Real-time three-dimensional visualization system of wind power plant

Abstract

The invention relates to the field of image data processing, in particular to a real-time three-dimensional visualization system of a wind power plant. The elastic data fusion module is used for aggregating fan SCADA, weather station and weather forecast data in real time, inputting the data into a data fusion cascade architecture, supplementing dynamic parameters and generating data streams. The three-dimensional visualization module is used for obtaining a basic geographic scene through a three-dimensional modeling tool in combination with the oblique photography image; constructing three-dimensional digital assets by adopting a three-dimensional modeling engine, and obtaining a static digital scene; updating by using the data stream to obtain a dynamic three-dimensional scene; and a visual prompt is sent through a mechanism of changing the visual attributes of the three-dimensional digital assets. And the rendering maintenance module is used for outputting a rendering maintenance updating prompt by analyzing the data stream, and carrying out coloring rendering updating by dynamic color coding. Through the data fusion cascade architecture and the three-dimensional modeling engine, the problem that a visual scene is disjointed from reality due to data interruption is solved, and the operation reliability of the system is remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of image data processing, specifically to a real-time 3D visualization system for wind power plants. Background Technology

[0002] With the digital transformation of industry, 3D visualization technology is widely used for centralized monitoring of large facilities, especially in the field of wind power. Existing technologies mainly achieve this through two paths: first, using wind turbine operating data to drive pre-built 3D digital assets, dynamically displaying equipment status; second, combining GIS geographic information and meteorological data to construct high-precision wind farm environment and airflow models. These technologies map the physical entities and operating status of wind farms to a virtual 3D space, enabling remote and intuitive monitoring and management of remote wind farms, and initially improving operation and maintenance efficiency.

[0003] However, in the process of digital transformation, existing technologies still have some common shortcomings. For example, they rely heavily on stable data transmission, and in remote environments such as wind farms, weak network signals and easy data loss affect the real-time performance and accuracy of the system. Furthermore, when multi-source data conflicts or anomalies occur, existing data fusion methods often employ simple fixed-priority strategies, which are insufficient for intelligently handling complex situations and can easily lead to model distortion. In addition, current system functions are mostly limited to status monitoring and basic alarms, lacking advanced applications such as intelligent predictive maintenance based on massive amounts of data, operational optimization analysis, and immersive interactive training, making it difficult to meet the higher requirements of refined and intelligent operation of wind farms.

[0004] In summary, what is needed now is a real-time 3D visualization system for wind farms that offers reliable data transmission, intelligent decision-making, and highly integrated functionality.

[0005] To address this, a real-time 3D visualization system for wind power plants is proposed. Summary of the Invention

[0006] The purpose of this invention is to provide a real-time three-dimensional visualization system for wind power plants to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a real-time three-dimensional visualization system for wind power plants, comprising:

[0008] The elastic data fusion module is a cascaded architecture that aggregates input data from wind turbine SCADA, weather stations, and weather forecasts in real time. When SCADA data is interrupted, it uses weather station data, weather forecasts, and data extrapolation algorithms to estimate dynamic parameters and generate a data stream to drive the generation of dynamic images of 3D scenes.

[0009] 3D Visualization Module: Generates 3D digital terrain using 3D modeling tools, and obtains a basic geographic scene by combining oblique photogrammetry images; constructs 3D digital assets using a procedural 3D modeling engine, and obtains a static digital scene by combining the basic geographic scene; drives the 3D digital assets using the data stream to obtain a dynamic 3D scene; selects parameters using a ray casting method, obtains corresponding operating parameters, compares them using a rule-based judgment engine, triggers a visualization alarm mechanism, and sends a visualization prompt.

[0010] Rendering maintenance module: By analyzing the data stream, it outputs rendering maintenance update prompts; it converts the rendering maintenance update prompts into color values ​​through a color gradient mapping function, and performs color rendering updates through a custom shader program.

[0011] Preferably, the specific implementation process of the elastic data fusion module includes:

[0012] The system receives and timestamps data from multiple sources, including wind turbine SCADA, local weather stations, and weather forecasts, to form a data foundation. Based on this foundation, a data validity detection mechanism continuously verifies the integrity of the SCADA data. When an anomaly is detected in the SCADA data, a cascading supplementation process is triggered, executing in a preset order: First, measured data from the local weather station is retrieved from the data foundation for parameter estimation; if the local weather station is invalid, the weather forecast is retrieved; if the weather forecast is invalid, a preset data extrapolation algorithm is used to estimate dynamic parameters; after standardization, a data stream with a unified structure suitable for 3D animation and rendering is generated.

[0013] Preferably, the specific implementation process of the data fusion cascade architecture includes:

[0014] The system receives and integrates inputs from multiple heterogeneous data sources, assigning them a predetermined priority order. All input data undergoes timestamp alignment and synchronization to form a unified data processing benchmark. Based on this benchmark, a continuously running verification mechanism monitors the status of the highest-priority data source, verifying its integrity, continuity, and quality. Upon successful verification, the highest-priority data source is directly selected and standardized. When the verification mechanism determines that the highest-priority data source is experiencing an anomaly, a cascading supplementary process is triggered, following a predetermined priority order to sequentially probe and verify the data from secondary data sources. This verification process is repeated. When a heterogeneous data source becomes unavailable, a predetermined data estimation algorithm is invoked as a data generation method. All generated data is then fed into a standardized processing pipeline, generating a data stream with a unified structure suitable for graphics processing pipelines.

[0015] Preferably, the specific implementation process of the static digital scene includes:

[0016] The digital elevation model data procured through commercial use is processed by computer graphics software. A polygon mesh generation algorithm is used to generate a three-dimensional digital terrain of the wind farm. Texture mapping is performed on the image content obtained by oblique photography to construct a basic geographic scene. Engineering parameters are read from preset engineering data. A procedural 3D modeling engine based on parametric templates is used to construct a three-dimensional digital asset containing the shape of specific equipment based on the engineering parameters. Combined with the basic geographic scene, a static digital scene is formed.

[0017] Preferably, the specific implementation process of the dynamic three-dimensional scene includes:

[0018] The data stream is received and parsed to extract real-time dynamic parameters, which include a unique identifier, device operating parameters, and operating status codes. The real-time dynamic parameters are bound to the 3D digital asset based on the unique identifier. Based on the binding relationship, the device operating parameters are used to drive the geometric position and posture of specific devices in the 3D digital asset in real time, generate dynamic behaviors, update the animation of the 3D digital asset, and obtain a dynamic 3D scene. The operating status codes are used to identify the estimated data in the real-time dynamic parameters and trigger a preset visual mode in the dynamic 3D scene.

