Hybrid-driven metal multi-fatigue crack propagation fusion prediction method

Abstract

The invention belongs to the technical field of aircraft strength experiments, and particularly relates to a hybrid-driven metal multi-fatigue crack propagation fusion prediction method. The method comprises the following steps: establishing a stress intensity factor expression under multiple fatigue cracks; initial distribution of material parameters in the Paris physical model is obtained, and posterior distribution of the material parameters is adjusted; based on the stress intensity factor expression under the multi-fatigue crack and the posterior distribution of the material parameters, obtaining an improved Paris physical model for describing the multi-fatigue crack propagation process; a depth encoder network model is established, and a network bottleneck layer is utilized to extract guided wave deep features with strong correlation with the fatigue crack length; establishing a guided wave data driven multi-fatigue crack length online monitoring model; and building a particle filter fusion framework, and correcting and improving a distribution interval of uncertain material parameters in the Paris physical model according to guided wave monitoring data to realize dynamic tracking and prediction of metal multi-fatigue crack propagation.

Description

Technical Field

[0001] The present application belongs to the technical field of aircraft strength experiments, and in particular relates to a hybrid-driven metal multi-fatigue crack extension fusion prediction method. Background Art

[0002] Aircraft metal structures are very prone to widespread fatigue cracks due to long-term fatigue loads, which in turn threaten structural safety and cause catastrophic accidents. Therefore, accurate tracking of the propagation process of metal multiple fatigue cracks has important application value and economic benefits in shortening the downtime of in-service aircraft, extending flight life and enhancing maintenance support.

[0003] At present, based on the micro-fracture failure mechanism, a large number of empirical equations describing the fatigue crack growth process have been proposed and developed, including the Paris model applicable to a single stress ratio, the Walker model for different stress ratios, and the Forman model considering variable stress ratio and fracture toughness. However, these classic formulas are only applicable to single fatigue cracks, and have not yet considered the mutual interference between multiple fatigue cracks in the widely distributed fatigue damaged structure, which leads to a surge in stress intensity factors and an acceleration of crack growth rate. At the same time, fatigue crack growth is a random process with multiple uncertainties, such as material properties, load and crack observation data, physical growth models, and other uncertainties. If a physical growth model based on deterministic parameters is used to predict the crack length, its prediction error will gradually accumulate, and the reliability level of the fatigue life prediction results will gradually decrease.

[0004] With the rapid development of sensor technology and information technology, data-driven metal crack length prediction methods have received widespread attention. Due to the long-distance propagation and sensitivity to tiny damage of guided wave damage monitoring technology, it has received widespread attention in the field of online monitoring technology of structural damage. The focus of the crack length prediction method based on guided wave data drive is to extract the characterization parameters related to the crack length, and then establish a mapping relationship between the characterization parameters and the crack length. However, the current guided wave feature extraction process relies on a lot of signal processing and expert experience knowledge, resulting in poor applicability of guided wave features, and intelligent guided wave parameter characterization has not yet been realized. At the same time, the metal crack length prediction method based on guided wave data drive focuses more on the perception of the current state of the crack, and it is difficult to complete the prediction of the crack extension trend.

[0005] Therefore, it is desired to have a technical solution to overcome or at least alleviate at least one of the above-mentioned defects of the prior art. Summary of the invention

[0006] The purpose of this application is to provide a hybrid-driven fusion prediction method for multi-fatigue crack propagation in metals, so as to solve the problems existing in the prior art, such as the non-linearity, uncertainty, and mutual interference of multi-fatigue crack propagation in metals, and the limitations of guided wave data-driven crack on-line monitoring relying on a large amount of signal processing and expert experience.

