A surface morphology evolution prediction method and system based on multi-source surface information and time condition modulation, a terminal and a storage medium

Abstract

The application relates to the technical field of data processing, and discloses a surface morphology evolution prediction method and system based on multi-source surface information and time condition modulation, a terminal and a storage medium.The method comprises the following steps: acquiring surface morphology data, performing data conversion and multi-modal fusion on the surface morphology data to obtain model input features; the model input features are input into a time modulation mechanism network of a conditional surface generation model for training to obtain a target prediction model; target time conditions and a to-be-measured surface are acquired; the target prediction model is used to predict the to-be-measured surface according to the target time conditions to obtain a predicted surface increment field; and the predicted surface increment field and the reference surface are updated in increments to obtain a surface morphology prediction result.The application is based on surface morphology data conversion and multi-modal fusion, a conditional surface generation model is trained, and a to-be-measured surface is predicted, so that the surface morphology can be accurately predicted.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method, system, terminal, and computer-readable storage medium for predicting surface morphology evolution based on multi-source surface information and temporal condition modulation. Background Technology

[0002] Sealing structures are widely used in nuclear power equipment, aero-engines, aerospace equipment, and high-end mechanical systems. To improve the wear resistance, corrosion resistance, sealing performance, and service reliability of the sealing interface, functional coatings are typically applied to the sealing surfaces. Under actual operating conditions, the sealing interface is often affected by vibration, thermal deformation, cyclic loads, or minute displacements. Even if it is nominally a static seal, fretting wear and surface degradation may occur in the contact area, leading to a decrease in sealing performance. Existing research mainly focuses on wear mechanism analysis, surface inspection, or single-moment characterization, lacking predictive methods for the time-varying evolution of surface morphology. Traditional empirical models and numerical simulation methods have limited ability to express complex surface textures, multi-scale statistical characteristics, and long-term evolution patterns; furthermore, some solutions depend on specific materials or specific operating conditions, lacking versatility and making it difficult to extend to multiple types of sealing coatings.

[0003] Traditional methods for predicting the surface morphology of sealing coatings (such as empirical models or numerical simulations) only focus on geometric reconstruction errors, lack joint constraints on surface height distribution, spectral characteristics, and multi-scale consistency, and rely on specific materials or operating conditions. Under complex service conditions such as fretting or contact wear, they are unable to accurately predict the evolution of the surface morphology of various sealing coatings over time, which is a problem that urgently needs to be solved.

[0004] Therefore, existing technologies still need to be improved and developed. Summary of the Invention

[0005] The main objective of this invention is to provide a method, system, terminal, and computer-readable storage medium for predicting surface morphology evolution based on multi-source surface information and time-condition modulation. This aims to solve the problem that existing technologies struggle to accurately predict the evolution of the surface morphology of various sealing coatings over time under complex service conditions such as micro-motion or contact wear, resulting in the inability to accurately predict the surface morphology of sealing coatings.

[0006] To achieve the above objectives, the present invention provides a surface morphology evolution prediction method based on multi-source surface information and temporal condition modulation. The surface morphology evolution prediction method based on multi-source surface information and temporal condition modulation includes the following steps: Surface topography data at different service times are acquired, and the surface topography data are transformed to obtain a reference surface, mask information, random noise and time interval. Multimodal fusion is performed on the reference surface, the mask information, the random noise and the time interval based on a network mapping function to obtain model input features. The input features of the model are fed into the temporal modulation mechanism network of the conditional surface generation model for training to obtain the target prediction model; The test surface and target time conditions are obtained. The test surface is predicted based on the target time conditions using the target prediction model to obtain the predicted surface increment field. The predicted surface increment field and the reference surface are incrementally updated to obtain the surface morphology prediction result.

[0007] Optionally, the surface morphology evolution prediction method based on multi-source surface information and time-conditional modulation, wherein acquiring surface morphology data at different service times, performing data transformation on the surface morphology data to obtain a reference surface, mask information, random noise, and time interval, and performing multi-modal fusion on the reference surface, the mask information, random noise, and the time interval based on a network mapping function to obtain model input features, specifically includes: Surface topography data at different service times are acquired, the surface topography data is extracted to obtain structured data, and the structured data is transformed to obtain a global effective height set; The global effective height set is logarithmically compressed and mapped to obtain the reference surface, mask information, random noise, and time interval. Based on the network mapping function, the reference surface, the mask information, the random noise, and the time interval are fused in a multimodal manner to obtain the model input features.

[0008] Optionally, the surface morphology evolution prediction method based on multi-source surface information and temporal condition modulation, wherein acquiring surface morphology data at different service times, extracting structured data from the surface morphology data, and transforming the structured data to obtain a globally effective height set, specifically includes: Surface topography data at different service times are acquired, the surface topography data is extracted to obtain structured data, and the structured data is identified to obtain effective pixels; Obtain the quantile function, and calculate the global height range of the effective pixels based on the quantile function to obtain the global effective height set; The structured data includes surface height file path data, grayscale image path data, PSD file path data, HPD file path data, and data integrity flag data.

