Steel wire rope fatigue life prediction method based on true stress-true strain curve

Abstract

The invention discloses a steel wire rope fatigue life prediction method based on a true stress-true strain curve. Firstly, comprehensive data acquisition and processing are performed on the steel wire rope, an accurate stress-strain model is established, and multiple factors such as a strand structure and a contact relation are considered. Then, an N-BEATS network model based on physical information is constructed, a stress-strain model is integrated, and a specific network structure and a calculation process are designed; the model is trained and optimized through the collected data, the state parameters of the steel wire rope are monitored in real time through a sensor, and then the fatigue life is predicted. And a layered full-life-cycle management platform is also constructed and covers the layers of data acquisition, transmission, storage, analysis processing, application service and the like, so that efficient state monitoring, life prediction and maintenance management are realized, and the use efficiency and safety of the steel wire rope are improved.

Description

Technical Field

[0001] The invention relates to the technical field of steel wire ropes, and in particular to a steel wire rope fatigue life prediction method based on a true stress-true strain curve. Background Art

[0002] In the modern industrial field, wire ropes, as an important load-bearing and transmission component, are widely used in many fields such as lifting equipment, elevators, mining machinery, bridges, etc. However, due to the complexity of its working environment and high-intensity use, wire ropes are prone to fatigue damage, which affects their service life and safety.

[0003] Traditional wire rope inspection methods mainly rely on manual periodic inspections, which is not only inefficient but also difficult to accurately detect early damage inside the wire rope. In addition, some physical inspection methods, such as magnetic particle inspection and ultrasonic inspection, can detect some surface and near-surface defects, but they still have great limitations in evaluating the fatigue life of the entire wire rope.

[0004] With the development of computer technology and sensor technology, data-driven wire rope fatigue life prediction methods have gradually attracted attention. However, existing methods often have deficiencies in the comprehensiveness and accuracy of data collection, and fail to fully consider the impact of the complex structure of the wire rope and the multi-factor working environment on its fatigue life. For example, in the data collection stage, detailed data on the material properties, structural parameters, mechanical properties and other aspects of the wire rope are not covered; when establishing the model, the structural details of the wire rope, such as strands, filaments, and rope cores, as well as the contact relationship with components such as pulleys and drums are not fully considered, resulting in the accuracy and reliability of the model being affected.

[0005] At the same time, the existing wire rope fatigue life prediction model needs to be improved in terms of computational efficiency and prediction accuracy when processing massive monitoring data, and it is difficult to meet the needs of real-time and accurate life prediction and health management of wire ropes in actual engineering projects. Summary of the invention

[0006] In order to solve the above problems, the present invention proposes a method for predicting the fatigue life of a steel wire rope based on a true stress-true strain curve.

[0007] The fatigue life prediction method of the steel wire rope based on the true stress-true strain curve has the following specific steps and is characterized by:

[0008] Step 1: Data collection and processing: collect data of wire ropes of different types and specifications, including material properties, structural parameters, mechanical properties, etc.; establish a stress-strain model for wire ropes, taking into account the structural details of the wire ropes, such as strands, wires, and rope cores, as well as the contact relationship between the wire ropes and components such as pulleys and drums;

[0009] Step 2: Establish a N-BEATS network model based on physical information, design the N-BEATS network structure for wire rope fatigue life prediction, integrate the wire rope stress-strain model in step 1, and select appropriate activation functions and loss functions to ensure that the model can effectively learn and predict;

[0010] Step 3: Model training and optimization: Use the collected true stress-true strain curve data to train the model and output the fatigue life prediction value of the wire rope. By adjusting the model parameters and optimizing the algorithm, the prediction accuracy and generalization ability of the model can be improved.

[0011] Step 4: Real-time monitoring of the wire rope status, using a sensor network to monitor the wire rope's stress, strain, temperature, vibration and other parameters in real time;

[0012] Step 5: Fatigue life prediction: according to the real-time monitored wire rope status data and working condition information in step 4, the stress-strain parameters are input into the trained neural network model to predict the fatigue life of the wire rope;

[0013] Step 6: Build a wire rope health management system and establish a wire rope full life cycle management platform to realize functions such as wire rope status monitoring, life prediction, and maintenance management to improve the efficiency and safety of wire rope use.