[0007] The technical solution of this application is as follows:

[0008] A hybrid-driven fusion prediction method for multi-fatigue crack propagation in metals, comprising:

[0009] Step 1: Establish a simulation model of a multi-fatigue crack structure in metal, obtain the stress intensity factors at different multi-fatigue crack lengths, and establish an expression for the stress intensity factors under multi-fatigue cracks;

[0010] Step 2: Obtain the initial distribution of material parameters in the Paris physical model and adjust the posterior distribution of the material parameters;

[0011] Step 3: Based on the expression of the stress intensity factor under multi-fatigue cracks and the posterior distribution of the material parameters, obtain an improved Paris physical model describing the multi-fatigue crack propagation process;

[0012] Step 4: Establish a deep encoder network model, and use the network bottleneck layer to extract the deep features of guided waves with strong correlation with the fatigue crack length;

[0013] Step 5: Use linear regression to fit the mapping relationship between the fatigue crack length and the deep features of guided waves, and establish an on-line monitoring model for the multi-fatigue crack length driven by guided wave data;

[0014] Step 6: Build a particle filter fusion framework, fuse the improved Paris physical model and the on-line monitoring model for the multi-fatigue crack length, and correct the distribution interval of the uncertain material parameters in the improved Paris physical model according to the guided wave monitoring data, so as to realize the dynamic tracking and prediction of multi-fatigue crack propagation in metals.

[0015] In at least one embodiment of this application, in Step 1, establishing a simulation model of a multi-fatigue crack structure in metal, obtaining the stress intensity factors at different multi-fatigue crack lengths, and establishing an expression for the stress intensity factors under multi-fatigue cracks, includes:

[0016] Establish a simulation model of a multi-fatigue crack structure in metal, and use the contour integral method to obtain the amplitude of the stress intensity factor at different multi-fatigue crack lengths;

[0017] Use polynomial fitting to establish the non-linear mapping relationship between different multi-fatigue crack lengths and the stress intensity factors, and establish an expression for the stress intensity factors under synchronous propagation of multi-fatigue cracks:

[0018] △Kβ = Δσh β (X) + ξ β = Δσh β (x1, x2, x3,..., x β ,..., x n ) + ξ β

[0019] where X = [x1, x2, x3,..., x β ,..., x n is the multi-fatigue crack length vector of n cracks, ΔK β is the stress intensity factor at the tip of the β-th crack, h β (·) is the polynomial fitting function between the stress intensity factor at the tip of the β-th crack and n cracks, ξ β is the fitting error of the polynomial fitting function of the β-th crack, and Δσ is the constant amplitude stress spectrum.

[0020] In at least one embodiment of the present application, in step two, obtaining the initial distribution of the material parameters in the Paris physical model and adjusting the posterior distribution of the material parameters includes:

[0021] Obtaining the initial distribution of the material parameters in the Paris physical model according to expert knowledge;

[0022] Adopting the Monte Carlo simulation method to adjust the posterior distribution of the material parameters through the prediction error between the predicted results of the multi-fatigue crack length and the real data:

[0023] p post (θ|X) ∝ p lhd (X|θ) × p prior (θ)

[0024] where θ is the vector of uncertain material parameters in the conventional Paris physical model, p prior (θ) is the initial distribution of the vector of uncertain material parameters, p lhd (X|θ) is the occurrence probability of the multi-fatigue crack length under a specific vector of material parameters, and p post (θ|X) is the posterior distribution of the vector of material parameters.

[0025] In at least one embodiment of the present application, in step three, the improved Paris physical model is:

[0026]

[0027] where X t is the multi-fatigue crack length at time t, X t+1 is the multi-fatigue crack length at time t + 1, C t , m tThe uncertain material parameters of the improved Paris physical model at time t, ΔK is the stress intensity factor vector corresponding to the lengths of n multi-fatigue cracks, and ΔN t is the number of fatigue cycles from time t to time t + 1, and δ t is the observation noise of the improved Paris physical model.

[0028] In at least one embodiment of the present application, in step four, after establishing a deep encoder network model and extracting the deep guided wave features strongly correlated with the fatigue crack length using the network bottleneck layer, it further includes:

[0029] Taking the online guided wave monitoring signal of the multi-fatigue crack length as the input, and using the norm of the reconstructed output signal and the input signal by the decoding layer and the strong correlation between the deep guided wave features extracted by the network bottleneck layer and the fatigue crack length to construct a model loss function, and globally optimizing the model parameters. Among them, the model loss function is:

[0030]

[0031] where Z is the input guided wave signal of the deep encoder, is the reconstructed output guided wave signal of the output layer of the deep encoder, y is the deep guided wave feature output by the deep encoder, x is the true fatigue crack length corresponding to the output guided wave signal, and ρ(·) represents calculating the Pearson correlation coefficient between the two.

