Wetland equipment structure life prediction method based on enhanced information transfer multi-task learning

Abstract

The invention discloses a method for predicting the service life of a wetland equipment structure based on enhanced information transfer multitask learning. The method comprises the following steps: acquiring a real-time load spectrum of the wetland equipment structure based on dynamic / static analysis; establishing a load spectrum fusion model based on multi-objective optimization adapted to a complex environment; obtaining an optimized load spectrum of fusion enhanced information transfer multi-task learning; conditioning adaptive parameters of the fatigue performance analysis of the wetland equipment structure; and predicting the fatigue life of the wetland equipment structure. According to the method, the service life of the wetland equipment structure is predicted through multi-task learning, on the basis of considering multi-task learning under different working conditions, transmission of multi-target information of a complex environment is enhanced, and prediction of the service life of the wetland equipment structure is achieved.

Description

Technical Field

[0001] The present invention relates to the field of equipment structure reliability, and in particular to a wetland equipment structure life prediction method based on enhanced information transfer multi-task learning. Background Art

[0002] Wetlands are one of the most important ecosystems on Earth, with multiple functions such as maintaining biodiversity and purifying water. Wetland ecosystems are similar to other things in nature, and are dynamic systems that are in constant motion and change. Therefore, conducting research on wetland environmental monitoring has important theoretical significance and application value. The wetland environment is complex, and predicting the service life of wetland equipment can help improve the reliability of the equipment, ensure that the equipment can operate normally during its service life, reduce downtime, and improve monitoring efficiency. The prediction of the equipment structure life not only helps to improve safety and economy, but also can effectively extend the service life of the equipment and optimize resource allocation.

[0003] At present, the life prediction of wetland equipment structures mainly focuses on the comprehensive analysis and evaluation of environmental factors, material aging, dynamic loads and fatigue damage, and has not considered the use of multi-task learning to predict the life of wetland equipment structures. Therefore, it is necessary to strengthen the transmission of multi-target information in complex environments and realize the prediction of wetland equipment structure life based on multi-task learning under different working conditions. Summary of the invention

[0004] Purpose of the invention: The purpose of the present invention is to provide a method for predicting the structural life of wetland equipment based on enhanced information transfer multi-task learning.

[0005] Technical solution: The wetland equipment structure life prediction method based on enhanced information transfer multi-task learning of the present invention comprises the following steps:

[0006] (1) Acquisition of real-time load spectrum of wetland equipment structure based on dynamic / static analysis;

[0007] (2) Establishment of a load spectrum fusion model based on multi-objective optimization that adapts to complex environments;

[0008] (3) Obtaining optimized load spectra by integrating multi-task learning with enhanced information transfer;

[0009] (4) Adjustment of adaptive parameters for fatigue performance analysis of wetland equipment structures;

[0010] (5) Fatigue life prediction of wetland equipment structures.

[0011] Furthermore, the step (1) comprises:

[0012] Based on the known wetland equipment, a three-dimensional model of the structure is established in the three-dimensional software SolidWorks to accurately describe the shape, size and relative position of each equipment component and structure; and the established three-dimensional model is imported into the Adams software to analyze its structural dynamics under actual working conditions, and obtain the load spectrum of important components of the wetland equipment structure; on this basis, the important components are extracted from the Adams software and imported into the Ansys software for static simulation, and the stress and strain distribution of these important components under static conditions is obtained by setting appropriate boundary conditions and load conditions in the Ansys software.

[0013] Furthermore, the equipment components and structures include a body, a suspension, wheels, a telescopic rod and a sensor chassis.

[0014] Furthermore, step (2) includes establishing, based on the analysis of step (1), a fusion model that can adapt to the complex environment of the wetland and perform multi-objective optimization of the load spectrum. By introducing a gating mechanism, it is possible to selectively remember or forget information, the core of which is three gates: an input gate, a forget gate, and an output gate.