[0009] Optionally, in the surface morphology evolution prediction method based on multi-source surface information and temporal condition modulation, the temporal modulation mechanism network includes an encoder layer, a bottleneck layer, a decoder layer, and a skip connection layer. The step of inputting the model input features into a temporal modulation mechanism network of a conditional surface generation model for training to obtain a target prediction model specifically includes: The reference surface, mask information, random noise, and time interval of the model input features are input into the encoder layer, bottleneck layer, decoder layer, and skip connection layer of the conditional surface generation model for modulation to obtain the modulation result: ; in, Indicates the modulation result. Represents the input features of the model. This represents element-wise multiplication. This indicates the mapping result at the first moment. This represents the result of the second time mapping; Multiple weight coefficients are obtained, and the conditional surface generation model is trained based on the modulation result and all the weight coefficients according to the multi-objective joint loss function to obtain the target prediction model.

[0010] Optionally, in the surface morphology evolution prediction method based on multi-source surface information and temporal condition modulation, the multi-objective joint loss function includes distribution difference loss and difference loss; The process of obtaining multiple weight coefficients, training the conditional surface generation model based on the modulation result and all the weight coefficients using a multi-objective joint loss function to obtain a target prediction model specifically includes: Obtain the surface height matrix and multiple weight coefficients at the actual target time. Based on the distribution difference loss and difference loss, train the conditional surface generation model according to the surface height matrix at the actual target time, the modulation result, and all the weight coefficients to obtain the target prediction model. ; in, Represents the total loss function. The weighting coefficients represent the height error of the mask region. This represents the surface height matrix at the actual target time. This represents element-wise multiplication. This represents the effective region mask matrix corresponding to the source surface. The weight coefficients representing the gradient consistency constraints. Represents the spatial gradient operator. Denotes the L1 norm of a matrix. The weighting coefficients represent the frequency domain PSD distribution error. The weighting coefficients representing the height distribution error. Indicates the loss due to distributional differences. Indicates difference loss, This indicates the predicted surface morphology.

[0011] Optionally, the surface morphology evolution prediction method based on multi-source surface information and time condition modulation, wherein the steps of acquiring the surface to be measured and the target time condition, predicting the surface to be measured using the target prediction model according to the target time condition to obtain a predicted surface increment field, and incrementally updating the predicted surface increment field and the reference surface to obtain the surface morphology prediction result, specifically include: The surface to be tested is acquired, and information is extracted from the surface to be tested using the target prediction model to obtain the source surface, target surface, PSD information, and HPD information. Obtain the target time conditions, and use the target prediction model to predict the source surface, the target surface, the PSD information, and the HPD information based on the target time conditions to obtain the predicted surface increment field. Incrementally update the predicted surface increment field and the reference surface to obtain the surface morphology prediction result. ; in, Indicates the reference surface. This indicates the predicted surface increment field.

[0012] Optionally, the surface morphology evolution prediction method based on multi-source surface information and time condition modulation, wherein the steps of acquiring the surface to be measured and the target time condition, predicting the surface to be measured according to the target time condition using the target prediction model to obtain a predicted surface increment field, incrementally updating the predicted surface increment field and the reference surface to obtain the surface morphology prediction result, further include: Obtain the actual measurement surface, compare the morphology of the actual measurement surface with that of the predicted surface, and calculate the error result. If the error result is a preset error value, then the surface height matrix corresponding to the predicted surface morphology is normalized and mapped according to the global fixed scale and the single map adaptive scale to obtain the prediction map visualization result. The error results include mean absolute error, mean square error, logarithmic error of PSD spectrum, and CDF distribution error.

[0013] Furthermore, to achieve the above objectives, the present invention also provides a surface morphology evolution prediction system based on multi-source surface information and temporal condition modulation, wherein the surface morphology evolution prediction system based on multi-source surface information and temporal condition modulation is as follows: The surface topography data collection module is used to acquire surface topography data at different service times, perform data transformation on the surface topography data to obtain a reference surface, mask information, random noise and time interval, and perform multimodal fusion on the reference surface, the mask information, the random noise and the time interval based on the network mapping function to obtain model input features; The prediction model training module is used to input the model input features into the temporal modulation mechanism network of the conditional surface generation model for training, so as to obtain the target prediction model. The surface morphology prediction module is used to acquire the surface to be measured and the target time conditions, predict the surface to be measured based on the target time conditions using the target prediction model, obtain the predicted surface increment field, and incrementally update the predicted surface increment field and the reference surface to obtain the surface morphology prediction result.

[0014] Furthermore, to achieve the above objectives, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a surface morphology evolution prediction program based on multi-source surface information and time-condition modulation, and when the surface morphology evolution prediction program based on multi-source surface information and time-condition modulation is executed by a processor, it implements the steps of the surface morphology evolution prediction method based on multi-source surface information and time-condition modulation as described above.

[0015] This invention acquires surface morphology data at different service times, performs data transformation on the surface morphology data to obtain a reference surface, mask information, random noise, and time intervals, and performs multimodal fusion on the reference surface, mask information, random noise, and time intervals based on a network mapping function to obtain model input features. These model input features are then input into a time modulation mechanism network of a conditional surface generation model for training, resulting in a target prediction model. The surface to be tested and target time conditions are acquired, and the target prediction model predicts the surface to be tested based on these target time conditions, obtaining a predicted surface increment field. The predicted surface increment field and the reference surface are incrementally updated to obtain the surface morphology prediction result. This invention, based on surface morphology data transformation and multimodal fusion, trains a conditional surface generation model and predicts the surface to be tested, achieving accurate surface morphology prediction. Attached Figure Description