[0032] In at least one embodiment of the present application, in step five, the multi-fatigue crack length online monitoring model is:

[0033] Y t+1 = q0X t+1 + q1 + υ t+1

[0034] where q0 and q1 are the fitting coefficients of the unary linear regression fitting function between the multi-fatigue crack length and the deep guided wave feature, and υ t+1 is the measurement noise of the multi-fatigue crack length online monitoring model.

[0035] The invention has at least the following beneficial technical effects:

[0036] The hybrid-driven metal multi-fatigue crack propagation fusion prediction method of the present application combines the physical model and the guided wave monitoring data, reduces the uncertainty of the physical model, improves the prediction accuracy of the metal multi-fatigue crack propagation, and provides accurate information input for the formulation of the metal structure maintenance and support plan. Description of the Drawings

[0037] Figure 1 ​Flowchart of the hybrid-driven metal multi-fatigue crack propagation fusion prediction method according to an embodiment of the present application;

[0038] Figure 2 Schematic diagram of the finite element simulation model for the propagation of multiple fatigue cracks in a metal opening according to an embodiment of the present application;

[0039] Figure 3 Schematic diagram of the guided wave deep feature extraction process based on a deep encoder according to an embodiment of the present application;

[0040] Figure 4 Schematic diagram of the metal opening test piece and sensor monitoring scheme according to an embodiment of the present application;

[0041] Figure 5 Schematic diagram of the expression for fitting the stress intensity factor using a binary cubic polynomial according to an embodiment of the present application;

[0042] Figure 6 Schematic diagram of the optimal deep encoder network architecture after cross-validation according to an embodiment of the present application;

[0043] Figure 7 Schematic diagram of fitting the corresponding relationship between the deep damage feature and the crack length using a unary linear relationship according to an embodiment of the present application;

[0044] Figure 8 Schematic diagram of the dynamic update process of the uncertain material parameters [C, m] according to an embodiment of the present application;

[0045] Figure 9 Schematic diagram of the fusion prediction result and its confidence interval of the double fatigue crack lengths of the metal opening structure according to an embodiment of the present application. Specific implementation mode

[0046] To make the purpose, technical solutions, and advantages of the implementation of the present application clearer, the technical solutions in the embodiments of the present application will be described in more detail below with reference to the accompanying drawings in the embodiments of the present application. In the accompanying drawings, the same or similar reference numerals denote the same or similar elements or elements with the same or similar functions throughout. The described embodiments are some, but not all, of the embodiments of the present application. The embodiments described below by referring to the accompanying drawings are exemplary and are intended to explain the present application and should not be construed as limiting the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts fall within the scope of protection of the present application. The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0047] In the description of the present application, it should be understood that the orientation or positional relationships indicated by the terms "center", "longitudinal", "transverse", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. are based on the orientation or positional relationships shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as limiting the protection scope of the present application.

[0048] The following will further elaborate on the present application in conjunction with the attached Figures 1 to 9 drawings.

[0049] The present application provides a hybrid-driven metal multi-fatigue crack propagation fusion prediction method, as Figure 1 shown, including the following steps:

[0050] Step 1: Establish a simulation model of the metal multi-fatigue crack structure, obtain the stress intensity factors at different multi-fatigue crack lengths, and establish an expression for the stress intensity factors under multi-fatigue cracks;

[0051] Step 2: Obtain the initial distribution of the material parameters in the Paris physical model and adjust the posterior distribution of the material parameters;

[0052] Step 3: Based on the expression for the stress intensity factors under multi-fatigue cracks and the posterior distribution of the material parameters, obtain an improved Paris physical model describing the multi-fatigue crack propagation process;

[0053] Step 4: Establish a deep encoder network model, and use the network bottleneck layer to extract the deep guided wave features with strong correlation with the fatigue crack length;

[0054] Step 5: Use linear regression to fit the mapping relationship between the fatigue crack length and the deep guided wave features, and establish an online monitoring model for multi-fatigue crack lengths driven by guided wave data;

[0055] Step 6: Build a particle filter fusion framework, fuse the improved Paris physical model and the online monitoring model for multi-fatigue crack lengths, and correct the distribution interval of the uncertain material parameters in the improved Paris physical model according to the guided wave monitoring data to achieve dynamic tracking and prediction of metal multi-fatigue crack propagation.

