Adaptive test parameter optimization method

Abstract

The invention discloses a self-adaptive test parameter optimization method, and relates to the technical field of parameter optimization, and the method comprises the steps: collecting an original sensing signal in a mechanical test system, and carrying out the preprocessing of the original sensing signal; based on the preprocessed data set, constructing a four-dimensional space-time tensor, and executing improved parallel factor tensor decomposition to obtain a decoupled core factor and space-time feature component matrix; performing cross-modal alignment on the decoupled core factor and the time-space feature component matrix through a wear feature channel and an acoustic emission feature channel to obtain a fused cross-modal feature vector; performing crack growth rate prediction on the fused cross-modal feature vectors to obtain a crack risk level, and adjusting a strategy through dynamic parameters to obtain an optimized parameter set; the problem of signal noise and time mismatch is solved through multi-mode signal preprocessing, and high signal-to-noise ratio input is provided for subsequent analysis in combination with wavelet denoising, space-time alignment and double-domain feature extraction.

Description

Technical Field

[0001] The present invention relates to the technical field of parameter optimization, in particular to an adaptive test parameter optimization method. Background Art

[0002] With the rapid advancement of industrial inspection technology toward intelligent and high-precision capabilities, the demand for adaptive test parameter optimization methods in the health management of complex equipment continues to grow. Current technology architectures primarily rely on three core modules: multi-source sensor data fusion, dynamic modeling, and closed-loop control. In terms of data acquisition and preprocessing, multi-sensor collaborative sensing technology has enabled the simultaneous acquisition of heterogeneous signals. Combined with wavelet threshold denoising and dynamic time warping algorithms, it significantly improves the signal-to-noise ratio and temporal consistency of the original signals. In the area of ​​feature modeling, multidimensional data representation methods based on tensor decomposition effectively decouple the multi-physics coupling characteristics of mechanical systems by constructing a multidimensional correlation model between spatiotemporal dimensions and physical modes, providing theoretical support for the analysis of failure modes such as wear and cracks. Furthermore, the combination of model predictive control and deep reinforcement learning provides a technical framework for online dynamic parameter adjustment, showing particular promise in load and speed optimization under complex operating conditions. However, existing methods still face significant technical bottlenecks in key areas such as multimodal feature correlation modeling, physical constraint embedding, and safety boundary collaborative control, limiting optimization accuracy and system reliability in complex scenarios.

[0003] Existing technologies are limited in three key areas: First, they lack the ability to model spatiotemporal correlations in multi-source data. Traditional tensor decomposition methods fail to fully account for dynamic variations in sensor signal-to-noise ratios, resulting in weak robustness to low-quality data and the introduction of noise during the decomposition process. Furthermore, the lack of spatiotemporal sparsity constraints limits the physical interpretability of feature components. Second, cross-modal feature fusion suffers from flaws in deep correlation modeling. Existing methods often rely on linear mapping or static attention mechanisms for feature alignment, which struggles to capture the nonlinear dynamic correlations between heterogeneous features. Furthermore, they fail to incorporate adversarial training constraints, causing the distribution of fused features to deviate from actual physical laws, impacting the generalization performance of fault prediction models. Finally, there is a lack of a coordinated mechanism between dynamic parameter optimization and the safe operating boundaries of equipment. Mainstream optimization algorithms are single-objective oriented and lack a multi-objective coordinated framework for efficiency, energy consumption, and safety thresholds. The parameter adjustment process lacks a real-time safety enhancement mechanism, which can easily lead to equipment overload or critical component failure. Therefore, an adaptive test parameter optimization method that deeply integrates dynamic feature decoupling, cross-modal adversarial alignment, and safety enhancement optimization is urgently needed to systematically overcome the bottlenecks of existing technologies. Summary of the Invention

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

[0005] Therefore, the present invention provides an adaptive test parameter optimization method to solve the problems of incomplete feature decoupling, failure of cross-modal feature adversarial alignment, and disconnection between dynamic parameter optimization and collaborative control of equipment safety boundaries caused by insufficient spatiotemporal correlation modeling of multi-source heterogeneous sensor signals.

