Cross-dimensional turbomachinery dynamic risk prediction and visualization system

Abstract

This invention relates to the field of turbomachinery monitoring technology and discloses a cross-dimensional turbomachinery dynamic risk prediction and visualization system. The system's dynamic data acquisition module acquires multi-source heterogeneous sensor data in real time, including vibration spectrum sequences, aerodynamic load distribution matrices, rotor axial displacement time series, and surface temperature field thermograms. The service degradation assessment module extracts vibration feature spectra based on a historical degradation pattern library, calculates instability risk levels by combining aerodynamic load deviations, and generates a service degradation trajectory matrix by fusing displacement abrupt change gradients and temperature difference anomalous areas. The multi-dimensional risk prediction module inputs the degradation trajectory matrix into a spatiotemporally coupled prediction network and outputs a dynamic risk assessment vector containing time-varying failure probability distributions, remaining life predictions, and a set of risk propagation paths. The risk visualization engine module constructs a three-dimensional risk thermodynamic field model and generates a dynamic thermodynamic mesh layer superimposed on a digital twin.

Description

Technical Field

[0001] This invention relates to the field of turbomachinery monitoring technology, specifically a cross-dimensional turbomachinery dynamic risk prediction and visualization system. Background Technology

[0002] Turbomachinery, as core equipment in energy, power, aerospace, and other fields, directly impacts the continuity and safety of industrial production. These machines typically operate under complex conditions of high temperature, high pressure, and high-speed rotation for extended periods, making them highly susceptible to malfunctions due to abnormal vibration, aerodynamic instability, rotor misalignment, or abnormal temperature, potentially leading to shutdowns and significant economic losses. Therefore, real-time monitoring and risk prediction of turbomachinery's operating status are crucial for ensuring the stable operation of industrial systems.

[0003] Current turbomachinery monitoring technologies often rely on single-type sensors for data acquisition, such as using only vibration sensors to obtain vibration signals or relying solely on temperature sensors to monitor equipment temperature. This makes it difficult to comprehensively capture the complex operating conditions of the equipment. The heterogeneous nature of multi-source data—including the temporal sequence of vibration spectra, the spatial distribution of aerodynamic loads, the dynamic changes in displacement data, and the image characteristics of temperature fields—poses challenges to the fusion and analysis of different data types. Traditional methods often process various types of data in isolation, failing to establish correlations between multiple parameters, leading to a one-sided assessment of equipment degradation.

[0004] In service degradation assessment, existing technologies mostly rely on threshold judgments for single parameters, such as determining equipment abnormality solely by whether the vibration amplitude exceeds a preset value, neglecting the coupling effects between parameters. For example, abnormal aerodynamic load distribution may be related to sudden changes in rotor axial displacement, while local anomalies in the temperature field may exacerbate the deterioration of vibration characteristics. Single-parameter assessments are insufficient to reflect the overall degradation trend of the equipment. Furthermore, the insufficient application of historical degradation patterns makes it impossible to combine long-term operating patterns for trend prediction, resulting in poor dynamic adaptability of the assessment results.

[0005] In the risk prediction phase, existing models mostly employ static analysis methods, failing to fully consider the time-varying and spatially correlated characteristics of equipment operating states. The identification of risk propagation paths lacks a systematic approach, making it difficult to accurately capture the diffusion patterns of faults across different components, resulting in insufficient timeliness and accuracy of prediction results. Furthermore, risk visualization methods are relatively simple, often presented in the form of numerical tables or two-dimensional curves, making it difficult for operators to intuitively understand the spatial distribution and dynamic changes of risks, thus affecting the timeliness and effectiveness of risk management.

[0006] With the advancement of industrial digitalization, the application of digital twin technology in equipment monitoring is gradually increasing. However, existing systems have not yet achieved deep integration of multi-source heterogeneous data with digital twins. There are still technological gaps in the dynamic overlay and intuitive display of risk information on the twin, which cannot provide operators with an immersive risk perception experience. Summary of the Invention

[0007] The purpose of this invention is to provide a cross-dimensional turbomachinery dynamic risk prediction and visualization system to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, this invention provides a cross-dimensional turbomachinery dynamic risk prediction and visualization system, the system comprising:

[0009] The dynamic data acquisition module is used to acquire multi-source heterogeneous sensor data of turbomachinery in real time. The multi-source heterogeneous sensor data includes vibration spectrum sequence, aerodynamic load distribution matrix, rotor axial displacement time sequence and surface temperature field thermogram.

[0010] The service degradation assessment module is used to extract the current vibration feature spectrum based on the historical degradation mode library of the vibration spectrum sequence, calculate the aerodynamic instability risk level according to the deviation between the aerodynamic load distribution matrix and the preset operating condition threshold, and generate the service degradation trajectory matrix by fusing the abrupt gradient of the rotor axial displacement time sequence and the area of ​​the abnormal temperature difference region of the surface temperature field thermogram.

[0011] The multidimensional risk prediction module is used to input the service degradation trajectory matrix into the spatiotemporal coupled prediction network and output a dynamic risk assessment vector containing the time-varying failure probability distribution, the predicted value of the remaining life of key components, and the set of risk propagation paths.

[0012] The risk visualization engine module is used to construct a three-dimensional spatial risk thermal field model based on the dynamic risk assessment vector, and generate a dynamic thermal mesh layer superimposed on the digital twin of the turbomachinery.

[0013] Preferably, the dynamic data acquisition module includes:

[0014] The vibration feature analysis unit is used to perform wavelet packet energy spectrum decomposition on the vibration spectrum sequence, extract the energy entropy value within a preset frequency band interval, and mark the abnormal frequency band positions where the energy entropy value exceeds the entropy value threshold.

[0015] The load distribution verification unit is used to compare the aerodynamic load distribution matrix with the reference load distribution template under standard working conditions, and to calculate the area ratio and maximum over-limit amplitude of the local load over-limit region.

[0016] The displacement co-analysis unit is used to correlate the rotor axial displacement timing with the speed pulse signal and identify the phase synchronization error between displacement abrupt points and speed step points.

[0017] The temperature field fusion unit is used to map the surface temperature field thermogram to the three-dimensional mesh model of the turbomachinery and locate the spatial coordinate set of the temperature difference abnormal area.

[0018] Preferably, the service degradation assessment module includes:

[0019] The degradation pattern matching unit is used to retrieve the degradation pattern that matches the abnormal frequency band location from the historical degradation pattern library and output the first risk level;

[0020] The instability risk quantification unit is used to calculate the aerodynamic instability risk level based on the over-limit amplitude gradient when the area ratio of the local load over-limit region exceeds the area threshold and the maximum over-limit amplitude exceeds the amplitude threshold.

