Sliding bearing life prediction method

Abstract

The invention discloses a sliding bearing service life prediction method, and relates to the technical field of sliding bearing intelligent monitoring, and the method comprises the steps: an edge calculation node uses a quantum decryption algorithm to decrypt multi-modal time series data, and carries out the fractional order chaotic dynamics analysis, thereby achieving the prediction of the service life of a sliding bearing. Performing phase-space reconstruction on the multi-modal time sequence data, extracting fractal dimensions of phase-space attractors, and fusing chemical and crystal data to a quantum entanglement space by adopting a quantum variational auto-encoder to generate a high-dimensional data set fusing macroscopic and microscopic features; constructing a quantum enhanced life prediction model, performing parameter calibration by adopting quantum Bayesian optimization, and predicting a bearing health index according to the high-dimensional data set; according to the invention, the quantum-enhanced life prediction model is constructed through the quantum variational auto-encoder, and the model parameters are dynamically calibrated in combination with the quantum Bayesian optimization algorithm, so that the accuracy and reliability of bearing life prediction are improved.

Description

Technical Field

[0001] The invention relates to the technical field of intelligent monitoring of sliding bearings, in particular to a method for predicting the life of sliding bearings. Background Art

[0002] Conventional sliding bearing life prediction methods are often based on vibration signal analysis and empirical model construction. A typical process involves acquiring time-domain vibration signals using accelerometers and extracting spectral features using fast Fourier transforms. Combined with temperature sensor data, the Arrhenius equation is used to establish a temperature-wear rate mapping relationship. Principal component analysis is used to integrate macroscopic parameters such as vibration and temperature, which are then input into a support vector regression model to predict remaining life. Data is securely transmitted using AES-256 encryption. Feature extraction relies on time-domain statistics (such as effective value and kurtosis), and model optimization uses grid search or Bayesian methods. These methods achieve life estimation by establishing an empirical relationship between vibration energy, temperature, and wear volume.

[0003] Existing methods still have some defects: First, the dynamic disturbance modeling and data security mechanisms are insufficient. Traditional vibration analysis ignores the influence of the microstructure of the friction pair (dislocation density, residual stress) on nonlinear dynamics, resulting in the fractal characteristics of the phase space attractor not being effectively quantified; AES encryption relies on classical computing complexity and cannot resist quantum computing attacks, and end-to-end quantum key distribution is not achieved, which poses a risk of data leakage at intermediate nodes. Second, the cross-scale feature fusion capability is weak. Traditional principal component analysis only extracts linearly correlated features and cannot characterize the nonlinear coupling relationship between metal element content (Fe, Cu) and lattice distortion data; the temperature acceleration model is based on the steady-state assumption and does not integrate the dynamic fractal scaling index, resulting in a correlation modeling error of more than 15% between the microscopic wear mechanism and macroscopic performance degradation. These problems seriously restrict the application of high-precision life prediction in strong disturbance industrial scenarios. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a sliding bearing life prediction method to solve the problem of weak cross-scale feature fusion capability in sliding bearing life prediction.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a method for predicting the life of a sliding bearing, which includes real-time collection of chemical composition, lattice distortion data and macroscopic operating status data of the bearing friction pair surface, and encrypting them through quantum keys to generate encrypted multimodal time series data; the edge computing node uses a quantum decryption algorithm to decrypt the multimodal time series data, and through fractional-order chaotic dynamics analysis, reconstructs the phase space of the multimodal time series data and extracts the fractal dimension of the phase space attractor, and at the same time uses a quantum variational autoencoder to fuse chemical and crystal data into a quantum entangled space to generate a high-dimensional data set that integrates macroscopic and microscopic features; constructs a quantum enhanced life prediction model, uses quantum Bayesian optimization for parameter calibration, and predicts the bearing health index based on the high-dimensional data set; evaluates the bearing life based on the bearing health index, and generates a bearing life prediction report.

[0008] As a preferred solution of the sliding bearing life prediction method of the present invention, wherein: the encryption by quantum key to generate encrypted multi-modal time series data is performed as follows:

[0009] The BB84 protocol is used to generate a quantum random number sequence as a one-time key to encrypt the chemical composition, lattice distortion data, and macroscopic operating status data of the bearing friction pair surface. A quantum hash verification tag is generated using a quantum collision-resistant hash function based on the SHA-3 algorithm.

