Spring operating mechanism stress calculation method for multiple fault couplings

Abstract

The application discloses a kind of spring operating mechanism stress calculation methods of multiple fault coupling, it is related to the technical field of mechanical simulation and fault analysis of electric power equipment, including: constructing spring operating mechanism N×N order fault coupling coefficient matrix;Stress exclusive simulation method is used to calculate the stress field data of component under single fault condition;From N×N order fault coupling coefficient matrix, the submatrix of corresponding order of target compound fault and coupling coefficient are extracted, and the stress field data of single fault in target compound fault are linearly superimposed with weighting, to obtain preliminary coupling stress field;Attention enhancement-multiple scale feature fusion neural network model is constructed and trained, and the network parameters are corrected by adaptive weight decay optimization algorithm, and the nonlinear distortion area in preliminary coupling stress field is accurately corrected.The application solves the problem that the existing technology has large calculation deviation under the stress nonlinear distortion of composite fault.

Description

Technical Field

[0001] This invention relates to the field of power equipment mechanical simulation and fault analysis technology, specifically to a method for calculating the stress of a spring operating mechanism with multiple fault coupling. Background Technology

[0002] The CT26 spring operating mechanism is the core power actuation component of a high-voltage circuit breaker, and its operating status directly determines the reliability of the circuit breaker's opening and closing. In long-term, complex operating environments, this mechanism does not only experience single faults, but often exhibits compound faults such as "cam wear + energy storage shaft jamming" and "closing spring fatigue + output shaft jamming." These faults are the superposition of multiple single faults, and there is mutual interference between the faults, resulting in a strong nonlinear distortion in the stress distribution of key components. Its stress field is not a simple superposition of the stress fields of a single fault.

[0003] In existing technologies, stress calculation methods for the CT26 mechanism mostly focus on single fault scenarios, only able to obtain stress data for independent faults such as cam wear and transmission jamming through simulation or experiment. For compound faults, existing methods typically use a simplified approach of "directly adding the stress fields of individual faults" without considering the coupling interference effect between faults. On the one hand, a single fault can change the stress transmission path of another fault; for example, cam wear can cause a shift in the force angle of the transmission mechanism, thereby exacerbating the stress concentration at the jamming point. On the other hand, stress distortion under compound faults exhibits nonlinear characteristics, and simple linear superposition can lead to a large deviation between the calculated results and the actual stress.

[0004] Furthermore, existing technologies lack effective means to quantify the coupling relationships between faults, making it impossible to accurately determine the impact of a single fault on the stress distribution of another fault. This results in highly subjective and poorly repeatable calculations of composite fault stress. Such inaccurate calculations can mislead maintenance personnel in assessing the severity of mechanical faults, leading to insufficient or excessive maintenance, or even causing mechanical failure and widespread power outages. Therefore, developing a method for calculating composite fault stress that quantifies fault coupling relationships and corrects nonlinear distortions has become a pressing technical challenge in the field of mechanical analysis of the CT26 type mechanism. Summary of the Invention

[0005] To address the aforementioned shortcomings in the existing technology, this invention provides a stress calculation method for a multi-fault coupled spring operating mechanism. This method solves the technical problem that the existing technology can only calculate the stress distribution of a single fault in the spring operating mechanism of the CT26 circuit breaker, and cannot accurately cope with complex fault scenarios such as "cam wear + transmission jamming", "closing spring fatigue + energy storage shaft jamming", and "energy storage shaft jamming + output shaft jamming" in actual operation. Furthermore, the nonlinear distortion of stress under complex faults leads to large calculation deviations.

[0006] To achieve the above-mentioned objectives, the technical solution adopted by this invention is as follows: a stress calculation method for a multi-fault coupled spring operating mechanism, comprising the following steps:

[0007] S1: Construct an N×N order fault coupling coefficient matrix for the spring operating mechanism, and quantify the stress influence weights among multiple faults through the coefficient matrix;

[0008] S2: The stress field data of the component under a single fault condition is calculated using a stress-specific simulation method.

[0009] S3: Extract the submatrix and coupling coefficients corresponding to the order of the target composite fault from the N×N order fault coupling coefficient matrix, and perform weighted linear superposition of the stress field data of the single fault in the target composite fault to obtain the preliminary coupled stress field.

