Engine turbine rotor machining method and system based on data processing

Abstract

The present invention provides a method and system for processing an engine turbine rotor based on data processing. The present invention relates to the technical field of turbine rotor processing. The method includes obtaining a turbine rotor operation video under different blade assembly spacings; determining the operation information under each blade assembly spacing using a rotor determination model based on the turbine rotor operation video under different blade assembly spacings; constructing a knowledge graph; processing the knowledge graph based on a graph convolutional network to determine a minimum safe blade assembly spacing and a maximum effective blade assembly spacing; determining a target blade assembly spacing based on the minimum safe blade assembly spacing and the maximum effective blade assembly spacing; and processing the turbine rotor based on the target blade assembly spacing. The method can accurately determine the optimal blade assembly spacing of the engine turbine rotor.

Description

Technical Field

[0001] The present invention relates to the technical field of turbine rotor processing, and in particular to a method and system for processing an engine turbine rotor based on data processing. Background Art

[0002] With the development of high-performance power units such as engines and gas turbines, the design and processing technology of turbine rotors have a decisive impact on the overall performance, reliability and life of the engines. Among them, the optimization of the blade assembly spacing of the engine turbine rotor is particularly critical. Traditional turbine rotor processing methods mainly rely on empirical formulas, finite element simulations or experimental tests to determine the blade assembly spacing. However, these methods have obvious limitations. The strong dependence on experience makes it difficult to adapt to different working conditions. Finite element simulation calculations are costly and time-consuming. Manual physical experimental testing requires not only the manufacture of a large number of samples, but also the construction of a dedicated test bench for long-term operation tests. The entire process is time-consuming, labor-intensive and costly.

[0003] Therefore, how to accurately determine the optimal blade assembly spacing of the engine turbine rotor is an urgent problem to be solved. Summary of the Invention

[0004] The main technical problem solved by the present invention is how to accurately determine the optimal blade assembly spacing of an engine turbine rotor.

[0005] According to a first aspect, the present invention provides an engine turbine rotor processing method based on data processing, comprising: obtaining turbine rotor operation videos under different blade assembly spacings; determining operation information under each blade assembly spacing using a rotor determination model based on the turbine rotor operation videos under different blade assembly spacings; constructing a knowledge graph, wherein the knowledge graph comprises a plurality of blade assembly spacing nodes and a plurality of edges between the plurality of blade assembly spacing nodes, the plurality of blade assembly spacing nodes being connected in ascending order of blade assembly spacings, the node feature of each blade assembly spacing node comprising operation information under each blade assembly spacing, and the feature of each edge being the difference in blade assembly spacings between two blade assembly spacing nodes; processing the knowledge graph based on a graph convolutional network to determine a minimum safe blade assembly spacing and a maximum effective blade assembly spacing; determining a target blade assembly spacing based on the minimum safe blade assembly spacing and the maximum effective blade assembly spacing; and processing the turbine rotor based on the target blade assembly spacing.

[0006] In one possible implementation, determining the target blade assembly spacing based on the minimum safe blade assembly spacing and the maximum effective blade assembly spacing includes: determining a plurality of preliminary blade assembly spacings based on the minimum safe blade assembly spacing, the maximum effective blade assembly spacing, and the operating information at each blade assembly spacing; obtaining turbine rotor operating videos at a plurality of preliminary blade assembly spacings; determining a plurality of preferred blade assembly spacings based on the turbine rotor operating videos at the plurality of preliminary blade assembly spacings; determining a turbine rotor operating simulation video for each preferred blade assembly spacing based on the plurality of preferred blade assembly spacings, the turbine rotor operating videos at different blade assembly spacings, and the turbine rotor operating videos at the plurality of preliminary blade assembly spacings; and determining the target blade assembly spacing based on the turbine rotor operating simulation video for each preferred blade assembly spacing.

[0007] In one possible implementation, the determining of each turbine rotor operation simulation video with a preferred blade assembly spacing based on the multiple preferred blade assembly spacings, the turbine rotor operation videos at different blade assembly spacings, and the turbine rotor operation videos at the multiple preliminary blade assembly spacings includes: determining each turbine rotor operation simulation video with a preferred blade assembly spacing based on the multiple preferred blade assembly spacings, the turbine rotor operation videos at different blade assembly spacings, and the turbine rotor operation videos at the multiple preliminary blade assembly spacings using a diffusion model.

[0008] In a possible implementation, the rotor determination model is a recurrent neural network model.

