A method and apparatus for analyzing the axial force of bolts

Abstract

This invention provides a method and apparatus for analyzing the axial force of bolts, relating to the field of computer-aided measurement technology. The invention uses a finite element model of the bolt to screen for weak stress areas, analyzes the effective stress area length and longitudinal wave transit time, establishes a preload calibration expression and performs temperature compensation to achieve high-precision measurement of bolt preload. It establishes a relationship between the transverse and longitudinal wave transit time ratio and axial stress to obtain the bolt's axial force component, calculates the bolt axial force based on the preload, constructs a risk prediction model, designs a composite loss function incorporating the effective stress area length, optimizes model performance, and finally utilizes a long short-term memory neural network to perform time-series health diagnosis and intelligent early warning on long-term monitoring data. This invention combines finite element analysis, dual-wave transit time ratio measurement, and a long short-term memory neural network to achieve high-precision online monitoring, risk prediction, and intelligent health management of wind turbine bolt axial force.

Description

Technical Field

[0001] This invention relates to the field of computer-aided measurement technology, specifically to a method and apparatus for analyzing the axial force of a bolt. Background Technology

[0002] In the past, wind farms commonly used the traditional torque method for bolt pre-tightening. This method relies on the operator's visual observation and manual control, making it difficult to accurately grasp the pre-tightening force and posing a risk of missed inspections. Furthermore, differences in the amount of different lubricants applied can cause uneven axial force, resulting in the actual bolt axial force often deviating from the design target value. When the axial force is insufficient, high-strength bolts are prone to loosening or even breakage under long-term vibration and alternating loads, seriously threatening the structural safety of the wind turbine and potentially leading to major accidents such as blade detachment or turbine collapse. In addition, most wind farms currently rely on manual, periodic inspections to check the bolt condition, which has a long inspection cycle and makes it difficult to promptly detect changes in the effective stress area length of bolts due to axial force variations. Moreover, the traditional method for measuring axial force is the single-wave measurement method, which also has limitations, as it cannot measure the axial force component of bolts in service and is difficult to perform temperature compensation.

[0003] In existing technologies, traditional methods for analyzing bolt axial force mostly employ the torque method for measurement. However, this is difficult to monitor online during actual bolt operation. Furthermore, the effective stress-bearing area of ​​a bolt changes due to variations in axial force during operation, which is often overlooked and leads to inaccurate measurements. Moreover, most measurements do not consider the influence of temperature and lack temperature compensation measures. Secondly, existing technologies often use the single-wave method to measure bolt axial force. Since the single-wave method requires prior measurement of the ultrasonic waves in an unstressed state, it is limited when dealing with axial force testing of bolts in service. It cannot measure bolts currently in operation. Therefore, the preload force obtained by the single-wave measurement method is often used to replace the axial force of the bolt in service, but this ignores the axial force component generated during actual bolt operation, leading to measurement deviations.

[0004] Therefore, it is necessary to provide a method and apparatus for analyzing the axial force of bolts to solve the aforementioned problem.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a method and apparatus for analyzing the axial force of bolts, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A method for analyzing the axial force of a bolt, comprising the following steps:

[0009] Step 1: Obtain the relevant material and physical and mechanical parameters of the bolts used to equip the wind turbine, and establish a finite element model of the bolts based on the finite element analysis method. Perform axial force distribution simulation analysis on the bolts and screen out the weak stress areas of the bolts.

[0010] Step 2: Real-time acquisition of bolt acoustic time data, strain signal and ambient temperature at the weak stress area of ​​the bolt. Combining Hooke's law and acoustoelastic theory, the effective stress area length and longitudinal wave acoustic time difference before and after bolt loading are determined. A bolt preload calibration expression is established, and temperature compensation is applied to the bolt preload calibration expression to obtain the bolt preload.

[0011] Step 3: Based on the dual-wave measurement method, establish the relationship between the transverse and longitudinal dual-wave acoustic time ratio and the bolt axial stress. Introduce a temperature correction coefficient to compensate for the temperature of the transverse and longitudinal dual-wave acoustic time ratio. Calculate the axial component of the bolt during operation based on the temperature-compensated transverse and longitudinal dual-wave acoustic time ratio, and combine it with the calculated bolt preload to obtain the bolt axial force.