[0010] The quantum hash verification tag is transmitted through the optical fiber quantum channel to the edge computing node equipped with a quantum decryption chip and a hash verification unit, and is parsed through quantum state tomography reconstruction to generate encrypted multimodal time series data containing a timestamp.

[0011] As a preferred solution of the sliding bearing life prediction method described in the present invention, the edge computing node uses a quantum decryption algorithm to decrypt the multimodal time series data, which means recovering the quantum key through the Grover search algorithm in the quantum decryption chip and performing a bit-by-bit decryption operation on the encrypted multimodal time series data.

[0012] As a preferred solution of the sliding bearing life prediction method of the present invention, wherein: the phase space reconstruction of multimodal time series data and the extraction of the fractal dimension of the phase space attractor are performed through fractional-order chaotic dynamics analysis, the steps are as follows:

[0013] The Caputo fractional differential operator is used to reconstruct the phase space of the multimodal data after integrity verification and generate the phase space attractor.

[0014] In the reconstructed phase space, the maximum Lyapunov exponent is calculated using the Wolf algorithm;

[0015] The box dimension of the phase space attractor is calculated to quantify the dynamic behavior as the fractal dimension of the phase space attractor.

[0016] As a preferred solution of the sliding bearing life prediction method of the present invention, the steps of generating a high-dimensional data set integrating macroscopic and microscopic features are as follows:

[0017] The metal element content and dislocation density data are mapped to an 8-qubit Hilbert space using a quantum variational autoencoder, and quantum entanglement gates are used to generate quantum feature vectors.

[0018] Based on the maximum Lyapunov exponent λ1, the fractal dimension of the phase space attractor, the quantum eigenvector and the temperature data, a high-dimensional data set integrating macroscopic and microscopic features is generated through the cascade architecture of quantum principal component analysis and radial basis function kernel support vector.

[0019] As a preferred solution of the sliding bearing life prediction method described in the present invention, the quantum enhanced life prediction model is constructed based on a quantum timing processing layer, a fractal evolution correction layer, a thermodynamic coupling constraint layer, and a quantum optimization and output layer.

[0020] As a preferred solution of the sliding bearing life prediction method described in the present invention, the parameter calibration using quantum Bayesian optimization refers to calibrating the parameters of the quantum enhanced life prediction model using quantum Bayesian optimization based on the temperature-accelerated equivalent wear time τ and wear correction coefficient α.

[0021] As a preferred embodiment of the sliding bearing life prediction method described in the present invention, the method of evaluating the bearing life according to the bearing health index refers to defining the critical threshold of bearing failure through Weibull distribution fitting and quantum cluster analysis, and evaluating the bearing life according to the comparison result between the critical threshold of bearing failure and the bearing health index.

[0022] In a second aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the sliding bearing life prediction method as described in the first aspect of the present invention is implemented.

[0023] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the sliding bearing life prediction method as described in the first aspect of the present invention is implemented.

[0024] The beneficial effects of the present invention are as follows: through fractional-order chaotic dynamics analysis, based on the Caputo differential operator and phase space reconstruction technology, the intrinsic relationship between the nonlinear dynamic behavior of the bearing friction pair and the wear evolution is accurately quantified; a quantum-enhanced life prediction model is constructed through a quantum variational autoencoder, and the model parameters are dynamically calibrated in combination with the quantum Bayesian optimization algorithm, thereby improving the accuracy and reliability of bearing life prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0026] Figure 1 This is a flow chart of the sliding bearing life prediction method in Example 1.

[0027] Figure 2 Schematic diagram of the quantum encryption transmission process in Example 1.

[0028] Figure 3 This is a flow chart of phase space reconstruction and fractal analysis in Example 1.

[0029] Figure 4 Flowchart for constructing the quantum enhanced lifetime prediction model in Example 1. DETAILED DESCRIPTION

[0030] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0031] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0032] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0033] Example 1, with reference to Figures 1 to 4 This embodiment provides a method for predicting the life of a sliding bearing, comprising the following steps:

[0034] S1. Real-time collection of the chemical composition, lattice distortion data, and macroscopic operating status data of the bearing friction pair surface, and encryption through quantum key to generate encrypted multi-modal time series data.