[0010] S4: Construct and train an attention-enhanced multi-scale feature fusion neural network model, correct network parameters through an adaptive weight decay optimization algorithm, accurately correct the nonlinear distortion region in the initial coupled stress field, and output the stress distribution results of the composite fault of the spring operating mechanism.

[0011] Furthermore, the construction of the N×N order fault coupling coefficient matrix of the spring operating mechanism in S1 includes the following steps:

[0012] S11: Construct N types of single-fault test platforms and multiple types of composite fault test platforms respectively, and collect stress data of key components under each platform;

[0013] S12: For each type of composite fault, calculate the absolute value of the sum of the measured stress value of the composite fault and the stress value of the single fault, and use it as the stress disturbance intensity.

[0014] S13: Establish a hierarchical model. The target layer is to determine the coupling coefficient. The criterion layer includes stress interference intensity, failure probability, and structural correlation. The scheme layer is a combination of various failures. Construct a judgment matrix and perform consistency verification. After normalization, obtain the subjective weight vector.

[0015] S14: Standardize the composite fault sample data to obtain a normalized matrix, calculate the information entropy of the fault, and calculate the objective weight vector based on the information entropy.

[0016] S15: The subjective and objective weights are fused using variable weighting coefficients to obtain a mixed weight;

[0017] S16: Generate an N×N fault coupling coefficient matrix. This matrix is ​​symmetric, with diagonal elements equal to 1 and off-diagonal elements equal to 0. for:

[0018] = × ×

[0019] in, For the first Hybrid weights for fault classes For the first Hybrid weights for fault classes For fault and The intensity of stress interference, ≠ , 0.1≤ ≤0.8.

[0020] Furthermore, the information entropy of the fault for:

[0021] = ln( +ε)

[0022]

[0023] in, The number of composite fault sample data sets. It is a positive indicator, and ε is used to prevent the logarithmic term from having meaningless positive values. For the first The first group of composite fault samples The original measured stress data for this type of fault. For all composite fault samples, the first A column vector composed of the original measured stress data of the fault type.

[0024] Furthermore, the variable weighting coefficients for:

[0025]

[0026] in, Objective weight vector coefficient of variation, Subjective weight vector coefficient of variation;

[0027] The mixed weight for:

[0028] = +

[0029] in, For the first Subjective weighting of fault types For the first Objective weights for different types of faults.

[0030] Furthermore, the off-diagonal elements of the N×N order fault coupling coefficient matrix satisfy the following: the coupling coefficient of cam wear to energy storage shaft jamming is 0.5–0.7; the coupling coefficient of cam wear to closing spring fatigue is 0.2–0.4; the coupling coefficient of energy storage shaft jamming to closing spring fatigue is 0.1–0.3; the coupling coefficient of cam wear to output shaft jamming is 0.4–0.6; the coupling coefficient of energy storage shaft jamming to output shaft jamming is 0.3–0.5; and the coupling coefficient of output shaft jamming to closing spring fatigue is 0.5–0.7.

[0031] Furthermore, in S3, the stress field data of the individual faults in the target composite fault are weighted and linearly superimposed, as shown in the formula:

[0032]

[0033]

[0034] in, The data is the superimposed stress field data. For the first Fault-like stress values, For normalized mixed weights, For fault and The coupling coefficient, This refers to the type of fault.

[0035] Furthermore, the structure of the attention-enhanced multi-scale feature fusion neural network model in S4 is as follows:

[0036] The input layer nodes include the peak stress after linear superposition, the median stress distribution interval, the slope of stress change, fault level 1 parameters, fault level 2 parameters, fault combination type coding, and multi-fault coupling strength comprehensive factor.

[0037] In the hidden layer, the first layer is a multi-scale feature extraction layer, the second layer is a spatial attention module, and the third layer is a feature fusion layer;

[0038] The output layer nodes include the corrected peak stress, the area ratio of high-stress regions, the nonlinear distortion correction coefficient, and the regional distortion level label.