[0021] The displacement risk correlation unit is used to generate an axial displacement risk coefficient based on the amplitude increment of the phase synchronization error and the displacement abrupt change point.

[0022] The fusion calculation unit is used to weight and fuse the first risk level, the aerodynamic instability risk level, and the axial displacement risk coefficient to generate the service degradation trajectory matrix.

[0023] Preferably, the multidimensional risk prediction module includes:

[0024] The spatiotemporal feature extraction unit is used to perform spatiotemporal convolution operations on the service degradation trajectory matrix to extract cross-scale risk feature vectors.

[0025] The fault probability prediction unit is used to process the cross-scale risk feature vector through a gated recurrent network and output the time-varying fault probability distribution within a future time window.

[0026] The life decay calculation unit is used to calculate the predicted value of the remaining life of the key component based on the mapping relationship between the cross-scale risk feature vector and the fatigue curve of the component material.

[0027] The path inference unit is used to simulate the diffusion process of risk in the topology of turbomachinery based on graph neural networks and generate the set of risk propagation paths.

[0028] Preferably, the risk visualization engine module includes:

[0029] A spatial mapping unit is used to map the time-varying fault probability distribution to the corresponding mesh unit of the turbomachinery three-dimensional mesh model;

[0030] The thermal field generation unit is used to calculate the color gradient value of the grid cell based on the predicted remaining life of the key component, and combine it with the spatial coordinates of the risk propagation path set to generate the three-dimensional spatial risk thermal field model.

[0031] The dynamic rendering unit is used to synchronously update the three-dimensional spatial risk thermal field model with the turbomachinery operating status and output the dynamic thermal mesh layer.

[0032] Preferably, the system further includes:

[0033] The anomaly tracing module is used to extract all paths in the risk propagation path set that pass through the grid cell when the probability value of a specific grid cell in the time-varying fault probability distribution exceeds the probability threshold, and trace them back to the source grid cell.

[0034] The anomaly tracing module includes a path filtering unit, which calculates the propagation strength weight of each path and filters paths whose weights exceed a weight threshold to form the main propagation chain.

[0035] Preferably, the system further includes: an adaptive monitoring module, used to adjust the acquisition frequency of corresponding sensor data according to the spatial location of the source grid unit;

[0036] The adaptive monitoring module includes a frequency decision unit, which is used to calculate the incremental coefficient of the acquisition frequency based on the path length and propagation intensity weight of the main propagation chain.

[0037] Preferably, the system further includes: a thermal grid optimization module, used to dynamically adjust the grid density of the dynamic thermal grid layer;

[0038] The thermal grid optimization module includes:

[0039] Risk gradient analysis unit, used to calculate the risk probability gradient value between adjacent grid cells;

[0040] A density controller unit is used to insert a subdivided grid layer in the corresponding region when the risk probability gradient value exceeds a gradient threshold.

[0041] Transition layer generation unit, used to configure a gradient transition mesh between the subdivision mesh layer and the original mesh layer.

[0042] Preferably, the system further includes: a multi-dimensional data cleaning module, used to calculate the redundancy of historical data storage based on the update frequency of the dynamic risk assessment vector;

[0043] The multidimensional data cleaning module includes:

[0044] A redundancy assessment unit is used to count the number of historical versions of the service degradation trajectory matrix.

[0045] The cleanup decision unit is used to delete the earliest version data in reverse chronological order of storage time when the number of historical versions exceeds the version threshold.

[0046] Preferably, the system further includes: a cloud collaboration module, used for encrypted uploading of the dynamic risk assessment vector to a cloud knowledge base;

[0047] The cloud-based collaboration module includes:

[0048] An encrypted retrieval unit is used to initiate similar case retrieval requests to the cloud and obtain a set of matching historical risk assessment cases.

[0049] The local correction unit is used to compare the deviation between the historical risk assessment case set and the current dynamic risk assessment vector, and correct the weight parameters of the spatiotemporal coupled prediction network.

[0050] Compared with the prior art, the beneficial effects of the present invention are:

[0051] This multi-dimensional turbomachinery dynamic risk prediction and visualization system achieves comprehensive perception, accurate prediction, and intuitive presentation of operational risks through the collaborative operation of multiple modules. The dynamic data acquisition module overcomes the limitations of traditional single-data acquisition, incorporating multi-source heterogeneous data such as vibration spectrum sequences, aerodynamic load distribution matrices, rotor axial displacement time series, and surface temperature field thermograms into its monitoring scope. This comprehensively captures the equipment's operational characteristics across different dimensions, providing rich foundational data support for subsequent risk assessment. This collaborative acquisition of multi-source data more completely reflects the equipment's true operating status, avoiding misjudgments of risk due to incomplete data.

[0052] The service degradation assessment module constructs a more comprehensive degradation assessment system by integrating historical degradation patterns with real-time operating parameters. Based on a historical degradation pattern library of vibration spectrum sequences, it extracts the current vibration characteristic spectrum, enabling the identification of potential anomalies by combining the long-term operating patterns of the equipment. It calculates the aerodynamic instability risk level by calculating the deviation between the aerodynamic load distribution matrix and preset operating condition thresholds, achieving dynamic monitoring of aerodynamic performance. Furthermore, it generates a service degradation trajectory matrix by integrating the abrupt gradient of the rotor axial displacement time series and the area of ​​abnormal temperature difference regions from the surface temperature field thermogram. This comprehensively considers the coupled influence of mechanical displacement and thermal characteristics, making the degradation assessment results more closely match the actual degradation process of the equipment and more accurately depicting the evolution trajectory of the equipment from normal operation to potential failure.

[0053] The multidimensional risk prediction module introduces a spatiotemporally coupled prediction network, fully considering the dynamic changes in equipment operating status over time and the spatial correlation characteristics. After inputting the service degradation trajectory matrix into this network, the output dynamic risk assessment vector includes a time-varying failure probability distribution, predicted remaining lifespan of key components, and a set of risk propagation paths. This not only updates risk probabilities in real time but also clearly defines the diffusion paths of risks among various components of the equipment. This allows risk prediction to move beyond static assessments at a single point in time, forming a dynamic and evolving risk panorama. This spatiotemporally coupled prediction method can more accurately grasp risk development trends and provide more targeted information for risk early warning.