[0009] According to a second aspect, the present invention provides an engine turbine rotor processing system based on data processing, comprising: an acquisition module for acquiring turbine rotor operation videos under different blade assembly spacings; an information determination module for determining the operation information under each blade assembly spacing using a rotor determination model based on the turbine rotor operation videos under different blade assembly spacings; a construction module for constructing a knowledge graph, wherein the knowledge graph comprises multiple blade assembly spacing nodes and multiple edges between the multiple blade assembly spacing nodes, the multiple blade assembly spacing nodes are connected in order of blade assembly spacing from small to large, the node feature of each blade assembly spacing node comprises the operation information under each blade assembly spacing, and the feature of each edge is the difference in blade assembly spacings between two blade assembly spacing nodes; a range determination module for processing the knowledge graph based on a graph convolutional network to determine the minimum safe blade assembly spacing and the maximum effective blade assembly spacing; a target determination module for determining the target blade assembly spacing based on the minimum safe blade assembly spacing and the maximum effective blade assembly spacing; and a processing module for processing the turbine rotor based on the target blade assembly spacing.

[0010] In one possible implementation, the target determination module is also used to: determine multiple preliminary blade assembly spacings based on the minimum safe blade assembly spacing, the maximum effective blade assembly spacing, and the operating information at each blade assembly spacing; obtain turbine rotor operation videos at multiple preliminary blade assembly spacings; determine multiple preferred blade assembly spacings based on the turbine rotor operation videos at the multiple preliminary blade assembly spacings; determine each preferred blade assembly spacing turbine rotor operation simulation video based on the multiple preferred blade assembly spacings, the turbine rotor operation videos at different blade assembly spacings, and the turbine rotor operation videos at the multiple preliminary blade assembly spacings; and determine the target blade assembly spacing based on each preferred blade assembly spacing turbine rotor operation simulation video.

[0011] In one possible implementation, the target determination module is also used to: use a diffusion model to determine each preferred blade assembly spacing turbine rotor operation simulation video based on the multiple preferred blade assembly spacings, the turbine rotor operation videos under different blade assembly spacings, and the turbine rotor operation videos under the multiple preliminary blade assembly spacings.

[0012] In a possible implementation, the rotor determination model is a recurrent neural network model.

[0013] According to a third aspect, an embodiment of the present invention provides an electronic device, comprising: a processor; a memory; and a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method as described above, the method comprising: obtaining a turbine rotor operation video under different blade assembly spacings; determining the operation information under each blade assembly spacing using a rotor determination model based on the turbine rotor operation video under different blade assembly spacings; constructing a knowledge graph, the knowledge graph comprising a plurality of blade assembly spacing nodes and a plurality of edges between the plurality of blade assembly spacing nodes, the plurality of blade assembly spacing nodes being connected in order of blade assembly spacing from small to large, the node feature of each blade assembly spacing node comprising the operation information under each blade assembly spacing, and the feature of each edge being the difference in blade assembly spacings between two blade assembly spacing nodes; processing the knowledge graph based on a graph convolutional network to determine a minimum safe blade assembly spacing and a maximum effective blade assembly spacing; determining a target blade assembly spacing based on the minimum safe blade assembly spacing and the maximum effective blade assembly spacing; and processing the turbine rotor based on the target blade assembly spacing.

[0014] According to the fourth aspect, this embodiment provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the aforementioned engine turbine rotor processing method based on data processing, the method comprising: obtaining a turbine rotor operation video under different blade assembly spacings; determining the operation information under each blade assembly spacing using a rotor determination model based on the turbine rotor operation video under different blade assembly spacings; constructing a knowledge graph, the knowledge graph comprising a plurality of blade assembly spacing nodes and a plurality of edges between the plurality of blade assembly spacing nodes, the plurality of blade assembly spacing nodes being connected in order of blade assembly spacing from small to large, the node feature of each blade assembly spacing node comprising the operation information under each blade assembly spacing, the feature of each edge being the difference in blade assembly spacings between two blade assembly spacing nodes; processing the knowledge graph based on a graph convolutional network to determine the minimum safe blade assembly spacing and the maximum effective blade assembly spacing; determining a target blade assembly spacing based on the minimum safe blade assembly spacing and the maximum effective blade assembly spacing; and processing the turbine rotor based on the target blade assembly spacing.

[0015] The present invention provides a method and system for processing an engine turbine rotor based on data processing, the method comprising: obtaining turbine rotor operation videos under different blade assembly spacings; determining operation information under each blade assembly spacing using a rotor determination model based on the turbine rotor operation videos under different blade assembly spacings; constructing a knowledge graph, the knowledge graph comprising multiple blade assembly spacing nodes and multiple edges between the multiple blade assembly spacing nodes, the multiple blade assembly spacing nodes being connected in ascending order of blade assembly spacings, the node features of each blade assembly spacing node comprising operation information under each blade assembly spacing, and the feature of each edge being the difference in blade assembly spacings between two blade assembly spacing nodes; processing the knowledge graph based on a graph convolutional network to determine a minimum safe blade assembly spacing and a maximum effective blade assembly spacing; determining a target blade assembly spacing based on the minimum safe blade assembly spacing and the maximum effective blade assembly spacing; and processing the turbine rotor based on the target blade assembly spacing. The method can accurately determine the optimal blade assembly spacing of the engine turbine rotor. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic flow chart of a method for machining an engine turbine rotor based on data processing provided by an embodiment of the present invention;