[0012] Step 4: Using historically calculated bolt axial force data as input and bolt risk assessment scores as output, establish a risk prediction model for bolt axial force, and construct a loss function related to the effective stress zone length of the bolt to optimize the risk prediction model for bolt axial force.

[0013] Step 5: Combining long-term historical monitoring data from the risk prediction model, apply a long short-term memory neural network to achieve time-series health diagnosis and intelligent early warning of bolt axial force, and link the model's prediction results with the digital operation and maintenance platform to achieve automatic early warning of abnormal bolt axial force, push maintenance suggestions, and full life cycle management.

[0014] Furthermore, a finite element model of the bolt was established, and the weak stress areas of the bolt were identified. The method used was as follows:

[0015] Collect relevant material and physical and mechanical parameters of the bolt, including: bolt material composition and grade, bolt density, elastic modulus, Poisson's ratio, Lamé elastic constant and fatigue performance parameters. The relevant material and physical and mechanical parameters of the bolt are obtained from experiments or supplier technical data.

[0016] Based on the specifications of actual wind turbine bolts, including bolt diameter, length, thread shape and clamping length, a three-dimensional geometric model of the bolt is created using CAD software, and constraints are applied to fix one end of the bolt to simulate the bolt installation clamping state, and apply a preload force equal to the axial stress of the bolt to the other end to simulate the actual preload condition of the bolt.

[0017] The three-dimensional geometric model of the bolt is divided into mesh elements, and an axial tensile force equal to the nominal preload of the bolt is applied to the top of the bolt. The nominal preload of the bolt is determined by multiplying the yield strength of the bolt material by a safety factor. The stress value in each mesh element of the three-dimensional geometric model is calculated, and 70% of the bolt yield strength is taken as the stress threshold. Mesh elements in the three-dimensional geometric model that exceed the stress threshold are counted. Spatial topology analysis is performed on the mesh elements that exceed the stress threshold using the adjacency method. Mesh elements with common edges or common nodes are grouped into the same category, forming several continuous stress concentration areas. Each continuous stress concentration area is defined as a weak stress region, and the number of meshes contained in it, its spatial location, and its maximum stress value are counted.

[0018] Furthermore, the effective stress zone length and longitudinal wave acoustic time difference before and after bolt loading were determined, and a calibration expression for the bolt preload was established. Temperature compensation was then applied to this calibration expression to obtain the bolt preload. The method used was as follows:

[0019] Based on Hooke's Law, the effective stress zone length of a bolt is analyzed. When the bolt has no axial stress, its effective stress zone, according to empirical formulas in mechanics of materials, is equal to the sum of the bolt clamping length and the bolt diameter. The formula is as follows:

[0020]

[0021] in, Indicates the effective stress area of ​​the bolt. This refers to the bolt clamping length. The diameter of the bolt;

[0022] When a bolt is under axial stress, the axial strain of the bolt under axial stress is obtained according to the generalized Hooke's law. Based on this axial strain, the increment of the bolt's effective length is calculated to obtain the overall effective stress-bearing region length of the bolt under axial stress. The formula used is as follows:

[0023]

[0024] in, This indicates the length of the effective stress-bearing area of ​​the bolt under load. It is the axial stress applied to the cross-sectional area of ​​the bolt, and It is known. The elastic modulus of the bolt;

[0025] Based on the acoustoelastic theory, the longitudinal wave time difference before and after bolt loading is analyzed. Since the axial stress on the bolt causes changes in ultrasonic wave velocity and bolt length, the ultrasonic wave velocity and bolt length before and after loading are obtained respectively to calculate the longitudinal wave time difference before and after bolt loading. The formula used is as follows:

[0026]

[0027]

[0028]

[0029] in, When representing the longitudinal wave sound before bolt loading, When representing the longitudinal wave sound after bolt loading, It is the length of the effective stress zone before bolt loading. It is the longitudinal wave velocity before bolt loading. It is the longitudinal wave velocity after bolt loading. It is the longitudinal wave acoustic time difference before and after bolt loading;