[0014] The present invention is further improved, and the stress-strain model of the steel wire rope established in step 1 can be expressed as:

[0015] Step 1.1, first, for a single steel wire in the wire rope, describe the stress-strain relationship in its elastic stage according to Hooke's law:

[0016] σ=Eε

[0017] Among them, σ is stress, E is elastic modulus, and ε is strain. However, since the wire rope is a complex structure composed of multiple strands of steel wires and a rope core, the interaction between strands and wires needs to be considered at the same time;

[0018] The wire rope is composed of individual strands, each with a cross-sectional area of ​​A. i (i=1,2,…,n), n represents the number of strands in the wire rope; the elastic modulus is E i , the equivalent cross-sectional area A of the entire wire rope can be expressed as:

[0019]

[0020] Among them, A i represents the cross-sectional area of ​​the i-th strand. When subjected to axial tension F, the stress σ for each strand is i It can be expressed as:

[0021]

[0022] Among them, F i is the tension on each share;

[0023] According to the force balance condition:

[0024]

[0025] The strain per share ε can be obtained i :

[0026]

[0027] Among them, E i Represents the elastic modulus of the i-th strand. The macroscopic strain ε of the entire wire rope can be obtained by weighted average, and the weight coefficient w i It is expressed as:

[0028]

[0029] Among them, j represents the serial number of the steel wire in each share, w i The weight coefficient of each share when the macroscopic strain of the wire rope is expressed, the macroscopic strain ε can be expressed as:

[0030]

[0031] Step 1.2, correct the contact stress. When the wire rope contacts the pulley, bending stress will be generated. The bending stress σ b It can be expressed as:

[0032]

[0033] M=TR

[0034] Where M is the bending moment when the wire rope contacts the pulley, T is the tension of the wire rope, R is the pulley radius, y is the distance to the neutral axis, and I is the cross-sectional inertia moment. When calculating the cross-sectional inertia moment I, the distribution of strands and wires is considered. Let the outer diameter of the wire rope be D and the inner diameter be d, and its cross-sectional inertia moment I can be approximated as:

[0035]

[0036] In the contact area, the total stress σ total :

[0037] σ total =σ+σ b

[0038] Where σ is the axial tensile stress; when the wire rope passes around the pulley, the friction force F f It can be expressed as:

[0039] F f =μT

[0040] Among them, μ is the friction coefficient. The friction will produce tangential stress inside the wire rope. The tangential stress σ t It is expressed as:

[0041]

[0042] The total stress at this time is expressed as σ total :

[0043] σ total =σ+σ b +σ t

[0044] Step 1.3, improve the model according to the test data, and obtain the true stress-true strain curve through the tensile testing machine; during the test, measure the elongation ΔL under different tensile forces F, and calculate the true stress σ t and strain ε t :

[0045]

[0046] Among them, A t is the actual cross-sectional area, L0 is the original length, and L is the length after deformation;

[0047] The data points obtained from the experiment (σ t ,ε t ) is compared with the stress-strain relationship calculated by the theoretical model; the least square method and other data fitting methods are used to correct the parameters in the theoretical model: elastic modulus E, friction coefficient μ, etc., so that the theoretical model can better fit the test data; after the fitting and correction process, the final wire rope stress-strain model is obtained:

[0048] σ total =Eε

[0049] The present invention is further improved, and the process of establishing the N-BEATS network model based on physical information in step 2 can be expressed as follows:

[0050] Step 2.1: Establish the model input layer. After appropriate preprocessing of the stress-strain data measured by the sensor, the input vector is expressed as X = [x in1 ,x in2 ,...,x int ], where t represents the tth moment, x i Represents the stress value, strain value, wire diameter, wire strand number, yield strength, lay length and elastic modulus calculated from the stress-strain model in step 1 at different times;

[0051] Step 2.2: Construct the block structure of the N-BEATS model. The input data first passes through a multi-layer fully connected neural network. Let the input of the block structure be x in , after L layers of fully connected neural network, the weight matrix of each layer is W l , with a bias of b l , the activation function is ReLU, then its output can be expressed as:

[0052] h l =ReLU(W l h l-1 +b l ),l=1,2,...,L

[0053] Among them, h0 is the input vector X of the input layer, h l It represents the output of the lth layer of fully connected neural network in the basic block after the activation function; after passing through the Lth layer of fully connected neural network, the output is divided into two parts, and passes through a fully connected layer to obtain the forward and backward prediction values ​​respectively: Assume that the two parts of the output after segmentation are h L1 and h L2 ,but:

[0054]