[0035] It should be noted that the chemical composition of the bearing friction pair surface includes the content of metal elements such as Fe and Cu. During the operation of the sliding bearing, the micro laser induced breakdown spectroscopy probe is used;

[0036] Lattice distortion data, including microstructural parameters such as dislocation density and residual stress, are collected by micro-area X-ray diffraction;

[0037] Furthermore, the macro-operation status data includes real-time monitoring indicators such as vibration acceleration signals and temperature changes. The vibration acceleration signal is collected by a piezoelectric acceleration sensor installed on the bearing seat, and the temperature changes are monitored in real time by an embedded platinum resistance temperature sensor. The temperature sensor fits tightly to the surface of the bearing outer ring and adopts a four-wire connection method to eliminate the influence of wire resistance. The analog signals of the two sensors are converted by a 24-bit high-precision ADC and recorded synchronously with the timestamp.

[0038] The quantum key distribution terminal uses the BB84 protocol to generate a quantum random number sequence as a one-time key to encrypt the chemical composition, lattice distortion data and macroscopic operating status data of the bearing friction pair surface. At the same time, it uses a quantum collision-resistant hash function based on the SHA-3 algorithm to generate a quantum hash verification tag.

[0039] The workflow of the quantum key distribution terminal is as follows: First, a quantum random number sequence is generated using the BB84 protocol. Specifically, a laser diode generates weakly coherent light pulses, which are encoded into four polarization states by a polarization modulator and then transmitted through a fiber optic channel. The receiving end uses a polarization beam splitter to randomly select a measurement basis for single-photon detection. After basis comparison and key negotiation, a one-time key is generated. The one-time key is directly used to encrypt the chemical composition spectrum data of the bearing friction pair surface, the electron backscatter diffraction pattern of lattice distortion, and the temperature-vibration time-domain signal of the macroscopic operating state. The encryption process uses a one-time pad with bit-by-bit exclusive-or (XOR) to ensure information-theoretic security. Simultaneously, the raw data is processed using the Keccak-f

[1600] permutation function of the SHA-3 algorithm, and a 512-bit quantum hash verification tag is generated through 12 rounds of absorption-squeezing.

[0040] The quantum hash verification tag is transmitted through the optical fiber quantum channel to the edge computing node equipped with a quantum decryption chip and a hash verification unit, and is parsed through quantum state tomography reconstruction to generate encrypted multimodal time series data containing a timestamp.

[0041] Furthermore, the quantum hash verification tag is transmitted through a single-mode optical fiber quantum channel, and polarization encoding and time division multiplexing technology are used to encode the 512-bit hash value and timestamp information into a quantum state sequence. The 1550nm wavelength laser pulse passes through a lithium niobate modulator to produce four polarization states, corresponding to the quantum state preparation process of the BB84 protocol. The quantum signal is transmitted to the edge computing node through the optical fiber, and the photoelectric conversion is completed by the superconducting nanowire single-photon detector array. The measurement results are recorded as quantum projection measurement data. The quantum state tomography reconstruction method uses the maximum likelihood estimation algorithm to process the measurement data, reconstructs the quantum state density matrix through Pauli basis measurement, and restores the encrypted data packet containing the precise timestamp. The timestamp information is synchronized with the original data through the optical fiber transmission delay compensation algorithm, with an accuracy of nanoseconds. The encrypted multimodal time series data finally generated contains monitoring parameters such as bearing vibration spectrum, temperature gradient, and acoustic emission signal. All data fields are encrypted using the AES-256 algorithm and stored in conjunction with the quantum hash verification tag.

[0042] S2. Edge computing nodes use quantum decryption algorithms to decrypt multimodal time series data, and through fractional-order chaotic dynamics analysis, reconstruct the phase space of multimodal time series data and extract the fractal dimension of the phase space attractor. At the same time, quantum variational autoencoders are used to fuse chemical and crystal data into quantum entanglement space to generate a high-dimensional data set that integrates macroscopic and microscopic features.

[0043] The edge computing node recovers the quantum key through the Grover search algorithm in the quantum decryption chip and performs bit-by-bit decryption operations on the encrypted multimodal time series data.

[0044] Furthermore, after receiving the encrypted information transmitted by the quantum key distribution terminal, the quantum decryption chip in the edge computing node first captures the quantum state signal using a superconducting quantum interference device. The Grover search algorithm is implemented on a quantum logic gate array, using Hadamard gates to create a superposition state. The target key state is marked using an oracle black box, and then a diffusion operator is applied to amplify the target amplitude. The quantum key recovery process is performed on a 20-qubit processor, with each iteration increasing the probability amplitude of the correct key by the square. The decryption operation uses a one-time pad method, performing a bit-by-bit XOR operation on the recovered quantum random number sequence with the AES-256-encrypted multimodal time series data. The decryption process synchronously verifies the quantum hash verification tag, recalculates the 512-bit hash value of the received data using the SHA-3 algorithm, and compares it with the transmitted quantum hash verification tag to verify data integrity. The decrypted data contains timing parameters such as the bearing vibration spectrum, temperature gradient, and acoustic emission signal, with nanosecond-level timestamp synchronization.