[0056] The hybrid-driven metal multi-fatigue crack propagation fusion prediction method of the present application, as Figure 2 shown, in Step 1, establishing a simulation model of the metal multi-fatigue crack structure, obtaining the stress intensity factors at different multi-fatigue crack lengths, and establishing an expression for the stress intensity factors under multi-fatigue cracks, includes:

[0057] A simulation model of a metal multi-fatigue crack structure was established, and the contour integral method was used to obtain the stress intensity factor amplitude at different multi-fatigue crack lengths;

[0058] The non-linear mapping relationship between different multi-fatigue crack lengths and the stress intensity factor was fitted by polynomials, and an expression for the stress intensity factor under the synchronous propagation of multi-fatigue cracks was established:

[0059] △K β =△σh β (X)+ξ β =△σh β (x1,x2,x3,...,x β ,...,x n )+ξ β (1)

[0060] where X = [x1,x2,x3,...,x β ,...,x n is the vector of n multi-fatigue crack lengths, ΔK β is the stress intensity factor at the tip of the β-th crack, h β (·) is the polynomial fitting function between the stress intensity factor at the tip of the β-th crack and n cracks, ξ β is the fitting error of the polynomial fitting function of the β-th crack, and Δσ is the constant amplitude stress spectrum.

[0061] In step two, the initial distribution of the material parameters in the Paris physical model was obtained, and the posterior distribution of the material parameters was adjusted, including:

[0062] The initial distribution of the material parameters in the Paris physical model was obtained according to expert knowledge;

[0063] The Monte Carlo simulation method was used to adjust the posterior distribution of the material parameters through the prediction error between the predicted results and the real data of the multi-fatigue crack lengths:

[0064] p post (θ|X) ∝ p lhd (X|θ) × p prior (θ) (2)

[0065] where θ is the vector of uncertain material parameters in the conventional Paris physical model, p prior (θ) is the initial distribution of the vector of uncertain material parameters, p lhd (X|θ) is the occurrence probability of the multi-fatigue crack lengths under a specific vector of material parameters, which can be obtained through the prediction error between the predicted results and the real data of the multi-fatigue crack lengths, and p post (θ|X) is the posterior distribution of the vector of material parameters.

[0066] In step 3, based on the stress intensity factor expression and the posterior distribution of material parameters under multiple fatigue cracks, the improved Paris physical model describing the multi-fatigue crack growth process is obtained as follows:

[0067]

[0068] Among them, X t is the length of multiple fatigue cracks at time t, X t+1 is the length of multiple fatigue cracks at time t+1, C t 、m t is the uncertain material parameter of the improved Paris physical model at time t, ΔK is the stress intensity factor vector corresponding to the length of n multiple fatigue cracks, ΔN t is the number of fatigue cycles from time t to time t+1, δ t To improve the observation noise of the Paris physical model, its variance is the fatigue crack length prediction error of the posterior distribution of material parameters.

[0069] like Figure 3 As shown, in step 4, a deep encoder network model is established, and the network bottleneck layer is used to extract the deep waveguide features that are strongly correlated with the fatigue crack length, and then the following is included:

[0070] The online guided wave monitoring signal of multiple fatigue crack lengths is used as input, and the decoding layer is used to reconstruct the output signal and the input signal. The model loss function is constructed based on the strong correlation between the deep characteristics of the guided wave extracted from the bottleneck layer of the network and the fatigue crack length, and the model parameters such as the number of neurons, training batches and learning rate are globally optimized. The model loss function is:

[0071]

[0072] Where Z is the input waveguide signal of the depth encoder, The output waveguide signal is reconstructed for the output layer of the deep encoder, y is the waveguide deep layer feature output by the deep encoder, x is the true fatigue crack length corresponding to the output waveguide signal, and ρ(·) represents the calculation of the Pearson correlation coefficient between the two.

[0073] In step 5, the online monitoring model of multiple fatigue crack lengths is:

[0074] Y t+1 =q0X t+1 +q1+υ t+1 (5)

[0075] Among them, q0 and q1 are the fitting coefficients of the univariate linear regression fitting function between the length of multiple fatigue cracks and the deep characteristics of guided waves, υ t+1is the measurement noise of the online monitoring model for multiple fatigue crack lengths, and its variance is the sum of the generalization error of the depth encoder and the variance of the unary linear regression fitting.