[0055] Among them, W forecast and b forecast Indicates the value used to generate the prediction The weight matrix and bias of the fully connected layer, W backcast and b backcast Indicates the value used to generate the prediction The weight matrix and bias of the fully connected layer;

[0056] Step 2.3, double residual stacking, in a stack, the input x of the i-th basic block i Backcast output with the previous basic block There is a residual connection between them, and its calculation formula is:

[0057]

[0058] Among them, x1 is the original input of stack, is the predicted value of the i-1th stack; and the forecast output of the block It is the sum of forecast outputs of all basic blocks in the stack:

[0059]

[0060] Where N represents the number of basic blocks in a stack. is the predicted value of the i-th stack;

[0061] Step 2.4, calculate trend block and seasonal block:

[0062] The trend block is obtained by removing the last fully connected layer from the basic block and multiplying the output with two matrices respectively. The output formula is:

[0063]

[0064] Among them, g trend is the activation function, which is used to transform the weighted results to obtain the final trend prediction value, V trend and U trend Represents the matrix in the trend block multiplied by the output of the basic block. x and y are the outputs of the basic block after removing the last fully connected layer, which are feature representations of different dimensions of the input time series data.

[0065] The seasonal block is after removing the last layer of the basic block, and the output is multiplied by two other matrices respectively. The formula is:

[0066]

[0067] Among them, g season is the activation function used to generate the final seasonal forecast value, V season and U season Represents the matrix in the seasonal block that is multiplied with the base block output;

[0068] Step 2.5, construct the model output layer: The N-BEATS model consists of multiple stacks. The final output of the model is the sum of the forecast outputs of all stacks. The final output formula is:

[0069]

[0070] in, is the final output of the model, i.e. the predicted value of fatigue life of the wire rope, and M is the number of all stacks.

[0071] The present invention is further improved, and the construction of the wire rope health management system in step 6 can be expressed as follows:

[0072] Establish a wire rope full life cycle management platform based on cloud computing platform, adopting a layered architecture design, including data collection layer, data transmission layer, data storage layer, data analysis and processing layer, and application service layer;

[0073] Data acquisition layer: Various types of sensors are deployed at the wire rope operation site, such as high-precision stress sensors, strain sensors, temperature sensors, and vibration sensors. These sensors are connected to the data acquisition nodes via wired or wireless means to collect real-time information on the stress value, strain data, temperature changes, vibration frequency, amplitude, and other parameter information of the wire rope during operation.

[0074] Data transmission layer: Use industrial-grade wireless networks to transmit collected data from sensor nodes to data center servers; use data encryption technology to encrypt the transmitted data to prevent the data from being stolen or tampered with during transmission; at the same time, set up a data cache mechanism. When the network is briefly interrupted, the sensor node can temporarily store the collected data in the local cache and automatically upload it after the network is restored to avoid data loss;

[0075] Data storage layer: Build a distributed database system to store massive amounts of wire rope monitoring data. The database uses a combination of relational and non-relational databases. The relational database is used to store structured data such as basic information of the wire rope, equipment information, and user information. The non-relational database is used to store a large amount of monitoring data to meet the needs of fast data reading and writing and high scalability. At the same time, back up the data regularly to prevent data loss, and adopt a data archiving strategy to archive and store historical data for subsequent query and analysis.

[0076] Model-based life prediction: Integrate the physical information-based N-BEATS network model trained in step 2 into the health management system; when the system receives the real-time monitored stress-strain parameters of the wire rope and other relevant working condition information, input these data into the N-BEATS network model, and the model predicts the remaining fatigue life of the wire rope based on the learned stress-strain law and fatigue damage mechanism of the wire rope;

[0077] Evaluation and update of life prediction results: Regularly compare and evaluate the actual wire rope fatigue life data with the model prediction results; if it is found that there is a large deviation between the prediction results and the actual situation, retrain and optimize the model and adjust the model parameters to improve the prediction accuracy and adaptability of the model;

[0078] Maintenance plan formulation: Based on the condition monitoring results and life prediction information of the wire rope, the system automatically generates a maintenance plan; the maintenance plan includes maintenance tasks, maintenance schedule, maintenance personnel allocation, and a list of required maintenance tools and materials to ensure the safe operation of the equipment.

[0079] The present invention provides a method for predicting fatigue life of a steel wire rope based on a true stress-true strain curve. Beneficial effects: The technical effects of the present invention are:

[0080] 1. The present invention can more accurately predict the fatigue life of the steel wire rope, reduce prediction errors, and provide a reliable basis for maintenance and replacement decisions by comprehensively collecting various data of the steel wire rope, including material properties, structural parameters, mechanical properties, etc., and establishing a stress-strain model and a complex network model that considers multiple factors.