[0019] Preferably, the specific process for generating the visual prompt includes:

[0020] The system detects user directional selection of 3D digital assets by performing raycasting in the dynamic 3D scene and obtains the unique identifier corresponding to the 3D digital assets. It retrieves relevant real-time and historical operating parameters from the data stream, transmits them to the dynamic 3D scene, automatically generates an information label panel for the 3D digital assets, and displays it in the dynamic 3D scene using a 2D image overlay method. A built-in rule-based judgment engine continuously monitors the data stream and performs uninterrupted comparisons with preset alarm thresholds. When the rule-based judgment engine determines that the real-time operating parameters exceed the alarm threshold, it generates a warning event, triggers the visual alarm mechanism, changes the rendering visualization state of the corresponding 3D digital assets, generates descriptive visual prompts, and displays the visual prompts on the information label panel.

[0021] Preferably, the specific process for generating the rendering maintenance update prompt includes:

[0022] The data stream is filtered and extracted to extract operating parameters related to the wear of key components, and a real-time feature vector is constructed. The real-time feature vector is then input into a remaining service life model trained using historical fault data and operating data. The remaining service life model performs model inference calculations to generate a rendering maintenance update prompt for a specific key component, which includes a unique identifier and a corresponding remaining service life estimate.

[0023] Preferably, the specific implementation process of performing the shading rendering update includes:

[0024] Upon receiving a rendering maintenance update prompt, the system uses a unique identifier in the prompt to retrieve and locate the 3D digital assets requiring visual state updates within the dynamic 3D scene. The remaining lifetime estimate in the prompt is input into a preset color gradient mapping function, mapping the lifetime estimate from its numerical range to continuously changing color values. These color values ​​are then passed as parameters to a custom shader program applied to the 3D digital assets. The color value parameters are applied to the surface material properties of the 3D digital assets, changing the final output color in the rendering pipeline in real time.

[0025] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0026] 1. To ensure the reliability and integrity of the data stream, a flexible data fusion module is constructed to aggregate wind turbine SCADA, meteorological station, and weather forecast data in real time. Its core data fusion cascading architecture sets clear data source priorities: when the highest priority SCADA data is detected to be interrupted or abnormal, the system triggers a cascading supplementation process, strictly following the order of "local meteorological station measured data, weather forecast data, and data extrapolation algorithm" to sequentially call secondary data sources for parameter supplementation. This mechanism ensures that regardless of the availability of a single data source, the system can generate a structurally unified and uninterrupted data stream to drive the 3D scene, thereby solving the problem of the visualization scene becoming disconnected from reality due to data interruption in traditional systems and improving the system's operational robustness.

[0027] 2. Enhancing the efficiency and visual intuitiveness of operation and maintenance decisions: Through the rendering maintenance module, the system can output rendered maintenance update prompts for key wind turbine components. These prompts are based on estimates calculated from analyzing operating parameters such as vibration and temperature and inputting them into the Remaining Useful Life (RUL) model. This estimate is then converted into specific color values ​​using a preset color gradient mapping function, and a custom shader program dynamically colors and renders the corresponding 3D digital assets in the dynamic 3D scene. This transforms abstract equipment health data into intuitive visual states, allowing maintenance personnel to directly perceive the urgency of equipment maintenance through color, quickly identify high-risk units, and thus efficiently formulate maintenance plans.

[0028] 3. The system enables interactive querying and intelligent alarms. It not only dynamically displays 3D scenes but also provides interactive capabilities. Users can select any 3D digital asset in the scene using raycasting. The system immediately retrieves and displays its corresponding real-time and historical operating parameters in an information tab panel. Simultaneously, the built-in rule-based judgment engine continuously compares real-time data with preset alarm thresholds. If parameters exceed limits, a visual alarm mechanism is triggered, issuing a clear visual prompt by changing the visual attributes of the 3D digital asset (such as changing color or highlighting). The alarm information is then simultaneously displayed on the information tab panel, achieving an intelligent management upgrade from passive observation to proactive early warning. Attached Figure Description

[0029] Figure 1 This is a structural diagram of a real-time three-dimensional visualization system for a wind power plant proposed in an embodiment of this invention.

[0030] Figure 2 This is a flowchart illustrating the elastic data fusion and dynamic scene generation process proposed in an embodiment of this invention.

[0031] Figure 3 This is a flowchart illustrating the predictive maintenance and visualization rendering process proposed in an embodiment of this invention. Detailed Implementation

[0032] 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.

[0033] Please see Figures 1-3 The present invention provides a real-time three-dimensional visualization system for wind power plants, the specific modules of which are as follows:

[0034] The elastic data fusion module is a cascaded architecture that aggregates input data from wind turbine SCADA, weather stations, and weather forecasts in real time. When SCADA data is interrupted, it uses weather station data, weather forecasts, and data extrapolation algorithms to estimate dynamic parameters and generate a data stream to drive the generation of dynamic images of 3D scenes.

[0035] 3D Visualization Module: Generates 3D digital terrain using 3D modeling tools, and obtains a basic geographic scene by combining oblique photogrammetry images; constructs 3D digital assets using a procedural 3D modeling engine, and obtains a static digital scene by combining the basic geographic scene; drives the 3D digital assets using the data stream to obtain a dynamic 3D scene; selects parameters using a ray casting method, obtains corresponding operating parameters, compares them using a rule-based judgment engine, triggers a visualization alarm mechanism, and sends a visualization prompt.

[0036] Rendering maintenance module: By analyzing the data stream, it outputs rendering maintenance update prompts; it converts the rendering maintenance update prompts into color values ​​through a color gradient mapping function, and performs color rendering updates through a custom shader program.

[0037] The technical solution of the present invention will be further described in detail below with reference to specific embodiments.