[0045] The decrypted multimodal time series data is verified for integrity by a hash verification unit and then separated into the chemical composition of the bearing friction pair surface, lattice distortion data, and macroscopic operating status data;

[0046] Furthermore, the decrypted multimodal time series data is first input into a hash verification unit, where the SHA-3 algorithm is used to perform a hash operation on the complete data block, generating a 512-bit verification code that is then compared bit-by-bit with the transmitted quantum hash verification tag. The verified data stream enters the data separation process and is parsed according to a predefined frame structure: chemical composition data on the bearing friction pair surface is extracted from the X-ray energy spectrum characteristic peak intensity values, and the elemental composition is matched using the NIST standard reference database; lattice distortion data is parsed from the electron backscatter diffraction pattern, and Kikuchi bands are identified and lattice orientation errors are calculated using the Hough transform; macroscopic operating status data is separated into triaxial vibration acceleration time-domain waveforms, infrared thermal image temperature matrices, and acoustic emission signal envelope curves. The data separation process strictly maintains synchronization with the original timestamps, and all data types are stored in IEEE 754 double-precision floating-point format.

[0047] It should be noted that frame structure parsing is defined based on the IEEE 1451.4 standard and quantum key encryption, and is used to achieve synchronous separation and integrity verification of encrypted multimodal time series data. For example, a 256-bit session key generated by quantum key distribution (such as the shared secret key of NIST P-521 elliptic curve cryptography) is used to implement AES-256 encryption on each frame of multimodal time series data, and anti-replay attack is achieved through the quantum random number (QRNG generation) embedded in the frame header.

[0048] Based on the macroscopic operating status data, the Caputo fractional differential operator is used to reconstruct the phase space of the multimodal data after integrity verification to generate the phase space attractor.

[0049] Furthermore, after preprocessing the macroscopic operating state data, the Caputo fractional-order differential operator is used to perform 0.5-order differential processing on the vibration acceleration time-domain signal, eliminating measurement noise while retaining nonlinear characteristics. The phase space reconstruction process strictly follows the Takens embedding theorem, calculating the time delay parameter through the mutual information function, and using the false nearest neighbor method to determine the minimum embedding dimension. The reconstruction algorithm combines the original sampling values ​​of the vibration signal with its fractional-order differential results into a state vector. The temperature gradient data is converted into state points through a two-dimensional phase space mapping, and the acoustic emission signal envelope curve is Hilbert transformed to form phase space coordinates. The resulting phase space attractor exhibits a three-dimensional manifold structure, with each state point precisely corresponding to the original data timestamp, fully characterizing the nonlinear dynamic characteristics of the bearing operating state.

[0050] In the reconstructed phase space, the maximum Lyapunov exponent λ1 is calculated by the Wolf algorithm, and the expression is:

[0051]

[0052] Among them, L(t k-1 ) is the distance between adjacent points in the k-1th step, L'(t k ) is the new spacing after the evolution time Δt, n is the total number of iterations, Δt is the time step of trajectory evolution, k is the number of current iterations, t k represents the kth discrete time point starting from the initial moment, t k-1 Represents the previous discrete time point of the current evolution step.

[0053] It should be noted that the process of calculating the maximum Lyapunov exponent λ1 is as follows: First, select the initial reference point and its nearest neighbor on the phase space attractor and record the initial distance. During the trajectory evolution, the new distance L'(t k ), when the distance exceeds the preset threshold, the adjacent points are reselected and the evolution direction is adjusted, and the logarithmic divergence rate is accumulated. The calculation process continues iteratively until the entire phase space attractor is covered. The maximum Lyapunov exponent λ1 is finally obtained by averaging the accumulated values, reflecting the exponential divergence of the trajectory. The phase space reconstruction parameters remain exactly the same as in the previous processing. The phase space coordinates of adjacent points are strictly recorded at each iteration. The original timestamp information is retained during the trajectory evolution process, and the distance measurement is based on the Euclidean distance formula.