[0081] 2. The present invention monitors the stress, strain, temperature, vibration and other parameters of the wire rope in real time, and formulates a maintenance plan based on accurate life prediction, which can timely discover potential faults and safety hazards, effectively avoid accidents caused by sudden failure of the wire rope, and ensure the safety of personnel and equipment.

[0082] 3. The wire rope full life cycle management platform constructed by the present invention realizes comprehensive control and scientific management of the wire rope status. The use of the wire rope can be reasonably arranged according to the prediction results to avoid overuse or premature replacement, thereby improving the use efficiency of the wire rope and reducing operating costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Figure 1 It is a flow chart of the present invention.

[0084] Figure 2 This is the N-BEATS model diagram of the present invention. DETAILED DESCRIPTION

[0085] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments:

[0086] The present invention proposes a method for predicting the fatigue life of wire ropes based on the true stress-true strain curve. First, the wire rope data is comprehensively collected and processed to establish a stress-strain model. Then, an N-BEATS network model is built that incorporates the model and is trained and optimized using the collected data. Multiple parameters of the wire rope are then monitored in real time to predict fatigue life. Finally, a full life cycle management platform is built to realize functions such as status monitoring, life prediction, and maintenance management, thereby improving efficiency and safety. The invention flowchart is shown below: Figure 1 shown.

[0087] Step 1: Data collection and processing, collect data of wire ropes of different types and specifications, including material properties, structural parameters, mechanical properties, etc. Establish a stress-strain model for wire ropes, taking into account the structural details of the wire ropes, such as strands, wires, and rope cores, as well as the contact relationship between the wire ropes and components such as pulleys and drums.

[0088] Step 1.1, first, for a single steel wire in the wire rope, describe the stress-strain relationship in its elastic stage according to Hooke's law:

[0089] σ=Eε

[0090] Among them, σ is stress, E is elastic modulus, and ε is strain. However, since the wire rope is a complex structure composed of multiple steel wires and a rope core, the interaction between strands and wires needs to be considered simultaneously.

[0091] The wire rope is composed of individual strands, each with a cross-sectional area of ​​A. i (i=1,2,...,n), n represents the number of strands in the wire rope. The elastic modulus is E i , the equivalent cross-sectional area A of the entire wire rope can be expressed as:

[0092]

[0093] Among them, A i represents the cross-sectional area of ​​the i-th strand. When subjected to axial tension F, the stress σ for each strand is i It can be expressed as:

[0094]

[0095] Among them, F i It is the tension on each share.

[0096] According to the force balance condition:

[0097]

[0098] The strain per share ε can be obtained i :

[0099]

[0100] Among them, E i Represents the elastic modulus of the i-th strand. The macroscopic strain ε of the entire wire rope can be obtained by weighted average, and the weight coefficient w i It is expressed as:

[0101]

[0102] Among them, j represents the serial number of the steel wire in each share, w i The weight coefficient of each share when the macroscopic strain of the wire rope is expressed, the macroscopic strain ε can be expressed as:

[0103]

[0104] Step 1.2, correct the contact stress. When the wire rope contacts the pulley, bending stress will be generated. The bending stress σ b It can be expressed as:

[0105]

[0106] M=TR

[0107] Where M is the bending moment when the wire rope contacts the pulley, T is the tension of the wire rope, R is the pulley radius, y is the distance to the neutral axis, and I is the cross-sectional inertia moment. When calculating the cross-sectional inertia moment I, the distribution of strands and wires is considered. Let the outer diameter of the wire rope be D and the inner diameter be d, and its cross-sectional inertia moment I can be approximated as:

[0108]

[0109] In the contact area, the total stress σ total :

[0110] σ total =σ+σ b

[0111] Where σ is the axial tensile stress. When the wire rope passes around the pulley, the friction force F f It can be expressed as:

[0112] F f =μT

[0113] Among them, μ is the friction coefficient. The friction will produce tangential stress inside the wire rope. The tangential stress σ t It is expressed as:

[0114]

[0115] The total stress at this time is expressed as σ total :

[0116] σ total =σ+σ b +σ t

[0117] Step 1.3, improve the model according to the test data, and obtain the true stress-true strain curve through the tensile testing machine. During the test, measure the elongation ΔL under different tensile forces F, and calculate the true stress σ t and strain ε t :

[0118]

[0119] Among them, A t is the actual cross-sectional area, L0 is the original length, and L is the length after deformation.