[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 an adaptive test parameter optimization method, which comprises:

[0008] The original sensor signals in the mechanical testing system are collected and preprocessed. Based on the preprocessed data set, a four-dimensional space-time tensor is constructed, and an improved parallel factor tensor decomposition is performed to obtain the decoupled core factors and space-time feature component matrices. The decoupled core factors and space-time feature component matrices are cross-modally aligned through the wear feature channel and the acoustic emission feature channel to obtain a fused cross-modal feature vector. The fused cross-modal feature vector is used to predict the crack propagation rate to obtain the crack risk level and a dynamic parameter adjustment strategy is used to obtain the optimized parameter set.

[0009] As a preferred solution of the adaptive test parameter optimization method of the present invention, the preprocessing is performed in the following specific steps:

[0010] The original sensor data is denoised using wavelet transform to obtain denoised sensor data; a time warp function is defined, and a spatiotemporal alignment model is constructed through nonlinear mapping combined with spatiotemporal variation parameter optimization. The denoised sensor data is input into the spatiotemporal alignment model to obtain a multimodal signal pair with time axis alignment; a dual-domain feature extraction module is constructed through the technical means of joint time-frequency analysis and cross-modal feature engineering, and the multimodal signal pair with time axis alignment is input into the dual-domain feature extraction module to obtain the preprocessed data set.

[0011] As a preferred solution of the adaptive test parameter optimization method of the present invention, the specific steps of constructing the four-dimensional space-time tensor are as follows:

[0012] Based on the preprocessed data set, the spatial resolution and temporal resolution are determined by analyzing the minimum spatial unit and event interval of the data sampling interval; the monitoring area is gridded according to the spatial resolution to generate a three-dimensional coordinate system, and the data are arranged by timestamp based on the temporal resolution to generate a data sequence that changes with time; the spatial coordinates of the three-dimensional coordinate system are combined with the data sequence that changes with time through space-time tensorization, and a four-dimensional space-time tensor containing three-dimensional and time dimensions is created through dimensional expansion.

[0013] As a preferred solution of the adaptive test parameter optimization method of the present invention, wherein: the improved parallel factor tensor decomposition is performed, the specific steps are as follows:

[0014] The four-dimensional space-time tensor is converted into a weighted combination of four factor matrices: time, longitude, latitude, and variable. The constraints of time smoothness, space continuity, and variable sparsity are added to form a regularized objective function. Based on the regularized objective function, a random matrix is ​​generated by random initialization method through tensor structure analysis. The initial guess is generated using high-order singular value decomposition, and the random matrix and the results of the initial guess are fused through alternating projection to generate an initial factor matrix. The initial factor matrix is ​​updated through hybrid parallel factors to obtain an updated factor matrix. According to the updated factor matrix, the rank contribution is calculated and the redundant ranks are deleted to obtain the decoupled core factor.

[0015] As a preferred solution of the adaptive test parameter optimization method of the present invention, wherein: the cross-modal alignment is performed through the wear characteristic channel and the acoustic emission characteristic channel, and the specific steps are as follows:

[0016] The mechanical wear components in the decoupled core factors and the wear dimensions in the spatiotemporal feature component matrix are used to extract local wear features through a convolutional neural network, and the dynamic evolution law of the wear features is captured using Transformer to obtain the wear modal feature vector. The acoustic emission energy components in the decoupled core factors and the high-frequency vibration signals in the spatiotemporal feature matrix are decomposed and clustered by wavelet packets to obtain the acoustic emission modal feature vector. The wear modal feature vector and the acoustic emission modal feature vector are projected into a unified low-dimensional space to eliminate the differences in modal scale and distribution. The low-dimensional features are then spliced ​​and fused with the original high-dimensional modal features through residual connections to obtain the fused cross-modal feature vector.

[0017] As a preferred solution of the adaptive test parameter optimization method of the present invention, wherein: the crack risk level is obtained by predicting the crack growth rate, the steps are as follows:

[0018] The fused cross-modal feature vector is spliced ​​with the operating condition parameters, and the crack length is predicted through the physical-data fusion model to obtain the crack length prediction value; based on the equipment material properties and historical failure data, the crack length threshold is set through experimental calibration and statistical distribution analysis; the crack length prediction value is compared with the preset threshold, and the risk level is divided to obtain the crack risk level.