[0038] Example 1

[0039] This application discloses a real-time 3D visualization system for wind power plants. In wind power plant A, this system is used to display current data. (See attached document.) Figure 1 The specific modules proposed in this invention include: an elastic data fusion module: a real-time aggregation and cascading architecture for wind turbine SCADA, meteorological station, and weather forecast input data. When SCADA data is interrupted, dynamic parameter estimates are calculated step-by-step using meteorological station, weather forecast, and data extrapolation algorithms to generate a data stream for driving the generation of dynamic images of the 3D scene; a 3D visualization module: generating 3D digital terrain through 3D modeling tools, combining it with oblique photogrammetry images to obtain a basic geographic scene; constructing 3D digital assets using a procedural 3D modeling engine, and obtaining a static digital scene by combining it with the basic geographic scene; the data stream drives the 3D digital assets to obtain a dynamic 3D scene; selecting corresponding operating parameters through ray casting, comparing them with a rule-based judgment engine, triggering a visualization alarm mechanism, and sending visualization prompts; and a rendering and maintenance module: analyzing the data stream to output rendering and maintenance update prompts for key wind turbine components; the rendering and maintenance update prompts use dynamic color coding to colorize and render the corresponding 3D digital assets.

[0040] Furthermore, real-time aggregation of wind turbine SCADA, meteorological station, and weather forecast data is input into a data fusion cascade architecture. When SCADA data is interrupted, dynamic parameters are supplemented step-by-step using meteorological station data, weather forecast data, and data extrapolation algorithms to generate a data stream used to drive the generation of dynamic images of the 3D scene; this corresponds to the aforementioned elastic data fusion module; see [link / reference]. Figure 2 The specific implementation process includes:

[0041] The core workflow of this elastic data fusion module begins with processing inputs from multiple heterogeneous data sources. These data sources are arranged in a preset priority order. For example, the highest priority is given to the wind turbine SCADA system, which provides high-frequency operating parameters such as "blade angle," "yaw angle," "generator speed," and "output power." Next are locally deployed weather stations, which provide measured environmental data such as "wind speed," "wind direction," "ambient temperature," and "air pressure." The lowest priority is given to weather forecast services. The system unifies and synchronizes the timestamps of all data sources, creating a unified data processing benchmark.

[0042] Based on a unified data processing benchmark, the verification mechanism continuously monitors the validity of SCADA data. During verification, the mechanism ensures data availability through multiple dimensions. Integrity verification checks whether the data frame contains all preset parameter fields; for example, if a SCADA data packet lacks the "generator speed" field or its value is invalid, the data packet is deemed incomplete. Continuity verification monitors the timestamp sequence based on the data source's reporting frequency; if there is a time interval exceeding the tolerance limit between timestamps, a data interruption is determined. Simultaneously, quality verification checks the reasonableness of data values ​​using a preset rule base. For example, physical boundary rules reject abnormal readings with a "blade angle" of 100 degrees (the actual physical range is -5 to 90 degrees); operating condition association rules determine that data points reporting 2MW power output at a wind speed of 2m / s (below the cut-in wind speed) are logically incorrect.

[0043] In terms of continuity verification, the system achieves this by setting a dynamic time tolerance limit (T_tolerance). This threshold can be configured by the user, and its value is generally between 1.5 and 3 times the theoretical reporting cycle of the data source. For example, for SCADA data with a theoretical reporting cycle of 1 second, T_tolerance can be set to 2 seconds. When the system is running, if the difference between the timestamp of the current data frame and the timestamp of the previous valid data frame exceeds T_tolerance, then it will be determined that the data has been interrupted. In terms of quality verification, the key lies in the "preset rule base". This rule base must contain at least the following three types of logical rules that can be executed by the computer: The first type is physical boundary rules, which set an effective value range for each key parameter that conforms to its engineering significance. For example, the rule base may specify that the effective range for "blade angle" is [-5.0, 90.0] degrees, and all data outside this range will be marked as invalid. The second category is operating condition association rules, which define the logical relationships between multiple parameters based on the fundamental physical principles of wind turbine operation. These rules are primarily used to identify "dirty data" that violates operational logic. The third category is rate-of-change rules, which monitor the magnitude of change of key parameters per unit time and filter out sudden noise caused by momentary sensor malfunctions. For example, the rule base specifies that the absolute value of the second-level rate of change of "generator speed" cannot exceed 1.5 RPM / s.

[0044] When the aforementioned verification mechanism determines any anomalies in the SCADA data, the system automatically triggers a cascading supplementary process. For example, if the "output power" of a wind turbine is deemed invalid due to a logical error, the first step is for the system to attempt to acquire real-time wind speed data collected by the local weather station. If this acquisition is successful, a massive amount of high-precision simulation data is generated by performing a systematic simulation of the three-dimensional digital terrain of the wind farm under various wind speed, wind direction, and atmospheric stability conditions using computational fluid dynamics (CFD). Based on this dataset, a lightweight surrogate model for wind farm flow mechanisms, such as a radial basis function network, is trained and constructed. This surrogate model can learn and quickly reproduce the nonlinear relationships revealed by complex CFD simulations. When supplementary data is needed, real-time wind speed and direction data from the weather station are used as input. The surrogate model performs rapid inference calculations to obtain the estimated wind speed at the target wind turbine hub height. This is then combined with the wind turbine's standard power curve model to estimate a theoretical output power reference value as a supplement. If valid data cannot be obtained in the first step, the process will automatically proceed to the second step, retrieving wind speed data from the weather forecast and using a microscale wind resource model to estimate a theoretical output power reference value as a supplement. If the weather forecast service is also interrupted or the data is invalid, then the process will proceed to the third step, activating a pre-set data extrapolation algorithm. The specific logic is as follows: First, the duration of the data interruption is judged. If the duration is less than a pre-set short-term threshold (e.g., 3 minutes), then the 'previous value filling' algorithm is used by default, directly using the last valid data point as the current estimate. If the duration of the data interruption exceeds the short-term threshold, then the process will switch to a pre-trained autoregressive integral moving average model (ARIMA(p, d, q)) to perform parameter extrapolation. This ARIMA model is trained offline solely based on the wind turbine's past operating data (e.g., the power output sequence sampled every minute over the past 24 hours). In this process, the key parameters p (number of autoregressive terms), d (difference order), and q (number of moving average terms) of the model are automatically optimized and determined within a predefined parameter space (e.g., p∈[0, 5], d∈[0, 2], q∈[0, 5]) by evaluating historical data using the Akaike Information Criterion (AIC). To ensure the reliability of the model, its construction process further includes the following steps: 1) Training data preparation: Selecting the minute-level power output sequence of the target wind turbine over the past 30 days after data cleaning as the training dataset. 2) Model training and validation: Dividing the prepared dataset into training and validation sets in an 8:2 ratio.The model is trained on the training set, and the root mean square error (RMSE) is used as the evaluation metric on the validation set to test the model's prediction accuracy. The model is considered valid only when the RMSE is below a preset threshold (e.g., below 5% of rated power). 3) Model Deployment and Update: The validated model is deployed in the system. The system continuously monitors the error between the calculated results and the subsequently recovered SCADA data. When the error continuously exceeds the threshold, the model retraining process is automatically triggered. Based on the inherent trend of recent data, the current dynamic parameter estimates are calculated to provide estimates that better conform to the operating logic. All data generated through this process is fed into a standardized processing pipeline and encapsulated into a data stream with a unified structure. For example, the final output JSON object will always contain fixed keys: {timestamp: ..., turbine_id: ..., power_output: 0.5, source_flag: extrapolated}.