[0120] The data points obtained from the experiment (σ t ,ε t) is compared with the stress-strain relationship calculated by the theoretical model. The least squares method and other data fitting methods are used to correct the parameters in the theoretical model: elastic modulus E, friction coefficient μ, etc., so that the theoretical model can better fit the test data. After the fitting and correction process, the final wire rope stress-strain model is obtained:

[0121] σ total =Eε

[0122] Step 2: Establish a N-BEATS network model based on physical information, design the N-BEATS network structure for wire rope fatigue life prediction, and integrate the wire rope stress-strain model in step 1, select appropriate activation function and loss function to ensure that the model can effectively learn and predict. The N-BEATS model is shown in the figure Figure 2 shown.

[0123] Step 2.1: Establish the model input layer. After appropriate preprocessing of the stress-strain data measured by the sensor, the input vector is expressed as X = [x in1 ,x in2 ,...,x int ], where t represents the tth moment, x i Represents the stress value, strain value, wire diameter, number of wire strands, yield strength, lay length and elastic modulus calculated from the stress-strain model in step 1 at different times.

[0124] Step 2.2: Construct the block structure of the N-BEATS model. The input data first passes through a multi-layer fully connected neural network. Let the input of the block structure be x in , after L layers of fully connected neural network, the weight matrix of each layer is W l , with a bias of b l , the activation function is ReLU, then its output can be expressed as:

[0125] h l =ReLU(W l h l-1 +b l ),l=1,2,...,L

[0126] Among them, h0 is the input vector X of the input layer, h l It represents the output of the lth layer of fully connected neural network in the basic block after the activation function. After L layers of fully connected neural network, the output is divided into two parts, and each part passes through a fully connected layer to obtain the forward and backward prediction values: Assume that the two parts of the output after segmentation are h L1 and h L2 ,but:

[0127]

[0128] Among them, W forecast and b forecast Indicates the value used to generate the prediction The weight matrix and bias of the fully connected layer, W backcast and b backcast Indicates the value used to generate the prediction The weight matrix and bias of the fully connected layer.

[0129] Step 2.3, double residual stacking, in a stack, the input x of the i-th basic block i Backcast output with the previous basic block There is a residual connection between them, and its calculation formula is:

[0130]

[0131] Among them, x1 is the original input of stack, is the predicted value of the i-1th stack. The forecast output of the block is It is the sum of forecast outputs of all basic blocks in the stack:

[0132]

[0133] Where N represents the number of basic blocks in a stack. is the predicted value of the i-th stack.

[0134] Step 2.4, calculate trend block and seasonal block:

[0135] The trend block is obtained by removing the last fully connected layer from the basic block and multiplying the output with two matrices respectively. The output formula is:

[0136]

[0137] Among them, g trend is the activation function, which is used to transform the weighted results to obtain the final trend prediction value, V trend and U trend It represents the matrix multiplied by the output of the basic block in the trend block. x and y are the outputs of the basic block after removing the last fully connected layer, which are feature representations of different dimensions of the input time series data.

[0138] The seasonal block is after removing the last layer of the basic block, and the output is multiplied by two other matrices respectively. The formula is:

[0139]

[0140] Among them, g season is the activation function used to generate the final seasonal forecast value, V season and U season Represents the matrix in the seasonal block that is multiplied with the base block output.

[0141] Step 2.5, construct the model output layer: The N-BEATS model consists of multiple stacks. The final output of the model is the sum of the forecast outputs of all stacks. The final output formula is:

[0142]

[0143] in, is the final output of the model, i.e. the predicted value of fatigue life of the wire rope, and M is the number of all stacks.

[0144] Step 3: Model training and optimization. Use the collected true stress-true strain curve data to train the model and output the fatigue life prediction value of the wire rope. By adjusting the model parameters and optimizing the algorithm, the prediction accuracy and generalization ability of the model can be improved.

[0145] Use a tensile testing machine to perform stress-strain tests on the wire rope material to obtain the true stress-true strain curve. The true stress-true strain curve data obtained by the tensile testing machine is sorted and preprocessed according to the data format of the input layer in step 2 to obtain training data, and the network model is trained according to the following steps:

[0146] Initialize weights: Randomly initialize the weights and biases in the network.