[0016] Figure 1 This is a flowchart of a preferred embodiment of the surface morphology evolution prediction method based on multi-source surface information and time-condition modulation of the present invention; Figure 2 This is a schematic diagram of the overall process of a preferred embodiment of the surface morphology evolution prediction method based on multi-source surface information and time condition modulation of the present invention. Figure 3 This is a schematic diagram of the overall model of a preferred embodiment of the surface morphology evolution prediction method based on multi-source surface information and temporal condition modulation of the present invention. Figure 4 This is a schematic diagram of a 15-minute wear test of a preferred embodiment of the surface morphology evolution prediction method based on multi-source surface information and time condition modulation of the present invention. Figure 5 This is a schematic diagram of 15-minute wear prediction, which is a preferred embodiment of the surface morphology evolution prediction method based on multi-source surface information and time condition modulation of the present invention. Figure 6 This is a schematic diagram of a 70-minute wear test of a preferred embodiment of the surface morphology evolution prediction method based on multi-source surface information and time condition modulation of the present invention. Figure 7 This is a schematic diagram of 70-minute wear prediction, which is a preferred embodiment of the surface morphology evolution prediction method based on multi-source surface information and time condition modulation of the present invention. Figure 8 This is a schematic diagram of a 140-minute wear test of a preferred embodiment of the surface morphology evolution prediction method based on multi-source surface information and time condition modulation of the present invention. Figure 9 This is a schematic diagram of 140-minute wear prediction based on a preferred embodiment of the surface morphology evolution prediction method based on multi-source surface information and time condition modulation of the present invention. Figure 10 This is a structural diagram of a preferred embodiment of the surface morphology evolution prediction system based on multi-source surface information and time-condition modulation of the present invention; Figure 11 This is a structural diagram of a preferred embodiment of the terminal of the device of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0018] Traditional methods for predicting the surface morphology of sealing coatings (such as empirical models or numerical simulations) focus only on geometric reconstruction errors, lack joint constraints on surface height distribution, spectral characteristics, and multi-scale consistency, and rely on specific materials or operating conditions. Under complex service conditions such as fretting or contact wear, they struggle to accurately predict the evolution of surface morphology over time for various types of sealing coatings, resulting in inaccurate predictions. Therefore, a surface morphology evolution prediction method based on multi-source surface information and temporal condition modulation is needed. This method should utilize surface morphology data transformation and multimodal fusion to incrementally update training conditions, generate a model, and predict the surface under test, thereby achieving accurate prediction of coating surface morphology.

[0019] The surface morphology evolution prediction method based on multi-source surface information and temporal condition modulation described in the preferred embodiment of the present invention, such as... Figure 1 and Figure 2 As shown, the surface morphology evolution prediction method based on multi-source surface information and temporal condition modulation includes the following steps: Step S10: Obtain surface topography data at different service times, perform data transformation on the surface topography data to obtain reference surface, mask information, random noise and time interval, and perform multimodal fusion on the reference surface, the mask information, the random noise and the time interval based on the network mapping function to obtain model input features.

[0020] Step S10 includes: Step S11: Obtain surface topography data at different service times, extract the surface topography data to obtain structured data, and transform the structured data to obtain a global effective height set; Step S12: Perform logarithmic compression mapping on the global effective height set to obtain the reference surface, mask information, random noise and time interval. Based on the network mapping function, perform multimodal fusion on the reference surface, the mask information, the random noise and the time interval to obtain the model input features.

[0021] Specifically, surface topography data at different service times are acquired, and the surface topography data is extracted to obtain structured data. The structured data is then transformed to obtain a global effective height set. The global effective height set is logarithmically compressed and mapped to obtain a reference surface, mask information, random noise, and time interval. Based on a network mapping function, multimodal fusion is performed on the reference surface, the mask information, the random noise, and the time interval to obtain model input features. The model input features can be represented as: ; in, Represents the initial surface height matrix (reference surface); This represents the effective region mask matrix (mask information) corresponding to the source surface, where the effective region has a value of 1 and the invalid region has a value of 0; This represents random perturbation noise variables (random noise), used to enhance the model's generative and generalization capabilities; An embedding vector (time interval) representing the target time is used to characterize wear time condition information; Indicates that there are parameters, The conditional generation network mapping function is represented, where, These are the learnable parameters of the network; This represents the input features of the model, i.e., the amount of change of the target surface relative to the source surface.

[0022] For example, such as Figure 2 As shown, (MAE, Mean Absolute Error) measures the average difference in height values ​​between the predicted and actual surfaces, reflecting the overall reconstruction accuracy; (PSD, Power Spectral Density) describes the energy distribution of the surface at different spatial frequencies, used to characterize the frequency domain characteristics of surface roughness; (HPD, Height Probability Distribution) describes the statistical distribution of the probability of surface height values ​​occurring, used to characterize the statistical properties of surface morphology; and (CDF, ​​Cumulative Distribution Function) is obtained by integrating the height probability distribution, representing the cumulative probability that the height value is less than a certain threshold.

[0023] Step S11 includes: Step S111: Obtain surface morphology data at different service times, extract the surface morphology data to obtain structured data, and identify the structured data to obtain effective pixels; Step S112: Obtain the quantile function, and calculate the global height range of the effective pixels according to the quantile function to obtain the global effective height set; Specifically, surface topography data at different service times is acquired, the surface topography data is extracted to obtain structured data, the structured data is identified to obtain effective pixels, a quantile function is obtained, and the global height range of the effective pixels is calculated based on the quantile function to obtain a global effective height set. The structured data includes surface height file path data, grayscale image path data, PSD file path data, HPD file path data, and data integrity flag data (surface height file path, grayscale image path (if it exists), PSD file path, HPD file path, and data integrity flags (has_psd, has_hpd, etc.). The original discrete data is transformed into a structured data list to provide a unified data interface for subsequent training and inference.