[0019] As a preferred solution of the adaptive test parameter optimization method of the present invention, wherein: the optimization parameter set is obtained through the dynamic parameter adjustment strategy, the steps are as follows:

[0020] An optimization model is established based on the optimization objectives and constraints for suppressing crack propagation and reducing energy consumption defined according to the crack risk level and current operating parameters. The current equipment status is input into the optimization model to predict future crack trends, obtain a preliminary optimized parameter candidate set, and perform safety enhancement verification to obtain a set of optimized parameters.

[0021] In the second aspect, the present invention provides an adaptive test parameter optimization system, including: a preprocessing module: collecting the original sensor signals in the mechanical testing system and performing preprocessing. A four-dimensional tensor decomposition module: constructing a four-dimensional space-time tensor based on the preprocessed data set, and performing an improved parallel factor tensor decomposition to obtain the decoupled core factor and space-time feature component matrix. A cross-modal feature fusion module: performing cross-modal alignment on the decoupled core factor and space-time feature component matrix through the wear feature channel and the acoustic emission feature channel to obtain a fused cross-modal feature vector. A crack risk control optimization module: using the fused cross-modal feature vector to predict the crack propagation rate to obtain the crack risk level and obtaining the optimized parameter set through a dynamic parameter adjustment strategy.

[0022] In a third 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 adaptive test parameter optimization method described in the first aspect of the present invention is implemented.

[0023] In a fourth 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 adaptive test parameter optimization method as described in the first aspect of the present invention is implemented.

[0024] The beneficial effects of the present invention are as follows: the present invention solves the problems of signal noise and time mismatch through multimodal signal preprocessing, and combines wavelet denoising, space-time alignment and dual-domain feature extraction to provide high signal-to-noise ratio input for subsequent analysis; the improved tensor decomposition is based on four-dimensional space-time modeling and constraint optimization algorithm, explicitly associates space-time, modal and working condition dimensions, improves the physical interpretability of feature decoupling and suppresses noise interference; cross-modal adversarial alignment uses CNN, Transformer and wavelet models to collaboratively analyze wear and acoustic emission characteristics, eliminates modal scale differences through low-dimensional projection, and constructs fusion features driven by physical laws; crack risk control optimization integrates physical models and data-driven methods, balances equipment safety and efficiency through multi-objective optimization and safety verification, and realizes the coordination of crack suppression and energy consumption control; the entire link forms a "data purification-feature decoupling-modal fusion-dynamic decision-making" closed loop, which significantly improves the health management accuracy and reliability of complex mechanical systems. 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 Flowchart of the adaptive test parameter optimization method in Example 1.

[0027] Figure 2 This is a flowchart of the cross-modal feature fusion module flow chart in Example 1.

[0028] Figure 3 This is a flow chart of the crack risk control optimization module flow chart in Example 1.

[0029] Figure 4 This is a flow chart of the adaptive test parameter optimization system 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 Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides an adaptive test parameter optimization method, comprising the following steps:

[0034] S1: Collect the raw sensor signals from the mechanical testing system and perform preprocessing.

[0035] S1.1: Use wavelet transform to denoise the original sensor data to obtain denoised sensor data; define the time warp function, construct a spatiotemporal alignment model through nonlinear mapping combined with spatiotemporal variation parameter optimization, input the denoised sensor data into the spatiotemporal alignment model to obtain a multimodal signal pair with time axis alignment; construct a dual-domain feature extraction module through the technical means of joint time-frequency analysis and cross-modal feature engineering, and input the multimodal signal pair with time axis alignment into the dual-domain feature extraction module to obtain the preprocessed data set.

[0036] The vibration signal is decomposed into five layers using the sym5 wavelet basis and soft threshold denoising is performed; the acoustic emission signal is decomposed using the db4 wavelet basis to eliminate high-frequency random noise.

[0037] Define the dynamic compensation function and establish the time warp function model. The formula of the dynamic compensation function is as follows:

[0038] ;

[0039] in, is the time-varying compensation coefficient, and the local delay difference between sensor signals is calculated by the dynamic time warping algorithm; It is a static offset value used to correct the baseline delay caused by the fixed installation position of the sensor; is the post-compensation time, which indicates the equivalent time point after dynamic compensation.

