeneous modal data. The first computing unit is used to calculate the mean of multi-source heterogeneous modal data in real time. The first preset unit is used to preset the monitoring threshold for abnormal fluctuations in multi-source heterogeneous modal data and to learn parameters. The second analysis and judgment unit is used to output abnormal mode fluctuation data when abnormal fluctuations are detected in multi-source heterogeneous modal data; and to output abnormal fluctuation data of associated multi-source heterogeneous modal data when abnormal fluctuations are detected. The second computational unit is used to construct a dynamically updated matrix variable element model, inputting a multi-source heterogeneous modal data set, modal anomaly fluctuation data, correlated modal anomaly fluctuation data, and learning parameters. The model is as follows: in For the elements of the new matrix variables, These are the elements of the original matrix variables. For learning parameters, To correlate modal anomaly fluctuation data, This is modal anomaly fluctuation data.

[0038] Preferred options also include: The first analysis unit is used to fuse the matrix based on the multivariate heterogeneous data containing the elements of the new matrix variables; The seventh receiving unit is used to receive feature vectors of each mode and has preset vector-specific operators; The third calculation unit is used to construct a multi-source heterogeneous modal data feature fusion calculation formula. It inputs a multi-source heterogeneous data fusion matrix, each modal feature vector, and vector-specific operators to fuse the multi-source heterogeneous modal data features. The calculation formula is as follows: *[F_v,F_i,F_a,F_n]*T in To enhance the post-fused feature vector, The matrix variable elements are: F_v, which is the vibration feature vector; F_i, which is the infrared feature vector; F_a, which is the acoustic feature vector; F_n, which is the environmental feature vector; and T, which is the vector-specific operator.

[0039] Preferred options also include: The eighth receiving unit is used to receive the current health status of the equipment and the overall environmental stress. The second preset unit is used to preset the device's environmental tolerance threshold and digital metabolic coefficient. The third calculation unit is used to construct a digital metabolic rate calculation formula. It inputs the current health status of the device, comprehensive environmental stress, the device's environmental tolerance threshold, and the digital metabolic coefficient to calculate the digital metabolic rate. The formula is as follows: in For digital metabolic rate, Digital metabolic coefficient To take into account environmental stress, This refers to the equipment's environmental tolerance threshold. This indicates the current health status of the device.

[0040] Preferred options also include: The third preset unit is used to preset the prediction time step and attenuation coefficient; The fourth calculation unit is used to construct a health prediction model. It inputs the device's current health level, decay coefficient, prediction time step, and digital metabolic rate into the health prediction model to calculate the predicted health level of the device. The model is as follows: in Predict the health status of the equipment. The current health status of the device. The attenuation coefficient is... For digital metabolic rate, To predict the time step.

[0041] Preferred options also include: The fourth preset unit is used to preset the intelligent hierarchical early warning system. The system is divided into hierarchical levels, including safety level, attention level, risk level, and high level. The fifth preset unit is used to preset the self-healing strategy knowledge base; The third analysis and judgment unit is used to analyze and judge the predicted health of the equipment based on the intelligent hierarchical early warning system and obtain the early warning level; The second processing output unit is used to access the self-healing measurement library according to the warning level and obtain self-healing strategy instructions.

[0042] The steps of the methods or algorithms described in this application can be directly embedded in hardware, software units executed by a processor, or a combination of both. Exemplarily, a storage medium can be connected to a processor so that the processor can read information from and write information to the storage medium. Optionally, the storage medium can also be integrated into the processor. Optionally, the processor and the storage medium can also be located in different components within a terminal. These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0043] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely illustrative examples of this application and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Thus, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A method for fusing multi-source heterogeneous data in an equipment operation and maintenance safety pre-control system, characterized in that, The method includes: Receive multi-source heterogeneous modal data sets and perform time-series alignment and quality assessment on the multi-source heterogeneous modal data; Construct a multivariate heterogeneous data fusion matrix and calculate the matrix variable elements; Based on the matrix variable elements, construct a multi-source heterogeneous modal data feature fusion calculation formula to fuse the features of multi-source heterogeneous modal data; Construct a dynamic metabolic model between device health status and the external environment to obtain digital metabolic rate; Build a health prediction model, analyze digital metabolic rate, and obtain health prediction data; Based on the predicted health status of the equipment, analyze and judge the predicted health status of the equipment, obtain the warning level, and generate a self-healing strategy.

2. The method according to claim 1, characterized in that, The received multi-source heterogeneous modal data set includes: Vibration data from the receiving equipment is used as the first multi-source heterogeneous modal data; The surface temperature distribution data of the receiving device is used as the second multi-source heterogeneous mode data; The noise data from the receiving equipment is used as the third multi-source heterogeneous modal data. Receive ambient temperature and humidity as the fourth multi-source heterogeneous modal data; Dust concentration was received as the fifth multi-source heterogeneous mode data; The first, second, third, fourth, and fifth multi-source heterogeneous modal data are used as a set of multi-source heterogeneous modal data.