[0045] This ensures that the system can still generate continuous and complete data streams even under poor network conditions. Through a cascaded supplementary architecture, it effectively fills data gaps caused by communication interruptions, overcoming the limitations of traditional systems that rely on stable networks. This improves the reliability and data trustworthiness of the visualization system in remote areas, providing an uninterrupted data foundation for upper-layer applications.

[0046] Furthermore, the specific process of the data fusion cascade architecture includes:

[0047] The core of the data fusion cascade architecture of this invention lies in establishing a deterministic, priority-based heterogeneous data source processing framework. In this architecture, all available data sources are pre-defined with a clear priority order. In this embodiment, the order is as follows: first priority is the wind turbine's own SCADA system; second priority is the local weather station deployed at the wind farm; third priority is the regional weather forecast service obtained from the network; and fourth priority, as a final supplementary means, is a pre-defined data extrapolation algorithm. Based on the existing fixed-priority cascade architecture, a reinforcement learning agent can be deployed to manage the entire cascade strategy. This agent aims for "highest long-term data stream quality and lowest overall cost," continuously trying different combinations of data sources and supplementary logic. In actual operation, A / B testing is conducted on different wind turbine groups, with one group using a standard cascade strategy and the other using an experimental strategy. By comparing the final operational benefits, the agent can continuously learn and evolve, improving the reliability and accuracy of the data stream. In the initial stage of architecture operation, all data sources, regardless of their original format or update frequency, undergo timestamp alignment and synchronization processing to form a unified data processing benchmark. The continuously running verification mechanism will monitor the status of the highest priority data source without interruption, verifying its integrity, continuity, and quality, based on this benchmark. If a data source passes all verification rules, its data will be directly selected and sent to the standardized processing pipeline.

[0048] When the verification mechanism determines that the highest priority data source is interrupted or its data quality is unacceptable, a cascading supplementary process is triggered, automatically probing lower-level data sources according to a preset priority order. First, the second-priority local weather station is checked to determine if the acquired data passes all verification rules; if so, it fails. Next, the third-priority weather forecast is checked, again verifying if the acquired data passes all verification rules; if still, it fails. Finally, the process calls the fourth-priority data extrapolation algorithm. In this embodiment, this algorithm can be a statistical model based on historical operating conditions, calculating data change trends to infer the most likely reasonable value as a supplement. Regardless of the priority of the data ultimately selected, it will be sent to the final standardized processing pipeline and encapsulated into a uniformly structured data packet.

[0049] A priority-based decision-making framework was established. The selection of data sources was transformed from random or simple aggregation into a rigorous, ordered logical judgment process, ensuring that the system always prioritizes the most reliable data. This architectural design guarantees the clarity of data processing logic and the stability of output results, thus improving the overall reliability of the system.

[0050] Furthermore, 3D digital terrain is generated using 3D modeling tools and combined with oblique photogrammetry images to obtain a basic geographic scene; a procedural 3D modeling engine is used to construct 3D digital assets, and combined with the basic geographic scene to obtain a static digital scene; corresponding to the 3D visualization module; see [link / reference]. Figure 2 The specific implementation process includes:

[0051] To construct the basic geographic scene, the system processes commercially purchased digital elevation model (DEM) data covering the wind farm's location area using computer graphics software. The software's polygon mesh generation algorithm generates a 3D digital terrain reflecting the undulating terrain based on this DEM data. Subsequently, the system applies texture mapping technology to the surface of the 3D digital terrain using images acquired by drones at different angles and oblique photography. This imbues the terrain with natural textures including vegetation, soil, and rocks, and clearly presents essential man-made facilities for operation and maintenance, such as winding maintenance roads, hoisting platforms for each turbine, drainage ditches, and the concrete foundations of the substation, thus constructing the basic geographic scene.

[0052] Building upon this foundational geographic scenario, the system further employs a procedural 3D modeling engine to construct 3D digital assets based on engineering parameters. This engine can handle complex situations where a wind farm may contain multiple equipment models from different manufacturers, or even from the same manufacturer but with slight variations due to different batches. For example, when constructing a wind turbine unit, the system reads unique engineering parameters from the engineering bill of materials or design drawings, such as "tower type is flexible tower, height 94 meters," "blade model is S55, length 55 meters," and "specific dimensions and heat dissipation layout of the nacelle cover." These parameters are then input into the procedural 3D modeling engine, which, based on a parametric template, constructs a 3D digital asset that perfectly matches the engineering data. Specifically, the parametric template is a JSON-formatted configuration file that defines a series of basic geometries (such as cylinders and airfoil sections) and their controllable parameters (such as radius, height, and stretching path). Upon receiving the engineering parameters, the procedural 3D modeling engine first maps text descriptions such as "tower type is flexible tower" to specific template IDs and basic geometry selections using a built-in rule base. Subsequently, the engine passes numerical parameters such as 'height 94 meters' to the geometry generation function in the corresponding template. This function generates the tower mesh through a procedural method (such as segmented stretching of a circular cross-section along the Y-axis). For complex components such as blades, the engine loads the corresponding airfoil section curve data based on 'blade model S55' and performs a lofting operation along a preset length and torsion angle path. Finally, all generated component meshes are assembled according to their relative positions in the engineering drawings to form a complete 3D digital asset. This process is repeatedly applied to all equipment on site, including various types of wind turbines, weather towers, box-type transformers, and even the entire substation's structure and equipment. Each generated asset is automatically placed onto the corresponding location in the basic geographic scene based on its geographic coordinates (longitude, latitude, altitude) from the as-built survey data, ultimately forming a static digital scene.