[0040] The kurtosis and impulse factor of the vibration signal are extracted in the time domain. The energy entropy and resonance band energy ratio of the acoustic emission signal are extracted in the frequency domain through short-time Fourier transform, and a preprocessed sensing dataset containing 28-dimensional features is constructed.

[0041] S2: Based on the preprocessed dataset, a four-dimensional space-time tensor is constructed and an improved parallel factor tensor decomposition is performed to obtain the decoupled core factors and space-time feature component matrices.

[0042] S2.1: Based on the preprocessed data set, determine the spatial resolution and temporal resolution by analyzing the minimum spatial unit and event interval of the data sampling interval; grid the monitoring area according to the spatial resolution to generate a three-dimensional coordinate system, and arrange the data by timestamp based on the temporal resolution to generate a data sequence that changes over time; combine the spatial coordinates of the three-dimensional coordinate system with the data sequence that changes over time through space-time tensorization, and create a four-dimensional space-time tensor containing three-dimensional and time dimensions through dimensional expansion.

[0043] According to the horizontal spacing, vertical spacing and height difference of measuring points in the spatial information of the preprocessed sensor data set, the measuring point spacing is calculated to obtain the spatial resolution parameter; according to the temporal information of the preprocessed sensor data set, the time window step and window length are calculated to obtain the temporal resolution parameter.

[0044] A spatial index map is established according to the spatial resolution parameter, and the coordinates of the measurement points are arranged in grid index order to generate a three-dimensional coordinate tensor.

[0045] The time resolution parameter is evenly divided into time windows by window, and the timing deviation of the time resolution parameter is aligned using the interpolation method to obtain the time series data.

[0046] The spatial resolution parameters are arranged according to the original grid to form a continuous spatial dimension, the time windows of the time series data are arranged in chronological order as the time axis, and the 28-dimensional features are evenly split into four modes of vibration, acoustic emission, strain and temperature as the modal axes.

[0047] Taking space as the X-axis and Y-axis, and the time axis and modal axis as the final output, a four-dimensional space-time tensor is generated.

[0048] S2.2: Convert the four-dimensional space-time tensor into a weighted combination of four factor matrices: time, longitude, latitude, and variable, and add constraints of time smoothness, space continuity, and variable sparsity to form a regularized objective function; based on the regularized objective function, generate a random matrix using a random initialization method through tensor structure analysis; use high-order singular value decomposition to generate an initial guess, and fuse the random matrix and the results of the initial guess through alternating projection to generate an initial factor matrix; update the initial factor matrix through hybrid parallel factors to obtain an updated factor matrix; based on the updated factor matrix, calculate the rank contribution and delete redundant ranks to obtain the decoupled core factor.

[0049] The original timestamps in the original four-dimensional space-time tensor are divided into fixed time windows, and the total window covers the complete observation period to form the time dimension; the measurement point coordinates in the original four-dimensional space-time tensor are mapped to longitude and latitude grids according to the actual physical position to form the space dimension; the multimodal sensor modes and working conditions in the original four-dimensional space-time tensor are merged to form the variable dimension.

[0050] The time dimension, space dimension and variable dimension are standardized, and a structured four-dimensional tensor is formed with the standardized time dimension, space dimension and variable dimension as the structure. The four-dimensional tensor structure is time × longitude × latitude × variable.

[0051] S3: The decoupled core factors and spatiotemporal feature component matrices are cross-modally aligned through the wear feature channel and the acoustic emission feature channel to obtain the fused cross-modal feature vector.

[0052] S3.1 uses a convolutional neural network to extract local wear features from the mechanical wear components in the decoupled core factors and the wear dimensions in the spatiotemporal feature component matrix, and uses Transformer to capture the dynamic evolution law of the wear features to obtain the wear modal feature vector; the acoustic emission energy components in the decoupled core factors and the high-frequency vibration signals in the spatiotemporal feature matrix are decomposed and clustered by wavelet packets to obtain the acoustic emission modal feature vector; the wear modal feature vector and the acoustic emission modal feature vector are projected into a unified low-dimensional space to eliminate the modal scale and distribution differences, and the low-dimensional features are spliced ​​and fused with the original high-dimensional modal features through residual connections to obtain the fused cross-modal feature vector.