[0015] Furthermore, the forget gate controls how much memory information of the previous moment needs to be discarded, and the formula is:

[0016] f t =σ(W f ·[h t-1 , x t ]+b f )

[0017] Among them, f t is the output of the forget gate, σ is the Sigmoid activation function, W f is the weight matrix of the forget gate, h t-1 is the hidden state of the previous moment, x t is the input at the current moment, b f is the bias term of the forget gate.

[0018] Furthermore, the input gate controls how much of the input information at the current moment can be written into the cell state, and determines the cell content that needs to be updated through the Sigmoid activation function and the tanh activation function, so as to integrate the new information into the cell state. The formula is:

[0019] i t =σ(W i ·[h t-1 , x t ]+b i )

[0020]

[0021]

[0022] Where: i t is the output of the input gate, W i is the weight matrix of the input gate, b i is the bias term of the input gate, tanh is the hyperbolic tangent activation function, W c is the weight matrix of the candidate cell state, b c is the bias term of the candidate cell state, is the candidate cell state.

[0023] Furthermore, the output gate determines the output of the cell state at the current moment. The output state of the current cell is first evaluated by the Sigmoid activation function, and then the current memory state is processed by the tanh activation function. The two are multiplied and output. The formula is:

[0024] o t =σ(W o ·[h t-1 , x t ]+b o )

[0025] h t =o t tanh(c t )

[0026] Among them: t is the output of the output gate, W o is the weight matrix of the output gate, b o is the bias term of the output gate, h t is the hidden state at the current moment.

[0027] Furthermore, the step (3) comprises:

[0028] Based on the analysis in step (2), in order to enhance the transmission of multi-target information in complex environments, it is necessary to introduce a multi-task learning model to optimize the load spectrum, so that the model focuses on certain parts of the input data and assigns different weights to different parts to highlight the features or information that are more important to the task.

[0029] The additive attention mechanism is used to calculate the weighted sum of the inputs of the decoder at each time step, as follows:

[0030]

[0031] Among them, q is the hidden state of the decoder at the current moment, k i is the hidden state generated by the encoder, w a , W q , Wk is the learnable weight;

[0032] Then, the attention weight α is calculated by soffmax normalization i , and use these weights to weight the encoder hidden state to get the final context vector, the formula is:

[0033]

[0034] Among them, α i is the attention weight, v i is the hidden state produced by the encoder.

[0035] Furthermore, the step (4) comprises:

[0036] Based on the analysis in steps (1)-(3), the fatigue life parameters of important components under specific working conditions are calculated by substituting the following formula:

[0037]

[0038] Among them, L i is the fatigue life of important components, S max is the maximum stress value during the test of important components, S avg is the average stress value of important components during the test, σ max is the maximum strain value of important components, σ avg is the average strain of important components, E max E is the maximum energy density of important components during fatigue under working conditions; avg is the average energy density of important components during fatigue under working conditions, f max is the maximum response frequency observed in the test of important components, f avg is the average frequency of the important components during the test, k m is the fatigue performance parameter of the material, which is related to the material properties and working conditions, τ t is the total test time.

[0039] Furthermore, the step (5) comprises:

[0040] Based on the analysis of steps (1)-(4), the fatigue life of important components of wetland equipment structure is solved by substituting the following formula:

[0041]

[0042] Among them, L i is the fatigue life of wetland equipment structure, D j is the amplitude parameter of fatigue load, K jis the fatigue life influence coefficient, which is related to material properties and working environment. ijmax is the maximum stress value observed during fatigue loading, S ij is the average stress observed during fatigue loading, σ ijmax is the maximum strain value, σ ij is the average strain value, E ijmax is the maximum energy density measured during fatigue, E ij is the average energy density during fatigue, f ijmax is the maximum response frequency observed in real-time monitoring, f ij is the average response frequency in real-time monitoring, n m is the fatigue performance parameter of the material, which is related to the material properties and working conditions. q is the nonlinear response index of fatigue load under environmental conditions.