[0147] Batch processing: Divide the training data into multiple batches, each batch contains a certain number of samples.

[0148] Forward propagation: Input sample data into the input layer and calculate the output value through the network.

[0149] Loss function calculation: The mean squared error loss function is used to calculate the difference between the predicted value and the true value.

[0150] Backpropagation: Update the weights and biases in the network based on the gradient of the loss function.

[0151] Parameter update: Use the Adam optimization algorithm to adjust the network parameters.

[0152] Step 4: Monitor the wire rope status in real time, and use the sensor network to monitor the wire rope's stress, strain, temperature, vibration and other parameters in real time.

[0153] Step 5: Fatigue life prediction. According to the real-time monitored wire rope status data and working condition information in step 4, the stress-strain parameters are input into the trained neural network model to predict the fatigue life of the wire rope.

[0154] Step 6: Build a wire rope health management system and establish a wire rope full life cycle management platform to realize functions such as wire rope status monitoring, life prediction, and maintenance management to improve the efficiency and safety of wire rope use.

[0155] Establish a wire rope full life cycle management platform based on cloud computing platform, adopting a layered architecture design, including data acquisition layer, data transmission layer, data storage layer, data analysis and processing layer, and application service layer.

[0156] Data acquisition layer: Various types of sensors are deployed at the wire rope operation site, such as high-precision stress sensors, strain sensors, temperature sensors, and vibration sensors. These sensors are connected to the data acquisition nodes via wired or wireless means to collect real-time parameter information such as stress value, strain data, temperature change, and vibration frequency and amplitude of the wire rope during operation.

[0157] Data transmission layer: The collected data is transmitted from the sensor node to the data center server using an industrial-grade wireless network. Data encryption technology is used to encrypt the transmitted data to prevent the data from being stolen or tampered with during transmission. At the same time, a data cache mechanism is set up. When the network is briefly interrupted, the sensor node can temporarily store the collected data in the local cache and automatically upload it after the network is restored to avoid data loss.

[0158] Data storage layer: Build a distributed database system to store massive amounts of wire rope monitoring data. The database uses a combination of relational and non-relational databases. The relational database is used to store structured data such as basic information of the wire rope, equipment information, and user information; the non-relational database is used to store a large amount of monitoring data to meet the needs of fast data reading and writing and high scalability. At the same time, the data is backed up regularly to prevent data loss, and a data archiving strategy is adopted to archive and store historical data for subsequent query and analysis.

[0159] Model-based life prediction: Integrate the physical information-based N-BEATS network model trained in step 2 into the health management system. When the system receives the real-time monitored stress-strain parameters of the wire rope and other related working condition information, the data is input into the N-BEATS network model. The model predicts the remaining fatigue life of the wire rope based on the learned stress-strain law and fatigue damage mechanism of the wire rope.

[0160] Evaluation and update of life prediction results: Regularly compare and evaluate the actual wire rope fatigue life data with the model prediction results. If it is found that there is a large deviation between the prediction results and the actual situation, the model will be retrained and optimized, and the model parameters will be adjusted to improve the prediction accuracy and adaptability of the model.

[0161] Maintenance plan formulation: Based on the condition monitoring results and life prediction information of the wire rope, the system automatically generates a maintenance plan. The maintenance plan includes maintenance tasks, maintenance schedule, maintenance personnel allocation, and a list of required maintenance tools and materials to ensure the safe operation of the equipment.

[0162] The above description is only a preferred embodiment of the present invention and does not constitute any other form of limitation to the present invention. Any modification or equivalent change made based on the technical essence of the present invention still falls within the scope of protection required by the present invention.

Claims

1. A method for predicting fatigue life of a steel wire rope based on true stress-true strain curve, the specific steps are as follows, and the characteristics are as follows: Step 1: Data collection and processing: collect data of wire ropes of different types and specifications, including material properties, structural parameters, mechanical properties, etc.; establish a stress-strain model for wire ropes, taking into account the structural details of the wire ropes, such as strands, wires, and rope cores, as well as the contact relationship between the wire ropes and components such as pulleys and drums; Step 2: Establish a N-BEATS network model based on physical information, design the N-BEATS network structure for wire rope fatigue life prediction, integrate the wire rope stress-strain model in step 1, and select appropriate activation functions and loss functions to ensure that the model can effectively learn and predict; Step 3: Model training and optimization: Use the collected true stress-true strain curve data to train the model and output the fatigue life prediction value of the wire rope. By adjusting the model parameters and optimizing the algorithm, the prediction accuracy and generalization ability of the model can be improved. Step 4: Real-time monitoring of the wire rope status, using a sensor network to monitor the wire rope's stress, strain, temperature, vibration and other parameters in real time; Step 5: Fatigue life prediction: according to the real-time monitored wire rope status data and working condition information in step 4, the stress-strain parameters are input into the trained neural network model to predict the fatigue life of the wire rope; Step 6: Build a wire rope health management system and establish a wire rope full life cycle management platform to realize functions such as wire rope status monitoring, life prediction, and maintenance management to improve the efficiency and safety of wire rope use.

