Defect feature guide-based bearing full life cycle digital twin vibration signal generation method and related equipment

Abstract

The invention discloses a bearing full life cycle digital twin vibration signal generation method based on defect feature guidance and related equipment, and the method comprises the steps: obtaining a target defect feature of a target bearing, inputting a pre-trained defect feature guidance condition diffusion model, and generating a corresponding digital twin vibration signal. The model training process comprises the following steps: acquiring simulation vibration signals under different defect features in different stages of the full life cycle of the bearing, extracting vibration features, and training a mapping model to establish a mapping relation between the vibration features and the defect features; processing the actually measured vibration signal by using the mapping model to generate a defect feature pseudo label; based on actually measured and simulated vibration signals and corresponding defect features, a conditional diffusion model is constructed and trained, and during training, the defect features serve as conditional input, the vibration signals serve as output, an envelope spectrum loss function is adopted, and parameters are optimized through back propagation. The invention aims to solve the contradiction between scarcity of real degraded samples and dependence of an intelligent model on massive labeled data in bearing fault diagnosis, and overcome the problem that the existing method cannot generate corresponding vibration signals according to characteristic parameters.

Description

Technical Field

[0001] This invention belongs to the field of mechanical fault prediction technology, specifically relating to a method and related equipment for generating digital twin vibration signals of bearings throughout their entire life cycle based on defect features. Background Technology

[0002] With the continuous and rapid development of the aerospace industry, fault diagnosis and online monitoring of key aerospace equipment are not only crucial for ensuring the safe operation of engines, but also essential for promoting the independent control and high reliability of aerospace equipment. Existing intelligent diagnostic algorithms typically rely on neural networks and machine learning, which offer high diagnostic accuracy but require substantial data support. Fault diagnosis of liquid rocket engine bearings still faces a contradiction between the scarcity of real fault samples due to high-reliability design and the reliance of intelligent models on massive amounts of labeled data. The high cost of a single test and the difficulty in controlling the reproduction of faults result in an extremely limited number of real degradation samples, becoming a major obstacle to the engineering application of intelligent diagnostic methods.

[0003] To address the aforementioned issues, existing technologies employ digital twin methods to augment real-world degradation samples. The core idea is to optimize the simulated vibration signals generated by the constructed dynamic model using neural network technology, thereby establishing a mapping relationship between the simulated vibration signals and the digital twin vibration signals. While this method can augment real-world degradation samples, actual equipment operation often involves changes in characteristic parameters. For example, in liquid rocket engine bearings, dynamic changes in the bearing's characteristic parameters affect its vibration signal response. Therefore, existing methods cannot generate corresponding vibration signals based on these characteristic parameters. Summary of the Invention

[0004] To address the problems existing in the prior art, this invention provides a method and related equipment for generating digital twin vibration signals of bearings throughout their entire life cycle based on defect features. The purpose is to resolve the contradiction between the scarcity of real degradation samples and the reliance of intelligent models on massive amounts of labeled data in bearing fault diagnosis, and to overcome the problem that existing digital twin methods cannot generate corresponding vibration signals based on feature parameters. This invention enables the generation of corresponding digital twin vibration signals of bearings throughout their entire life cycle based on feature parameters, thereby meeting the engineering application requirements of intelligent diagnostic methods.

[0005] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution: According to a first aspect of the present invention, a method for generating digital twin vibration signals of a bearing throughout its entire life cycle based on defect features is provided, comprising: Obtain the target defect characteristics of the target bearing; The target defect features are input into a pre-trained defect feature-guided conditional diffusion model to generate a digital twin vibration signal corresponding to the target defect features; wherein, the defect feature-guided conditional diffusion model is trained through the following process: The simulated vibration signals obtained from the simulation of different defect characteristics at different stages of the bearing's entire life cycle are obtained. Vibration features are extracted from simulated vibration signals, and a mapping model is trained based on the vibration features and corresponding defect features to establish the mapping relationship between vibration features and defect features. The measured vibration signal is processed using a mapping model to generate pseudo-labels for the defect features corresponding to the measured vibration signal; Based on measured vibration signals and their defect feature pseudo-labels, as well as simulated vibration signals and their corresponding defect features, a conditional diffusion model is constructed and trained. During training, the conditional diffusion model takes defect features as conditional input and vibration signals as output, and employs an envelope spectrum loss function. This envelope spectrum loss function is used to calculate the difference between the envelope spectrum of the digital twin vibration signal output by the conditional diffusion model and the envelope spectrum of the measured vibration signal, and optimizes the model parameters through backpropagation.