[0054] The risk visualization engine module constructs a three-dimensional spatial risk thermal field model and overlays it onto the digital twin of the turbomachinery to generate a dynamic thermal mesh layer, thus achieving a visual presentation of risk information. Operators can intuitively observe the spatial distribution and dynamic changes of risks in different parts of the equipment through the digital twin, quickly understanding the risk status without relying on complex numerical analysis. This visualization method transforms abstract risk data into a concrete spatial thermal distribution, reducing the difficulty of interpreting risk information and helping operators to grasp equipment operational risks more promptly and accurately, thereby improving the efficiency of risk management.

[0055] The collaborative work of each module forms a complete closed loop from data acquisition, degradation assessment, risk prediction to visualization, realizing dynamic management of the entire process of turbomachinery risk, adapting to the actual needs of equipment monitoring under complex working conditions, and providing comprehensive technical support for the safe and stable operation of turbomachinery. Attached Figure Description

[0056] Figure 1 This is a timeline diagram of the cross-dimensional turbomachinery dynamic risk prediction and visualization system described in this invention;

[0057] Figure 2 Workflow diagram for the service degradation assessment module;

[0058] Figure 3 A flowchart of the multidimensional risk prediction module;

[0059] Figure 4 This is a flowchart of the exception tracing module. Detailed Implementation

[0060] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0061] Please see Figure 1 This invention provides a cross-dimensional turbomachinery dynamic risk prediction and visualization system, the system comprising:

[0062] The dynamic data acquisition module continuously acquires real-time data streams from various sensors installed on the turbomachinery. This data includes vibration spectrum sequences characterizing the dynamic properties of the machinery, aerodynamic load distribution matrices reflecting the aerodynamic load distribution, rotor axial displacement time series indicating the rotor's axial position, and surface temperature field thermograms depicting the surface temperature distribution of the machinery. This data constitutes a multi-source heterogeneous sensor dataset.

[0063] The service degradation assessment module receives the aforementioned multi-source heterogeneous sensor data and performs a comprehensive evaluation. This module utilizes a stored historical degradation pattern library to extract features and perform pattern matching on the currently acquired vibration spectrum sequence. Simultaneously, it calculates the deviation between the aerodynamic load distribution matrix and a preset standard operating condition threshold to assess aerodynamic stability. Furthermore, it analyzes the abrupt changes in the rotor axial displacement time series and their synchronization with speed changes, combining this with the area information of abnormal temperature difference regions in the surface temperature field thermogram. By integrating the above multi-faceted assessment results, a service degradation trajectory matrix comprehensively characterizing the current health status and development trend of the equipment is generated.

[0064] The multidimensional risk prediction module receives the service degradation trajectory matrix as its core input. This module uses a spatiotemporally coupled prediction network to process the input matrix. The network analyzes the spatiotemporal evolution patterns contained in the trajectory matrix and outputs a dynamic risk assessment vector. This vector contains three key predictive pieces of information: the probability distribution of failures in different components or regions within a specific future time period; the expected remaining operating time of key components before failure; and the possible propagation direction and path of potential faults within the equipment's internal structure.

[0065] The risk visualization engine module receives dynamic risk assessment vectors and transforms them into intuitive visualizations. Based on a 3D digital model of the turbomachinery, this module spatially maps the predicted risk information. Specifically, it constructs a 3D spatial risk thermal field model, which uses color gradients and transparency variations to represent the risk probability, remaining lifespan, and risk propagation direction at different locations on a 3D grid. Finally, this dynamically changing risk thermal field model is overlaid on the turbomachinery's digital twin as a dynamic thermal grid layer, achieving a real-time, three-dimensional, and intuitive display of the risk.

[0066] Example 1: See Figure 2The dynamic data acquisition module, serving as the system's data entry point, is primarily responsible for real-time capture and preliminary processing of multi-source heterogeneous sensor data from the turbomachinery. This module comprises four functional units, each handling a specific type of data stream. The vibration feature analysis unit is responsible for processing continuous vibration spectrum sequence data. This unit first performs wavelet packet transform decomposition on the input vibration spectrum sequence. This decomposition process breaks down the original spectral signal into a series of preset, mutually orthogonal frequency bands. After decomposition, the unit calculates the energy entropy value within each specified frequency band. Energy entropy is a quantitative indicator used to measure the complexity or disorder of signal energy distribution within a specific frequency band. After calculation, the unit compares the energy entropy value of each frequency band with a preset entropy threshold. When the energy entropy value of a frequency band consistently or significantly exceeds its corresponding entropy threshold, the frequency band is marked as an abnormal frequency band, and its specific location information is recorded. These marked abnormal frequency band locations indicate possible abnormal frequency components in the vibration signal.

[0067] The load distribution verification unit focuses on processing aerodynamic load distribution matrix data. This unit internally stores or accesses a pre-established benchmark load distribution template under standard operating conditions. This template represents the ideal or typical spatial distribution of aerodynamic loads under normal, stable operating conditions. The real-time acquired aerodynamic load distribution matrix is ​​compared point-by-point with this benchmark template. During the comparison, the difference between the real-time load value at each spatial location and the corresponding threshold value in the benchmark template is calculated. All spatial points where the real-time load value exceeds their corresponding threshold value are identified; the regions formed by these points are the local load over-limit areas. The unit further calculates the proportion of the area of ​​these over-limit areas to the total area of ​​the entire analysis region. Simultaneously, among all over-limit points, the maximum amplitude of the load value exceeding its threshold is identified. The area proportion and the maximum over-limit amplitude are key indicators for quantifying the degree to which the current aerodynamic load distribution deviates from the normal state.

[0068] The displacement coordination analysis unit processes two related signals: the rotor axial displacement timing signal and the speed pulse signal. The unit first analyzes the rotor axial displacement timing curve, detecting points where the displacement value changes abruptly and significantly, i.e., displacement abrupt change points. These abrupt change points typically correspond to abnormal changes in the rotor's operating state. Simultaneously, the unit analyzes the speed pulse signal, identifying the moments when the speed changes abruptly, i.e., speed step points. Speed ​​step points typically reflect changes in equipment control commands or load conditions. To assess the coordination between changes in rotor axial displacement and speed, the unit calculates the time difference between each detected displacement abrupt change point and its nearest speed step point. This time difference is the phase synchronization error. A large phase synchronization error indicates that the displacement change has not kept pace with the speed change, potentially suggesting mechanical lag or underlying problems. Furthermore, the unit also records the amplitude increment of the displacement abrupt change points, i.e., the magnitude of the displacement abrupt change.