[0017] Figure 2 A schematic diagram of a process for determining a target blade assembly spacing provided by an embodiment of the present invention;

[0018] Figure 3A schematic diagram of an engine turbine rotor processing system based on data processing provided by an embodiment of the present invention; DETAILED DESCRIPTION

[0019] The present invention will be further described in detail below by means of specific embodiments in conjunction with the accompanying drawings. Similar elements in different embodiments are numbered with associated similar elements. In the following embodiments, many detailed descriptions are provided to enable the present invention to be better understood. However, those skilled in the art will readily appreciate that some of the features may be omitted under different circumstances, or may be replaced by other elements, materials, or methods. In some cases, some operations related to the present invention are not shown or described in the specification. This is to avoid the core of the present invention being overwhelmed by excessive descriptions, and for those skilled in the art, it is not necessary to describe these related operations in detail. They can fully understand the related operations based on the description in the specification and the general technical knowledge in the art.

[0020] In an embodiment of the present invention, there is provided Figure 1 A method for machining an engine turbine rotor based on data processing is shown, and the method for machining an engine turbine rotor based on data processing includes steps S1 to S6:

[0021] Step S1, obtaining turbine rotor operation videos under different blade assembly spacings.

[0022] The blade assembly spacing is the spatial distance between adjacent blades in the engine turbine rotor. The blade assembly spacing is a key parameter affecting the performance of the turbine rotor. The size of this spacing is directly related to the mechanical strength and operating stability of the turbine.

[0023] Turbine rotor operation videos with different blade spacings are captured with a high-speed camera at high frame rates while the turbine rotor operates under the same speed, load, and airflow conditions at different blade spacings (e.g., d1, d2, d3, etc.). These videos capture the turbine rotor's operating status, including blade trajectory, vibration, and overall operating smoothness.

[0024] Step S2: determining the operation information at each blade assembly pitch using a rotor determination model based on the turbine rotor operation videos at different blade assembly pitches.

[0025] The rotor determination model is a recurrent neural network model, the input of the rotor determination model is the turbine rotor operation video under the different blade assembly spacings, and the output of the rotor determination model is the operation information under each blade assembly spacing.

[0026] Recurrent neural network models include recurrent neural networks (RNNs). RNNs are a type of artificial neural network that takes sequential data as input and can recursively repeat itself in the direction of the sequence's evolution, with all recurrent unit nodes connected in a chain-like fashion to form a closed loop. RNNs are capable of processing data with time series characteristics by memorizing previous input information and then analyzing and processing the current input, thereby capturing temporal dependencies in the data.

[0027] The operational information for each blade spacing is a multi-dimensional, dynamic collection of status data and performance indicators generated during the operation of the turbine rotor at a specific blade spacing. This information includes conventional parameters such as blade speed, vibration frequency, and stress distribution. It also covers the blade vibration displacement sequence and the dynamic variation trajectory of blade spacing at this blade spacing. This information can depict the operating status of the turbine rotor from multiple perspectives, including mechanical motion and structural changes.

[0028] The blade vibration displacement sequence is the spatial displacement trajectory of the blade changing with time during operation, which can intuitively show the vibration amplitude and frequency characteristics of the blade.

[0029] The dynamic change trajectory of blade spacing records the real-time fluctuation of the spacing between adjacent blades in the engine turbine rotor under high-speed rotation and airflow impact. The dynamic change trajectory of blade spacing can be used to evaluate the interference risk between blades.

[0030] Turbine rotor operation videos record the dynamic changes of the turbine rotor during operation in the form of continuous frames. These changes contain rich physical information. Each frame reflects the spatial position and posture of the blade at a specific moment, as well as its relative relationship with surrounding components. Combining multiple frames together forms a complete sequence of blade motion and vibration changes. By analyzing this visual image data, operational characteristics such as blade speed fluctuations and vibration amplitude can be captured, and operational information corresponding to the blade assembly spacing can be extracted.

[0031] The recurrent neural network model has powerful time-series data processing capabilities. Turbine rotor operation videos are essentially chronological image sequences, and recurrent neural networks can transfer the calculation results of the previous moment to the next moment through the connections between recurrent units, thereby realizing the memory and utilization of historical information. When processing turbine rotor operation video data, the recurrent neural network can continuously combine the features extracted from the previous frame with the image information of the current frame, and continuously update its understanding of the turbine rotor operation status, thereby excavating the complex patterns of time-varying changes in the video, and ultimately accurately determining the corresponding operation information for each blade assembly spacing.

[0032] Step S3, constructing a knowledge graph, wherein the knowledge graph includes multiple blade assembly spacing nodes and multiple edges between the multiple blade assembly spacing nodes, the multiple blade assembly spacing nodes are connected from small to large according to the blade assembly spacing, the node feature of each blade assembly spacing node includes the operation information under each blade assembly spacing, and the feature of each edge is the difference between the blade assembly spacings of two blade assembly spacing nodes.