[0076] In an embodiment of the present application, the propagation prediction of double fatigue cracks at the hole edge of an aluminum alloy center plate hole test piece was carried out. The material of the aluminum alloy center plate hole test piece is 7050 aluminum alloy, with dimensions of 400mm×168mm×3mm, and there is a circular through hole with a diameter of 25mm at the center position. A fatigue test was carried out using an MTS-type fatigue testing machine. The test loading load spectrum is a sine wave, the maximum load is 40kN, the stress ratio is 0.1, and the loading frequency is 8Hz. During the test, 2 P-51 piezoelectric sensors were adhesively bonded to the surface of the test piece with 401 glue to form 2 scanning channels. The diameter of the sensor is 8mm, and the thickness is 0.45mm. The excitation signal of the guided wave is a five-peak sine excitation signal with a center frequency of 230kHz, the sampling frequency of the signal is 10MHz, and the sampling length is 4000 data points. The metal open-hole test piece and the sensor monitoring scheme are as Figure 4 shown.

[0077] According to the hybrid-driven metal multi-fatigue crack propagation fusion prediction method of the present application, the double crack propagation prediction is carried out. The specific process is as follows:

[0078] First, the stress intensity factor amplitude corresponding to different lengths of double cracks in the open-hole structure is obtained by using the extended finite element numerical simulation method, and the expression of the stress intensity factor is fitted by using a binary cubic polynomial, as Figure 5 shown; furthermore, the Monte Carlo simulation method is used to statistically analyze the double fatigue crack propagation data of the existing test pieces to obtain the distribution of material parameters; finally, the improved Paris physical propagation evolution model of double fatigue cracks is determined, as shown in Equation (6).

[0079]

[0080] Then, the optimal depth encoder network architecture after cross-validation (as Figure 6 shown) is used to extract the deep features of the guided wave monitoring signal, and the corresponding relationship between the deep damage feature and the crack length is fitted by using a unary linear relationship, as Figure 7 shown; finally, an online monitoring model of crack length based on the guided wave monitoring data is established, as shown in Equation (7).

[0081] Finally, the particle filter algorithm is used to fuse the improved Paris physical model and the guided wave online monitoring model, and the uncertain parameters of the improved Paris model are updated online according to the guided wave monitoring results of the double crack length. The dynamic update process of the uncertain material parameters [C, m] is as Figure 8As shown, the fusion prediction results of the double fatigue crack lengths of the metal open-hole structure and their confidence intervals are obtained, as Figure 9 shown.

[0082] The hybrid-driven metal multi-fatigue crack propagation fusion prediction method of the present application has the following

[0083] beneficial effects:

[0084] 1. An improved Paris physical model for describing the dynamic synchronous propagation of multi-fatigue cracks is proposed, revealing the mutual interference mechanism of multi-crack lengths;

[0085] 2. A crack length guided wave intelligent monitoring model based on a deep encoder is constructed, clarifying the quantitative relationship between crack lengths and on-line guided wave data;

[0086] 3. A fusion prediction method for physical models and guided wave data is proposed, improving the prediction accuracy of metal fatigue crack propagation and directly quantifying the uncertainty of the predicted remaining life.

[0087] The above are only specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed in the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claimed rights.

Claims

1. A hybrid driven metal multi-fatigue crack propagation fusion prediction method, characterized in that: include: Step 1: Establish a metal multi-fatigue crack structure simulation model, obtain the stress intensity factor under multiple fatigue crack lengths, and establish an expression for the stress intensity factor under multiple fatigue cracks; Step 2: Obtain the initial distribution of material parameters in the Paris physical model and adjust the posterior distribution of material parameters; Step 3: Based on the stress intensity factor expression under multiple fatigue cracks and the posterior distribution of material parameters, an improved Paris physical model describing the multiple fatigue crack propagation process is obtained; Step 4: Establish a deep encoder network model and use the network bottleneck layer to extract the deep waveguide features that are strongly correlated with the fatigue crack length; Step 5: Use linear regression to fit the mapping relationship between fatigue crack length and guided wave deep characteristics, and establish a guided wave data-driven online monitoring model for multiple fatigue crack lengths; Step 6: Build a particle filter fusion framework, fuse the improved Paris physical model and the online monitoring model of multiple fatigue crack lengths, correct the distribution range of uncertain material parameters in the improved Paris physical model according to the guided wave monitoring data, and realize dynamic tracking and prediction of metal multiple fatigue crack extension.