[0069] The temperature field fusion unit processes surface temperature field thermographic data. Its core task is to correlate two-dimensional temperature distribution image information with the three-dimensional geometry of the turbomachinery. First, it maps the acquired surface temperature field thermographic data onto the surface of the turbomachinery's three-dimensional mesh model. This mapping process ensures a correspondence between temperature data points and surface mesh cells in the three-dimensional mesh model. After mapping, the unit judges the temperature value of each mesh cell based on the normal temperature range threshold set for the current operating conditions of the equipment. Mesh cells whose temperature values ​​significantly deviate from their normal temperature range are identified as temperature anomaly points. The continuous area formed by the aggregation of these anomaly points is the temperature anomaly region. Finally, the unit outputs the set of spatial coordinates of all mesh cells covered by these temperature anomaly regions, accurately identifying the spatial location of the abnormally high temperature on the equipment surface.

[0070] The service degradation assessment module receives the processing results from the four data acquisition units, performs a comprehensive status assessment, and generates a service degradation trajectory matrix. The degradation pattern matching unit utilizes the abnormal frequency band location information output by the vibration feature analysis unit. This unit accesses a historical degradation pattern library, a knowledge base storing characteristic abnormal frequency band patterns and their corresponding risk levels that appeared in the vibration spectrum when the equipment experienced various typical failure modes throughout history. The unit searches the historical library for the historical degradation pattern that best matches or is most similar to the currently detected abnormal frequency band location. The matching process may involve pattern similarity calculation or rule matching. Upon successful matching, the unit outputs the first risk level associated with that historical pattern, which characterizes the degree of failure risk implied by the current vibration anomaly.

[0071] The instability risk quantification unit processes the area proportion and maximum over-limit amplitude data provided by the load distribution verification unit. This unit makes judgments based on preset logic: only when the area proportion of a local load over-limit region exceeds a set area threshold, and simultaneously the maximum over-limit amplitude also exceeds a set amplitude threshold, is the current aerodynamic state considered to have a significant instability risk. If both conditions are met, the unit outputs a specific, quantified aerodynamic instability risk level based on the distribution characteristics of the over-limit region and the gradient of the over-limit amplitude, using preset quantification rules or a calculation model. This level reflects the likelihood of system instability caused by abnormal aerodynamic load distribution.

[0072] The displacement risk correlation unit processes the phase synchronization error and displacement abrupt change amplitude increment data provided by the displacement collaborative analysis unit. Based on the magnitude of the phase synchronization error and the amplitude increment of the displacement abrupt change, this unit calculates and generates an axial displacement risk coefficient using preset correlation rules or functions. Generally, a larger phase synchronization error indicates a more severe asynchrony between displacement and rotational speed changes, resulting in a higher risk; a larger amplitude increment of the displacement abrupt change indicates a more severe degree of displacement anomaly, also resulting in a higher risk. This coefficient comprehensively reflects the risk level inherent in rotor axial displacement anomalies.

[0073] The fusion calculation unit is the core output unit of the service degradation assessment module. It receives the first risk level from the degradation pattern matching unit, the aerodynamic instability risk level from the instability risk quantification unit, and the axial displacement risk coefficient from the displacement risk correlation unit. Additionally, spatial coordinate set information of the temperature difference abnormal region may also be used as input or to assist in weight allocation. This unit performs a weighted fusion calculation on these three main risk indicators according to a preset weighting coefficient scheme. The weighting coefficients can be determined based on expert experience, historical data analysis, or equipment criticality analysis; for example, under specific operating conditions, more attention may be paid to vibration risk or aerodynamic risk. The result of the fusion calculation is a multi-dimensional matrix, namely the service degradation trajectory matrix. The structure of this matrix is ​​designed to comprehensively characterize the degradation state and trend of the equipment. Rows in the matrix may represent different assessment dimensions or indicators, and columns may represent different time points or different equipment sub-regions / components. Each element value in the matrix integrates the degradation degree information of the corresponding dimension, time point, or region. This service degradation trajectory matrix provides core, multi-source integrated input data for the subsequent risk prediction module.

[0074] Example 2: See Figure 3The core function of the multidimensional risk prediction module lies in using deep learning technology to perform in-depth analysis of the service degradation trajectory matrix and predict the future risk status of the equipment. This module relies on an architecture called a spatiotemporally coupled prediction network, which consists of multiple functional units working together. The spatiotemporal feature extraction unit is the starting point for processing input data. This unit receives the service degradation trajectory matrix from the service degradation assessment module as input. This matrix is ​​typically a multidimensional data structure, with dimensions including spatial information, temporal information, and multiple evaluation metrics. The core operation of the spatiotemporal feature extraction unit is to perform spatiotemporal convolution. This involves using a specially designed convolution kernel to perform sliding computation along the spatiotemporal dimension of the matrix. Convolution operations can simultaneously capture the correlation of states in spatially adjacent regions and evolution patterns over time. For example, a three-dimensional convolution kernel can slide simultaneously along two spatial dimensions and the temporal dimension to extract feature patterns within the local spatiotemporal neighborhood. Another approach is to use separate convolutions: first, a two-dimensional spatial convolution kernel is used to extract spatial correlation features between different regions at the same point in time in the spatial dimension; then, a one-dimensional temporal convolution kernel is used to extract temporal features of the state of the same spatial location changing over time in the temporal dimension. The purpose of these convolutional operations is to extract abstract, cross-scale risk feature patterns that reflect the essence of risk evolution from complex service degradation trajectories. After multi-layer convolution processing, the unit outputs a highly condensed cross-scale risk feature vector, which contains key information characterizing the risk evolution trend of the equipment's current and historical states.

[0075] The failure probability prediction unit receives the cross-scale risk feature vector output by the spatiotemporal feature extraction unit. Since the risk feature vector typically contains time-series information, this unit employs a gated recurrent network structure for processing. Gated recurrent networks, such as Long Short-Term Memory (LSTM) networks or Gated Recurrent Units (GRUs), are chosen because they can effectively learn and remember long-term dependencies in time series. The network structure of this unit learns the dynamic patterns of the risk feature vector changing over time. Based on the learned patterns, the network predicts the probability of failure in various components or spatial regions of the equipment within a set future time window. The output is a time-varying failure probability distribution. This distribution is a complex data structure, typically containing spatial and temporal dimensions. Spatially, it identifies each grid cell or component on the equipment model; temporally, it provides an estimate of the probability of failure for each spatial cell at a series of discrete future time points. For example, it can be represented as a three-dimensional array: dimension one represents the spatial location, dimension two represents future time points, and dimension three stores the predicted failure probability value for that location at that time point. This distribution dynamically depicts the evolution of risk probability across the spatial structure of the equipment over time.