[0039] The beneficial effects of this invention are:

[0040] (1) This invention uses “experimental quantification + hybrid weighting modeling” to accurately define the stress influence weights between multiple faults, break through the coupling limitation of dual faults, and avoid subjective judgment of coupling relationship; the matrix dimension is extended to N×N order (N≥2), which can cover N typical faults such as cam wear, energy storage shaft jamming, output shaft jamming, closing spring fatigue, opening spring relaxation, bearing wear, and fastener loosening. The coupling coefficient is determined by a hybrid weighting model of the analytic hierarchy process (AHP) and the entropy weight method (EWM), taking into account the dual constraints of subjective engineering experience and objective experimental data.

[0041] (2) This invention corrects nonlinear distortion by using an improved neural network. A multi-fault coupling strength comprehensive factor is added to the input layer to strengthen the correlation between coupling characteristics and stress correction. The hidden layer adopts a three-level structure of "multi-scale feature extraction - spatial attention enhancement - feature fusion". The first layer extracts multi-dimensional stress features through convolutional kernels of different scales. The second layer dynamically allocates weights through the attention mechanism to focus on high distortion regions. The third layer fuses deep and shallow features through residual connections to avoid gradient vanishing. The output layer adds regional distortion level labels (mild / moderate / severe) to achieve the dual functions of stress correction and distortion normalization.

[0042] (3) The present invention is adapted to the structural characteristics and typical compound fault types of the CT26 mechanism, with high calculation accuracy and strong adaptability, providing reliable data support for mechanism operation and maintenance and structural optimization under compound fault conditions. Attached Figure Description

[0043] Figure 1 This is a flowchart of the stress calculation method for a spring operating mechanism with multiple fault coupling according to the present invention. Detailed Implementation

[0044] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0045] like Figure 1 As shown, a method for calculating the stress of a spring-operated mechanism with multiple fault coupling includes the following steps:

[0046] S1: Construct an N×N order fault coupling coefficient matrix for the spring operating mechanism, and quantify the stress influence weights among multiple faults through the coefficient matrix;

[0047] This step uses typical faults of the CT26 type mechanism (such as cam wear, transmission mechanism jamming, energy storage shaft jamming, output shaft jamming, closing spring fatigue, etc.) as the matrix dimension, defining the matrix elements as "stress influence weighting coefficients of fault A on fault B". Through stress comparison experiments of single and compound faults and a hybrid weighting model of Analytic Hierarchy Process (AHP)-Entropy Weighting Method (EWM), the values ​​of each coupling coefficient are determined. The matrix element values ​​range from 0.1 to 0.8, characterizing the degree of interference of a single fault on the stress distribution of another fault. The specific implementation is as follows:

[0048] Constructing the N×N order fault coupling coefficient matrix of the spring operating mechanism includes the following steps:

[0049] S11: Construct N types of single-fault test platforms and multiple types of composite fault test platforms respectively, and collect stress data of key components under each platform;

[0050] Design comparative experiments of single faults and compound faults, and build single fault test platforms for cam wear, transmission jamming, energy storage shaft jamming, output shaft jamming, and closing spring fatigue, as well as multiple compound fault test platforms such as "cam wear + transmission jamming", "closing spring fatigue + energy storage shaft jamming", "energy storage shaft jamming + output shaft jamming", and "transmission jamming + closing spring fatigue + energy storage shaft jamming".

[0051] S12: For each type of composite fault, calculate the absolute value of the sum of the measured stress value of the composite fault and the stress value of the single fault, and use it as the stress interference intensity to quantify the degree of stress interference between multiple faults.

[0052] S13: Establish a hierarchical model. The target layer determines the coupling coefficients; the criterion layer includes stress disturbance intensity, failure probability, and structural correlation; and the scheme layer consists of various failure combinations. Pairwise comparisons are performed on the failure combinations under each criterion layer indicator. A judgment matrix is ​​constructed using a 1-9 scale and a consistency test (CI < 0.1) is conducted. After normalization, the subjective weight vector is obtained. =[ , ,..., ];

[0053] S14: Standardize the composite fault sample data to obtain the normalized matrix Y=( ) m×N (Positive indicator) (The negative index is the opposite), calculate the information entropy of the fault. And calculate the objective weight vector based on information entropy. =[ , ,..., ];

[0054] Information entropy of faults for:

[0055] = ln( +ε)

[0056]

[0057] in, The number of composite fault sample data sets. It is a positive indicator, and ε is used to prevent the logarithmic term from having a very small positive value that is meaningless. For the first The first group of composite fault samples The original measured stress data for this type of fault. For all composite fault samples, the first A column vector composed of the original measured stress data of the fault type.