[0033] A knowledge graph is a data structure that can be used to represent relationships between entities. It is composed of nodes (vertices) and edges (edges). Knowledge graphs use nodes and edges to model entities and their relationships. They can be used to store and represent complex domain knowledge and support logical reasoning and data mining.

[0034] In some embodiments, the knowledge graph takes blade assembly spacing as an entity and as a node, the node represents the value of the blade assembly spacing, the node feature is the operating information under each blade assembly spacing, the edge represents the difference relationship under each blade assembly spacing, and the edge feature is the difference in the blade assembly spacing of two blade assembly spacing nodes.

[0035] The knowledge graph can be used to structure the discrete blade assembly spacing data and its operating information, and can intuitively present the relationship between different blade assembly spacings and rotor operating information, thereby providing a reasonable knowledge network for subsequent analysis.

[0036] Step S4: Process the knowledge graph based on a graph convolutional network to determine the minimum safe blade assembly spacing and the maximum effective blade assembly spacing.

[0037] A graph convolutional network (GCN) is a deep learning model that can be used to process knowledge graph data. It learns representations of knowledge graphs through message passing between nodes and aggregating feature information from adjacent nodes. In some embodiments, the GCN can iteratively update node embedding vectors to capture the underlying relationship patterns between blade assembly spacing nodes.

[0038] The minimum safe blade spacing is the minimum allowable spacing between adjacent blades while ensuring safe operation of the turbine rotor, as output by a graph convolutional network. If the spacing is below the minimum safe blade spacing, blades in the turbine rotor may collide due to factors such as vibration and thermal expansion, leading to mechanical failure. This minimum safe blade spacing serves as a lower-limit constraint for turbine rotor design, ensuring reliable and durable operation of the engine's turbine rotor.

[0039] The maximum effective blade spacing, output by the graph convolutional network, is the maximum allowable spacing between adjacent blades while maintaining the turbine rotor's aerodynamic efficiency. This spacing ensures the engine's energy conversion efficiency. Exceeding this spacing increases airflow leakage, significantly reducing the engine's turbine efficiency.

[0040] Through the synergistic effect of nodes, node features, and edge features, the knowledge graph can provide systematic data support for determining the minimum safe blade assembly spacing and the maximum effective blade assembly spacing. Its nodes serve as the core carriers for storing different blade assembly spacings, and the nodes convert the continuous parameter space into computable entity units. The node features give each blade assembly spacing node rich physical connotations, and by recording operating information, the abstract spacing values are deeply bound to the actual operating status. The edge features quantify the relationship between nodes, and they use the spacing difference as a link to construct a gradient network of parameter changes, thereby making the performance differences between adjacent blade assembly spacings explicit.

[0041] Graph convolutional networks (GCNs) possess powerful feature learning capabilities and a knowledge graph processing mechanism. The knowledge graph stores different blade assembly spacings and their corresponding operating information, as well as the differential relationships between different blade assembly spacings, in the form of nodes and edges, thereby forming a complex turbine rotor performance correlation network. GCNs, through a message-passing mechanism, aggregate feature information from adjacent nodes along the edges of the knowledge graph and, during the iterative update process, capture the nonlinear relationship between blade assembly spacing changes and turbine rotor operating performance. For example, by analyzing different blade assembly spacings, it is possible to analyze the mutation patterns of blade vibration amplitude and stress distribution when the spacing decreases to a certain range, thereby identifying the minimum safe blade assembly spacing. By evaluating the changing trends in turbine efficiency under different blade assembly spacings, GCNs can identify the threshold at which efficiency begins to significantly decline, thereby determining the maximum effective blade assembly spacing.

[0042] Step S5: determining a target blade assembly spacing based on the minimum safe blade assembly spacing and the maximum effective blade assembly spacing.

[0043] In some embodiments, Figure 2 A schematic diagram of a process for determining a target blade assembly spacing is provided in an embodiment of the present invention. The process for determining the target blade assembly spacing includes steps S21 to S25:

[0044] Step S21, determining a plurality of preliminary blade assembly spacings based on the minimum safe blade assembly spacing, the maximum effective blade assembly spacing, and the operation information at each blade assembly spacing;

[0045] In some embodiments, a plurality of preliminary blade assembly spacings can be determined based on the minimum safe blade assembly spacing, the maximum effective blade assembly spacing, and the operating information at each blade assembly spacing through a deep neural network model. The input of the deep neural network model is the minimum safe blade assembly spacing, the maximum effective blade assembly spacing, and the operating information at each blade assembly spacing, and the output of the deep neural network model is a plurality of preliminary blade assembly spacings.

[0046] Deep neural network models are built on the Deep Neural Networks (DNN) architecture. DNNs are a type of machine learning model based on artificial neural network architecture. They automatically extract data features and learn complex patterns by constructing a network structure consisting of multiple hidden layers. Unlike shallow neural networks, DNNs are able to mine deep, abstract feature relationships in data through layer-by-layer nonlinear transformations.