[0054] The box dimension of the phase space attractor is calculated to quantify the dynamic behavior as the fractal dimension of the phase space attractor, which is expressed as:

[0055]

[0056] Where D is the fractal dimension of the attractor in phase space, N(∈) is the minimum number of grids required to completely cover the attractor at scale ∈;

[0057] It should be noted that the calculation process of the fractal dimension of the phase space attractor is as follows: in the reconstructed phase space coordinate system, the attractor is spatially covered using a gradually refined cubic grid. The initial grid size is set to half of the maximum characteristic length of the attractor, and then the grid size is gradually reduced according to the bisection method. After each grid division, the minimum number of grids containing the attractor points is accurately counted, and the logarithmic value of the grid size and the number of grids is recorded. The fractal dimension of the phase space attractor is obtained by fitting the data points in the double logarithmic coordinate system using the least squares method. After each grid adjustment, the spatial distribution of the attractor points is rescanned to ensure the accuracy of the grid count. The fractal dimension finally obtained accurately quantifies the geometric complexity of the phase space attractor. The larger the value, the more irregular the attractor structure. The fractal dimension of the phase space attractor directly characterizes the nonlinear dynamic characteristics of the bearing operating state.

[0058] The metal element content data and dislocation density data are mapped to the 8-qubit Hilbert space through a quantum variational autoencoder, and the quantum entanglement gate is used to generate the quantum feature vector;

[0059] Furthermore, the metal element content and dislocation density data are first normalized and preprocessed before being input into the encoder portion of the quantum variational autoencoder. The encoder is implemented using a parameterized quantum circuit and consists of eight qubits, each of which receives input data via an Ry rotation gate. The quantum circuit includes multiple entanglement layers and parameterized rotation gates, with the entanglement layers using CZ gates to entangle adjacent qubits. The data mapping process uses variational optimization to match the classical distribution of the input data with the statistical distribution of the quantum state measurement results. Quantum eigenvectors are generated by constructing correlations in Hilbert space using quantum entanglement gates. Stable quantum eigenvectors are obtained through repeated preparation and measurement of quantum states. The decoder portion of the quantum variational autoencoder uses a mirror-symmetric quantum circuit structure to verify the encoding effect.

[0060] Based on the maximum Lyapunov exponent λ1, the fractal dimension of the phase space attractor, the quantum eigenvector and the temperature data, a high-dimensional data set integrating macroscopic and microscopic features is generated through the cascade architecture of quantum principal component analysis and radial basis function kernel support vector.

[0061] Furthermore, quantum principal component analysis is first used to reduce the dimensionality of high-dimensional quantum eigenvectors, and the main characteristic components are extracted through quantum computing and converted into HDF5 format; then the quantum eigenvectors are input together with the maximum Lyapunov exponent and the fractal dimension of the phase space attractor into a support vector machine based on the radial basis function kernel, and the nonlinear mapping capability is used to realize feature fusion, and finally a high-dimensional data set of macroscopic and microscopic features is output.

[0062] S3. Build a quantum-enhanced life prediction model, use quantum Bayesian optimization for parameter calibration, and predict the bearing health index based on high-dimensional data sets.

[0063] Based on the quantum timing processing layer, fractal evolution correction layer, thermodynamic coupling constraint layer and quantum optimization and output layer, a quantum enhanced lifetime prediction model is constructed.

[0064] The quantum timing processing layer is as follows:

[0065] Quantum amplitude coding is used to encode high-dimensional data sets into quantum amplitudes, and then processed in parallel through rotation gates and entanglement gates in a 6-layer parameterized quantum circuit to generate the weight matrix of the forget gate.

[0066] Furthermore, high-dimensional datasets are mapped to quantum states through quantum amplitude encoding, converting each data point into a qubit's probability amplitude. A six-layer parameterized quantum circuit consists of alternating rotation gates and entanglement gates. The rotation gates use Ry and Rz gates to transform single qubits, while the entanglement gates use CZ gates to establish correlations between qubits. The parameters of each quantum gate layer are adjusted using a classical optimization algorithm, allowing the quantum state evolution to capture the data characteristics. Finally, the weight matrix of the forget gate is extracted from the quantum state measurement. This weight matrix represents the nonlinear transformation of the high-dimensional data in Hilbert space.

[0067] The weight matrix of the forget gate is input into the variational quantum eigensolver (VQE), and the quantum circuit parameters are optimized by gradient descent to generate temperature-adaptive quantum states.