[0058] S15: The subjective and objective weights are fused using variable weighting coefficients to obtain a mixed weight;

[0059] Variable weight coefficient for:

[0060]

[0061] in, Objective weight vector coefficient of variation, Subjective weight vector coefficient of variation;

[0062] Mixed weights for:

[0063] = +

[0064] in, For the first Subjective weighting of fault types For the first Objective weights for different types of faults.

[0065] S16: Generate an N×N fault coupling coefficient matrix. This matrix is ​​symmetric, with diagonal elements... =1.0 (Single fault self-weight coefficient), off-diagonal element for:

[0066] = × ×

[0067] in, For the first Hybrid weights for fault classes For the first Hybrid weights for fault classes For fault and The intensity of stress interference, ≠ , 0.1≤ ≤0.8.

[0068] The off-diagonal elements of the N×N order fault coupling coefficient matrix satisfy the following: the coupling coefficient of cam wear to energy storage shaft jamming is 0.5–0.7; the coupling coefficient of cam wear to closing spring fatigue is 0.2–0.4; the coupling coefficient of energy storage shaft jamming to closing spring fatigue is 0.1–0.3; the coupling coefficient of cam wear to output shaft jamming is 0.4–0.6; the coupling coefficient of energy storage shaft jamming to output shaft jamming is 0.3–0.5; and the coupling coefficient of output shaft jamming to closing spring fatigue is 0.5–0.7.

[0069] The core of this step is to accurately define the stress influence weights among multiple faults through "experimental quantification + hybrid weighted modeling," breaking through the limitations of dual-fault coupling and avoiding subjective judgment of coupling relationships. The matrix dimension is expanded to N×N order (N≥2), which can cover N typical faults such as cam wear, energy storage shaft jamming, output shaft jamming, closing spring fatigue, opening spring relaxation, bearing wear, and fastener loosening. The coupling coefficient is determined by a hybrid weighted model of the Analytic Hierarchy Process (AHP) and the Entropy Weight Model (EWM), taking into account both subjective engineering experience and objective experimental data constraints.

[0070] S2: The stress field data of the component under a single fault condition is calculated using a stress-specific simulation method.

[0071] In this embodiment, a dedicated simulation method for stress of key components of the CT26 type mechanism is used to calculate the stress field data of components such as cam, transmission shaft, and transmission connecting rod under various single fault conditions, such as cam wear, transmission mechanism jamming, energy storage shaft jamming, output shaft jamming, and closing spring fatigue.

[0072] The dedicated simulation method for stress in key components of the CT26 mechanism is based on the solid dimensions of the CT26 mechanism. It constructs a multibody dynamics model that retains details of stress-sensitive areas, imports the mechanical parameters of components such as alloy structural steel and spring steel, loads the actual opening and closing load curves, sets targeted friction coefficients (0.1 to 0.15 for the cam-roller contact pair, 0.09 to 0.13 for the energy storage shaft bearing contact pair, and 0.08 to 0.12 for the output shaft hinge), and uses a step-by-step iterative algorithm to simulate the stress distribution under a single fault condition across all operating conditions.

[0073] In this step, a CT26-type mechanism failure test platform was built to simulate N types of single failures and multiple types of compound failures (including combinations of double failures, triple failures, and more). Data was collected according to the severity level of the failures to ensure coverage of the entire failure evolution stage. Stress data of key parts were collected using micro stress gauges, and multiple sets of data were collected for each type of working condition and the average value was taken to reduce experimental errors.

[0074] S3: Extract the submatrix and coupling coefficients corresponding to the order of the target composite fault from the N×N order fault coupling coefficient matrix, and perform weighted linear superposition of the stress field data of the single fault in the target composite fault to obtain the preliminary coupled stress field.

[0075] The linear superposition calculation object is the stress value at each coordinate point of the key component. Following the principle of "point-by-point superposition and full-domain coverage", the integrity of the initial coupled stress field is ensured.