[0024] In this embodiment, each surface topography data is read sequentially, and invalid placeholders (represented by different symbols on different devices) are identified as missing regions. Only valid height pixels are retained for statistical analysis. Simultaneously, abnormal samples with an excessively low proportion of valid pixels are removed. To avoid a few large samples having an overly dominant effect on the overall statistics, only a preset number of valid pixels are randomly sampled for each valid height map, and the results are aggregated across all samples. Finally, a robust quantile method is used to determine the global height reference interval. The height range determination method can be expressed as: ; in, Indicates the lowest point. Indicates the highest point. This indicates that the set of globally valid heights is in the lower quantile. This indicates that the set of global effective heights is at the high quantile.

[0025] As an example, for the surface height data of all samples, valid pixels are extracted (missing values ​​are removed), and robust statistics are performed to calculate the global height range: ; in, This represents the set of surface height data for all samples, which is a one-dimensional data set formed by summing the height values ​​of the valid regions in all samples. This represents the lower bound of the global height, used to characterize the low-value cutoff range of the surface height; This represents the upper bound of the global height, used to characterize the high-value cutoff range of the surface height; This represents the quantile function, here. This means extracting the height value of the 0.5th quantile from the set. This indicates that the height value of the 99.5th percentile is extracted from the set. Step S20: Input the model input features into the temporal modulation mechanism network of the conditional surface generation model for training to obtain the target prediction model.

[0026] Step S20 includes: Step S21: Input the reference surface, mask information, random noise and time interval of the model input features into the encoder layer, bottleneck layer, decoder layer and skip connection layer of the conditional surface generation model for modulation to obtain the modulation result; Step S22: Obtain multiple weight coefficients, and based on the multi-objective joint loss function, train the conditional surface generation model according to the modulation result and all the weight coefficients to obtain the target prediction model.

[0027] Specifically, the reference surface, mask information, random noise, and time interval of the model input features are input into the encoder layer, bottleneck layer, decoder layer, and skip connection layer of the conditional surface generation model for modulation to obtain the modulation result: ; in, Indicates the modulation result. Represents the input features of the model. This represents element-wise multiplication. This indicates the mapping result at the first moment. This represents the result of the second time mapping.

[0028] In this embodiment, as Figure 3As shown, in terms of network structure, this module adopts the U-Net generative network with temporal conditional modulation mechanism. U-Net is a typical encoder-decoder convolutional neural network, originally used for image segmentation tasks. Its core features are multi-scale feature extraction and high-resolution reconstruction capabilities. Specifically, U-Net has the following four main characteristics: Encoder path: The encoder extracts features from the input surface step by step through multiple convolutions and downsampling operations, mapping the original surface height information into multi-scale feature representations. As the network depth increases, the features gradually transform from local geometric information to global structural information. Bottleneck layer: A bottleneck layer is set between the encoder and decoder to fuse global feature information, achieving a high-level expression of the overall surface morphology and structure. Decoder path: The decoder gradually restores the spatial resolution through upsampling operations and combines the features of the corresponding layers of the encoder (skip connections) to reconstruct detailed information, so that the output surface maintains both overall structural consistency and local texture details. Skip Connection Mechanism: U-Net introduces same-level feature connections between the encoder and decoder, allowing shallow high-resolution information to directly participate in the reconstruction process, thereby effectively avoiding information loss and improving detail recovery capabilities. The feature map is modulated using FiLM (Feature-wise Linear Modulation).

[0029] In this embodiment, multiple weight coefficients are obtained, and the conditional surface generation model is trained based on the modulation result and all the weight coefficients according to the multi-objective joint loss function to obtain the target prediction model.

[0030] Step S22 includes: Step S221: Obtain the surface height matrix and multiple weight coefficients at the actual target time. Based on the distribution difference loss and difference loss, train the conditional surface generation model according to the surface height matrix at the actual target time, the modulation result and all the weight coefficients to obtain the target prediction model.

[0031] Specifically, the surface height matrix and multiple weight coefficients at the actual target time are obtained. Based on the distribution difference loss and difference loss, the conditional surface generation model is trained according to the surface height matrix at the actual target time, the modulation result, and all the weight coefficients to obtain the target prediction model. ; in, Represents the total loss function. The weighting coefficients represent the height error of the mask region. This represents the surface height matrix at the actual target time. This represents element-wise multiplication. This represents the effective region mask matrix corresponding to the source surface. The weight coefficients representing the gradient consistency constraints. Represents the spatial gradient operator. Denotes the L1 norm of a matrix. The weighting coefficients represent the frequency domain PSD distribution error. The weighting coefficients representing the height distribution error. Indicates the loss due to distributional differences. This indicates the difference loss.

[0032] As an example, using a mask constraint, it can be written as: ; in, This represents the surface height matrix at the actual target time predicted by the model. This represents the surface height matrix obtained from actual measurements at the target time. This represents the effective region mask at the target time. This indicates the surface height reconstruction loss within the effective area.