[0053] By using procedural generation technology, three-dimensional digital assets that match engineering parameters are constructed, breaking through the limitations of traditional fixed models that cannot reflect equipment specifications, improving the fidelity and engineering accuracy of digital scenes, and, due to its automated and parameterized characteristics, enabling rapid and low-cost deployment of different models of equipment, thus enhancing the scalability of the system.

[0054] Furthermore, the data stream is used to process the 3D digital assets to obtain a dynamic 3D scene; corresponding to the 3D visualization module; see below. Figure 2 The specific implementation process includes:

[0055] The system receives and parses the data stream. The data structure within this stream includes explicit identification of the data source. For example, when the SCADA system of Wind Turbine 7 (WT07) is communicating normally, the system continuously receives JSON objects similar to the following: {id: WT07, params: {rpm: 15.2, yaw: 285, pitch: 2.5, temp_gen: 65}, status_code: 1, status_desc: real-time data}. This data contains the device's unique identifier (WT07), detailed device operating parameters (such as speed, yaw angle, pitch angle, and generator temperature), and an operating status code indicating the data status (status_code: 1 represents real-time data). When the SCADA data of wind turbine WT12 is interrupted due to network issues, the elastic data fusion module will initiate a supplementary process to generate a JSON object containing estimated parameters, for example: {id: WT12, params: {power_output_est: 2055.0, rpm_est: 14.5}, status_code: 101, status_desc: status estimate}. The running status code (status_code: 101) here is predefined as the "data estimation" status, and its parameters (params) only contain core performance parameters estimated based on algorithms such as the power curve.

[0056] The system utilizes unique identifiers (such as "WT07" or "WT12") in the data packet to efficiently retrieve data from the 3D scene asset registry and binds real-time dynamic parameters to the corresponding 3D digital assets in the scene. After establishing the binding relationship, the system provides differentiated driving for the 3D digital assets based on the binding relationship and the running status code: For assets receiving real-time data (such as WT07), the system drives the dynamic behavior of the 3D digital assets in real time based on their detailed equipment operating parameters. For example, for the "yaw angle" parameter, maintenance personnel will see the nacelle of WT07 smoothly rotate to align with the wind direction; for the "pitch angle" parameter, they can clearly see the blades making fine adjustments to control power. These parameter-driven geometric transformations are calculated and applied in each rendering frame, generating a dynamic 3D scene that is highly synchronized with the physical entity's operating state. For an asset receiving estimated data (such as WT12), after the system identifies it as a "state estimation" data packet, it uses the estimated parameters in the data packet (such as rpm_est: 14.5) to drive the core animation of the 3D digital asset. For example, it makes the blades of WT12 rotate at a speed coordinated with the estimated rotational speed to ensure visual continuity. However, the system will not perform inaccurate simulations of its yaw, pitch, or other attitudes that cannot be predicted. At the same time, this status code will trigger a preset visual mode. In this embodiment, the visual mode is set to render the 3D digital asset corresponding to the WT12 wind turbine as a semi-transparent blue and overlay a clear text label next to it, displaying "SCADA signal interrupted, current state estimation".

[0057] This achieves a high-fidelity real-time mapping from the physical world's operational state to the dynamic behavior of the virtual scene, ensuring the visual continuity and physical accuracy of dynamic behaviors such as wind turbine speed and orientation. This provides maintenance personnel with an immersive, highly synchronized remote monitoring experience, improving situational awareness efficiency.

[0058] Furthermore, selection is made using a ray casting method to obtain corresponding operating parameters, which are then compared by a rule-based judgment engine to trigger a visual alarm mechanism and send a visual prompt; this corresponds to the 3D visualization module; the specific implementation process includes:

[0059] In terms of user interaction queries, when a user performs a directional selection operation in a dynamic 3D scene using an input device such as a mouse, the system detects the user's intent using raycasting. Specifically, the system launches a ray from the virtual camera (the user's viewpoint) into the 3D scene, passing through the location of the mouse cursor. When this ray collides with a geometric object such as a wind turbine model, the system determines that the asset has been selected and then retrieves its unique identifier. In addition to traditional raycasting, the system uses multiple input methods to achieve more efficient human-computer interaction: eye-tracking technology can automatically preload key information when a user stares at a component for an extended period; it can also parse user voice commands; for example, when a user looks directly at a wind turbine and says "isolate this device," the corresponding safety operation can be executed; simultaneously, through gesture recognition, users can directly grasp and rotate equipment models in 3D space, creating a high-bandwidth, low-latency, and intuitive interactive experience, improving work efficiency and decision-making quality. Then, the system uses this unique identifier as a keyword to search for related real-time and historical operating parameters in the data stream. For example, the system will extract its real-time power, wind speed, and other parameters, while also querying its power output curve data for the past 24 hours. This data will be transmitted to the dynamic 3D scene and used to automatically generate a unique information label panel for the 3D digital asset, which will be presented in an overlay manner in the 3D scene (usually floating next to the selected asset).

[0060] Meanwhile, to achieve proactive risk warning, the built-in rule-based engine continuously monitors the data flow of the entire wind farm in the background, constantly comparing the real-time operating parameters of all equipment with pre-set alarm thresholds. For example, a pre-set alarm rule in the engine states that if the bearing temperature of any wind turbine generator continuously exceeds 85 degrees Celsius for 60 seconds, a level one warning will be triggered. When the condition of this rule-based engine is met, the system immediately generates a warning event. This event triggers a visual alarm mechanism, issuing a clear prompt by changing the rendering visualization state of the corresponding 3D digital asset, for example, by covering the surface of its nacelle model with a striking, continuously flashing red highlight effect. At the same time, a descriptive visual prompt (such as the text "Warning: Generator bearing temperature exceeds limit") is also generated and simultaneously displayed on the asset's information label panel.

[0061] By using ray casting technology, intuitive and immersive data interaction in a three-dimensional scene is achieved, which simplifies the data retrieval process, improves the efficiency and intuitiveness of monitoring and analysis of complex power plants, and transforms abstract alarm events into clear and intuitive visual signals in three-dimensional space, helping maintenance personnel to quickly locate problematic equipment and judge its severity, thereby improving the speed and accuracy of risk perception.