2. The method for predicting fatigue life of a steel wire rope based on true stress-true strain curve according to claim 1 is characterized in that: The stress-strain model of the wire rope established in step 1 can be expressed as: Step 1.1, first, for a single steel wire in the wire rope, describe the stress-strain relationship in its elastic stage according to Hooke's law: σ=Eε Among them, σ is stress, E is elastic modulus, and ε is strain. However, since the wire rope is a complex structure composed of multiple strands of steel wires and a rope core, the interaction between strands and wires needs to be considered at the same time; The wire rope is composed of individual strands, each with a cross-sectional area of ​​A. i (i=1,2,…,n), n represents the number of strands in the wire rope; the elastic modulus is E i , the equivalent cross-sectional area A of the entire wire rope can be expressed as: Among them, A i represents the cross-sectional area of ​​the i-th strand. When subjected to axial tension F, the stress σ for each strand is i It can be expressed as: Among them, F i is the tension on each share; According to the force balance condition: The strain per share ε can be obtained i : Among them, E i Represents the elastic modulus of the i-th strand. The macroscopic strain ε of the entire wire rope can be obtained by weighted average, and the weight coefficient w i It is expressed as: Among them, j represents the serial number of the steel wire in each share, w i The weight coefficient of each share when the macroscopic strain of the wire rope is expressed, and the macroscopic strain ε can be expressed as: Step 1.2, correct the contact stress. When the wire rope contacts the pulley, bending stress will be generated. The bending stress σ b It can be expressed as: M=TR Where M is the bending moment when the wire rope contacts the pulley, T is the tension of the wire rope, R is the pulley radius, y is the distance to the neutral axis, and I is the cross-sectional inertia moment. When calculating the cross-sectional inertia moment I, the distribution of strands and wires is considered. Let the outer diameter of the wire rope be D and the inner diameter be d, and its cross-sectional inertia moment I can be approximated as: In the contact area, the total stress σ total : s total =s+s b Where σ is the axial tensile stress; when the wire rope passes around the pulley, the friction force F f It can be expressed as: F f =μT Where μ is the friction coefficient. Friction will generate tangential stress inside the wire rope. The tangential stress σ t It is expressed as: The total stress at this time is expressed as σ total : s total =s+s b +s t Step 1.3, improve the model according to the test data, and obtain the true stress-true strain curve through the tensile testing machine; during the test, measure the elongation ΔL under different tensile forces F, and calculate the true stress σ t and strain ε t : Among them, A t is the actual cross-sectional area, L0 is the original length, and L is the length after deformation; The data points obtained from the experiment (σ t ,ε t ) is compared with the stress-strain relationship calculated by the theoretical model; the least square method and other data fitting methods are used to correct the parameters in the theoretical model: elastic modulus E, friction coefficient μ, etc., so that the theoretical model can better fit the test data; after the fitting and correction process, the final wire rope stress-strain model is obtained: s total =Ee 3. The method for predicting fatigue life of a steel wire rope based on true stress-true strain curve according to claim 1, characterized in that: The process of establishing the N-BEATS network model based on physical information in step 2 can be expressed as follows: Step 2.1: Establish the model input layer. After appropriate preprocessing of the stress-strain data measured by the sensor, the input vector is expressed as X = [x in1 ,x in2 ,…,x int ], where t represents the tth moment, x i Represents the stress value, strain value, wire diameter, wire strand number, yield strength, lay length and elastic modulus calculated from the stress-strain model in step 1 at different times; Step 2.2: Construct the block structure of the N-BEATS model. The input data first passes through a multi-layer fully connected neural network. Let the input of the block structure be x in , after L layers of fully connected neural network, the weight matrix of each layer is W l , with a bias of b l , the activation function is ReLU, then its output can be expressed as: h l =ReLU(W l h l-1 +b l ),l=1,2,…,L Among them, h0 is the input vector X of the input layer, h l It represents the output of the lth layer of fully connected neural network in the basic block after the activation function; after passing through the Lth layer of fully connected neural network, the output is divided into two parts, and passes through a fully connected layer to obtain the forward and backward prediction values ​​respectively: Assume that the two parts of the output after segmentation are h L1 and h L2 ,but: Among them, W forecast and b forecast Indicates the value used to generate the prediction The weight matrix and bias of the fully connected layer, W backcast and b backcast Indicates the value used to generate the prediction The weight matrix and bias of the fully connected layer; Step 2.3, double residual stacking, in a stack, the input x of the i-th basic block i Backcast output with the previous basic block There is a residual connection between them, and its calculation formula is: Among them, x1 is the original input of stack, is the predicted value of the i-1th stack; and the forecast output of the block It is the sum of forecast outputs of all basic blocks in the stack: Where N represents the number of basic blocks in a stack. is the predicted value of the i-th stack; Step 2.4, calculate trend block and seasonal block: The trend block is obtained by removing the last fully connected layer from the basic block and multiplying the output with two matrices respectively. The output formula is: Among them, g trend is the activation function, which is used to transform the weighted results to obtain the final trend prediction value, V trend and U trend Represents the matrix in the trend block multiplied by the output of the basic block. x and y are the outputs of the basic block after removing the last fully connected layer, which are feature representations of different dimensions of the input time series data. The seasonal block is after removing the last layer of the basic block, and the output is multiplied by two other matrices respectively. The formula is: Among them, g season is the activation function used to generate the final seasonal forecast value, V season and U season Represents the matrix in the seasonal block that is multiplied with the base block output; Step 2.5, construct the model output layer: The N-BEATS model consists of multiple stacks. The final output of the model is the sum of the forecast outputs of all stacks. The final output formula is: in, is the final output of the model, i.e. the predicted value of fatigue life of the wire rope, and M is the number of all stacks.