[0006] In one possible implementation of the first aspect, the conditional diffusion model, in the process of generating a digital twin vibration signal corresponding to the target defect feature, further includes: A data matching method is used to match the defect feature most relevant to the target defect feature in the defect feature library, and the simulated vibration signal corresponding to the most relevant defect feature is obtained as a reference signal; the defect feature library stores the simulated vibration signal and its corresponding defect feature. Extract time-domain guidance features from the reference signal; The target defect features and the time-domain guidance features are fused together and used as conditions input into the conditional diffusion model to generate the digital twin vibration signal.

[0007] In one possible implementation of the first aspect, the data matching method is specifically expressed as follows:

[0008]

[0009] In the formula, An indicator used to measure the defect features most relevant to the target defect features; Target defect characteristics; To simulate vibration signals and their corresponding defect characteristics; For time smoothing; for The digital twin vibration signal generated in real time; for The digital twin vibration signal generated in real time; for The front of the digital twin vibration signal generated at any time One point; for The post-digital twin vibration signal generated at any time One point; and These are used to balance the prediction accuracy term. With time smoothing term The weight.

[0010] In one possible implementation of the first aspect, the conditional diffusion model employs a U-Net structure, which guides the generation of digital twin vibration signals by utilizing the defect features in the following manner: The defect features are mapped into high-dimensional defect embedding vectors through an embedding layer; The defect embedding vector is fused with the time step embedding vector of the diffusion process to form a unified condition vector; The unified conditional vector is input into multiple residual blocks of the U-Net network to conditionally control the denoising process through one-dimensional convolutional layers and multi-head self-attention layers within the residual blocks, thereby generating digital twin vibration signals corresponding to the defect features.

[0011] In one possible implementation of the first aspect, the envelope spectrum loss function is defined as:

[0012] In the formula, The difference between the envelope spectrum of the digital twin vibration signal output by the conditional diffusion model and the envelope spectrum of the measured vibration signal; To the time step in the diffusion model Measured vibration signal and noise items The mathematical expectation; For envelope spectrum operators; In the diffusion model Digital twin vibration signals generated at time steps; This is the measured vibration signal.

[0013] In one possible implementation of the first aspect, obtaining the simulated vibration signal obtained from simulations of different defect characteristics corresponding to different stages of the bearing's entire life cycle specifically involves: Construct a two-degree-of-freedom dynamic model of the bearing; Optimize the parameters of the two-degree-of-freedom dynamic model; Based on the optimized two-degree-of-freedom dynamic model, the simulated vibration signal is generated by numerically simulating the dynamic response of the bearing under different defect characteristics.

[0014] In one possible implementation of the first aspect, the optimization of the parameters of the two-degree-of-freedom dynamic model specifically involves: The ant colony algorithm is used to optimize and calibrate the stiffness and damping parameters in the two-degree-of-freedom dynamic model with the goal of minimizing the error between the simulated vibration signal and the measured vibration signal in the time domain waveform.

[0015] According to a second aspect of the present invention, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the aforementioned method for generating digital twin vibration signals of a bearing throughout its entire life cycle based on defect features.

[0016] According to a third aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program, which, when executed by a processor, implements the aforementioned method for generating digital twin vibration signals of a bearing throughout its entire life cycle based on defect features.

[0017] According to a fourth aspect of the present invention, a computer program product is provided, which, when executed by a processor, implements the aforementioned method for generating digital twin vibration signals for the entire life cycle of a bearing based on defect features.

[0018] Compared with the prior art, the present invention has at least the following beneficial effects: This invention provides a method for generating digital twin vibration signals for the entire lifecycle of bearings based on defect feature guidance. By constructing a conditional diffusion model guided by defect features, it can directly generate corresponding digital twin vibration signals based on given target defect features. This effectively solves the problem that existing digital twin methods cannot dynamically generate corresponding vibration signals based on feature parameters, thus achieving a precise mapping from defect features to vibration signals. When training the conditional diffusion model, an envelope spectrum loss function is constructed as the optimization objective, ensuring that the generated digital twin vibration signals are highly consistent with the measured vibration signals in the frequency domain, thereby improving the accuracy of the generated digital twin vibration signals. This invention can simulate the vibration response of bearings at different stages of their entire lifecycle and under different defect features, generating digital twin vibration signals covering multiple degradation states, enabling better prediction and health management of bearing operating conditions.