[0068] Furthermore, the forget gate's weight matrix is ​​fed into a variational quantum eigensolver, which constructs the target optimization problem using the Hamiltonian. The quantum circuit employs a parameterized unitary transformation and uses a gradient descent algorithm to adjust the rotation gate angle to minimize the expected value. Each iteration updates the quantum gate parameters, bringing the quantum state closer to the target eigenstate. Upon completion of the optimization, a temperature-adaptive quantum state is generated, which encodes the coupling characteristics of the temperature field and high-dimensional data.

[0069] The temperature-adaptive quantum state is reduced in dimension through Pauli-Z basis projection to generate a time series eigenvector.

[0070] Furthermore, the temperature-adaptive quantum state is projected using Pauli-Z basis measurements. The measurement result of each qubit corresponds to a dimension of the eigenvector. The quantum state is prepared and measured multiple times, and the expected value of the Z-basis projection results is calculated to form a reduced-dimensional time series eigenvector. The time series eigenvector preserves the dynamic evolution characteristics of the original data while eliminating redundant information. The projection process strictly maintains timestamp synchronization to ensure the temporal correlation between the time series eigenvector and the bearing operating status.

[0071] The fractal evolution correction layer is as follows:

[0072] Based on the fractal dimension and time series eigenvector of the phase space attractor, the fractal oscillation function fitting and regularization optimization are used to calculate the dynamic fractal scaling index, which is expressed as:

[0073]

[0074] Where s is the dynamic fractal scaling index, D is the fractal dimension of the phase space attractor, M(W(T)) represents the logarithm of the grid coverage number of the fractal oscillation signal W(T) at the discretized grid unit size, W(T) represents the fractal oscillation signal, β is the material wear coefficient calibrated by the ASTM G65 test, & is the regularization weight, s0 is the mean value obtained based on the historical dynamic fractal scaling index, d is the box dimension of the phase space attractor, and T represents the normalized time series of the vibration signal;

[0075] Furthermore, the dynamic fractal scaling index is calculated as follows: the fractal dimension and time series eigenvector of the phase space attractor are used as input. The fractal oscillation signal is discretized on a multiscale grid, and the logarithm of the grid coverage number is calculated. The L-BFGS optimization algorithm is used to search for the optimal solution within a preset interval. The objective function includes a difference measure between the fractal dimension and the derivative term of the phase space attractor, as well as a regularization constraint between the dynamic fractal scaling index and the historical mean. The material wear coefficient is kept constant during the optimization process, and the normalized time series of the vibration signal is used as the independent variable. The grid coverage characteristics of the fractal oscillation signal are updated and the derivative term is recalculated at each iteration. The final result is the dynamic fractal scaling index, which accurately quantifies the multiscale dynamic characteristics of the bearing degradation process.

[0076] It should be noted that the fractal oscillation signal is a non-stationary time domain signal with self-similar characteristics that is generated by a fractal oscillation function;

[0077] The thermodynamic coupling constraint layer is as follows:

[0078] Based on the dynamic fractal scaling index and the fractal dimension of the phase space attractor, a correction pair is constructed by the least squares method, and then the correction pair is mapped to generate the wear correction coefficient α through a three-layer fully connected neural network.

[0079] Furthermore, the dynamic fractal scaling exponent and the fractal dimension of the phase space attractor are first constructed into a correction pair using the least squares method, establishing a linear relationship between the two. This correction pair is then processed into a three-layer fully connected neural network. The first layer uses the ReLU activation function to nonlinearly map the input features to the latent space. The second layer uses the Sigmoid function to extract high-order correlations between features. The third layer uses a linear transformation to map the latent features to a scalar output. The neural network training process uses a backpropagation algorithm to update weights, and the loss function is defined as the mean squared error between the predicted wear correction coefficient α and the experimental calibration value. After training is complete, the neural network maps the correction pair of the dynamic fractal scaling exponent and the fractal dimension of the phase space attractor to generate the wear correction coefficient α. This current coefficient is then used to calibrate the accuracy of the subsequent life prediction model.