[0033] Furthermore, the gradient field consistency loss can be written as: ; in, This represents the gradient field consistency loss. This represents the gradient operator along the horizontal direction. This represents the gradient operator in the vertical direction.

[0034] Furthermore, the roughness statistical loss can be written as: ; in, This represents the roughness statistical constraint. Indicates the deviation of the arithmetic mean. Indicates root mean square deviation; and The calculation formula is as follows: ; ; in, This indicates the input value. This indicates the average height.

[0035] Furthermore, the multi-scale pyramid detail loss is calculated using the following formula: ; in, This indicates the loss of detail in the multi-scale pyramid. Indicates the number of layers in the pyramid. Indicates the first Predicting surfaces at various scales This represents the pyramid detail components on the actual surface.

[0036] Furthermore, we adopt a logarithmic form: ; in, Indicates standard deviation constraint. Indicates the predicted surface within the effective area. It represents the standard deviation of the actual surface.

[0037] Furthermore, statistical moment constraints: ; in, Indicates statistical constraints. Indicates the predicted surface, This represents the mean of the effective area of ​​the real surface.

[0038] Furthermore, the PSD distribution matching loss: ; in, This represents the matching loss of the frequency domain PSD distribution. This indicates the normalized horizontal forecast. This indicates the normalized horizontal forecast. This represents the actual horizontal direction after normalization. This represents the actual horizontal direction after normalization.

[0039] Furthermore, a highly cumulative distribution matching loss: ; in, This represents the highly cumulative distribution matching loss. Represents the cumulative distribution curve. express Transform the data.

[0040] By using a multi-objective joint loss function (surface height reconstruction loss, gradient field consistency loss, roughness statistical constraint, multi-scale pyramid detail loss, pyramid decomposition layer number, standard deviation constraint, statistical moment constraint, frequency domain PSD distribution matching loss, and cumulative height distribution matching loss), the model not only learns the surface geometry itself, but also learns multi-dimensional attributes such as local texture, statistical roughness, spectral structure, and height distribution, thereby improving the physical rationality and representation consistency of the predicted surface.

[0041] Step S30: Obtain the surface to be tested and the target time conditions. Predict the surface to be tested based on the target time conditions using the target prediction model to obtain the predicted surface increment field. Incrementally update the predicted surface increment field and the reference surface to obtain the surface morphology prediction result.

[0042] like Figure 3 As shown, step S30 includes: Step S31: Obtain the surface to be tested, and extract information from the surface to be tested using the target prediction model to obtain the source surface, target surface, PSD information and HPD information; Step S32: Obtain the target time conditions. Using the target prediction model, predict the source surface, the target surface, the PSD information, and the HPD information based on the target time conditions to obtain the predicted surface increment field. Incrementally update the predicted surface increment field and the reference surface to obtain the surface morphology prediction result. ; in, Indicates the reference surface. This indicates the predicted surface increment field.

[0043] Specifically, after step S30, the method further includes acquiring the actual measurement surface, comparing and calculating the predicted surface morphology with the actual measurement surface to obtain an error result. If the error result is a preset error value, the surface height matrix corresponding to the predicted surface morphology is normalized and mapped according to the global fixed scale and the single-image adaptive scale to obtain the prediction image visualization result. The error result includes mean absolute error, mean square error, PSD spectral logarithmic error, and CDF distribution error. The method also involves acquiring the surface to be measured, extracting information from the surface to be measured using the target prediction model, and obtaining the source surface, target surface, PSD information, and HP. D information (constructing fixed validation sample pairs according to the time stratification principle consistent with the training phase, and reading the source surface, target surface and their PSD and HPD information one by one to form inference input), obtain the target time conditions, and predict the source surface, target surface, PSD information and HPD information according to the target time conditions through the target prediction model to obtain the predicted surface increment field (after calculation by the generative model, the predicted surface increment field at the target time is output, and then superimposed with the reference surface to obtain the surface morphology prediction result), and incrementally update the predicted surface increment field and the reference surface to obtain the surface morphology prediction result.

[0044] In this embodiment, the predicted surface morphology is compared with the actual measured surface, and indices such as the mean absolute error, mean square error, PSD spectral logarithmic error, and CDF distribution error of the mask region are calculated. In this application example, the model achieves an average CDF L1 error of 0.038, indicating good agreement on the height distribution; the average PSD log-L1 error is 0.184, indicating satisfactory multi-scale spectral consistency; and the mean absolute error of the wear increment is approximately 0.051 μm, confirming the reliability of the geometric prediction accuracy. It is evident that the predicted surface exhibits high consistency with the actual surface in terms of overall morphology and maintains good statistical consistency in height distribution, effectively reproducing the scale changes caused by wear.

[0045] For example, such as Figure 4 , Figure 5 , Figure 6 , Figure 7 , Figure 8 and Figure 9 As shown, Figure 4 In the figure, (a) and (b) represent the two-dimensional and three-dimensional planes of the measured morphology of the wear surface corresponding to 15 minutes. Figure 5 In the figure, (a) and (b) represent the two-dimensional and three-dimensional planes of the predicted morphology of the wear surface at 15 minutes. Figure 6 In the figure, (a) and (b) represent the two-dimensional and three-dimensional planes of the measured morphology of the wear surface corresponding to 70 minutes. Figure 7 In the figure, (a) and (b) represent the two-dimensional and three-dimensional planes of the predicted morphology of the wear surface corresponding to 70 minutes. Figure 8 In the figure, (a) and (b) represent the two-dimensional and three-dimensional planes of the measured morphology of the wear surface corresponding to 140 minutes. Figure 9 In the diagram, (a) and (b) represent the two-dimensional and three-dimensional planes of the predicted morphology of the wear surface at 140 minutes, enabling a visual display of the surface wear area and comparative analysis at different time points. Furthermore, the prediction results can be used for sealing performance degradation analysis, contact condition assessment, life prediction, and maintenance decision-making.