[0047] Multiple preliminary blade assembly spacings are a set of candidate blade assembly spacing values output by integrating multi-dimensional constraints such as safety and effectiveness through a deep neural network model, while ensuring the safety of turbine rotor operation and compliance with performance standards.

[0048] The deep neural network model can determine multiple preliminary blade spacings due to its powerful feature learning and multi-objective optimization capabilities. Deep neural networks can fuse and transform input information from different dimensions and, through multi-layer neural computation, automatically extract key features valuable for blade spacing decisions. Then, by designing a multi-objective loss function that incorporates safety constraints (such as vibration thresholds) and effectiveness constraints (such as efficiency thresholds), the model, during training, searches for solutions that meet multiple performance objectives within the parameter space defined by the minimum safe blade spacing and the maximum effective blade spacing. This output then generates multiple preliminary blade spacings that meet both safe operation requirements and guaranteed equipment performance.

[0049] Step S22, obtaining turbine rotor operation videos at multiple pre-selected blade assembly spacings;

[0050] The turbine rotor operation video at the preliminary blade assembly spacing is a high-frame rate video obtained by shooting the operation process of the turbine rotor at the preliminary blade assembly spacing under uniform working conditions with a high-speed camera.

[0051] Step S23, determining a plurality of preferred blade assembly spacings based on the turbine rotor operation videos under the plurality of preliminarily selected blade assembly spacings;

[0052] In some embodiments, a plurality of preferred blade assembly spacings can be determined based on the turbine rotor operation videos under the plurality of preliminary blade assembly spacings using a preferred spacing determination model, wherein the preferred spacing determination model is a Transformer model, the input of the preferred spacing determination model is the turbine rotor operation videos under the plurality of preliminary blade assembly spacings, and the output of the preferred spacing determination model is a plurality of preferred blade assembly spacings.

[0053] The Transformer model is a deep learning model based on an attention mechanism. It uses multi-head attention to enable parallel processing and global correlation analysis of information at different positions in the input sequence. When processing time series data, the Transformer model uses self-attention to focus on keyframe features, effectively extracting dynamic patterns and contextual information from the video.

[0054] The optimal blade spacing is a set of blade spacings that better meets actual project requirements, selected through a selection model based on multiple preliminary blade spacings. Compared to the preliminary blade spacing, the optimal spacing further comprehensively considers the turbine's operating performance under real-world operating conditions and is an optimized result that balances equipment safety, stability, and efficiency.

[0055] Turbine rotor operation videos contain dynamic information such as blade motion trajectories and vibration changes. Essentially, they are image sequence data with time-series characteristics. The Transformer model, through its multi-head attention mechanism, can simultaneously focus on features at different moments and regions in the video, capturing key information such as subtle changes in blades under complex operating conditions, such as sudden changes in vibration frequency and dynamic fluctuations in blade spacing. Furthermore, the Transformer model can perform a global comparative analysis of turbine rotor operation video data at multiple pre-selected blade assembly spacings. It then identifies differences and patterns in turbine rotor operating performance at each pre-selected blade assembly spacing. This allows the model to select the spacing combination with the best overall performance from the pre-selected spacings while meeting safety and efficiency standards, and output multiple preferred blade assembly spacings.

[0056] In some embodiments, the preferred spacing determination model includes a time series analysis layer, an index calculation layer, and a preferred spacing determination layer. The input of the time series analysis layer is a turbine rotor operation video at multiple preliminary blade assembly spacings, and the output of the time series analysis layer is a blade vibration trajectory sequence and a rotor speed change sequence corresponding to each blade assembly spacing; the input of the index calculation layer is a blade vibration trajectory sequence and a rotor speed change sequence corresponding to each blade assembly spacing, and the output of the index calculation layer is blade vibration amplitude information, blade average vibration frequency, and rotor speed fluctuation data at each blade assembly spacing; the input of the preferred spacing determination layer is blade vibration amplitude information, blade average vibration frequency, and rotor speed fluctuation data at each blade assembly spacing, and the output of the preferred spacing determination layer is multiple preferred blade assembly spacings.

[0057] The vibration trajectory sequence is formed by analyzing the turbine rotor operation video frame by frame, extracting the motion coordinates of the key points of the blade edge in space, and arranging them in chronological order to form continuous trajectory data, which reflects the spatial path and time evolution process of the blade vibration.

[0058] The rotor speed change sequence is a sequence of data formed by real-time monitoring of the periodic motion of the rotor mark points in the video to record the speed values at different time points. It is used to characterize the dynamic changes of the rotor speed over time.

[0059] Blade vibration amplitude information is a characteristic parameter of blade vibration, including maximum excursion distance, vibration direction, and temporal variation trend. Blade vibration amplitude information is used to assess blade stability and fatigue risk.