[0043] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: The present invention uses multi-task learning to predict the life of wetland equipment structures. On the basis of considering multi-task learning under different working conditions, it strengthens the transmission of multi-target information in complex environments and realizes the prediction of the life of wetland equipment structures. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0045] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0046] like Figure 1 As shown, the present invention comprises the following steps:

[0047] (1) Acquisition of real-time load spectrum of wetland equipment structure based on dynamic / static analysis:

[0048] Based on the known wetland equipment, a 3D model of the structure is established in the 3D software SolidWorks to accurately describe the shape, size and relative position of each component and structure. The equipment components include the body, suspension, wheels, telescopic rod, sensor chassis, etc. The established 3D model is imported into the Adams software to analyze its structural dynamics under actual working conditions and obtain the load spectrum of important components of the wetland equipment structure.

[0049] On this basis, important components are extracted from Adams software and imported into Ansys software for static simulation. By setting appropriate boundary conditions and load conditions in Ansys software, the stress and strain distribution of these important components under static conditions can be obtained. Ansys static simulation can reveal the deformation, stress concentration and potential structural weaknesses of components under stress.

[0050] (2) Establishment of load spectrum fusion model based on multi-objective optimization adapted to complex environments:

[0051] Based on the analysis in step (1), a fusion model is established that can adapt to the complex environment of the wetland and perform multi-objective optimization of the load spectrum.

[0052] By introducing a gating mechanism, information can be selectively remembered or forgotten. The core of the mechanism is three gates: input gate, forget gate, and output gate.

[0053] The forget gate controls how much memory information from the previous moment needs to be discarded. The formula is:

[0054] f t =σ(W f ·[h t-1 , x t ]+b f )

[0055] Among them, f t is the output of the forget gate, σ is the Sigmoid activation function, W f is the weight matrix of the forget gate, h t-1 is the hidden state of the previous moment, x t is the input at the current moment, b f is the bias term of the forget gate.

[0056] The input gate controls how much of the current input information can be written into the cell state. The Sigmoid activation function and the tanh activation function are used to determine the cell content that needs to be updated, so that the new information can be integrated into the cell state. The formula is:

[0057] i t =σ(W i ·[h t-1 , x t ]+b i )

[0058]

[0059]

[0060] Where: i t is the output of the input gate, W i is the weight matrix of the input gate, b i is the bias term of the input gate, tanh is the hyperbolic tangent activation function, W c is the weight matrix of the candidate cell state, b c is the bias term of the candidate cell state, is the candidate cell state.

[0061] The output gate determines the output of the cell state at the current moment. First, the output state of the current cell is evaluated through the Sigmoid activation function, and then the current memory state is processed through the tanh activation function. The two are multiplied and then output. The formula is:

[0062] o t =σ(W o ·[h t-1 , x t ]+b o )

[0063] h t =o t tanh(c t )

[0064] Among them: t is the output of the output gate, W o is the weight matrix of the output gate, b o is the bias term of the output gate, h t is the hidden state at the current moment.

[0065] (3) Obtaining optimized load spectrum by integrating multi-task learning with enhanced information transfer:

[0066] Based on the analysis in step (2), in order to enhance the transmission of multi-target information in complex environments, it is necessary to introduce a multi-task learning model to optimize the load spectrum so that the model focuses on certain parts of the input data and assigns different weights to different parts to highlight the features or information that are more important to the task.

[0067] The additive attention mechanism is used to calculate the weighted sum of the inputs of the decoder at each time step, as follows:

[0068]

[0069] Among them, q is the hidden state of the decoder at the current moment, k i is the hidden state generated by the encoder, w a , W q , W k are learnable weights.

[0070] Then, the attention weight α is calculated by soffmax normalization i , and use these weights to weight the encoder hidden state to get the final context vector, the formula is:

[0071]

[0072] Among them, α i is the attention weight, v i is the hidden state produced by the encoder.