[0030] Based on the fundamental assumptions of the single-wave method combined with acoustoelastic theory, when the ultrasonic longitudinal wave propagates along the stress direction, the relationship between the longitudinal wave velocity and the stress is expressed as:

[0031]

[0032] in, This indicates the material density of the bolt. Let the wave velocity be the longitudinal wave. Let Lame be the elastic constant. For fatigue performance parameters, , All elastic constant, Let be the acoustic elastic calibration coefficient, and ;

[0033] When the stress is zero, that is The expression for the propagation speed of longitudinal waves in a solid medium is: Substituting this into the above equation relating longitudinal wave velocity and stress, we obtain the following simultaneous formula:

[0034]

[0035] The above equation is processed using the first-order Taylor approximation to obtain... and The relation is:

[0036]

[0037]

[0038] in, It is the ultrasonic acoustoelastic coefficient, which is the ultrasonic velocity of a material under stress-free conditions.

[0039] Obtain the cross-sectional area of ​​the bolt and calculate the bolt preload using empirical formulas from mechanics of materials. The formulas used are:

[0040]

[0041] in, Indicates the preload of the bolt. This represents the cross-sectional area of ​​the bolt.

[0042] The expression for the longitudinal wave acoustic time difference before and after bolt loading is compared with... and The relationships are combined, and the expression for the longitudinal wave acoustic time difference before and after bolt loading is further written as:

[0043]

[0044] because The order of magnitude is ,and The order of magnitude is ,therefore Therefore, the above formula can be simplified to:

[0045]

[0046] Based on the simplified expression for the longitudinal wave acoustic time difference before and after bolt loading, the expression for bolt preload is substituted into it to obtain the bolt preload calibration expression related to the longitudinal wave acoustic time difference:

[0047]

[0048]

[0049] in, The calibration factor representing the bolt preload;

[0050] To correct for the influence of temperature, based on the temperature compensation method, the ultrasonic transit time caused by temperature is substituted into the bolt preload calibration expression. The resulting formula for calculating the bolt preload after temperature compensation is as follows:

[0051]

[0052] in, This indicates the time difference of ultrasonic waves caused by temperature. This is the bolt preload after temperature compensation.

[0053] Furthermore, a relationship between the transverse and longitudinal double-wave acoustic time ratio and the bolt axial stress is established, and temperature compensation is applied to the transverse and longitudinal double-wave acoustic time ratio. Based on the temperature-compensated transverse and longitudinal double-wave acoustic time ratio, the axial component of the bolt during operation is calculated, and the bolt axial force is calculated in conjunction with the obtained bolt preload. The method used is as follows:

[0054] Based on the acoustoelastic coefficients of the bolt's transverse and longitudinal waves obtained using the dual-wave detection method, the acoustic time of the transverse wave and the acoustic time of the longitudinal wave are calculated separately when the bolt is under non-zero stress. The ratio of the transverse wave acoustic time to the longitudinal wave acoustic time is the transverse-longitudinal dual-wave acoustic time ratio, based on the following formula:

[0055]

[0056]

[0057]

[0058] in, , These represent the acoustic times of the longitudinal and transverse waves of the bolt under non-zero stress conditions, respectively. , These represent the acoustic times of the longitudinal and transverse waves of the bolt under stress-free conditions, respectively. The transverse and longitudinal dual-wave acoustic time ratio under non-zero stress conditions. , These are the acoustic elastic coefficients of the longitudinal and transverse waves, respectively. This represents the axial component of the working load acting on the bolt cross-section;

[0059] The acoustic time of the transverse wave and the acoustic time of the longitudinal wave of the bolt under stress-free conditions are obtained using the same method described above, so as to obtain the acoustic time ratio of the bolt under stress-free conditions. Furthermore, based on the obtained acoustoelastic coefficients of the bolt's longitudinal and transverse waves, the difference between the two is the bolt's acoustoelastic anisotropy coefficient. ;