[0046] Furthermore, such as Figure 10 As shown, based on the above-mentioned surface morphology evolution prediction method based on multi-source surface information and temporal condition modulation, the present invention also provides a surface morphology evolution prediction system based on multi-source surface information and temporal condition modulation, wherein the surface morphology evolution prediction system based on multi-source surface information and temporal condition modulation includes: The surface topography data collection module 51 is used to acquire surface topography data at different service times, perform data conversion on the surface topography data to obtain a reference surface, mask information, random noise and time interval, and perform multimodal fusion on the reference surface, the mask information, the random noise and the time interval based on the network mapping function to obtain model input features; Prediction model training module 52 is used to input the model input features into the temporal modulation mechanism network of the conditional surface generation model for training, so as to obtain the target prediction model; The surface morphology prediction module 53 is used to acquire the surface to be measured and the target time conditions, predict the surface to be measured according to the target time conditions through the target prediction model, obtain the predicted surface increment field, and incrementally update the predicted surface increment field and the reference surface to obtain the surface morphology prediction result.

[0047] Furthermore, such as Figure 11 As shown, based on the above-mentioned surface morphology evolution prediction method and system based on multi-source surface information and time condition modulation, the present invention also provides a terminal, which includes a processor 10, a memory 20 and a display 30. Figure 11 Only some of the terminal components are shown; however, it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0048] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as a hard disk or memory. In other embodiments, the memory 20 may be an external storage device of the terminal, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc. Further, the memory 20 may include both internal and external storage devices. The memory 20 is used to store application software and various types of data installed on the terminal, such as program code installed on the terminal. The memory 20 may also be used to temporarily store data that has been output or will be output. In one embodiment, the memory 20 stores a surface morphology evolution prediction program 40 based on multi-source surface information and time-conditional modulation. This surface morphology evolution prediction program 40 can be executed by the processor 10 to implement the surface morphology evolution prediction method based on multi-source surface information and time-conditional modulation in this application.

[0049] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor, or other data processing chip, used to run program code stored in the memory 20 or process data, such as executing the surface morphology evolution prediction method based on multi-source surface information and time condition modulation.

[0050] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. The display 30 is used to display information on the terminal and to display a visual user interface. The terminals communicate with each other via a system bus.

[0051] In one embodiment, when the processor 10 executes the surface morphology evolution prediction program 40 based on multi-source surface information and time-conditional modulation in the memory 20, the following steps are performed: Surface topography data at different service times are acquired, and the surface topography data are transformed to obtain a reference surface, mask information, random noise and time interval. Multimodal fusion is performed on the reference surface, the mask information, the random noise and the time interval based on a network mapping function to obtain model input features. The input features of the model are fed into the temporal modulation mechanism network of the conditional surface generation model for training to obtain the target prediction model; The test surface and target time conditions are obtained. The test surface is predicted according to the target time conditions using the target prediction model to obtain the predicted surface increment field. The predicted surface increment field and the reference surface are incrementally updated to obtain the surface morphology prediction result. The process of acquiring surface topography data at different service times, performing data transformation on the surface topography data to obtain a reference surface, mask information, random noise, and time interval, and performing multimodal fusion on the reference surface, mask information, random noise, and time interval based on a network mapping function to obtain model input features specifically includes: Surface topography data at different service times are acquired, the surface topography data is extracted to obtain structured data, and the structured data is transformed to obtain a global effective height set; The global effective height set is logarithmically compressed and mapped to obtain the reference surface, mask information, random noise, and time interval. Based on the network mapping function, the reference surface, the mask information, the random noise, and the time interval are fused in a multimodal manner to obtain the model input features.

[0052] The process of acquiring surface topography data at different service times, extracting structured data from the surface topography data, and transforming the structured data to obtain a global effective height set specifically includes: Surface topography data at different service times are acquired, the surface topography data is extracted to obtain structured data, and the structured data is identified to obtain effective pixels; Obtain the quantile function, and calculate the global height range of the effective pixels based on the quantile function to obtain the global effective height set; The structured data includes surface height file path data, grayscale image path data, PSD file path data, HPD file path data, and data integrity flag data.

[0053] The time modulation mechanism network includes an encoder layer, a bottleneck layer, a decoder layer, and a skip connection layer. The step of inputting the model input features into a temporal modulation mechanism network of a conditional surface generation model for training to obtain a target prediction model specifically includes: The reference surface, mask information, random noise, and time interval of the model input features are input into the encoder layer, bottleneck layer, decoder layer, and skip connection layer of the conditional surface generation model for modulation to obtain the modulation result: ; in, Indicates the modulation result. Represents the input features of the model. This represents element-wise multiplication. This indicates the mapping result at the first moment. This represents the result of the second time mapping; Multiple weight coefficients are obtained, and the conditional surface generation model is trained based on the modulation result and all the weight coefficients according to the multi-objective joint loss function to obtain the target prediction model.