[0076] The formula for weighted linear superposition is:

[0077]

[0078]

[0079] in, The data is the superimposed stress field data. For the first Fault-like stress values, For normalized mixed weights, For fault and The coupling coefficient, This refers to the type of fault.

[0080] In this embodiment, the target composite fault (including Class of faults, ≤N), extract the corresponding from the N×N order fault coupling coefficient matrix. × The order submatrix and coupling coefficients will The stress field data of a single fault are weighted and linearly superimposed according to the above formula to obtain the preliminary coupled stress field.

[0081] S4: Construct and train an attention-enhanced multi-scale feature fusion neural network model, correct network parameters through an adaptive weight decay optimization algorithm, accurately correct the nonlinear distortion region in the initial coupled stress field, and output the stress distribution results of the composite fault of the spring operating mechanism.

[0082] The structure of the attention-enhanced multi-scale feature fusion neural network model is as follows:

[0083] The input layer nodes include the peak stress after linear superposition, the median stress distribution interval, the slope of stress change, fault level 1 parameters, fault level 2 parameters, fault combination type coding, and multi-fault coupling strength comprehensive factor.

[0084] In the hidden layer, the first layer is a multi-scale feature extraction layer (12 nodes, ELU activation function), the second layer is a spatial attention module (10 nodes, adaptive weight allocation mechanism), and the third layer is a feature fusion layer (8 nodes, residual connection mechanism).

[0085] The output layer nodes include the corrected peak stress, the area ratio of high-stress regions, the nonlinear distortion correction coefficient, and the regional distortion level label.

[0086] The activation function used is the ELU function, the training algorithm adopts the AdamW adaptive weight decay optimization strategy, combined with early stopping mechanism and Dropout technology, the number of iterations is set to 1000-2000 times, and the convergence error threshold is set to 0.001.

[0087] The neural network training optimization constructs a stress sample set covering different combinations of fault numbers and fault levels, which is proportionally divided into training and test sets. The training algorithm adopts an adaptive weight decay optimization strategy (AdamW), combined with an early stopping mechanism and Dropout technology to suppress overfitting, ensuring that the model converges quickly and has stable generalization ability. After training, the model is verified through experimental data to ensure that the correction deviation meets the engineering accuracy requirements.

[0088] This step, which uses an improved neural network to correct nonlinear distortion, is the core of enhancing creativity and computational accuracy: Input layer optimization: Based on the original stress parameters, fault levels, and fault combination encodings, a new "multi-fault coupling strength comprehensive factor" (obtained by weighted summation of coupling coefficients between k types of faults) is added, forming a 7-dimensional input feature to strengthen the correlation between coupling characteristics and stress correction; Innovative hidden layer architecture: A three-level structure of "multi-scale feature extraction - spatial attention enhancement - feature fusion" is adopted. The first layer extracts multi-dimensional stress features through convolutional kernels of different scales, the second layer dynamically allocates weights through an attention mechanism to focus on high-distortion regions, and the third layer fuses deep and shallow layer features through residual connections to avoid gradient vanishing; Output layer expansion: The corrected stress peak value, high-stress region proportion, and distortion correction coefficient are retained, and a new "regional distortion level label" (mild / moderate / severe) is added to achieve the dual functions of stress correction and distortion normalization.

[0089] Perform nonlinear correction: Input the initial coupled stress field parameters into the trained neural network, output correction coefficients and distortion level labels, adjust the stress values ​​at each coordinate point according to the formula, and perform secondary fine correction on the severely distorted region:

[0090]

[0091] Finally, the final stress distribution cloud map is output, along with the stress peak value, coordinates of high-stress areas, stress change time series curves, and distortion level labels for each area, providing comprehensive data support for operation and maintenance and structural optimization.

[0092] The necessary experimental data of the present invention will be further described in detail below with reference to specific embodiments, taking the combined fault of "energy storage shaft jamming + closing spring fatigue" in the CT26 type mechanism as an example:

[0093] Composite fault type: a combination of two faults, denoted as [Fault 1 ( = Energy storage shaft jamming], [Fault 2 ( = Closing spring fatigue], fault dimension N=2, a 2×2 order fault coupling coefficient matrix needs to be constructed.