[0060] The rotor speed fluctuation data is an indicator of rotor speed stability, which includes the speed fluctuation range and frequency components per unit time to reflect the rotor operation smoothness and potential fault hazards.

[0061] Different layers undertake differentiated data processing tasks. The timing analysis layer focuses on extracting the time series data of blade vibration trajectory and rotor speed changes frame by frame from the turbine rotor operation video, ensuring the complete recording of dynamic information of the operation process; the indicator calculation layer converts the original video information into quantifiable physical parameters; the optimal spacing determination layer conducts a comprehensive screening of multi-dimensional indicators, and finally outputs the optimal blade assembly spacing that meets stability and performance requirements. Through this layered architecture, each link can focus on optimizing specific functions. The timing analysis layer ensures data integrity, the indicator calculation layer realizes feature quantification, and the optimal decision-making layer completes multi-objective screening, thereby improving the efficiency and accuracy of the overall processing, while enhancing the interpretability and modularity of the model.

[0062] Step S24, determining a turbine rotor operation simulation video for each preferred blade assembly spacing based on the multiple preferred blade assembly spacings, the turbine rotor operation videos at different blade assembly spacings, and the turbine rotor operation videos at the multiple pre-selected blade assembly spacings;

[0063] In some embodiments, a diffusion model may be used to determine each preferred blade assembly spacing turbine rotor operation simulation video based on the multiple preferred blade assembly spacings, the turbine rotor operation videos at different blade assembly spacings, and the turbine rotor operation videos at the multiple preliminary blade assembly spacings.

[0064] The Diffusion Model is a generative deep learning model inspired by the physical diffusion process. Its core concept is to generate high-quality samples by learning data distribution through the process of gradually adding (forward diffusion) and removing (backward diffusion) noise. In the forward diffusion phase, the diffusion model gradually adds Gaussian noise to the original data (such as images or videos) until the data degenerates into pure noise. In the backward diffusion phase, a neural network performs reverse predictions and removes noise, gradually restoring the clear target data. Diffusion models excel at capturing detailed features and complex distributions in data, and are particularly powerful in video generation. Diffusion models can generate realistic content that meets specific requirements.

[0065] Each turbine rotor operation simulation video for the optimal blade assembly spacing is a virtual video generated based on the diffusion model to simulate the operating state of the turbine rotor at each optimal blade assembly spacing. The turbine rotor operation simulation video uses real-life turbine rotor operation videos at different blade assembly spacings and turbine rotor operation videos at the preliminary blade assembly spacing as training data, and is then generated through the reverse diffusion process of the diffusion model in combination with the optimal blade assembly spacing parameters. The turbine rotor operation simulation video content includes dynamic operation images of the turbine rotor at the theoretically calculated optimal blade assembly spacing, such as blade vibration amplitude, mechanical stress distribution and other details, and can intuitively present the expected operating state of the turbine at this spacing. The turbine rotor operation simulation video can be used to verify the effect of the optimal blade assembly spacing.

[0066] The diffusion model can be trained through forward diffusion of real turbine rotor operation videos shot at different blade spacings. This allows it to learn the dynamic characteristic distribution patterns of complex physical phenomena such as blade vibration and mechanical stress distribution during turbine rotor operation as blade spacing is adjusted. In the reverse diffusion stage, the model uses a specific optimal blade spacing as a conditional constraint and combines it with the distribution patterns of turbine rotor operation video data to gradually remove noise from random noise to reconstruct the turbine rotor operation image under the optimal blade spacing condition. The advantage of the diffusion model lies in its ability to capture the subtle motion patterns and physical connections in real turbine rotor operation videos and use a data-driven approach to generate turbine rotor operation simulation videos that better reflect actual operating conditions.

[0067] In some embodiments, a turbine rotor operation simulation video with each preferred blade assembly spacing can also be determined based on the multiple preferred blade assembly spacings, the turbine rotor operation videos at different blade assembly spacings, and the turbine rotor operation videos at the multiple preliminary blade assembly spacings by using a generative adversarial network.

[0068] A Generative Adversarial Network (GAN) is a deep learning framework consisting of a generator and a discriminator, which generate data through an adversarial training mechanism. The generator attempts to generate realistic samples from random noise, while the discriminator distinguishes between generated samples and real samples. The two continuously improve their capabilities in a competitive process, ultimately enabling the generator to produce synthetic content that is indistinguishable from real data. The generator learns features such as blade vibration frequency from real videos, conditional on the optimal blade assembly spacing, and attempts to generate operating images at that blade assembly spacing. The discriminator, by comparing the real videos with the generated videos, forces the generator to continuously optimize details, ultimately producing a simulated video of the turbine rotor operation that is both physically consistent and close to actual operating conditions.

[0069] Step S25 , determining a target blade assembly spacing based on the turbine rotor operation simulation video of each preferred blade assembly spacing.