[0076] The life decay calculation unit also receives a cross-scale risk feature vector as input. This unit focuses on predicting the remaining service life of critical components before failure. Its working principle is based on the correlation between physical mechanisms such as material fatigue, wear, or aging and the operating state of the equipment. The unit internally stores or accesses material fatigue characteristic curves or empirical degradation models of critical components. The cross-scale risk feature vector contains information reflecting the current stress state, cumulative damage level, or degradation rate of the component. The life decay calculation unit establishes a mapping relationship between these risk features and material fatigue curves or degradation models. This mapping can be achieved in several ways: directly predicting the remaining life using a trained regression model based on the input features; or calculating the cumulative damage level using information in the feature vector based on a physical damage accumulation model, and then estimating the remaining life. The unit outputs a predicted remaining life value for the critical component, which is typically a time quantity representing the length of time the component is expected to operate safely under the predicted degradation path.

[0077] The path inference unit is responsible for simulating and predicting the propagation paths of potential faults within the internal structure of the equipment. This unit builds a model based on a graph neural network. First, a topology graph of the turbomachinery needs to be constructed. This graph abstracts the physical structure of the equipment: nodes in the graph represent key components, functional units, or divided spatial regions of the equipment; edges in the graph represent potential fault propagation paths or influence relationships between nodes. Edges are established based on physical connections, functional dependencies, or known fault propagation mechanisms. Each edge can be assigned a weight or attribute to indicate the ease or strength of propagation. The path inference unit inputs cross-scale risk feature vectors as the initial states or features of the nodes into the graph neural network. The core operation of the graph neural network is message passing: each node updates its own state based on its current state, the states of its neighboring nodes, and the attributes of the connecting edges. Through multiple rounds of message passing iterations, the network simulates how risk features diffuse and propagate along the edges of the graph structure from the source node to other nodes. After simulation and inference, the unit outputs a set of risk propagation paths. This set contains multiple predicted path sequences that start from potential risk source nodes, are connected by edges in the graph, and propagate to other nodes. Each path records an ordered sequence of nodes, representing a chain of possible directions for risk propagation. This set reveals how a failure can evolve from a local problem into a global one.

[0078] The task of the risk visualization engine module is to transform the abstract prediction results output by the multi-dimensional risk prediction module into intuitive and understandable 3D visualizations, which are then overlaid on the digital twin model of the turbomachinery. The spatial mapping unit handles the time-varying failure probability distribution. The core of this unit is to map the predicted failure probability values ​​to their corresponding spatial locations within the turbomachinery's 3D mesh model. Specifically, for each mesh cell in the 3D mesh model, based on its position within the equipment structure, the predicted probability value for that location at a specific future time point is found in the time-varying failure probability distribution. Then, based on the magnitude of the probability value, a visual attribute value is assigned to each mesh cell. The most commonly used visual channels are color and transparency. For example, using a color mapping scheme from cool to warm colors, the higher the probability value, the more reddish the color and the less transparent it may be. In this way, the spatial distribution of probability values ​​is intuitively presented through the color and transparency of the mesh cells.

[0079] The thermal field generation unit is responsible for integrating various risk information to construct the final 3D thermal field model. This unit processes the predicted remaining lifespan of critical components and the set of risk propagation paths. First, for critical components, the corresponding color gradient value is calculated based on their predicted remaining lifespan. A color mapping scheme different from the failure probability is typically used here to avoid confusion. For example, a long remaining lifespan might be represented by green, a medium lifespan by yellow, and a short lifespan by orange or red. The color gradient value directly reflects the length of the remaining lifespan. Next, the risk propagation path set is processed. This unit extracts the spatial coordinates corresponding to the sequence of nodes traversed by each path in the set. In the 3D scene, visual path indicators are generated along these spatial coordinate sequence points. Common indicators include: connecting nodes on the path with lines of different colors; adding special marks or highlighting effects to the mesh cells traversed by the path; or adding icons at the starting point and critical nodes of the path. Finally, the thermal field generation unit fuses and overlays the failure probability distribution processed by the spatial mapping unit, the color gradient values ​​of the remaining lifespan of critical components, and the spatial indicators of the risk propagation paths onto the same 3D mesh model. Visual conflicts need to be addressed during the fusion process; for example, path markers need to be sufficiently conspicuous to distinguish them from probability heatmaps. The final output is a comprehensive three-dimensional spatial risk heatfield model containing rich risk information. This model simultaneously displays in three-dimensional space: the probability of failure at different locations, the lifespan status of critical components, and the potential directions and paths of risk diffusion.

[0080] The dynamic rendering unit is the output stage of the visualization engine, ensuring that the visualization results are synchronized with the real-time or predicted state of the equipment. This unit receives continuous updates from the multi-dimensional risk prediction module or directly receives sensor data streams to drive state updates of the digital twin. When a new dynamic risk assessment vector arrives, the dynamic rendering unit triggers an update to the entire visualization process: the spatial mapping unit updates the mesh color / transparency based on the new time-varying failure probability distribution; the thermal field generation unit updates the colors and path indicators of key components based on the new remaining life prediction values ​​and risk propagation path set; and the thermal field generation unit reconstructs or updates the 3D spatial risk thermal field model. Simultaneously, this unit is responsible for synchronizing the updated 3D spatial risk thermal field model with the current state of the turbomachinery digital twin. The rendering engine calculates the effects of lighting, perspective, and model transformations in the scene in real time. Finally, the dynamic rendering unit outputs the rendered dynamic thermal mesh layer to the graphics display device. This layer is dynamically changing; as the prediction results are updated or the equipment state changes, the risk heatmap, life indicator color blocks, and risk propagation path lines on the layer are updated in real time, providing users with an immediate and intuitive understanding of the equipment's current and future risk status.

[0081] Example 3: See Figure 4 When the failure probability value corresponding to a specific grid cell in the time-varying failure probability distribution generated by the multidimensional risk prediction module exceeds a preset critical threshold, the system triggers an anomaly tracing mechanism. This mechanism analyzes the set of risk propagation paths to trace the potential origin of the high-risk location in reverse. The anomaly tracing module first filters out all propagation paths containing the high-risk grid cell from the risk propagation path set. These paths record the possible trajectory of risk spreading from the source to the current high-risk location. Each path consists of a series of ordered grid cells, representing the sequence of intermediate nodes in the risk propagation.