[0094] Key calculation components: Select [energy storage shaft] and [output shaft]. Specific parameters are shown in Table 1.

[0095] Table 1. Peak stress of key components under typical failure conditions

[0096]

[0097] Step 1: Construct a 2×2 order fault coupling coefficient matrix

[0098] 1.1 Calculation of stress disturbance intensity

[0099] definition: = Characterizes the degree of stress interference between faults ( =1, =2, because the matrix is ​​symmetric. = ).

[0100] =

[0101] =

[0102] Take the maximum interference intensity of the key component: =max(431.51, 342.98)=431.51MPa.

[0103] 1.2 Determining Subjective Weights using AHP (Analytic Hierarchy Process)

[0104] (1) Construction of hierarchical structure model

[0105] Target layer: Determine and Subjective weight

[0106] Criterion layer: (Degree of stress interference) (Probability of failure) (Institutional structure correlation)

[0107] Solution layer: (Energy storage shaft jamming) (Fatigue of the closing spring)

[0108] (2) Constructing the judgment matrix and calculating the weights

[0109] ① Judgment matrix A of the criterion layer to the target layer

[0110] The parameter values ​​for the criterion layer are shown in Table 2:

[0111] Table 2. Criterion Layer Parameter Values

[0112]

[0113] Weight calculation: =1+3+3=7, =1 / 3 + 1 + 1 = 7 / 3 ≈ 2.333 =7 / 3≈2.333

[0114] Criterion layer weights: =7 / (7+7 / 3+7 / 3)=0.6, =0.2, =0.2

[0115] Consistency test: Calculate the largest eigenvalue A× =

[0116] = = =3

[0117] Consistency Index (CI) = = =0<0.1, the test is passed.

[0118] ② Judgment matrix of scheme layer to criterion layer ( , , , , , )

[0119] Below : The interference is slightly stronger than The scale is 1.2.

[0120] = Normalized weights: =0.4545, =0.5455

[0121] Below : and Similar probabilities, scale 1

[0122] = Normalized weights: =0.5, =0.5

[0123] Below : Slightly more correlated with springs, scale 1.1

[0124] = Normalized weights: =0.4545, =0.5

[0125] ③ Integration of subjective weights at the scheme level

[0126] = × + × + × =0.4545×0.6+0.5×0.2+0.4545×0.2≈0.4636

[0127] =0.5455×0.6+0.5×0.2+0.5×0.2≈0.5273

[0128] Final subjective weight vector: =[0.4636,0.5273]

[0129] 1.3 Determining Objective Weights using EWM (Entropy Weight Method)

[0130] (1) Sample data standardization

[0131] Take m=3 sets of composite fault samples and construct a sample matrix. X=

[0132] Standardized formula:

[0133] Column (j=1): min=515.2, max=520.3, =0.5098, =1, =0

[0134] Column (j=2): min=489.7, max=495.1, =0.5019, =1, =0

[0135] Standardized matrix: Y=

[0136] (2) Calculate information entropy =- ln( +ε)

[0137] :∑ ln( +ε)=0.5098×ln0.5098+1×ln1+0×ln ≈ 0.343

[0138] = ×( 0.343)≈0.312

[0139] :∑ ln( +ε)=0.5019×ln0.5019+1×ln1+0×ln ≈ 0.347

[0140] = ×( 0.347)≈0.316

[0141] (3) Calculation of objective weights

[0142] =0.688, =0.684, =1.372

[0143] =1.3720.688≈0.5015, =1.3720.684≈0.4985

[0144] Final objective weight vector: =[0.5015,0.4985]

[0145] 1.4 Dynamic Fusion of Hybrid Weights

[0146] (1) Calculate the coefficient of variation

[0147] Subjective weight of Mean μA = ≈0.4955, standard deviation σA≈0.0323

[0148] = ≈0.0652

[0149] Objective weight of Mean μE = 0.5, Standard deviation σE ≈ 0.0015

[0150] = =0.003

[0151] (2) Calculation of mixed weights

[0152] =0.4 + 0.2 × 0.0652 + 0.0030.003 ≈ 0.4088

[0153] = × +(1 )× ≈0.4088×0.4636+0.5912×0.5015≈0.486

[0154] = × +(1 )× ≈0.4088×0.5273+0.5912×0.4985≈0.5103

[0155] Normalized mixed weights: ω=[0.486,0.5103]