[0053] Mechanical wear components in the decoupled core factors , and the wear dimension in the spatiotemporal feature component matrix , capturing multi-scale local patterns through dilated convolution, 3×3 convolutions with dilation rates of 1 / 2 / 4 are applied in parallel to generate multi-scale feature maps, which are then concatenated and compressed to the dimension through pooling.

[0054] By fixing the sine-cosine encoding and spatial dimension expansion, the temporal order information is explicitly injected into the features, and the global evolution law is modeled through the Transformer self-attention mechanism. After the local and global feature residuals are connected, they are averaged and pooled along the spatiotemporal dimensions to output the wear mode feature vector. The formula is as follows:

[0055] ;

[0056] in, Represented as a set of query vectors generated by linear transformation; Represented as a collection of key vectors generated by linear transformation, represents a set of value vectors generated by a linear transformation; represents the dimension of the feature vector, Represents global features.

[0057] The db4 wavelet was used to perform a three-layer decomposition of the core factor acoustic emission component and the high-frequency signal of the space-time matrix to generate 8 sub-band coefficients, and the energy at each position was normalized.

[0058] The signal is constructed according to the similarity matrix, and the first three eigenvectors are extracted by spectral clustering. The eigenvectors are expanded into cluster features, and the wavelet cluster features are fused with the core factors after linear projection to output the acoustic emission modal eigenvector.

[0059] The wear modal eigenvector and the acoustic emission modal eigenvector are minimized by the maximum mean difference and the shared projection matrix is ​​optimized synchronously. The formula is as follows:

[0060] ;

[0061] in, represents the alignment loss, which is used to measure the difference between the features of the two modalities; Indicates the mean calculation of sample features; represents the wear clustering characteristics; represents the acoustic emission clustering characteristics; represents the reproducing kernel Hilbert space, Indicates passing The original features Projection to a higher-dimensional or more abstract representation space.

[0062] Calculate the cross-gating weights and fuse them, concatenate the fused features with the original features, and then perform secondary projection, and finally output the cross-modal feature vector. The calculation formula is as follows:

[0063]

[0064] in, represents the learnable weight matrix; Indicates the feature concatenation operation of connecting two Outai feature vectors in the specified dimension; Indicates the wear mode characteristics after preliminary projection or transformation; Represents the acoustic emission modal characteristics after preliminary projection or transformation, Represents the Sigmoid function (S-shaped function).

[0065] S4: The fused cross-modal feature vector is used to predict the crack growth rate to obtain the crack risk level and the optimized parameter set is obtained through the dynamic parameter adjustment strategy.

[0066] S4.1: The fused cross-modal feature vector is spliced ​​with the operating condition parameters, and the crack length is predicted through the physical-data fusion model to obtain the crack length prediction value; based on the equipment material properties and historical failure data, the preset threshold of the crack length is pre-set through experimental calibration and statistical distribution analysis; the crack length prediction value is compared with the preset threshold, and the risk level is divided to obtain the crack risk level.

[0067] The fused feature vector and the working condition parameters collected directly in real time by physical sensors or equipment controllers are directly connected in the dimensional direction, and the multimodal information is jointly input into the physical-data fusion model. Based on the improvement of the integral form of the Paris formula, the neural network correction term is introduced to obtain the crack length prediction value. The formula is as follows:

[0068] ;

[0069] in, Indicates the predicted crack length corresponding to time or number of cycles; represents the initial crack length; Indicates the current or target number of cycles, Indicates the initial cycle number; and Represents material constants, which are related to material type and environmental conditions; represents the stress intensity factor range; represents the correction term based on the neural network, which is implemented by a multi-layer perceptron; is the joint eigenvector; Expressed as a differential increment of the number of load cycles.

[0070] The predicted crack length value is compared with the preset threshold, and the risk level is obtained based on the difference between the predicted crack length value and the predicted threshold.

[0071] S4.2: Define the optimization objectives and constraints for suppressing crack propagation and reducing energy consumption based on the crack risk level and current operating parameters, and establish an optimization model; input the current equipment status into the optimization model to predict future crack trends, obtain a preliminary optimized parameter candidate set, and perform safety enhancement verification to obtain a set of optimized parameters.