3. The method according to claim 1, characterized in that, The construction of the multivariate heterogeneous data fusion matrix and the calculation of matrix variable elements include: Construct a multi-source heterogeneous data fusion matrix, where the matrix variable elements represent the degree of influence of multi-source heterogeneous modal data anomalies on the anomalies of associated multi-source heterogeneous modal data; The mean value of multi-source heterogeneous modal data is obtained in real time; Preset thresholds for monitoring abnormal fluctuations in multi-source heterogeneous modal data, and learn parameters; When abnormal fluctuations are detected in multi-source heterogeneous modal data, output abnormal modal fluctuation data; when abnormal fluctuations are detected in correlated multi-source heterogeneous modal data, output correlated modal abnormal fluctuation data. A dynamically updated matrix variable element model is constructed, incorporating a multi-source heterogeneous modal data set, modal anomaly fluctuation data, correlated modal anomaly fluctuation data, and learning parameters. The model is as follows: Where Ω n Ω is the element of the new matrix variable. o ΔI represents the original matrix variable elements, μ represents the learning parameters, ΔI represents the associated modal anomaly fluctuation data, and ΔJ represents the modal anomaly fluctuation data.

4. The method according to claim 1, characterized in that, The step of constructing a multi-source heterogeneous modal data feature fusion calculation formula based on matrix variable elements, and fusing multi-source heterogeneous modal data features, includes: Based on the multivariate heterogeneous data fusion matrix where the elements of the new matrix variables are located; Receives feature vectors for each modality, with preset vector-specific operators; A calculation formula for feature fusion of multi-source heterogeneous modal data is constructed. Substituting the multivariate heterogeneous data fusion matrix, each modal feature vector, and vector-specific operators, the features of the multi-source heterogeneous modal data are fused. The calculation formula is as follows: F e =Ω*[F_v,F_i,F_a,F_n]*T Where F e To enhance the fused feature vector, the elements of the Ω matrix are: F_v, the vibration feature vector; F_i, the infrared feature vector; F_a, the acoustic feature vector; F_n, the environmental feature vector; and T, the vector-specific operator.

5. The method according to claim 1, characterized in that, The construction of a dynamic metabolic model between the health status of the device and the external environment to obtain a digital metabolic rate includes: Current health status of the receiving equipment, and overall environmental stress; Preset device environmental tolerance threshold, digital metabolic coefficient; A formula for calculating digital metabolic rate is constructed, and by inputting the current health status of the device, comprehensive environmental stress, the device's environmental tolerance threshold, and the digital metabolic coefficient, the digital metabolic rate is calculated. The formula is as follows: M=β*(E e -E t )*(1-H c ) Where M is the numerical metabolic rate, β is the numerical metabolic coefficient, and E is... e To take into account environmental stress, E t H is the environmental tolerance threshold of the equipment. c This indicates the current health status of the device.

6. The method according to claim 1, characterized in that, The construction of a health prediction model, analysis of digital metabolic rate, and acquisition of health prediction data include: Preset prediction time step and decay coefficient; A health prediction model is constructed by inputting the device's current health status, decay coefficient, prediction time step, and digital metabolic rate into the model to calculate the predicted health status of the device. The model is as follows: H f =H c -k*M*Δt Where H f H predicts device health. c The current health status of the device is given by k, the decay coefficient is given by M, the digital metabolic rate is given by M, and the predicted time step is given by Δt.

7. The method according to claim 1, characterized in that, The process of analyzing and judging the predicted health status of the device, obtaining an early warning level, and generating a self-healing strategy based on the device's predicted health status includes: The system is designed with an intelligent hierarchical early warning system, which is divided into four levels: safety level, attention level, risk level, and high level. Pre-defined self-healing strategy knowledge base; Based on the intelligent hierarchical early warning system, the predicted health status of the equipment is analyzed and judged to obtain the early warning level; Based on the warning level, access the self-healing measurement library to obtain self-healing strategy instructions.

8. A multi-source heterogeneous data fusion system for equipment operation and maintenance safety pre-control system, the system being used to implement the above method, the system comprising: The first receiving unit is used to receive a multi-source heterogeneous modal data set; The first calculation module is used to construct a multi-dimensional heterogeneous data fusion matrix and calculate the matrix variable elements. The second calculation module is used to construct a multi-source heterogeneous modal data feature fusion calculation formula based on the matrix variable elements, and to fuse the multi-source heterogeneous modal data features. The third calculation module is used to construct a dynamic metabolic model between the device's health status and the external environment to obtain the digital metabolic rate. The fourth calculation module is used to build a health prediction model, analyze digital metabolic rate, and obtain health prediction data. The first analysis and judgment unit analyzes and judges the predicted health status of the equipment based on the predicted health status, obtains the warning level, and generates a self-healing strategy.

Images