[0080] The standard Arrhenius equation is used to calculate the temperature-accelerated equivalent wear time, which is expressed as:

[0081]

[0082] Where τ is the temperature-accelerated equivalent wear time, τ0 is the baseline wear rate, E is the wear activation energy obtained by differential scanning calorimetry, B is the Boltzmann constant, and V(t1) is the bearing temperature data at time t1;

[0083] Furthermore, the calculation process of the temperature-accelerated equivalent wear time is as follows: First, the wear activation energy E of the bearing material is obtained through differential scanning calorimetry experiments, and the baseline wear rate τ0 under standard operating conditions is recorded. During the operation of the bearing, the temperature data V(t1) is collected in real time, and the temperature data and the Boltzmann constant B are substituted into the standard Arrhenius equation for calculation. When calculating, the temperature data V(t1) is first converted to the thermodynamic temperature scale, and then the exponential term is calculated. The value of is then multiplied by the baseline wear rate τ0 to obtain the temperature-accelerated equivalent wear time τ. The entire calculation process strictly maintains the consistency of physical units, with the units of the wear activation energy E matching those of the Boltzmann constant B. Temperature data V(t1) is updated using a sliding time window to ensure that the calculation results reflect the real-time temperature impact of the bearing. The final output, equivalent wear time τ, is used to correct actual operating time to accurately quantify the accelerated effect of temperature on the bearing wear process.

[0084] It should be noted that a pin-on-disc wear tester conforming to ASTM G99 was used to prepare specimens made of the same material as the bearings. Continuous wear testing was performed under rated load and fixed speed. The wear volume change was measured regularly using a surface profilometer. The baseline wear rate τ0 under standard operating conditions was obtained by dividing the wear volume per unit time by the contact area.

[0085] Based on the temperature-accelerated equivalent wear time τ and wear correction coefficient α, quantum Bayesian optimization is used to calibrate the parameters of the quantum enhanced life prediction model.

[0086] Furthermore, the quantum Bayesian optimization process uses the temperature-accelerated equivalent wear time τ and the wear correction coefficient α as inputs. The process first defines a Gaussian process prior distribution in parameter space. A quantum sampler generates candidate parameter combinations in Hilbert space, and the quantum circuit is constructed using parameterized rotation gates and entanglement gates. In each iteration, the quantum processor evaluates the lifetime prediction errors of multiple quantum-enhanced lifetime prediction model parameter combinations in parallel, while the classical processor updates the Gaussian process posterior distribution. The optimization goal is to minimize the mean squared error of the parameters of the quantum-enhanced lifetime prediction model. The final output is the calibrated quantum-enhanced lifetime prediction model parameters, specifically the temperature-accelerated equivalent wear time τ, the wear correction coefficient α, the forget gate weight matrix, the dynamic fractal scaling exponent s, and the fractal dimension D of the phase space attractor.

[0087] The high-dimensional data set is input into the quantum enhanced life prediction model to predict the bearing health index, which is expressed as:

[0088]

[0089] Where H(t1) is the bearing health index at t1, s(t1) is the dynamic fractal scaling index at t1, and R max is the maximum allowed feature offset, Y0 is the initial healthy reference quantum state, and Y(t1) is the high-dimensional dataset at t1.

[0090] S4. Evaluate the bearing life according to the bearing health index and generate a bearing life prediction report.

[0091] Based on historical life cycle data, the critical threshold H1 of bearing failure is defined through Weibull distribution fitting and quantum cluster analysis;

[0092] When H(t1)≥H1, it is considered that the current bearing life is long;

[0093] When H(t1)

[0094] It should be noted that the historical full life cycle data contains multi-dimensional information such as vibration spectrum, temperature time series, and phase space reconstruction characteristics. The Weibull distribution is used to fit the time distribution characteristics of the bearing from the initial state to complete failure. At the same time, quantum cluster analysis is used to perform pattern recognition on the degradation trajectory in the high-dimensional feature space. The vibration energy mutation point, temperature acceleration inflection point and micro-morphology electron microscope scanning results are cross-validated to finally determine the critical failure threshold H1 of the bearing. The quantum cluster analysis is implemented based on the quantum K-means algorithm of IBM Qiskit. The value range of the critical failure threshold H1 of the bearing is limited to the range of 15%-25% of the bearing health index HI. This range also meets the condition that the Weibull cumulative failure probability reaches 92%±3%.

[0095] Generate a bearing life prediction report based on the bearing life assessment results.

[0096] It should be noted that the bearing life prediction report includes the bearing health index, bearing life assessment results and maintenance recommendations. For example, when H(t1) is lower than H1 for three consecutive times, it is recommended to replace the bearing immediately; when the vibration amplitude exceeds the ISO standard, it is recommended to stop the machine and inspect the bearing immediately.

[0097] This embodiment also provides a computer device suitable for the sliding bearing life prediction method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the sliding bearing life prediction method proposed in the above embodiment.​

[0098] The computer device may be a terminal, comprising a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner may be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a button, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse.

[0099] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the sliding bearing life prediction method proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, disk or optical disk.