2. The hybrid-driven metal multi-fatigue crack extension fusion prediction method according to claim 1 is characterized in that: In step 1, a simulation model of a metal multi-fatigue crack structure is established to obtain the stress intensity factor under multiple fatigue crack lengths, and an expression for the stress intensity factor under multiple fatigue cracks is established, including: A simulation model of metal multi-fatigue crack structure is established, and the stress intensity factor amplitude under multi-fatigue crack length is obtained by using the contour integration method; The nonlinear mapping relationship between different multiple fatigue crack lengths and stress intensity factors is fitted by polynomials, and the stress intensity factor expression under the simultaneous extension of multiple fatigue cracks is established: △K β =△σh β (X)+ξ β =△σh β (x1,x2,x3,...,x β ,...,x n )+ξ β Where X = [x1, x2, x3, ..., x β ,...,x n ] is the length vector of n multiple fatigue cracks, ΔK β is the stress intensity factor at the β crack tip, h β (·) is the polynomial fitting function between the stress intensity factor at the β-th crack tip and the n cracks, ξ β is the fitting error of the β-crack polynomial fitting function, and Δσ is the constant amplitude stress spectrum.

3. The hybrid-driven metal multi-fatigue crack propagation fusion prediction method according to claim 2 is characterized in that: In step 2, the initial distribution of material parameters in the Paris physical model is obtained, and the posterior distribution of material parameters is adjusted, including: Obtain the initial distribution of material parameters in the Paris physical model based on expert knowledge; The Monte Carlo simulation method is used to adjust the posterior distribution of material parameters through the prediction error between the predicted results of multiple fatigue crack lengths and the actual data: p post (θ|X)∝p lhd (X|θ)×p prior (i) Where θ is the uncertain material parameter vector in the conventional Paris physical model, p prior (θ) is the initial distribution of the uncertain material parameter vector, p lhd (X|θ) is the probability of occurrence of multiple fatigue crack lengths under a specific material parameter vector, p post (θ|X) is the posterior distribution of the material parameter vector.

4. The hybrid-driven metal multi-fatigue crack propagation fusion prediction method according to claim 3 is characterized in that: In step 3, the improved Paris physical model is: Among them, X t is the length of multiple fatigue cracks at time t, X t+1 is the length of multiple fatigue cracks at time t+1, C t 、m t is the uncertain material parameter of the improved Paris physical model at time t, ΔK is the stress intensity factor vector corresponding to the length of n multiple fatigue cracks, ΔN t is the number of fatigue loading cycles from time t to time t+1, δ t To improve the observation noise of the Paris physical model.

5. The hybrid-driven metal multiple fatigue crack propagation fusion prediction method according to claim 4 is characterized in that: In step 4, a deep encoder network model is established, and the network bottleneck layer is used to extract the deep waveguide features that are strongly correlated with the fatigue crack length, and then the following steps are included: Taking the online guided wave monitoring signal of multiple fatigue crack lengths as input, the model loss function is constructed based on the l2 norm of the output signal reconstructed by the decoding layer and the input signal, as well as the strong correlation between the deep characteristics of the guided wave extracted by the network bottleneck layer and the fatigue crack length, and the model parameters are globally optimized. The model loss function is: Where Z is the input waveguide signal of the depth encoder, The output waveguide signal is reconstructed for the output layer of the deep encoder, y is the waveguide deep layer feature output by the deep encoder, x is the true fatigue crack length corresponding to the output waveguide signal, and ρ(·) represents the calculation of the Pearson correlation coefficient between the two.

6. The hybrid-driven metal multiple fatigue crack propagation fusion prediction method according to claim 5 is characterized in that: In step 5, the online monitoring model of multiple fatigue crack lengths is: Y t+1 =q0X t+1 +q1+υ t+1 Among them, q0 and q1 are the fitting coefficients of the univariate linear regression fitting function between the length of multiple fatigue cracks and the deep characteristics of guided waves, υ t+1 Measurement noise for the online monitoring model of multiple fatigue crack lengths.

Images