[0073] (4) Adjustment of adaptive parameters for fatigue performance analysis of wetland equipment structures:

[0074] Based on the analysis of steps (1) to (3), the fatigue life parameters of important components under specific working conditions are calculated by substituting the following formula:

[0075]

[0076] Among them, L i is the fatigue life of important components, S max is the maximum stress value during the test of important components, S avg is the average stress value of important components during the test, σ max is the maximum strain value of important components, σ avg is the average strain of important components, E max E is the maximum energy density of important components during fatigue under working conditions; avg is the average energy density of important components during fatigue under working conditions, f max is the maximum response frequency observed in the test of important components, f avg is the average frequency of the important components during the test, k m is the fatigue performance parameter of the material, which is related to the material properties and working conditions, τ t is the total test time.

[0077] (5) Fatigue life prediction of wetland equipment structures:

[0078] Based on the analysis of steps (1) to (4), the fatigue life of important components of wetland equipment structures is solved by substituting the following formula:

[0079]

[0080] Among them, L i is the fatigue life of wetland equipment structure, D j is the amplitude parameter of fatigue load, K j is the fatigue life influence coefficient, which is related to material properties and working environment. ijmaxis the maximum stress value observed during fatigue loading, S ij is the average stress observed during fatigue loading, σ ijmax is the maximum strain value, σ ij is the average strain value, E ijmax is the maximum energy density measured during fatigue, E ij is the average energy density during fatigue, f ijmax is the maximum response frequency observed in real-time monitoring, f ij is the average response frequency in real-time monitoring, n m is the fatigue performance parameter of the material, which is related to the material properties and working conditions. q is the nonlinear response index of fatigue load under environmental conditions.

Claims

1. A method for predicting the structural life of wetland equipment based on multi-task learning with enhanced information transfer, characterized in that: The steps include: (1) Acquisition of real-time load spectrum of wetland equipment structure based on dynamic / static analysis; (2) Establishment of a load spectrum fusion model based on multi-objective optimization that adapts to complex environments; (3) Obtaining optimized load spectra by integrating multi-task learning with enhanced information transfer; (4) Adjustment of adaptive parameters for fatigue performance analysis of wetland equipment structures; (5) Fatigue life prediction of wetland equipment structures.

2. The wetland equipment structure life prediction method based on enhanced information transfer multi-task learning according to claim 1 is characterized in that: The step (1) comprises: Based on the known wetland equipment, a three-dimensional model of the structure is established in the three-dimensional software SolidWorks to accurately describe the shape, size and relative position of each equipment component and structure; and the established three-dimensional model is imported into the Adams software to analyze its structural dynamics under actual working conditions, and obtain the load spectrum of important components of the wetland equipment structure; on this basis, the important components are extracted from the Adams software and imported into the Ansys software for static simulation, and the stress and strain distribution of these important components under static conditions is obtained by setting appropriate boundary conditions and load conditions in the Ansys software.

3. The wetland equipment structure life prediction method based on enhanced information transfer multi-task learning according to claim 2 is characterized in that: The equipment components and structures include a body, a suspension, wheels, a telescopic rod and a sensor chassis.

4. The wetland equipment structure life prediction method based on enhanced information transfer multi-task learning according to claim 1 is characterized in that: The step (2) includes establishing a fusion model that can adapt to the complex environment of the wetland and perform multi-objective optimization of the load spectrum based on the analysis of step (1). By introducing a gating mechanism, it is possible to selectively remember or forget information. The core of the gating mechanism is three gates: an input gate, a forget gate, and an output gate.

5. The method for predicting the structural life of wetland equipment based on enhanced information transfer multi-task learning according to claim 4 is characterized in that: The forget gate controls how much memory information from the previous moment needs to be discarded, and the formula is: f t =σ(W f ·[h t-1 ,x t ]+b f ) Among them, f t is the output of the forget gate, σ is the Sigmoid activation function, W f is the weight matrix of the forget gate, h t-1 is the hidden state at the previous moment, x t is the input at the current moment, b f is the bias term of the forget gate.