[0019] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the specific embodiments of the present invention, the drawings used in the description of the specific embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0021] Figure 1 This is a flowchart of a method for generating digital twin vibration signals for the entire life cycle of a bearing based on defect features, according to the present invention.

[0022] Figure 2 This is a comparison graph in the time domain between the digital twin vibration signal and the measured vibration signal in the simulation case. Figure 3 This is a comparison graph in the frequency domain between the digital twin vibration signal and the measured vibration signal in the simulation case. Figure 4 A comparison chart of RUL prediction curves under different methods. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] like Figure 1 As shown, this invention provides a method for generating digital twin vibration signals for the entire lifecycle of bearings based on defect features. This method combines simulation data and measured data, utilizes a conditional diffusion model to generate high-fidelity vibration signals, and then uses these signals for bearing condition monitoring and life prediction. Specifically, it includes the following steps: Step 1: Obtain the target defect characteristics of the target bearing.

[0025] It should be understood that the target defect features can be parameters set by the user according to the specific application scenario, such as defect size, location or type.

[0026] Step 2: Input the target defect features into a pre-trained conditional diffusion model guided by the defect features to generate a digital twin vibration signal corresponding to the target defect features.

[0027] In other words, the target defect features are input into a pre-trained conditional diffusion model guided by the defect features. The model then generates a digital twin vibration signal corresponding to the target defect features through a denoising process. The generated digital twin vibration signal can be used for simulation testing, fault diagnosis algorithm verification, or predictive maintenance strategy formulation in bearing lifecycle management.

[0028] In detail, the training phase of the defect feature-guided conditional diffusion model is as follows: First, simulated vibration signals were obtained by simulating various defect characteristics corresponding to different stages of the bearing's entire life cycle. It should be understood that the simulated vibration signals are generated based on the bearing's physical model and operating conditions, covering typical scenarios from early defects to severe failures.

[0029] Subsequently, vibration features are extracted from the simulated vibration signals using conventional signal processing techniques, such as time-domain analysis, frequency-domain analysis, or time-frequency analysis. Based on the extracted vibration features and their corresponding defect features, a mapping model is trained to establish a nonlinear mapping relationship between the vibration features and the defect features. For example, the training process of the mapping model employs supervised learning, using an optimization algorithm to minimize the error between the predicted defect features and the actual defect features.

[0030] Next, the trained mapping model is used to process the measured vibration signals. The measured vibration signals are derived from data collected during the actual operation of the bearing, including vibration records under normal conditions and various defect conditions. By inferring from the measured vibration signals using the mapping model, pseudo-labels of defect features corresponding to the measured vibration signals are generated, thereby compensating for the problem of missing or insufficient defect labels in the measured data.

[0031] Finally, based on the measured vibration signal and its defect feature pseudo-labels, as well as the simulated vibration signal and its corresponding defect features, a defect feature-guided conditional diffusion model is constructed and trained. The conditional input of the conditional diffusion model is the defect feature, and the output is the vibration signal. During training, the model parameters are optimized using the gradient descent method. Due to the gradual injection of Gaussian noise during the diffusion process, the final signal often exhibits frequency domain characteristic degradation. This distortion may mask key diagnostic features, especially the fault-related harmonics and modulation sidebands typically captured in the envelope spectrum. Therefore, an envelope spectrum loss function is introduced to directly constrain the spectral similarity between the digital twin vibration signal and the measured vibration signal. The specific calculation method of the envelope spectrum loss function is as follows: First, envelope analysis is performed on the digital twin vibration signal and the measured vibration signal output by the conditional diffusion model to obtain their respective envelope spectra; then, the difference between the two envelope spectra is calculated; finally, the model parameters are updated using the backpropagation algorithm to reduce the envelope spectrum difference.

[0032] In this embodiment, the envelope spectrum loss function is defined as:

[0033] In the formula, The difference between the envelope spectrum of the digital twin vibration signal output by the conditional diffusion model and the envelope spectrum of the measured vibration signal; To the time step in the diffusion model Measured vibration signal and noise items The mathematical expectation is the average loss over different time steps and samples; For envelope spectrum operators; In the diffusion model Digital twin vibration signals generated at time steps; This is the measured vibration signal.

[0034] This implementation method achieves efficient generation of bearing vibration signals through the above steps, ensuring consistency between the digital twin model and the physical entity, while avoiding the problem of relying on a large amount of labeled measured data.