[0062] Furthermore, by analyzing the data stream, rendering maintenance update prompts for key wind turbine components are output; corresponding to the rendering maintenance module; see [link / reference]. Figure 3 The specific implementation process includes:

[0063] From the data stream, the system filters and extracts the multiple operating parameters most closely related to the main bearing wear. These parameters are pre-determined based on domain expert knowledge and data analysis. For example, the system retrieves and analyzes the bearing's historical load spectrum to calculate its equivalent fatigue load. Simultaneously, it performs a Fast Fourier Transform on the real-time acquired vibration acceleration signal to extract the complete spectrum, paying particular attention to the amplitude changes of fault characteristic frequencies (such as BPFI and BPFO) of various components like the bearing's inner and outer rings and rollers. To integrate low-frequency offline data such as lubricant analysis data, the system uses a forward-filling strategy and introduces the "Data Age" feature to reflect its timeliness. The system then combines the filled offline data, Data Age, and high-frequency real-time sensor data and performs normalization to construct a comprehensive and logically rigorous multi-dimensional real-time feature vector.

[0064] The real-time feature vectors are processed into a sequence with a time step of 60, and used as input samples to a pre-trained Remaining Life (RUL) prediction model. This model employs a Long Short-Term Memory (LSTM) architecture, specifically consisting of two LSTM layers and one fully connected output layer. Specifically, the first LSTM layer has 128 neurons, and the second LSTM layer has 64 neurons. Both use tanh as the activation function to handle potential negative inputs. To avoid overfitting, a dropout layer with a dropout rate of 0.2 is added after each LSTM layer. The final output layer is a single-neuron fully connected layer using a linear activation function, directly outputting the predicted RUL value. In this embodiment, the LSTM model is trained using the Adam optimizer with an initial learning rate of 0.001. The root mean square error (RMSE) is used as the loss function. The batch size during training is 128, the time step is 60, the total training period is 100 epochs, and an early stopping strategy is used to prevent overfitting. For example, training is stopped when the validation set loss does not decrease for 10 consecutive epochs.

[0065] This remaining useful life prediction model is trained by means of supervised learning using the historical operation data of all fans in Factory A over the past several years and detailed fault repair records, and a reference RUL value is assigned to each input sequence in the historical data. This reference RUL value is defined based on a non-linear RUL label generation method based on key physical events. Specifically, for each historical fault case, first, by analyzing its historical data such as high-frequency vibration and temperature, signal processing techniques (such as envelope spectrum analysis) are used to identify the exact time point (T_initial) when the initial fault characteristics appear, such as the moment when the characteristic frequency of the outer ring peeling of the bearing first appears. Thus, the RUL label is generated according to the following non-linear rules: 1) For the time point T, if T < T_initial, that is, before the initial physical fault characteristics appear, the component is considered to be in a healthy state, and its RUL label is set to a constant maximum value (RUL_max). 2) For the time point T, if T_initial ≤ T ≤ T_fail (the final fault shutdown time), it is considered that the component enters the degradation stage, and its RUL label is set to a non-linear decay function, that is, the estimated value of its remaining useful life will start from a maximum value and decrease rapidly exponentially as the time after the fault passes. The specific rate of decrease is controlled by a decay coefficient determined according to the typical degradation curve of this component type. In this way, RUL target values for model training can be generated for each input sequence in the historical dataset.When the model receives the real-time feature vector of the main bearing of the fan, it performs model inference calculations, processes the feature vector through its internal multi-layer neural network, and each layer of the network identifies the deep association between specific combination patterns of these operating condition parameters and the degradation of the component health state through complex weighted operations and non-linear transformations.

[0066] After the model expansion inference operation, it will output the quantified remaining useful life estimate of the main bearing of Fan A-07, such as giving a result like 'The predicted expected value of RUL is 4500 operating hours'. This prediction result will be immediately encapsulated into a structured data packet, which clearly contains the unique identifier of the component, such as 'A-07-MainBearing', and at the same time, the corresponding RUL estimate '4500h'.

[0067] By introducing a pre-trained RUL model, the system has acquired the prediction ability and changed from passive monitoring to active prediction. Moreover, the system can detect in advance the degradation trend presented by the health status of key components, and thus provides extremely crucial data support for implementing predictive maintenance. For managers, this undoubtedly helps them optimize the maintenance strategy and reduce the risk of unplanned shutdowns.

[0068] Furthermore, the rendering maintenance update prompt uses dynamic color coding to perform color rendering updates on the corresponding 3D digital assets; this corresponds to the rendering maintenance module; see below. Figure 3 The specific implementation process includes:

[0069] The rendering module receives a data packet containing the message: "Component identifier is A-07-MainBearing, estimated remaining useful life is 4500 operating hours." Upon receiving this data packet, the rendering module immediately uses the unique identifier "A-07-MainBearing" to quickly search the asset tree of the 3D digital scene and locate the target 3D digital asset that needs a visual state update, namely the specific geometric component representing the main bearing in the A-07 wind turbine model.

[0070] After locking onto the target asset, the system inputs the estimated remaining useful life of "4500 hours" into a preset color gradient mapping function. This function is specifically a piecewise linear interpolation function, its core function being to map the numerical value to a specific color. In this embodiment, the function predefines the entire lifespan range of the main bearing, for example, from 200,000 hours for a brand-new bearing to 0 hours for scrap, and sets a continuous color gradient from green (representing health) to yellow (representing warning) and then to red (representing danger). The input estimated remaining useful life is normalized, its value being "the current remaining useful life divided by the total design life in a brand-new state." Then, this normalized health level, ranging from 0 to 1, is used as an interpolation factor to perform linear interpolation in the color space defined by the three key points of green, yellow, and red, calculating the final color value. Furthermore, when the color code of a component enters "red," the system not only changes the color but also triggers a corresponding spatial sound effect. When a user's viewpoint is close to the wind turbine, simulated abnormal noise matching the fault mode can be heard, the volume of which is proportional to the degree of danger. For users equipped with haptic feedback devices, the system can also transmit low-frequency vibrations to simulate abnormal vibrations of the device. This cross-sensory coding method of "visual + auditory + tactile" enhances the immersive experience of risk perception and the effectiveness of alarms.