[0070] In some embodiments, a target spacing determination model can be used to determine the target blade assembly spacing based on each of the turbine rotor operation simulation videos with the preferred blade assembly spacing. The target spacing determination model is a Transformer model, the input of the target spacing determination model is each of the turbine rotor operation simulation videos with the preferred blade assembly spacing, and the output of the target spacing determination model is the target blade assembly spacing.

[0071] The target blade spacing is the optimal turbine rotor blade spacing parameter output by the target spacing determination model. This parameter enables the turbine rotor to achieve optimal performance during operation, such as minimum vibration amplitude and lowest mechanical stress.

[0072] Each turbine rotor simulation video for an optimal blade spacing captures the full dynamic characteristics of the turbine rotor at that specific optimal spacing, including key information such as blade vibration frequency and mechanical stress evolution. These videos accurately replicate the theoretical operating state of the turbine at different spacings, enabling a quantitative analysis of the direct correlation between optimal blade spacing parameters and performance indicators. By comparing and analyzing simulation videos for multiple optimal spacings, the model systematically evaluates turbine performance under various parameters, ultimately selecting the target blade spacing with the best overall performance.

[0073] The Transformer model efficiently extracts key information from simulated videos through temporal feature encoding and global cross-video comparison. Its self-attention mechanism can focus on the temporal evolution of blade motion within a single video segment, such as the transition from stable to sudden vibration at a given spacing. Furthermore, the Transformer model can compare feature differences across multiple videos in parallel to quantitatively evaluate the performance of each optimal spacing. Through end-to-end learning, the Transformer model directly extracts deep features related to target optimization metrics (such as vibration threshold and energy efficiency ratio) from video sequences, outputting the target blade assembly spacing with optimal overall performance, thereby avoiding the subjective bias and inefficiency of manual analysis.

[0074] Step S6: Processing the turbine rotor based on the target blade assembly spacing.

[0075] When the target blade assembly spacing is determined, the engine turbine rotor is processed and installed according to the target blade assembly spacing.

[0076] Based on the same inventive concept, Figure 3 A schematic diagram of an engine turbine rotor processing system based on data processing provided in an embodiment of the present invention, wherein the engine turbine rotor processing system based on data processing includes:

[0077] An acquisition module 31 is used to acquire turbine rotor operation videos at different blade assembly spacings;

[0078] An information determination module 32 is configured to determine operation information at each blade assembly spacing using a rotor determination model based on the turbine rotor operation videos at different blade assembly spacings;

[0079] A construction module 33 is configured to construct a knowledge graph, wherein the knowledge graph includes a plurality of blade assembly spacing nodes and a plurality of edges between the plurality of blade assembly spacing nodes, wherein the plurality of blade assembly spacing nodes are connected in ascending order of blade assembly spacing, wherein a node feature of each blade assembly spacing node includes operation information at each blade assembly spacing, and wherein a feature of each edge is a difference in blade assembly spacing between two blade assembly spacing nodes;

[0080] a range determination module 34, configured to process the knowledge graph based on a graph convolutional network to determine a minimum safe blade assembly spacing and a maximum effective blade assembly spacing;

[0081] a target determination module 35, configured to determine a target blade assembly spacing based on the minimum safe blade assembly spacing and the maximum effective blade assembly spacing;

[0082] The processing module 36 is configured to process the turbine rotor based on the target blade assembly spacing.

[0083] In addition, unless expressly stated in the claims, the order of the processing elements and sequences, the use of alphanumeric characters, or the use of other names described in this specification are not intended to limit the order of the processes and methods of this specification. Although the above disclosure discusses some of the invention embodiments currently considered useful through various examples, it should be understood that such details are for illustrative purposes only, and the appended claims are not limited to the disclosed embodiments. On the contrary, the claims are intended to cover all modifications and equivalent combinations that are consistent with the spirit and scope of the embodiments of this specification. For example, although the system components described above can be implemented by hardware devices, they can also be implemented only by software solutions, such as installing the described system on an existing server or mobile device.

[0084] Similarly, it should be noted that, in order to simplify the presentation of this specification and thus facilitate understanding of one or more embodiments of the invention, the foregoing descriptions of the embodiments of this specification sometimes combine multiple features into a single embodiment, figure, or description thereof. However, this disclosure method does not imply that the subject matter of this specification requires more features than those recited in the claims. In fact, an embodiment may have fewer features than all of the features of a single disclosed embodiment.

[0085] Finally, it should be understood that the embodiments described in this specification are intended only to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, by way of example and not limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly described and illustrated in this specification.

Claims

1. A method for processing an engine turbine rotor based on data processing, characterized in that: include: Obtain turbine rotor operation videos with different blade assembly spacings; Determine the operation information at each blade assembly spacing using a rotor determination model based on the turbine rotor operation video at different blade assembly spacings; Constructing a knowledge graph, the knowledge graph including a plurality of blade assembly spacing nodes and a plurality of edges between the plurality of blade assembly spacing nodes, the plurality of blade assembly spacing nodes being connected in ascending order of blade assembly spacing, a node feature of each blade assembly spacing node including operation information at each blade assembly spacing, and a feature of each edge being a difference in blade assembly spacing between two blade assembly spacing nodes; Processing the knowledge graph based on a graph convolutional network to determine a minimum safe blade assembly spacing and a maximum effective blade assembly spacing; Determining a target blade assembly spacing based on the minimum safe blade assembly spacing and the maximum effective blade assembly spacing; The turbine rotor is machined based on the target blade assembly spacing.