[0035] In a preferred embodiment, a conditional information enhancement step is further included before inputting the target defect features into the trained conditional diffusion model. Specifically: First, a data matching method is used to match the defect features most relevant to the target defect feature in the defect feature library. For example, this matching process can be achieved by calculating the Euclidean distance between feature vectors, and the defect feature with the smallest distance is determined to be the most relevant defect feature.

[0036] Subsequently, the simulated vibration signal corresponding to the most relevant defect feature is obtained from the defect feature library and used as a reference signal. Next, time-domain guidance features are extracted from the reference signal. These time-domain guidance features provide time-domain morphological guidance for the digital twin vibration signal generation process.

[0037] Finally, the target defect features are fused with the time-domain guidance features extracted from the reference signal, and the fused features are used as conditions input into the conditional diffusion model to generate the final digital twin vibration signal.

[0038] By introducing a reference signal and its time-domain guiding features, this preferred embodiment can provide more accurate generation conditions for the conditional diffusion model, thereby enabling the generated digital twin vibration signal to match the target features in the envelope spectrum and to be more realistic in the time-domain waveform.

[0039] In one possible implementation, the data matching method is specifically expressed as follows:

[0040]

[0041] In the formula, An indicator used to measure the defect features most relevant to the target defect features; Target defect characteristics; To simulate vibration signals and their corresponding defect characteristics; For time smoothing; for The digital twin vibration signal generated in real time; for The digital twin vibration signal generated in real time; for The front of the digital twin vibration signal generated at any time One point; for The post-digital twin vibration signal generated at any time One point; and These are used to balance the prediction accuracy term. With time smoothing term The weight.

[0042] In one possible implementation, the conditional diffusion model employs a U-Net network structure, which guides the generation of digital twin vibration signals by utilizing the defect features in the following manner: First, the defect features, which serve as conditional inputs, are processed through an embedding layer. The embedding layer maps the defect features into high-dimensional defect embedding vectors, representing defect information in a higher-dimensional feature space and providing rich semantic conditions for the model.

[0043] Subsequently, the generated defect embedding vector is fused with the time step embedding vector of the diffusion process. For example, the fusion method can be vector concatenation or element-wise addition to form a unified condition vector, which simultaneously contains target defect information and current diffusion time step information.

[0044] Finally, the unified conditional vector is input into multiple residual blocks of the U-Net network. In each residual block, the conditional information first interacts with the noisy vibration signal through a one-dimensional convolutional layer, thereby imposing conditional control on the local time scale of the signal. Next, the conditional information and signal features are jointly input into a multi-head self-attention layer, which calculates the long-range dependencies between different time points of the signal and adjusts the denoising process globally based on the conditional information. Through the one-dimensional convolution and multi-head self-attention mechanism performed within the residual blocks, the conditional diffusion model can guide the generation of a digital twin vibration signal that highly corresponds to the target defect features throughout the entire denoising iteration.

[0045] In one possible implementation, obtaining the simulated vibration signal obtained from simulations of different defect characteristics corresponding to different stages of the bearing's entire life cycle specifically involves: First, a two-degree-of-freedom dynamic model of the bearing is constructed, as follows:

[0046] in, The total mass of the inner ring and shaft. This refers to the internal damping coefficient of the bearing. This refers to the total contact stiffness between the rolling element and the inner and outer rings. For the inner circle Radial force in the direction, For the inner circle Radial force in the direction, This refers to the total deformation of the rolling element in contact with the raceway. and The outer rings are respectively in and Vibration displacement in the direction, and The inner circle is respectively in and Vibration displacement in the direction, and The outer rings are respectively in and Vibration velocity in the direction, and The inner circle is respectively in and Vibration velocity in the direction, and The outer rings are respectively in and The direction of vibration acceleration, The number of rolling elements. For the first The load zone coefficient of each ball, For the first The angle at which each sphere is located on the coordinate axis.

[0047] Next, the parameters of the two-degree-of-freedom dynamic model are optimized to make them closer to the measured vibration signal in the time domain. Specifically, this embodiment uses the ant colony algorithm to optimize and calibrate the stiffness and damping parameters in the model. During the optimization process, the goal is to minimize the error between the simulated vibration signal and the measured vibration signal in the time domain waveform, and the stiffness and damping parameters in the two-degree-of-freedom dynamic model are optimized and calibrated accordingly.

[0048] The core algorithm formula and optimization process of the ant colony algorithm are as follows:

[0049] In the formula, This represents the natural evaporation process of pheromones during the optimization process. For pheromone evaporation rate, For the reason The cumulative pheromone increment released by a single ant in the current iteration, where the th ant... The contribution of an ant is defined as , For the first The cost of searching parameters for a single ant, its reciprocal It reflects the quality of the path.