[0071] The calculated final color value, such as a yellowish-green representing a "sub-healthy" state, is passed in real-time as a parameter to a custom shader program specifically written for this system and applied to the 3D digital asset of the wind turbine main bearing. Specifically, this custom shader program is a surface shader that, upon receiving the color parameter, uses it as input to its self-illuminating color channel and enhances it according to a preset intensity (e.g., an intensity value of 1.5). Simultaneously, to make alarms more noticeable, the shader also includes a time-driven sine function that causes the self-illuminating color to pulse between a set maximum and minimum intensity. This shader program executes in the graphics hardware's rendering pipeline, receiving the yellowish-green color parameter and applying it to the surface material properties of the target 3D digital asset, thus overriding its original default color. This entire process, from receiving data to final rendering and coloring, is executed cyclically in every rendering frame of the system. Therefore, as the remaining service life of the A-07 wind turbine main bearing gradually decreases during its long operation, maintenance personnel will visually see its color smoothly and continuously transition from a healthy green to a dangerous red in the 3D scene.

[0072] By using custom shaders, complex Remaining Usable Life (RUL) data is transformed into easily understandable color information and directly displayed on 3D digital assets. This innovative visualization method lowers the barrier for operations and maintenance personnel to interpret professional data, enabling them to perceive the long-term health evolution trend of equipment through intuitive color changes, thus achieving efficient asset status monitoring.

[0073] This invention effectively solves the data interruption problem caused by unstable networks in remote areas by constructing a flexible data fusion and cascading architecture, ensuring the integrity and reliability of data flow. Combined with parametric 3D digital asset generation technology, it achieves high-fidelity, synchronous visualization of wind farms. Furthermore, the system integrates a remaining useful life prediction model and intuitively presents the abstract health status on the 3D digital assets using dynamic color coding, realizing a paradigm shift from "passive monitoring" to "proactive prediction" in intelligent management, significantly improving the efficiency of operation and maintenance decisions and the foresight of asset lifecycle management.

[0074] Example 2

[0075] In wind power plant B, the specific implementation of the flexible data fusion module is as follows:

[0076] In the actual operation of Wind Power Plant B, the system's data processing primarily begins with a flexible data fusion process. This process continuously aggregates heterogeneous data from multiple sources, including the wind turbine SCADA system, local weather stations, and third-party weather forecasts. For example, SCADA data packets {turbine_id: B-07, rotor_rpm: 15.2, power_output: 2150.7, ...} collected from the wind turbine (ID: B-07), environmental data {wind_speed: 12.8, wind_direction: 286.0, ...} reported by the weather tower (ID: MetMast-01) deployed at the plant center, and regional forecast data are all sent to the cascaded architecture for processing. All input data initially undergoes timestamp alignment and resampling to establish a unified data processing standard. A continuous verification mechanism then checks the integrity, continuity, and quality of the SCADA data according to this standard. If an anomaly is detected in the data, such as when the SCADA data of wind turbine No. 12 (ID: B-12) is interrupted, the cascaded supplementation process will automatically start. This process will strictly follow the preset priority, sequentially exploring and calling backup data sources. It will first try to make calculations based on the wind speed and direction from the local weather station; if the weather station data is invalid or insufficient, it will further call weather forecast data and, by consulting the wind turbine's unique performance curve file (B12_perf_curve.dat), calculate the possible missing output power (approximately 2055.0kW) and related core operating parameters such as generator speed (approximately 14.5RPM). All data that has been verified, filtered, or supplemented by calculation will eventually be encapsulated into a uniformly structured JSON data stream, with a source identifier attached, thus ensuring that downstream modules always receive complete and reliable drive data regardless of network conditions.

[0077] Example 3

[0078] In wind power plant B, the specific implementation of the 3D visualization module is as follows:

[0079] Following the data stream generated by the elastic data fusion module, this data stream drives a 3D visualization module running on a pre-built high-fidelity static digital scene. This static scene consists of a 5-meter resolution digital elevation model (DEM_B_WF_5m.tif) and a high-resolution oblique photogrammetric texture (Oblique_Img_Q2_2024). Each wind turbine within the scene, such as turbine B-07, is a 3D digital asset generated programmatically according to its engineering bill of materials (BOM_B-07_rev2.xlsx), perfectly matching the physical entity's dimensions, and positioned at its as-built survey coordinates. The visualization module continuously receives the data stream and binds detailed dynamic parameters (speed 15.2 RPM, yaw angle 285.3 degrees, etc.) to the corresponding 3D digital assets in real time, driving the model to produce dynamic behavior synchronized with the physical world. Simultaneously, the system provides maintenance personnel with intuitive interactive methods. When any wind turbine in the scene is selected by clicking with the mouse, the raycasting method immediately identifies the target and retrieves all relevant data, presenting it in chart form in an automatically popping-up information panel. The built-in rule-based judgment engine continuously monitors the operating parameters of all equipment in the background. Once it detects that the generator bearing temperature (85.5℃) of wind turbine No. 4 (ID: B-04) continuously exceeds the preset alarm threshold, the system will immediately trigger a visual alarm, causing the corresponding component in the 3D scene to flash a high-frequency red pulse, and push the alarm content to the information panel, realizing the transformation from passive query to proactive warning.

[0080] Example 4

[0081] In wind power plant B, the specific implementation of the rendering and maintenance module is as follows:

[0082] Building upon real-time visualization and interaction, the system reveals its deeper value through a rendering and maintenance module, which lies in its forward-looking rendering and maintenance capabilities. This process, using a background task approach, intelligently predicts the remaining useful life (RUL) of critical components (such as the main bearing ID of wind turbine B-07: B-07-MainBearing). The feature extraction module constructs a multi-dimensional real-time feature vector from recent data streams, including vibration frequency (25.1Hz), bearing temperature (75.4℃), load factor (0.91), and average rotational speed (15.1RPM). This vector is then input into a deep neural network model (RUL_MainBearing_v3.1.h5) trained using historical fault data from wind farm B. The model performs inference calculations and outputs a RUL estimate, such as 4500 operating hours. This estimate is then sent to the rendering engine. The color gradient mapping function within the engine compares the RUL value with its total design life (200,000 hours), calculates a normalized health index (0.02256), and maps it to a distinct RGBA color value (1.0, 0.2, 0.0, 1.0), i.e., a striking red. Finally, the custom shader program receives this color parameter and applies it to the surface material of the corresponding main bearing component in the 3D scene. In this way, the system transforms abstract, predictive data about the future into visually perceptible signals for maintenance personnel in 3D space, enabling intuitive and proactive monitoring of asset health degradation trends.