[0156] 1.5 Generate a 2×2 coupling coefficient matrix

[0157] Diagonal elements: the weight of a single fault. = =1.0

[0158] Non-diagonal elements: = × × ( = Ensure 0.1 ≤ ≤0.8)

[0159] = =0.486×0.5103× ≈0.247

[0160] Final matrix: K=

[0161] Step 2: Calculate the stress field of a single fault

[0162] (Energy storage shaft jamming): ,

[0163] (Fatigue of closing spring): ,

[0164] Step 3: Linear superposition of coupled stresses from multiple faults

[0165]

[0166] k =2 (double fault) (Normalized mixed weights) =0.2.

[0167] 3.1 Calculation of Normalized Mixed Weights

[0168] ≈0.4878, ≈0.5122.

[0169] 3.2 Superposition Calculation

[0170] (1) =(517.8+492.41)×0.5×1.2=1010.21×0.6=606.126MPa≈606.13MPa

[0171] (2) =(340.39+469.79)×0.6=810.18×0.6=486.108MPa≈486.11MPa

[0172] Step 4: Attention Enhancement - Nonlinear Correction of Multi-Scale Feature Fusion Neural Network

[0173] 4.1 Neural Network Model Parameters

[0174] Input layer: 7-dimensional features (stacked peak value 606.13MPa, median stress range 303MPa, stress slope 15MPa / ms, ... Level 1 Level 1, Fault Code 01, Coupling Strength Factor 0.2)

[0175] Hidden layers: 3 layers (12-node multi-scale extraction layer + 10-node spatial attention layer + 8-node feature fusion layer, ELU activation, residual connections)

[0176] Output layer: 4-dimensional results (corrected peak value, high stress percentage, distortion coefficient, distortion level)

[0177] Training parameters: AdamW optimization, 1500 iterations, convergence error 0.001, Dropout to suppress overfitting.

[0178] 4.2 Corrected formula, after training, the correction coefficient of the composite fault is 0.98.

[0179] (1) =606.13 × 0.98 = 594.007 MPa ≈ 594 MPa

[0180] (2) =486.11×0.98=476.388MPa≈476.39MPa

[0181] Step 5: Deviation Rate Calculation and Result Verification

[0182] 5.1 Deviation rate = ×100%, the measured values ​​are shown in Table 3:

[0183] Table 3 Measured values ​​of key components

[0184]

[0185] 5.2 Calculation Results

[0186] (1) = ×100%≈2.64%

[0187] (2) = ×100%≈1.97%

[0188] The following comparison between this method and conventional methods will provide a more detailed explanation of this solution, using the combined fault of "energy storage shaft jamming + closing spring fatigue" in the CT26 type mechanism as an example:

[0189] (1) Existing technology 1: direct superposition of stress from a single fault

[0190] formula: The calculation results are shown in Table 4.

[0191] Table 4 Calculation Results of Prior Art 1

[0192]

[0193] (2) Existing technology 2: Simplified linear superposition (fixed coupling coefficient 0.3, no correction)

[0194] formula: The calculation results are shown in Table 5.

[0195] Table 5 Calculation Results of Prior Art 2

[0196]

[0197] (3) Deviation rate calculation: to measure the accuracy of the calculation, the calculation results are shown in Table 6.

[0198] Table 6 Deviation Rate Calculation Results

[0199]

[0200] It can be seen that the technical solution of the present invention has significant progress compared with the prior art.

[0201] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of the invention.