[0054] The multi-objective joint loss function includes distribution difference loss and difference loss; The process of obtaining multiple weight coefficients, training the conditional surface generation model based on the modulation result and all the weight coefficients using a multi-objective joint loss function to obtain a target prediction model specifically includes: Obtain the surface height matrix and multiple weight coefficients at the actual target time. Based on the distribution difference loss and difference loss, train the conditional surface generation model according to the surface height matrix at the actual target time, the modulation result, and all the weight coefficients to obtain the target prediction model. ; in, Represents the total loss function. The weighting coefficients represent the height error of the mask region. This represents the surface height matrix at the actual target time. This represents element-wise multiplication. This represents the effective region mask matrix corresponding to the source surface. The weight coefficients representing the gradient consistency constraints. Represents the spatial gradient operator. Denotes the L1 norm of a matrix. The weighting coefficients represent the frequency domain PSD distribution error. The weighting coefficients representing the height distribution error. Indicates the loss due to distributional differences. Indicates difference loss, This indicates the predicted surface morphology.

[0055] Specifically, the process of acquiring the surface to be measured and the target time conditions, predicting the surface to be measured using the target prediction model based on the target time conditions to obtain a predicted surface increment field, and incrementally updating the predicted surface increment field and the reference surface to obtain a surface morphology prediction result includes: The surface to be tested is acquired, and information is extracted from the surface to be tested using the target prediction model to obtain the source surface, target surface, PSD information, and HPD information. Obtain the target time conditions, and use the target prediction model to predict the source surface, the target surface, the PSD information, and the HPD information based on the target time conditions to obtain the predicted surface increment field. Incrementally update the predicted surface increment field and the reference surface to obtain the surface morphology prediction result. ; in, Indicates the reference surface. This indicates the predicted surface increment field.

[0056] The process includes: acquiring the surface to be measured and the target time conditions; predicting the surface to be measured using the target prediction model based on the target time conditions to obtain a predicted surface increment field; incrementally updating the predicted surface increment field and the reference surface to obtain a surface morphology prediction result; and then further including: Obtain the actual measurement surface, compare the morphology of the actual measurement surface with that of the predicted surface, and calculate the error result. If the error result is a preset error value, then the surface height matrix corresponding to the predicted surface morphology is normalized and mapped according to the global fixed scale and the single map adaptive scale to obtain the prediction map visualization result. The error results include mean absolute error, mean square error, logarithmic error of PSD spectrum, and CDF distribution error.

[0057] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a surface morphology evolution prediction program based on multi-source surface information and time-condition modulation, and the surface morphology evolution prediction program based on multi-source surface information and time-condition modulation, when executed by a processor, implements the steps of the surface morphology evolution prediction method based on multi-source surface information and time-condition modulation as described above.

[0058] In summary, this invention provides a method, system, terminal, and storage medium for predicting surface morphology evolution based on multi-source surface information and temporal condition modulation. The method includes: acquiring surface morphology data at different service times; performing data transformation on the surface morphology data to obtain a reference surface, mask information, random noise, and time interval; performing multimodal fusion on the reference surface, mask information, random noise, and time interval based on a network mapping function to obtain model input features; inputting the model input features into a temporal modulation mechanism network of a conditional surface generation model for training to obtain a target prediction model; acquiring the surface to be tested and target time conditions; predicting the surface to be tested based on the target time conditions using the target prediction model to obtain a predicted surface increment field; and incrementally updating the predicted surface increment field and the reference surface to obtain a surface morphology prediction result. This invention, based on surface morphology data transformation and multimodal fusion, trains a conditional surface generation model and predicts the surface to be tested, achieving accurate prediction of surface morphology.

[0059] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal system that includes that element.

[0060] Of course, those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.). The program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The computer-readable storage medium can be a memory, magnetic disk, optical disk, etc.