2. The engine turbine rotor processing method based on data processing according to claim 1, characterized in that: Determining the target blade assembly spacing based on the minimum safe blade assembly spacing and the maximum effective blade assembly spacing includes: Determine a plurality of preliminary blade assembly spacings based on the minimum safe blade assembly spacing, the maximum effective blade assembly spacing, and the operation information under each blade assembly spacing; Obtaining videos of turbine rotor operation at multiple preliminary blade assembly pitches; determining a plurality of preferred blade assembly spacings based on the turbine rotor operation videos at the plurality of preliminary blade assembly spacings; Determine a turbine rotor operation simulation video for each preferred blade assembly spacing based on the multiple preferred blade assembly spacings, the turbine rotor operation videos at different blade assembly spacings, and the turbine rotor operation videos at the multiple preliminary blade assembly spacings; A target blade assembly pitch is determined based on the turbine rotor operation simulation video of each preferred blade assembly pitch.

3. The engine turbine rotor machining method based on data processing according to claim 2, characterized in that: Determining a turbine rotor operation simulation video with each preferred blade assembly spacing based on the multiple preferred blade assembly spacings, the turbine rotor operation videos at different blade assembly spacings, and the turbine rotor operation videos at the multiple pre-selected blade assembly spacings includes: Based on the multiple preferred blade assembly spacings, the turbine rotor operation videos at different blade assembly spacings, and the turbine rotor operation videos at the multiple preliminary blade assembly spacings, a diffusion model is used to determine each preferred blade assembly spacing turbine rotor operation simulation video.

4. The engine turbine rotor processing method based on data processing according to claim 1, characterized in that: The rotor determination model is a recurrent neural network model.

5. An engine turbine rotor processing system based on data processing, characterized in that: include: An acquisition module is used to acquire turbine rotor operation videos under different blade assembly spacings; An information determination module, configured to determine operation information at each blade assembly spacing using a rotor determination model based on the turbine rotor operation videos at different blade assembly spacings; a construction module for constructing a knowledge graph, wherein the knowledge graph includes a plurality of blade assembly spacing nodes and a plurality of edges between the plurality of blade assembly spacing nodes, wherein the plurality of blade assembly spacing nodes are connected in ascending order of blade assembly spacing, a node feature of each blade assembly spacing node includes operation information at each blade assembly spacing, and a feature of each edge is a difference in blade assembly spacing between two blade assembly spacing nodes; a range determination module, configured to process the knowledge graph based on a graph convolutional network to determine a minimum safe blade assembly spacing and a maximum effective blade assembly spacing; a target determination module, configured to determine a target blade assembly spacing based on the minimum safe blade assembly spacing and the maximum effective blade assembly spacing; A processing module is used to process the turbine rotor based on the target blade assembly spacing.

6. The engine turbine rotor machining system based on data processing according to claim 5, characterized in that: The target determination module is further configured to: Determine a plurality of preliminary blade assembly spacings based on the minimum safe blade assembly spacing, the maximum effective blade assembly spacing, and the operation information under each blade assembly spacing; Obtaining videos of turbine rotor operation at multiple preliminary blade assembly pitches; determining a plurality of preferred blade assembly spacings based on the turbine rotor operation videos at the plurality of preliminary blade assembly spacings; Determine a turbine rotor operation simulation video for each preferred blade assembly spacing based on the multiple preferred blade assembly spacings, the turbine rotor operation videos at different blade assembly spacings, and the turbine rotor operation videos at the multiple preliminary blade assembly spacings; A target blade assembly pitch is determined based on the turbine rotor operation simulation video of each preferred blade assembly pitch.

7. The engine turbine rotor processing system based on data processing according to claim 6, characterized in that: The target determination module is further configured to: Based on the multiple preferred blade assembly spacings, the turbine rotor operation videos at different blade assembly spacings, and the turbine rotor operation videos at the multiple preliminary blade assembly spacings, a diffusion model is used to determine each preferred blade assembly spacing turbine rotor operation simulation video.

8. The engine turbine rotor machining system based on data processing according to claim 5, characterized in that: The rotor determination model is a recurrent neural network model.

9. An electronic device, characterized in that: include: processor; Memory; and a computer program; wherein the computer program is stored in the memory and is configured to be executed by the processor to implement the engine turbine rotor processing method based on data processing according to any one of claims 1 to 4.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the engine turbine rotor machining method based on data processing as claimed in any one of claims 1 to 4 is implemented.

Images