Claims

1. A method for calculating the stress of a spring-operated mechanism with multiple fault coupling, characterized in that, Includes the following steps: S1: Construct an N×N order fault coupling coefficient matrix for the spring operating mechanism, and quantify the stress influence weights among multiple faults through the coefficient matrix; The construction of the N×N order fault coupling coefficient matrix of the spring operating mechanism in S1 includes the following steps: S11: Construct N types of single-fault test platforms and multiple types of composite fault test platforms respectively, and collect stress data of key components under each platform; S12: For each type of composite fault, calculate the absolute value of the sum of the measured stress value of the composite fault and the stress value of the single fault, and use it as the stress disturbance intensity. S13: Establish a hierarchical model. The target layer is to determine the coupling coefficient. The criterion layer includes stress interference intensity, failure probability, and structural correlation. The scheme layer is a combination of various failures. Construct a judgment matrix and perform consistency verification. After normalization, obtain the subjective weight vector. S14: Standardize the composite fault sample data to obtain a normalized matrix, calculate the information entropy of the fault, and calculate the objective weight vector based on the information entropy. S15: The subjective and objective weights are fused using variable weighting coefficients to obtain a mixed weight; S16: Generate an N×N fault coupling coefficient matrix. This matrix is ​​symmetric, with diagonal elements equal to 1 and off-diagonal elements equal to 0. for: = × × in, For the first Hybrid weights for fault classes For the first Hybrid weights for fault classes For fault and The intensity of stress interference, ≠ , 0.1≤ ≤0.8; S2: The stress field data of the component under a single fault condition is calculated using a stress-specific simulation method. S3: Extract the submatrix and coupling coefficients corresponding to the order of the target composite fault from the N×N order fault coupling coefficient matrix, and perform weighted linear superposition of the stress field data of the single fault in the target composite fault to obtain the preliminary coupled stress field. S4: Construct and train an attention-enhanced multi-scale feature fusion neural network model, correct network parameters through an adaptive weight decay optimization algorithm, accurately correct the nonlinear distortion region in the initial coupled stress field, and output the stress distribution results of the composite fault of the spring operating mechanism.

2. The stress calculation method for a multi-fault coupled spring operating mechanism according to claim 1, characterized in that, The information entropy of the fault for: = ln( +e) in, The number of composite fault sample data sets. It is a positive indicator, and ε is used to prevent the logarithmic term from having meaningless positive values. For the first The first group of composite fault samples The original measured stress data for this type of fault. For all composite fault samples, the first A column vector composed of the original measured stress data of the fault type.

3. The stress calculation method for a multi-fault coupled spring operating mechanism according to claim 1, characterized in that, The variable weighting coefficient for: in, Objective weight vector coefficient of variation, Subjective weight vector coefficient of variation; The mixed weight for: = + in, For the first Subjective weighting of fault types For the first Objective weights for different types of faults.

4. The stress calculation method for a multi-fault coupled spring operating mechanism according to claim 1, characterized in that, The off-diagonal elements of the N×N order fault coupling coefficient matrix satisfy the following: the coupling coefficient of cam wear to energy storage shaft jamming is 0.5–0.7; the coupling coefficient of cam wear to closing spring fatigue is 0.2–0.4; the coupling coefficient of energy storage shaft jamming to closing spring fatigue is 0.1–0.3; the coupling coefficient of cam wear to output shaft jamming is 0.4–0.6; the coupling coefficient of energy storage shaft jamming to output shaft jamming is 0.3–0.5; and the coupling coefficient of output shaft jamming to closing spring fatigue is 0.5–0.

7.

5. The stress calculation method for a multi-fault coupled spring operating mechanism according to claim 1, characterized in that, In S3, the stress field data of a single fault in the target composite fault are weighted and linearly superimposed, as shown in the formula: in, The data is the superimposed stress field data. For the first Fault-like stress values, For normalized mixed weights, For fault and The coupling coefficient, This refers to the type of fault.

6. The stress calculation method for a multi-fault coupled spring operating mechanism according to claim 1, characterized in that, The structure of the attention-enhanced multi-scale feature fusion neural network model in S4 is as follows: The input layer nodes include the peak stress after linear superposition, the median stress distribution interval, the slope of stress change, fault level 1 parameters, fault level 2 parameters, fault combination type coding, and multi-fault coupling strength comprehensive factor. In the hidden layer, the first layer is a multi-scale feature extraction layer, the second layer is a spatial attention module, and the third layer is a feature fusion layer; The output layer nodes include the corrected peak stress, the area ratio of high-stress regions, the nonlinear distortion correction coefficient, and the regional distortion level label.

Images