[0082] The path selection unit assesses the importance of the extracted propagation paths. This unit calculates a propagation intensity weight for each path, which comprehensively reflects the path's activity level and risk transmission capability. The propagation intensity weight is calculated using the following formula:

[0083] ;

[0084] Where: Wp represents the propagation intensity weight of path P; u represents the grid cell on path P; α u β is the probability value of grid cell u in the current time-varying fault probability distribution; u λ is the activity coefficient of grid cell u in the historical propagation record; P Indicates the number of hops (path length) of path P; This is a minimal constant introduced to avoid the denominator being zero. Activity coefficient β uThe coefficient is obtained by statistically analyzing the frequency with which the grid cell acts as a transit node in historical risk propagation; the higher the frequency, the larger the coefficient.

[0085] The path filtering unit filters out secondary paths with weights below a set threshold based on the calculated propagation strength weight Wp. The retained high-weight paths form the main propagation chain set, which is considered the key propagation channels most likely to lead to the current high-risk state. The anomaly tracing module traces the paths in the main propagation chain set in reverse. Starting from the identified high-risk grid cell, it checks the state of each predecessor node in reverse order of the path node sequence. The tracing process records the risk probability gradient changes of each node on the path, looking for inflection points where the probability value suddenly increases.

[0086] After the adaptive monitoring module locates the source grid cell, it analyzes the correspondence between its spatial position and the sensor deployment. The frequency decision unit calculates the adjustment amount of the acquisition frequency, mainly considering two characteristic parameters of the main propagation chain: path length L (the number of hops from the source to the high-risk location) and average propagation intensity weight. Shorter path lengths and higher average weights mean faster risk propagation and more direct impact, requiring more intensive monitoring. The incremental coefficient of the collection frequency has a non-linear relationship with these two parameters, prioritizing monitoring density in high-risk source areas.

[0087] In the dynamic data acquisition module, sensor nodes associated with the source grid cells receive frequency adjustment commands. The sampling rate of the vibration sensors is increased proportionally according to the increment coefficient to ensure the capture of transient vibration characteristics. The reading intervals of the temperature monitoring points are shortened accordingly to track local temperature rise trends. The data output frequency of the pneumatic load sensors is increased synchronously to monitor details of pressure fluctuations. These adjustments enable the system to acquire a more complete picture of the state evolution process in the source region.

[0088] The anomaly tracing module and the visualization engine module work together, feeding back the tracing results in real time to the 3D spatial risk thermal field model, highlighting the source grid cells with special markers. Paths on the main propagation chain are rendered with bright colors to enhance visual recognition. Dynamic rendering cells add a pulsating effect to the source area to attract the operator's attention. These visual elements, together with other status indicators of the digital twin, constitute a complete risk tracing display.

[0089] When the risk probability in the source region continues to rise, the system will initiate a secondary frequency adjustment. The new incremental coefficient is calculated based on the probability change rate ΔP:

[0090] ;

[0091] Where: η is the adjusted increment coefficient; η0 is the initial increment coefficient; γ is the sensitivity adjustment parameter; ΔP represents the change in the probability value of the source grid cell per unit time. This dynamic adjustment ensures that the monitoring intensity matches the rate of risk evolution.

[0092] Each successful tracing instance updates the correlation strength parameters between grid cells, optimizing the accuracy of subsequent path deduction. Through continuously accumulating tracing experience, the system gradually builds a knowledge base of device-specific risk propagation patterns. This knowledge is used to calibrate the edge weight parameters in the graph neural network, improving the targeting of risk prediction.

[0093] Users can select any high-risk grid cell through the interface to trigger immediate source tracing analysis. While rendering the main propagation chain, the system displays detailed status parameters of each node along the path. The timeline control allows users to trace the historical process of risk propagation and observe the diffusion dynamics within a specific time period. These interactive features enhance operators' understanding of complex risk evolution. When multiple high-risk sources need to be monitored simultaneously, the module allocates monitoring resources based on the severity of each area. Source areas of core components receive priority at higher sampling rates, while secondary areas have appropriately lower frequency requirements. This resource allocation algorithm balances monitoring accuracy with system load, maintaining stable operating performance. Identified source area features are added to the degradation pattern library, enriching the correspondence between vibration spectra and fault types. Newly discovered risk propagation paths are added to the topology graph, expanding the training samples for the graph neural network. This closed-loop learning mechanism enables the system to continuously improve its risk identification capabilities.

[0094] Example 4: The thermal grid optimization module continuously analyzes the spatial variation characteristics of risk distribution in the dynamic thermal grid layer. The risk gradient analysis unit is responsible for calculating the risk probability differences between adjacent grid cells. This unit traverses the entire grid model, and for each grid cell, calculates the absolute value of the difference between its risk probability value and that of all its directly adjacent cells. The maximum value among these differences is taken as the risk probability gradient value between that cell and its neighbors. This gradient value quantifies the degree of spatial variation of risk distribution in a local area.

[0095] Table 1: Examples of risk probability values ​​and calculated gradient values ​​for some mesh elements in the rotor region of a turbomachinery at a certain moment.

[0096]

[0097] This unit sets a risk probability gradient threshold. When the maximum risk probability gradient value between a grid cell and its neighboring cells exceeds this threshold (e.g., the gradient value of 0.62 between cells G-1027 and G-1101 in the table), it indicates that there is a significant risk boundary or high-risk concentration area in that region, and the current grid resolution may not be sufficient to accurately depict the details of the risk distribution. In this case, the density controller unit inserts a subdivision grid layer at the spatial location corresponding to the high-risk gradient region. The subdivision grid layer uses a smaller grid cell size; for example, dividing the original-sized grid cells into multiple sub-cells. The number of subdivision levels inserted can be dynamically determined based on how much the gradient value exceeds the threshold; the larger the gradient value, the more subdivision levels may be inserted. For example, for regions with gradient values ​​much higher than the threshold, a three-level subdivision grid may be inserted to form a very fine local grid structure.

[0098] Inserting a fine mesh directly next to a coarse mesh can result in noticeable jagged edges or discontinuities during rendering. This unit automatically generates one or more transition mesh layers with intermediate-sized mesh cells in the boundary region between the subdivision mesh layer and the original mesh layer. The cell size of these transition mesh layers smoothly transitions from the original mesh size to the subdivision mesh size. For example, a transition mesh layer with a cell size of approximately S / 1.5 is inserted between the original mesh (size S) and the first-level subdivision mesh (size S / 2); if a second-level subdivision (size S / 4) exists, another transition mesh layer with a size of approximately S / 3 is inserted between the second-level and first-level subdivisions. The existence of the transition mesh layer ensures that the entire mesh model is geometrically continuous and visually smooth and gradual, eliminating visual artifacts caused by abrupt changes in mesh density and improving the visualization quality and realism of the risk thermal field model.