[0061] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A surface morphology evolution prediction method based on multi-source surface information and temporal condition modulation, characterized in that, The surface morphology evolution prediction method based on multi-source surface information and time-condition modulation includes: Surface topography data at different service times are acquired, and the surface topography data are transformed to obtain a reference surface, mask information, random noise and time interval. Multimodal fusion is performed on the reference surface, the mask information, the random noise and the time interval based on a network mapping function to obtain model input features. The input features of the model are fed into the temporal modulation mechanism network of the conditional surface generation model for training to obtain the target prediction model; The test surface and target time conditions are obtained. The test surface is predicted according to the target time conditions using the target prediction model to obtain the predicted surface increment field. The predicted surface increment field and the reference surface are incrementally updated to obtain the surface morphology prediction result. The process involves acquiring surface topography data at different service times, performing data transformation on the surface topography data to obtain a reference surface, mask information, random noise, and time intervals, and then performing multimodal fusion on the reference surface, mask information, random noise, and time intervals based on a network mapping function to obtain model input features. Specifically, this includes: Surface topography data at different service times are acquired, the surface topography data is extracted to obtain structured data, and the structured data is transformed to obtain a global effective height set; The global effective height set is logarithmically compressed and mapped to obtain the reference surface, mask information, random noise and time interval. Based on the network mapping function, the reference surface, the mask information, the random noise and the time interval are fused in a multimodal manner to obtain the model input features. The time modulation mechanism network includes an encoder layer, a bottleneck layer, a decoder layer, and a skip connection layer; The step of inputting the model input features into a temporal modulation mechanism network of a conditional surface generation model for training to obtain a target prediction model specifically includes: The reference surface, mask information, random noise, and time interval of the model input features are input into the encoder layer, bottleneck layer, decoder layer, and skip connection layer of the conditional surface generation model for modulation to obtain the modulation result: ； in, Indicates the modulation result. Represents the input features of the model. This represents element-wise multiplication. This indicates the mapping result at the first moment. This represents the result of the second time mapping; Multiple weight coefficients are obtained, and the conditional surface generation model is trained based on the modulation result and all the weight coefficients according to the multi-objective joint loss function to obtain the target prediction model. The multi-objective joint loss function includes distribution difference loss and difference loss; The process of obtaining multiple weight coefficients, training the conditional surface generation model based on the modulation result and all the weight coefficients using a multi-objective joint loss function to obtain a target prediction model specifically includes: Obtain the surface height matrix and multiple weight coefficients at the actual target time. Based on the distribution difference loss and difference loss, train the conditional surface generation model according to the surface height matrix at the actual target time, the modulation result, and all the weight coefficients to obtain the target prediction model. ； in, Represents the total loss function. The weighting coefficients represent the height error of the mask region. This represents the surface height matrix at the actual target time. This represents element-wise multiplication. This represents the effective region mask matrix corresponding to the source surface. The weight coefficients representing the gradient consistency constraints. Represents the spatial gradient operator. Denotes the L1 norm of a matrix. The weighting coefficients represent the frequency domain PSD distribution error. The weighting coefficients representing the height distribution error. Indicates the loss due to distributional differences. Indicates difference loss, This indicates the surface morphology prediction results; The process of acquiring the surface to be tested and the target time conditions, predicting the surface to be tested based on the target time conditions using the target prediction model to obtain the predicted surface increment field, and incrementally updating the predicted surface increment field and the reference surface to obtain the surface morphology prediction result specifically includes: The surface to be tested is acquired, and information is extracted from the surface to be tested using the target prediction model to obtain the source surface, target surface, PSD information, and HPD information. Obtain the target time conditions, and use the target prediction model to predict the source surface, the target surface, the PSD information, and the HPD information based on the target time conditions to obtain the predicted surface increment field. Incrementally update the predicted surface increment field and the reference surface to obtain the surface morphology prediction result. ； in, Indicates the reference surface. This indicates the predicted surface increment field.

2. The surface morphology evolution prediction method based on multi-source surface information and temporal condition modulation according to claim 1, characterized in that, The process of acquiring surface topography data at different service times, extracting structured data from the surface topography data, and transforming the structured data to obtain a globally effective height set specifically includes: Surface topography data at different service times are acquired, the surface topography data is extracted to obtain structured data, and the structured data is identified to obtain effective pixels; Obtain the quantile function, and calculate the global height range of the effective pixels based on the quantile function to obtain the global effective height set; The structured data includes surface height file path data, grayscale image path data, PSD file path data, HPD file path data, and data integrity flag data.

3. The surface morphology evolution prediction method based on multi-source surface information and temporal condition modulation according to claim 1, characterized in that, The process of acquiring the surface to be measured and the target time conditions, predicting the surface to be measured using the target prediction model based on the target time conditions to obtain a predicted surface increment field, incrementally updating the predicted surface increment field and the reference surface to obtain a surface morphology prediction result, further includes: Obtain the actual measurement surface, compare the morphology of the actual measurement surface with that of the predicted surface, and calculate the error result. If the error result is a preset error value, then the surface height matrix corresponding to the predicted surface morphology is normalized and mapped according to the global fixed scale and the single map adaptive scale to obtain the prediction map visualization result. The error results include mean absolute error, mean square error, logarithmic error of PSD spectrum, and CDF distribution error.

4. A surface morphology evolution prediction system based on multi-source surface information and temporal condition modulation, characterized in that, The surface morphology evolution prediction system based on multi-source surface information and temporal condition modulation is applied to the surface morphology evolution prediction method based on multi-source surface information and temporal condition modulation as described in any one of claims 1-3, wherein the surface morphology evolution prediction system based on multi-source surface information and temporal condition modulation comprises: The surface topography data collection module is used to acquire surface topography data at different service times, perform data transformation on the surface topography data to obtain a reference surface, mask information, random noise and time interval, and perform multimodal fusion on the reference surface, the mask information, the random noise and the time interval based on the network mapping function to obtain model input features; The prediction model training module is used to input the model input features into the temporal modulation mechanism network of the conditional surface generation model for training, so as to obtain the target prediction model. The surface morphology prediction module is used to acquire the surface to be measured and the target time conditions, predict the surface to be measured based on the target time conditions using the target prediction model, obtain the predicted surface increment field, and incrementally update the predicted surface increment field and the reference surface to obtain the surface morphology prediction result.

5. A terminal, characterized in that, The terminal includes: a memory, a processor, and a surface morphology evolution prediction program based on multi-source surface information and time-condition modulation stored in the memory and executable on the processor. When the surface morphology evolution prediction program based on multi-source surface information and time-condition modulation is executed by the processor, it implements the steps of the surface morphology evolution prediction method based on multi-source surface information and time-condition modulation as described in any one of claims 1-3.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a surface morphology evolution prediction program based on multi-source surface information and time-condition modulation. When the surface morphology evolution prediction program based on multi-source surface information and time-condition modulation is executed by a processor, it implements the steps of the surface morphology evolution prediction method based on multi-source surface information and time-condition modulation as described in any one of claims 1-3.

Images