6. The method for predicting the structure life of wetland equipment based on enhanced information transfer multi-task learning according to claim 4 is characterized in that: The input gate controls how much of the input information at the current moment can be written into the cell state. The Sigmoid activation function and the tanh activation function are used to determine the cell content that needs to be updated, so as to integrate the new information into the cell state. The formula is: i t =σ(W i ·[h t-1 ,x t ]+b i ) Where: i t is the output of the input gate, W i is the weight matrix of the input gate, b i is the bias term of the input gate, tanh is the hyperbolic tangent activation function, W c is the weight matrix of the candidate cell state, b c is the bias term of the candidate cell state, is the candidate cell state.

7. The method for predicting the structure life of wetland equipment based on enhanced information transfer multi-task learning according to claim 4 is characterized in that: The output gate determines the output of the cell state at the current moment. The output state of the current cell is first evaluated by the Sigmoid activation function, and then the current memory state is processed by the tanh activation function. The two are multiplied and output. The formula is: the t =σ(W o ·[h t-1 ,x t ]+b o ) h t =o t fishy t ) Among them: t is the output of the output gate, W o is the weight matrix of the output gate, b o is the bias term of the output gate, h t is the hidden state at the current moment.

8. The method for predicting the structure life of wetland equipment based on enhanced information transfer multi-task learning according to claim 1 is characterized in that: The step (3) comprises: Based on the analysis in step (2), in order to enhance the transmission of multi-target information in complex environments, it is necessary to introduce a multi-task learning model to optimize the load spectrum, so that the model focuses on certain parts of the input data and assigns different weights to different parts to highlight the features or information that are more important to the task. The additive attention mechanism is used to calculate the weighted sum of the inputs of the decoder at each time step, as follows: Among them, q is the hidden state of the decoder at the current moment, k i is the hidden state generated by the encoder, w a , W q , W k is the learnable weight; Then, the attention weight α is calculated by softmax normalization i , and use these weights to weight the encoder hidden state to get the final context vector, the formula is: Among them, α i is the attention weight, v i is the hidden state produced by the encoder.

9. The method for predicting the structural life of wetland equipment based on enhanced information transfer multi-task learning according to claim 1, characterized in that: The step (4) comprises: Based on the analysis in steps (1)-(3), the fatigue life parameters of important components under specific working conditions are calculated by substituting the following formula: Among them, L i is the fatigue life of important components, S max is the maximum stress value during the test of important components, S avg is the average stress value of important components during the test, σ max is the maximum strain value of important components, σ avg is the average strain of important components, E max E is the maximum energy density of important components during fatigue under working conditions; avg is the average energy density of important components during fatigue under working conditions, f max is the maximum response frequency observed in the test of important components, f avg is the average frequency of the important components during the test, k m is the fatigue performance parameter of the material, which is related to the material properties and working conditions, τ t is the total test time.

10. The wetland equipment structure life prediction method based on enhanced information transfer multi-task learning according to claim 1 is characterized in that: The step (5) comprises: Based on the analysis of steps (1)-(4), the fatigue life of important components of wetland equipment structure is solved by substituting the following formula: Among them, L i is the fatigue life of wetland equipment structure, D j is the amplitude parameter of fatigue load, K j is the fatigue life influence coefficient, which is related to material properties and working environment. ijmax is the maximum stress value observed during fatigue loading, S ij is the average stress observed during fatigue loading, σ ijmax is the maximum strain value, σ ij is the average strain value, E ijmax is the maximum energy density measured during fatigue, E ij is the average value of energy density during fatigue, f ijmax is the maximum response frequency observed in real-time monitoring, f ij is the average response frequency in real-time monitoring, n m is the fatigue performance parameter of the material, which is related to the material properties and working conditions. q is the nonlinear response index of fatigue load under environmental conditions.

Images