[0072] Based on the risk level and the residual frequency, number of sensors and temperature controller in the operating parameters, an optimization model is established through the multi-objective weighted summation method.

[0073] The current state of the equipment is brought into the optimization model, and the NSGA-III algorithm is used to solve the Pareto frontier to generate a candidate parameter set. The candidate parameter set is then input into the digital twin model to predict the future evolution of crack length. The formula is as follows:

[0074] ;

[0075] in, Indicates Within the time, use The crack length predicted by the candidate parameter sets; Indicates the current time Crack length; Represents the total number of time steps for prediction; Indicates time The range of stress intensity factors when ; Indicates Joint eigenvector in time; Indicates the total number of time steps for prediction; Expressed as a calculation of the physical contribution to crack growth through a range of stress intensity factors and material constants, The parameter is expressed as Multilayer Perceptron, a feedforward neural network.

[0076] Eliminate the parameter combinations that lead to excessive cracks and keep those that meet the The candidate set is obtained by linear scalarization, and the parameter combination with the lowest energy consumption is selected from the candidate set to obtain the optimized parameter set.

[0077] This embodiment also provides an adaptive test parameter optimization system, including:

[0078] Preprocessing module: collects the original sensor signals from the mechanical testing system and performs preprocessing.

[0079] Four-dimensional tensor decomposition module: Based on the preprocessed dataset, it constructs a four-dimensional space-time tensor and performs improved parallel factor tensor decomposition to obtain the decoupled core factors and space-time feature component matrices.

[0080] Cross-modal feature fusion module: The decoupled core factors and spatiotemporal feature component matrices are cross-modally aligned through the wear feature channel and the acoustic emission feature channel to obtain the fused cross-modal feature vector.

[0081] Crack risk control optimization module: The fused cross-modal feature vector is used to predict the crack growth rate to obtain the crack risk level and the optimized parameter set is obtained through dynamic parameter adjustment strategy.

[0082] This embodiment also provides a computer device suitable for the case of an adaptive test parameter optimization 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 adaptive test parameter optimization method proposed in the above embodiment.

[0083] The computer device may be a terminal, comprising a processor, memory, a communication interface, a display, and an input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. The internal memory provides an environment for the operating system and computer programs stored in the non-volatile storage media. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication. Wireless communication may be achieved via Wi-Fi, a carrier network, NFC (near-field communication), or other technologies. The display of the computer device may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen overlay on the display, buttons, a trackball, or a touchpad on the computer device housing, or an external keyboard, touchpad, or mouse.

[0084] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the adaptive test parameter optimization 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, magnetic disk or optical disk.

[0085] 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. An adaptive test parameter optimization method, characterized by: include, Collect the original sensor signals from the mechanical testing system and perform preprocessing; Based on the preprocessed data set, a four-dimensional space-time tensor is constructed, and an improved parallel factor tensor decomposition is performed to obtain the decoupled core factors and space-time feature component matrices; The decoupled core factors and spatiotemporal feature component matrices are cross-modally aligned through the wear feature channel and the acoustic emission feature channel to obtain the fused cross-modal feature vector. The fused cross-modal feature vector is used to predict the crack growth rate to obtain the crack risk level, and the optimized parameter set is obtained through a dynamic parameter adjustment strategy.

2. The adaptive test parameter optimization method according to claim 1, wherein: The specific steps of the pretreatment are as follows: The original sensor data is denoised using wavelet transform to obtain denoised sensor data; Define a time warp function, build a spatiotemporal alignment model through nonlinear mapping combined with spatiotemporal parameter optimization, input the denoised sensor data into the spatiotemporal alignment model, and obtain a multimodal signal pair with time axis alignment. Through the technical means of joint time-frequency analysis and cross-modal feature engineering, a dual-domain feature extraction module is constructed, and the multimodal signal pair with time axis alignment is input into the dual-domain feature extraction module to obtain the preprocessed data set.