[0083] 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 real-time three-dimensional visualization system for wind power plants, characterized in that, include: Elastic data fusion module: Real-time aggregation of wind turbine SCADA, meteorological station and weather forecast input data fusion cascade architecture; When SCADA data is interrupted, dynamic parameter estimates are calculated step by step using meteorological stations, weather forecasts and data extrapolation algorithms to generate a data stream used to drive the generation of dynamic images of 3D scenes. 3D visualization module: Generates 3D digital terrain using 3D modeling tools, and combines it with oblique photogrammetry images to obtain a basic geographic scene; A procedural 3D modeling engine is used to construct 3D digital assets and combine them with a basic geographic scene to obtain a static digital scene; the data stream is used to drive the 3D digital assets to obtain a dynamic 3D scene; the ray casting method is used to select and obtain the corresponding running parameters, which are then compared by a rule-based judgment engine to trigger a visual alarm mechanism and send a visual prompt. Rendering maintenance module: By analyzing the data stream, output rendering maintenance update prompts; The rendering maintenance update prompts are converted into color values ​​using a color gradient mapping function, and then the color rendering update is performed using a custom shader program.

2. The real-time three-dimensional visualization system for wind power plants according to claim 1, characterized in that, The specific implementation process of the elastic data fusion module includes: receiving and aligning the timestamps of multi-source data from wind turbine SCADA, local meteorological stations, and weather forecasts to form a data foundation; based on the data foundation, continuously verifying the integrity of SCADA data through a data validity detection mechanism; when an anomaly is detected in the SCADA data, triggering a cascading supplementation process, which is executed in a preset order: first, calling the measured data from the local meteorological station from the data foundation for parameter estimation; if the local meteorological station is invalid, calling the weather forecast; if the weather forecast is invalid, using a preset data extrapolation algorithm to extrapolate dynamic parameter estimates; and generating a structured data stream suitable for 3D animation and rendering after standardization processing.

3. The real-time three-dimensional visualization system for wind power plants according to claim 2, characterized in that, The specific implementation process of the data fusion cascade architecture includes: receiving and integrating inputs from multiple heterogeneous data sources, and pre-setting a clear priority order for the heterogeneous data sources; all input data undergoing timestamp alignment and synchronization processing to form a unified data processing benchmark; based on the data processing benchmark, a continuously running verification mechanism monitors the status of the highest priority data source and verifies its integrity, continuity, and quality; upon verification, the highest priority data source is directly selected and standardized; when the verification mechanism determines that the highest priority data source has an anomaly, a cascaded supplementary process is triggered, following the preset priority order to sequentially explore and verify the data of secondary data sources; the verification process is repeated, and when the heterogeneous data source is unavailable, a preset data estimation algorithm is invoked as a data generation method; all generated data is fed into a standardized processing pipeline to generate a data stream with a unified structure suitable for graphics processing pipelines.

4. The real-time three-dimensional visualization system for wind power plants according to claim 1, characterized in that, The specific implementation process of the static digital scene includes: processing commercially purchased digital elevation model data using computer graphics software, generating a three-dimensional digital terrain of the wind farm using a polygon mesh generation algorithm, and performing texture mapping on the image content obtained from oblique photogrammetry to construct a basic geographic scene; reading engineering parameters from preset engineering data, using a procedural 3D modeling engine based on parametric templates, and constructing a three-dimensional digital asset containing the shape of specific equipment based on the engineering parameters, and combining it with the basic geographic scene to form a static digital scene.

5. The real-time three-dimensional visualization system for wind power plants according to claim 1, characterized in that, The specific implementation process of the dynamic 3D scene includes: receiving and parsing the data stream, extracting real-time dynamic parameters, wherein the real-time dynamic parameters include a unique identifier, device operating parameters, and operating status codes; binding the real-time dynamic parameters to the 3D digital asset according to the unique identifier; based on the binding relationship, the device operating parameters are used to drive the geometric position and posture of specific devices in the 3D digital asset in real time, generate dynamic behavior, update the animation of the 3D digital asset, and obtain the dynamic 3D scene; the operating status codes are used to identify the estimated data in the real-time dynamic parameters and trigger the preset visual mode in the dynamic 3D scene.

6. The real-time three-dimensional visualization system for wind power plants according to claim 1, characterized in that, The specific generation process of the visual prompt includes: detecting the user's directional selection operation on the 3D digital asset by performing a ray casting method in the dynamic 3D scene, and obtaining the unique identifier corresponding to the 3D digital asset; retrieving the associated real-time and historical operating parameters from the data stream, transmitting them to the dynamic 3D scene, automatically generating an information label panel for the 3D digital asset, and displaying it in the dynamic 3D scene by overlaying two-dimensional images; the built-in rule determination engine continuously monitors the data stream and continuously compares it with a preset alarm threshold; when the rule determination engine determines that the real-time operating parameter exceeds the alarm threshold, it generates a warning event, triggers the visual alarm mechanism, changes the rendering visualization state of the corresponding 3D digital asset, generates a descriptive visual prompt, and displays the visual prompt on the information label panel.

7. The real-time three-dimensional visualization system for wind power plants according to claim 1, characterized in that, The specific generation process of the rendering maintenance update prompt includes: filtering and extracting operating condition parameters related to the wear of key components from the data stream, and constructing a real-time feature vector; inputting the real-time feature vector into a remaining service life model trained with historical fault data and operating data; the remaining service life model performing model inference calculation to generate a rendering maintenance update prompt for a specific key component, including a unique identifier and a corresponding remaining service life estimate.

8. The real-time three-dimensional visualization system for wind power plants according to claim 1, characterized in that, The specific implementation process of performing shading and rendering updates includes: receiving a rendering maintenance update prompt; using the unique identifier in the rendering maintenance update prompt to retrieve and lock the 3D digital assets that need to be updated visually in the dynamic 3D scene; the remaining lifetime estimate in the rendering maintenance update prompt is input into a preset color gradient mapping function to map the remaining lifetime estimate from its numerical range to continuously changing color values; the color values ​​are passed as parameters to the custom shader program applied to the 3D digital assets; the color value parameters are applied to the surface material properties of the 3D digital assets to change the final output color in the rendering pipeline in real time.

Images