[0050] Finally, based on the optimized two-degree-of-freedom dynamic model, the simulated vibration signal is generated by numerically simulating the dynamic response of the bearing under different defect characteristics.

[0051] As a preferred implementation, the mapping model is implemented using a regression network. The regression network is preferably a multilayer perceptron or a backpropagation neural network. The regression network takes the extracted vibration features as input and the corresponding defect features as training targets. Through a supervised training process, the gradient descent algorithm is used to minimize the error between the network-predicted defect features and the actual defect features, thereby enabling the regression network to learn a complex nonlinear mapping function from the vibration feature space to the defect feature space.

[0052] To verify the usability of the proposed method for generating signals, verification was conducted from two perspectives. First, from the time and frequency domains of the vibration signal: specifically, by comparing the measured vibration signal and the generated digital twin vibration signal in the time domain, and by comparing the measured vibration signal and the generated digital twin vibration signal in the frequency domain after envelope transformation, the consistency between the measured vibration signal and the digital twin vibration signal in the time and frequency domains can be obtained. The verification results are as follows: Figure 2 and Figure 3As shown in the figure, Real Signal represents the measured vibration signal, Twin Signal represents the digital twin vibration signal generated by the method of the present invention, and Bearing1_1, Bearing1_2, and Bearing1_3 are the numbers of the data used. The digital twin vibration signal generated by the method of the present invention and the measured vibration signal have extremely high similarity in the frequency domain and time domain, indicating that the method of the present invention can indeed effectively generate the required vibration signal.

[0053] From the perspective of vibration signal availability, the same remaining lifetime prediction model was trained using measured vibration signals, digital twin vibration signals generated by the method of this invention, and twin signals generated by a comparison method. Subsequently, lifetime prediction was performed on the same set of data, and the interpolation RMSE between the predicted lifetime curve and the theoretical curve was calculated, thereby verifying the availability of the vibration signal. The RUL prediction curves under different comparison methods, after smoothing and monotonic processing, are shown below. Figure 4 As shown in Table 1, the RMSE of the RUL curves and actual lifetime curves calculated under different methods are shown in Table 1. CycleGAN, DCGAN, A-VAE, and DDPM are the comparison methods used. Proposed is the method of this invention, Real is the measured vibration signal, and Bearing1_5 is the measured data number used for lifetime prediction. As can be seen from the RMSE data calculated in Table 1, excluding the RMSE of the measured vibration signal, the method of this invention has the smallest RMSE. This indicates that the digital twin vibration signal generated by the method of this invention is closest to the measured vibration signal when used to train the lifetime prediction model, proving the usability of the digital twin vibration signal generated by the method of this invention.

[0054] Table 1 shows the RMSE data under different methods.

[0055] In another embodiment of the present invention, a computer device is provided, comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions in the computer storage medium to achieve a corresponding method flow or corresponding function. The processor described in this embodiment of the present invention can be used in the operation of a method for generating digital twin vibration signals for the entire life cycle of bearings based on defect features.

[0056] In another embodiment of the present invention, a storage medium is provided, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk storage device. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the above embodiment regarding a method for generating digital twin vibration signals of a bearing throughout its entire life cycle based on defect features.

[0057] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0058] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0059] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0060] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0061] This invention also provides a computer program product, which is used to execute any of the aforementioned methods for generating digital twin vibration signals for the entire life cycle of bearings based on defect features. Since the computer program product provided by this invention belongs to the same inventive concept as the aforementioned method for generating digital twin vibration signals for the entire life cycle of bearings based on defect features, it possesses all the advantages of the aforementioned method. Therefore, the beneficial effects of the computer program product provided by this invention will not be elaborated upon here.

[0062] In this invention, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to a specific feature, structure, material, or characteristic described in connection with that embodiment or example, which is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0063] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit them. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the scope of the technology disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention.