4. The method for predicting fatigue life of a steel wire rope based on true stress-true strain curve according to claim 1, characterized in that: The construction of the wire rope health management system in step 6 can be expressed as follows: Establish a wire rope full life cycle management platform based on cloud computing platform, adopting a layered architecture design, including data collection layer, data transmission layer, data storage layer, data analysis and processing layer, and application service layer; Data acquisition layer: Various types of sensors are deployed at the wire rope operation site, such as high-precision stress sensors, strain sensors, temperature sensors, and vibration sensors. These sensors are connected to the data acquisition nodes via wired or wireless means to collect real-time information on the stress value, strain data, temperature changes, vibration frequency, amplitude, and other parameter information of the wire rope during operation. Data transmission layer: Use industrial-grade wireless networks to transmit collected data from sensor nodes to data center servers; use data encryption technology to encrypt the transmitted data to prevent the data from being stolen or tampered with during transmission; at the same time, set up a data cache mechanism. When the network is briefly interrupted, the sensor node can temporarily store the collected data in the local cache and automatically upload it after the network is restored to avoid data loss; Data storage layer: Build a distributed database system to store massive wire rope monitoring data; The database adopts a combination of relational database and non-relational database. The relational database is used to store structured data such as basic information of wire rope, equipment information and user information; Non-relational databases are used to store large amounts of monitoring data to meet the needs of fast data reading and writing and high scalability. At the same time, data is backed up regularly to prevent data loss, and a data archiving strategy is adopted to archive and store historical data for subsequent query and analysis. Model-based life prediction: Integrate the physical information-based N-BEATS network model trained in step 2 into the health management system; when the system receives the real-time monitored stress-strain parameters of the wire rope and other relevant working condition information, input these data into the N-BEATS network model, and the model predicts the remaining fatigue life of the wire rope based on the learned stress-strain law and fatigue damage mechanism of the wire rope; Evaluation and update of life prediction results: Regularly compare and evaluate the actual wire rope fatigue life data with the model prediction results; if it is found that there is a large deviation between the prediction results and the actual situation, retrain and optimize the model and adjust the model parameters to improve the prediction accuracy and adaptability of the model; Maintenance plan formulation: Based on the condition monitoring results and life prediction information of the wire rope, the system automatically generates a maintenance plan; the maintenance plan includes maintenance tasks, maintenance schedule, maintenance personnel allocation, and a list of required maintenance tools and materials to ensure the safe operation of the equipment.

Images