[0100] In summary, the present invention uses: fractional-order chaotic dynamics analysis, based on the Caputo differential operator and phase space reconstruction technology, to accurately quantify the intrinsic relationship between the nonlinear dynamic behavior and wear evolution of bearing friction pairs; constructs a quantum-enhanced life prediction model through quantum variational autoencoders, and dynamically calibrates the model parameters in combination with the quantum Bayesian optimization algorithm, thereby improving the accuracy and reliability of bearing life prediction.

[0101] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for predicting the life of a sliding bearing, characterized by: include, Real-time collection of chemical composition, lattice distortion data, and macroscopic operating status data of the bearing friction pair surface, which are then encrypted using quantum keys to generate encrypted multi-modal time series data. Edge computing nodes use quantum decryption algorithms to decrypt multimodal time series data. They also use fractional-order chaotic dynamics analysis to reconstruct the phase space of multimodal time series data and extract the fractal dimension of the phase space attractor. At the same time, they use quantum variational autoencoders to fuse chemical and crystal data into quantum entanglement space, generating a high-dimensional dataset that integrates macroscopic and microscopic features. Build a quantum-enhanced life prediction model, use quantum Bayesian optimization for parameter calibration, and predict bearing health index based on high-dimensional data sets; Evaluate bearing life based on the bearing health index and generate a bearing life prediction report.

2. The method for predicting the life of a sliding bearing according to claim 1, wherein: The steps of encrypting with quantum key to generate encrypted multi-modal time series data are as follows: The BB84 protocol is used to generate a quantum random number sequence as a one-time key to encrypt the chemical composition, lattice distortion data, and macroscopic operating status data of the bearing friction pair surface. A quantum hash verification tag is generated using a quantum collision-resistant hash function based on the SHA-3 algorithm. The quantum hash verification tag is transmitted through the optical fiber quantum channel to the edge computing node equipped with a quantum decryption chip and a hash verification unit, and is parsed through quantum state tomography reconstruction to generate encrypted multimodal time series data containing a timestamp.

3. The method for predicting the life of a sliding bearing according to claim 2, wherein: The edge computing node uses a quantum decryption algorithm to decrypt multimodal time series data, which means recovering the quantum key through the Grover search algorithm in the quantum decryption chip and performing a bit-by-bit decryption operation on the encrypted multimodal time series data.

4. The method for predicting the life of a sliding bearing according to claim 1, wherein: The phase space of multimodal time series data is reconstructed and the fractal dimension of the phase space attractor is extracted by fractional-order chaotic dynamics analysis. The steps are as follows: The Caputo fractional differential operator is used to reconstruct the phase space of the multimodal data after integrity verification and generate the phase space attractor. In the reconstructed phase space, the maximum Lyapunov exponent λ1 is calculated by the Wolf algorithm; The box dimension of the phase space attractor is calculated to quantify the dynamic behavior as the fractal dimension of the phase space attractor.

5. The method for predicting the life of a sliding bearing according to claim 4, wherein: The steps of generating a high-dimensional dataset that integrates macroscopic and microscopic features are as follows: The metal element content and dislocation density data are mapped to an 8-qubit Hilbert space using a quantum variational autoencoder, and quantum entanglement gates are used to generate quantum feature vectors. Based on the maximum Lyapunov exponent λ1, the fractal dimension of the phase space attractor, the quantum eigenvector and the temperature data, a high-dimensional data set integrating macroscopic and microscopic features is generated through the cascade architecture of quantum principal component analysis and radial basis function kernel support vector.

6. The method for predicting the life of a sliding bearing according to claim 1, wherein: The quantum enhanced lifetime prediction model is constructed based on a quantum timing processing layer, a fractal evolution correction layer, a thermodynamic coupling constraint layer, and a quantum optimization and output layer.

7. The method for predicting the life of a sliding bearing according to claim 6, wherein: The parameter calibration using quantum Bayesian optimization refers to calibrating the parameters of the quantum enhanced life prediction model using quantum Bayesian optimization based on the temperature-accelerated equivalent wear time τ and the wear amount correction coefficient α.

8. The method for predicting the life of a sliding bearing according to claim 7, wherein: The method of evaluating the bearing life according to the bearing health index is to define a critical bearing failure threshold through Weibull distribution fitting and quantum cluster analysis, and evaluate the bearing life according to a comparison result between the critical bearing failure threshold and the bearing health index.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the sliding bearing life prediction method according to any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the sliding bearing life prediction method according to any one of claims 1 to 8 are implemented.

Images