Claims

1. A bearing full life cycle digital twin vibration signal generation method based on defect feature guidance, characterized by, The method comprises the following steps: obtaining a target defect feature of a target bearing; inputting the target defect feature into a pre-trained defect feature guided conditional diffusion model to generate a digital twin vibration signal corresponding to the target defect feature; wherein the defect feature guided conditional diffusion model is trained through the following process: obtaining simulation vibration signals simulated under different defect features corresponding to different stages of the life cycle of the bearing; extracting vibration features from the simulation vibration signals, and training a mapping model based on the vibration features and the corresponding defect features to establish a mapping relationship between the vibration features and the defect features; processing the measured vibration signal using the mapping model to generate a defect feature pseudo-label corresponding to the measured vibration signal; based on the measured vibration signal and its defect feature pseudo-label, as well as the simulation vibration signal and its corresponding defect feature, constructing and training a conditional diffusion model; wherein, during the training process, the conditional diffusion model inputs the defect feature as a condition and outputs the vibration signal, and uses an envelope spectrum loss function, which is used to calculate the difference between the envelope spectrum of the digital twin vibration signal output by the conditional diffusion model and the envelope spectrum of the measured vibration signal, and optimize the model parameters through back propagation.

2. The bearing full life cycle digital twin vibration signal generation method based on defect feature guidance according to claim 1, characterized in that, In the process of generating a digital twin vibration signal corresponding to the target defect feature, the conditional diffusion model further comprises: using a data matching method to match the most relevant defect feature in the defect feature library from the target defect feature, and obtaining the simulation vibration signal corresponding to the most relevant defect feature as a reference signal; the defect feature library stores simulation vibration signals and their corresponding defect features; extracting time domain guide features from the reference signal; fuse the target defect feature and the time domain guide feature together as a condition input into the conditional diffusion model to generate the digital twin vibration signal.

3. The bearing full life cycle digital twin vibration signal generation method based on defect feature guidance according to claim 2, characterized in that, The data matching method is specifically expressed as: wherein, is an index of defect features most relevant to the target defect feature; is the target defect feature; is the simulated vibration signal and its corresponding defect feature; is a time smoothing term; is the digital twin vibration signal generated at time t; is the digital twin vibration signal generated at time t; is the first p points of the digital twin vibration signal generated at time t; is the last p points of the digital twin vibration signal generated at time t; is the first p points of the digital twin vibration signal generated at time t; is is the weight for balancing the prediction accuracy term and the time smoothing term .

4. The bearing full life cycle digital twin vibration signal generation method based on defect feature guidance according to claim 1, characterized in that, The conditional diffusion model uses a U-Net structure to guide the generation of a digital twin vibration signal through the following method: map the defect feature to a high-dimensional defect embedding vector through an embedding layer; fuse the defect embedding vector and the time step embedding vector of the diffusion process to form a unified condition vector; input the unified condition vector into multiple residual blocks of the U-Net network to conditionally control the denoising process through one-dimensional convolution layers and multi-head self-attention layers in the residual blocks, thereby generating a digital twin vibration signal corresponding to the defect feature.

5. The bearing full life cycle digital twin vibration signal generation method based on defect feature guidance according to claim 1, characterized in that, The envelope spectrum loss function is defined as: wherein is the difference between the envelope spectrum of the digital twin vibration signal output by the conditional diffusion model and the envelope spectrum of the measured vibration signal; is the mathematical expectation of the diffusion model at time step , the measured vibration signal and the noise term ; is the envelope spectrum operator; is the digital twin vibration signal generated in the diffusion model at time step ; is the measured vibration signal.

6. The bearing full life cycle digital twin vibration signal generation method based on defect feature guidance according to claim 1, characterized in that, The simulation vibration signals simulated under different defect features corresponding to different stages of the life cycle of the bearing are specifically: construct a two-degree-of-freedom dynamic model of the bearing; optimize the parameters of the two-degree-of-freedom dynamic model; based on the optimized two-degree-of-freedom dynamic model, generate the simulation vibration signals by numerically simulating the dynamic response of the bearing under different defect features.

7. The bearing full life cycle digital twin vibration signal generation method based on defect feature guidance according to claim 6, characterized in that, The optimization of the parameters of the two-degree-of-freedom dynamic model is specifically: An ant colony algorithm is adopted to optimize and calibrate the stiffness parameters and damping parameters in the two-degree-of-freedom dynamic model by minimizing the error between the simulated vibration signal and the measured vibration signal in the time domain waveform.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor implements the computer program to realize a bearing full life cycle digital twin vibration signal generation method based on defect feature guidance according to any one of claims 1 to 7.

9. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 8. The computer program is executed by the processor to realize a bearing full life cycle digital twin vibration signal generation method based on defect feature guidance according to any one of claims 1 to 7.

10. A computer program product, characterised in that, The computer program product is executed by the processor to realize a bearing full life cycle digital twin vibration signal generation method based on defect feature guidance according to any one of claims 1 to 7.

Images