[0099] The multidimensional data cleaning module operates independently, responsible for managing historical data versions stored in the system and optimizing storage resource utilization. During continuous operation, this system periodically generates new service degradation trajectory matrices and dynamic risk assessment vectors, forming a historical data sequence. The redundancy assessment unit monitors the storage status of this core data. The core task of this unit is to count the number of historical versions of the currently stored service degradation trajectory matrix. For example, the system may generate one trajectory matrix version per hour, and the redundancy assessment unit records the total number of versions saved over a past period (e.g., 24 hours or one week).

[0100] This unit sets a maximum threshold for the number of historical versions. When the redundancy assessment unit reports that the number of historical versions of the stored service degradation trajectory matrix exceeds this threshold, the cleanup decision unit triggers a data deletion process. The deletion strategy strictly follows the principle of reverse chronological order. The system first identifies all stored historical versions and sorts them according to their generation or storage timestamps, starting deletion from the earliest stored version. The deletion operation continues until the number of remaining stored historical versions is equal to or lower than the set maximum threshold. For example, assuming the threshold is set to retain the most recent 50 versions, when the system detects that 60 versions are stored, the cleanup decision unit will delete the 10 earliest generated versions. This mechanism ensures that the system always retains the latest and most relevant historical data versions, providing necessary support for possible short-term retrospective analysis or model parameter fine-tuning, while effectively controlling the unlimited growth of storage space. Similar data cleanup strategies are also applicable to other accumulated historical data objects such as dynamic risk assessment vectors.

[0101] Example 5: The cloud collaboration module establishes a secure data channel between the local system and the remote cloud knowledge base, enabling predictive optimization based on collective experience. Upon startup, the encrypted retrieval unit first performs an encryption transformation on the locally generated latest dynamic risk assessment vector. The encryption process employs a preset cryptographic protocol to convert the numerical features in the vector into ciphertext that cannot be directly interpreted. The converted encrypted data packet is uploaded to the designated receiving interface of the cloud knowledge base via a secure transmission protocol. Simultaneously, the unit constructs and sends an encrypted similar case retrieval request, which embeds the encrypted dynamic risk assessment vector as the query basis.

[0102] After receiving encrypted data packets, the cloud-based knowledge base performs a retrieval operation while maintaining data encryption or performing secure decryption according to an agreed-upon strategy. The knowledge base internally stores a massive amount of historical cases, each containing a complete risk assessment record for a specific device within a specific time period, along with its contextual information. The retrieval algorithm analyzes the feature similarity between the query vector and the cases in the database; the calculation process may involve encrypted domain similarity calculations or decrypted feature comparisons. The algorithm outputs several historical cases with the highest matching degree to the query vector, forming an encrypted set of historical risk assessment cases. This case set, after encryption, is returned to the local system through a secure channel.

[0103] In a secure environment, the unit decrypts the case set and the original dynamic risk assessment vector stored locally. After decryption, the unit performs a multi-dimensional comparative analysis. The comparison includes: the difference between the currently predicted time-varying failure probability distribution and the actual failure probability distribution in historical cases; the deviation between the currently calculated remaining lifetime prediction of critical components and the actual remaining lifetime records of the same components in historical cases; and the degree of agreement between the currently generated set of risk propagation paths and the failure propagation paths actually observed in historical cases. These differences are quantified into specific deviation parameters.

[0104] Based on the calculated set of bias parameters, the local correction unit generates adjustment instructions for the weight parameters of the spatiotemporally coupled prediction network. The adjustment process employs an incremental learning strategy, modifying only the connection weights of specific layers within the network. For example, if comparisons reveal that the actual probability of a certain type of bearing failure in historical cases is generally higher than the current model's prediction, the output weights for the corresponding bearing failure category in the network's failure probability prediction branch are increased. If bias analysis shows that the remaining life prediction is systematically optimistic under high-temperature conditions, the weights of neurons processing temperature-related features in the life decay calculation unit are adjusted. The amount of weight adjustment is typically proportional to the magnitude of the bias parameters and is determined through a preset mapping function.

[0105] The system continuously records the degree to which the predicted performance of the corrected model matches the actual equipment status. These records form feedback data used to evaluate the actual effectiveness of cloud-based collaborative optimization. When a new dynamic risk assessment vector is generated, the cloud-based collaborative module automatically initiates a new round of encrypted upload, retrieval, and local correction processes, forming a closed loop of continuous optimization.

[0106] After obtaining user authorization, the local system can choose to encrypt and upload the anonymized data of the current case to the cloud knowledge base. The uploaded data undergoes verification and integration by the knowledge base, becoming a new historical case that can be retrieved by other devices. This mechanism allows the cloud knowledge base to continuously accumulate diverse risk assessment experience from different devices and operating conditions. Each time the spatiotemporal coupling prediction network undergoes weight correction, the system automatically saves a snapshot version of the network parameters. The snapshot is marked with the corresponding cloud retrieval timestamp and case identifier. The system retains the snapshot versions of the most recent corrections. When new prediction results show abnormal fluctuations, operators can roll back to a historical version of the network parameters to maintain system stability.

[0107] Operators can set limits on the scope of similar case searches, such as searching only for cases of the same model of equipment, or limiting the similarity threshold of operating parameters. The search request can include feature codes of the equipment's current operating environment, and the cloud-based knowledge base will prioritize matching historical cases with similar environmental characteristics. The search results will include a matching score and a summary of the source equipment for the case.

[0108] The logs record the trigger time of each correction, details of the involved deviation parameters, the adjusted weight positions, and the correction amounts. These logs are used to analyze the long-term effects of cloud-based collaborative optimization and identify the correlation between specific deviation patterns and the optimal correction strategy. The analysis results are fed back to the correction algorithm to optimize subsequent weight adjustment decision logic.