3. The adaptive test parameter optimization method according to claim 1, wherein: The specific steps of constructing the four-dimensional space-time tensor are as follows: Based on the preprocessed data set, the spatial resolution and temporal resolution are determined by analyzing the minimum spatial unit and event interval of the data sampling interval; The monitoring area is gridded according to the spatial resolution to generate a three-dimensional coordinate system, and the data are arranged by timestamp based on the temporal resolution to generate a data series that changes over time; The spatial coordinates of the three-dimensional coordinate system are combined with the time-varying data sequence through space-time tensorization, and a four-dimensional space-time tensor containing three-dimensional and time dimensions is created through dimensional expansion.

4. The adaptive test parameter optimization method according to claim 1, wherein: The improved parallel factor tensor decomposition is performed in the following steps: The four-dimensional space-time tensor is converted into a weighted combination of four factor matrices: time, longitude, latitude, and variable. The constraints of time smoothness, space continuity, and variable sparsity are added to form a regularized objective function. Based on the regularized objective function, a random matrix is ​​generated by random initialization method through tensor structure analysis; Use high-order singular value decomposition to generate an initial guess, and fuse the random matrix and the results of the initial guess through alternating projection to generate an initial factor matrix; Update the initial factor matrix through hybrid parallel factors to obtain an updated factor matrix; According to the updated factor matrix, the rank contribution is calculated and the redundant ranks are deleted to obtain the decoupled core factors.

5. The adaptive test parameter optimization method according to claim 1, wherein: The cross-modal alignment is performed through the wear characteristic channel and the acoustic emission characteristic channel. The specific steps are as follows: The mechanical wear component in the decoupled core factor and the wear dimension in the spatiotemporal feature component matrix are used to extract local wear features through a convolutional neural network. The dynamic evolution of the wear features is captured using a Transformer to obtain the wear modal feature vector. The acoustic emission energy component in the decoupled core factor and the high-frequency vibration signal in the spatiotemporal feature matrix are decomposed by wavelet packet and clustering algorithm to obtain the acoustic emission modal feature vector. The wear modal eigenvector and acoustic emission modal eigenvector are projected into a unified low-dimensional space to eliminate the differences in modal scale and distribution. The low-dimensional features are then concatenated and fused with the original high-dimensional modal features through residual connections to obtain the fused cross-modal feature vector.

6. The adaptive test parameter optimization method according to claim 1, wherein: The crack risk level is obtained by predicting the crack growth rate. The steps are as follows: The fused cross-modal feature vector is concatenated with the working condition parameters, and the crack length is predicted by the physical-data fusion model to obtain the crack length prediction value; Based on the equipment material properties and historical failure data, a preset threshold value for crack length is pre-set through experimental calibration and statistical distribution analysis; The crack length prediction value is compared with the preset threshold value, and the risk level is divided to obtain the crack risk level.

7. The adaptive test parameter optimization method according to claim 1, wherein: The steps to obtain the optimized parameter set through the dynamic parameter adjustment strategy are as follows: Define the optimization objectives and constraints of suppressing crack growth and reducing energy consumption according to the crack risk level and current operating parameters, and establish an optimization model; The current equipment status is input into the optimization model to predict future crack trends, obtain a preliminary optimized parameter candidate set, and perform safety enhancement verification to obtain a set of optimized parameters.

8. An adaptive test parameter optimization system, based on the adaptive test parameter optimization method according to any one of claims 1 to 7, characterized in that: include, Preprocessing module: collects the original sensor signals from the mechanical testing system and performs preprocessing; Four-dimensional tensor decomposition module: Based on the preprocessed data set, it constructs a four-dimensional space-time tensor and performs improved parallel factor tensor decomposition to obtain the decoupled core factors and space-time feature component matrices; Cross-modal feature fusion module: The decoupled core factors and spatiotemporal feature component matrices are aligned across modalities through the wear feature channel and the acoustic emission feature channel to obtain the fused cross-modal feature vector. Crack risk control optimization module: The fused cross-modal feature vector is used to predict the crack growth rate to obtain the crack risk level and the optimized parameter set is obtained through dynamic parameter adjustment strategy.

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 adaptive test parameter optimization method according to any one of claims 1 to 7 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 adaptive test parameter optimization method according to any one of claims 1 to 7 are implemented.

Images