[0109] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0110] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A cross-dimensional turbomachinery dynamic risk prediction and visualization system, characterized in that, include: The dynamic data acquisition module is used to acquire multi-source heterogeneous sensor data of turbomachinery in real time. The multi-source heterogeneous sensor data includes vibration spectrum sequence, aerodynamic load distribution matrix, rotor axial displacement time sequence and surface temperature field thermogram. The service degradation assessment module is used to extract the current vibration feature spectrum based on the historical degradation mode library of the vibration spectrum sequence, calculate the aerodynamic instability risk level according to the deviation between the aerodynamic load distribution matrix and the preset operating condition threshold, and generate the service degradation trajectory matrix by fusing the abrupt gradient of the rotor axial displacement time sequence and the area of ​​the abnormal temperature difference region of the surface temperature field thermogram. The multidimensional risk prediction module is used to input the service degradation trajectory matrix into the spatiotemporal coupled prediction network and output a dynamic risk assessment vector containing the time-varying failure probability distribution, the predicted value of the remaining life of key components, and the set of risk propagation paths. The risk visualization engine module is used to construct a three-dimensional spatial risk thermal field model based on the dynamic risk assessment vector, and generate a dynamic thermal mesh layer superimposed on the digital twin of the turbomachinery. The multidimensional risk prediction module includes: The spatiotemporal feature extraction unit is used to perform spatiotemporal convolution operations on the service degradation trajectory matrix to extract cross-scale risk feature vectors. The fault probability prediction unit is used to process the cross-scale risk feature vector through a gated recurrent network and output the time-varying fault probability distribution within a future time window. The life decay calculation unit is used to calculate the predicted value of the remaining life of the key component based on the mapping relationship between the cross-scale risk feature vector and the fatigue curve of the component material. The path inference unit is used to simulate the diffusion process of risk in the topology of turbomachinery based on graph neural networks and generate the set of risk propagation paths.

2. The cross-dimensional turbomachinery dynamic risk prediction and visualization system as described in claim 1, characterized in that, The dynamic data acquisition module includes: The vibration feature analysis unit is used to perform wavelet packet energy spectrum decomposition on the vibration spectrum sequence, extract the energy entropy value within a preset frequency band interval, and mark the abnormal frequency band positions where the energy entropy value exceeds the entropy value threshold. The load distribution verification unit is used to compare the aerodynamic load distribution matrix with the reference load distribution template under standard working conditions, and to calculate the area ratio and maximum over-limit amplitude of the local load over-limit region. The displacement co-analysis unit is used to correlate the rotor axial displacement timing with the speed pulse signal and identify the phase synchronization error between displacement abrupt points and speed step points. The temperature field fusion unit is used to map the surface temperature field thermogram to the three-dimensional mesh model of the turbomachinery and locate the spatial coordinate set of the temperature difference abnormal area.

3. The cross-dimensional turbomachinery dynamic risk prediction and visualization system as described in claim 2, characterized in that, The service degradation assessment module includes: The degradation pattern matching unit is used to retrieve the degradation pattern that matches the abnormal frequency band location from the historical degradation pattern library and output the first risk level; The instability risk quantification unit is used to calculate the aerodynamic instability risk level based on the over-limit amplitude gradient when the area ratio of the local load over-limit region exceeds the area threshold and the maximum over-limit amplitude exceeds the amplitude threshold. The displacement risk correlation unit is used to generate an axial displacement risk coefficient based on the amplitude increment of the phase synchronization error and the displacement abrupt change point. The fusion calculation unit is used to weight and fuse the first risk level, the aerodynamic instability risk level, and the axial displacement risk coefficient to generate the service degradation trajectory matrix.

4. The cross-dimensional turbomachinery dynamic risk prediction and visualization system as described in claim 1, characterized in that, The risk visualization engine module includes: A spatial mapping unit is used to map the time-varying fault probability distribution to the corresponding mesh unit of the turbomachinery three-dimensional mesh model; The thermal field generation unit is used to calculate the color gradient value of the grid cell based on the predicted remaining life of the key component, and combine it with the spatial coordinates of the risk propagation path set to generate the three-dimensional spatial risk thermal field model. The dynamic rendering unit is used to synchronously update the three-dimensional spatial risk thermal field model with the turbomachinery operating status and output the dynamic thermal mesh layer.

5. The cross-dimensional turbomachinery dynamic risk prediction and visualization system as described in claim 4, characterized in that, The system also includes: The anomaly tracing module is used to extract all paths in the risk propagation path set that pass through the grid cell when the probability value of a specific grid cell in the time-varying fault probability distribution exceeds the probability threshold, and trace them back to the source grid cell. The anomaly tracing module includes a path filtering unit, which calculates the propagation strength weight of each path and filters paths whose weights exceed a weight threshold to form the main propagation chain.

6. The cross-dimensional turbomachinery dynamic risk prediction and visualization system as described in claim 5, characterized in that, The system also includes an adaptive monitoring module, used to adjust the acquisition frequency of corresponding sensor data according to the spatial location of the source grid unit; The adaptive monitoring module includes a frequency decision unit, which is used to calculate the incremental coefficient of the acquisition frequency based on the path length and propagation intensity weight of the main propagation chain.

7. The cross-dimensional turbomachinery dynamic risk prediction and visualization system as described in claim 6, characterized in that, The system also includes a thermal grid optimization module, used to dynamically adjust the grid density of the dynamic thermal grid layer; The thermal grid optimization module includes: Risk gradient analysis unit, used to calculate the risk probability gradient value between adjacent grid cells; A density controller unit is used to insert a subdivided grid layer in the corresponding region when the risk probability gradient value exceeds a gradient threshold. Transition layer generation unit, used to configure a gradient transition mesh between the subdivision mesh layer and the original mesh layer.

8. The cross-dimensional turbomachinery dynamic risk prediction and visualization system as described in claim 7, characterized in that, The system also includes a multi-dimensional data cleaning module, used to calculate the redundancy of historical data storage based on the update frequency of the dynamic risk assessment vector; The multidimensional data cleaning module includes: A redundancy assessment unit is used to count the number of historical versions of the service degradation trajectory matrix. The cleanup decision unit is used to delete the earliest version data in reverse chronological order of storage time when the number of historical versions exceeds the version threshold.

9. The cross-dimensional turbomachinery dynamic risk prediction and visualization system as described in claim 8, characterized in that, The system also includes: a cloud collaboration module, used to encrypt and upload the dynamic risk assessment vector to a cloud knowledge base; The cloud-based collaboration module includes: An encrypted retrieval unit is used to initiate similar case retrieval requests to the cloud and obtain a set of matching historical risk assessment cases. The local correction unit is used to compare the deviation between the historical risk assessment case set and the current dynamic risk assessment vector, and correct the weight parameters of the spatiotemporal coupled prediction network.

Images