[0060] Based on the analysis of the effective region length of the bolt under axial stress and the effective region length without axial stress, the ratio of these two lengths is calculated and denoted as the effective region length deformation ratio of the bolt. The axial force sensitivity coefficient of the bolt is then calculated comprehensively based on the acoustic time ratio of the bolt under stress-free conditions, the acoustic elastic anisotropy coefficient of the bolt, the effective region length deformation ratio of the bolt, and the cross-sectional area of ​​the bolt thread. The formula used is as follows:

[0061]

[0062]

[0063] in, This represents the sensitivity coefficient of the bolt's axial force component. The effective area length deformation ratio of the bolt;

[0064] Based on the bolt's axial force sensitivity coefficient, bolt preload, and the acoustic time ratio under stress-free conditions, a relationship is established between the transverse and longitudinal dual-wave acoustic time ratio under non-zero stress conditions and the axial force of the bolt under test. The formula used is as follows:

[0065]

[0066] in, This represents the axial component of the force on the bolt to be measured.

[0067] Temperature compensation is applied to the relationship between the transverse and longitudinal dual-wave acoustic time ratio and the axial component force of the bolt under test under non-zero stress conditions. Based on the temperature compensation, the calibration relationship between the transverse and longitudinal dual-wave acoustic time ratio and the axial component force of the bolt under test under non-zero stress conditions is obtained:

[0068]

[0069] in, This is the temperature correction factor. This refers to the change in temperature.

[0070] Based on the calibration relationship between the transverse and longitudinal dual-wave acoustic time ratio under non-zero stress and the axial component of the bolt under test, the axial component of the bolt under test, i.e., the axial component of the bolt during operation, is obtained. Combined with the obtained temperature-compensated preload of the bolt, the axial force of the bolt is calculated using the following formula:

[0071]

[0072] in, This indicates the axial force of the bolt.

[0073] Furthermore, a risk prediction model for bolt axial force is established, and a loss function related to the effective stress-bearing region length of the bolt is constructed to optimize the risk prediction model for bolt axial force. The method used is as follows:

[0074] A risk prediction model for bolt axial force is constructed based on a neural network. The model structure includes an input layer, a hidden layer, and an output layer. The input layer receives bolt axial force data calculated from historical bolt data. By statistically analyzing the axial force measurement results of bolts under different service conditions, representative feature values ​​are extracted. An expert scoring method is introduced, and experts in relevant fields are organized to comprehensively evaluate and score the historical axial force data to form an expert score label for bolt axial force, which is output by the output layer as the output target of the model training. The hidden layer is responsible for nonlinear feature extraction and transformation of the input historical bolt axial force data.

[0075] The loss function consists of a traditional error term and a physical consistency error term. The traditional error term is the mean square error between the historical calculated axial force data sequence and the actual axial force data sequence. The physical consistency error term is the mean square error between the calculated value and the actual value of the effective stress zone length of the bolt. The true axial force value and the true length value of the effective stress area of ​​the bolt sample are used to calculate the traditional error term and the physical consistency error term. A linear weighted summation method is then used to assign weights to each term, which together form the loss function. The formula upon which the loss function is based is:

[0076]

[0077] in, This represents the construction of a loss function, the value of which is used to measure the overall error between the model's predictions and the actual results. This is the weighting coefficient for the mean square error of axial force. This is the weighting coefficient for the mean square error of the effective force-bearing length, and , This is the total number of bolt samples. Indicates the first The true axial force value of a sample bolt. Indicates the first The axial force prediction values ​​for each bolt sample are obtained from the model prediction. Indicates the first The actual length of the effective stress zone for each bolt sample Indicates the first The effective stress zone length of each bolt sample is predicted by the model. For the index of the bolt sample, and ;

[0078] The risk prediction model for bolt axial force is optimized based on the constructed loss function. The gradient of the loss function relative to the network output is calculated, the error signal is propagated in the back, and the gradient of the loss function with respect to the weights and biases of each layer of the network is calculated using the chain rule. The gradient of the physical consistency error term is propagated through the mapping link from the predicted axial force to the effective force-bearing area length to ensure the influence of physical constraints on parameter updates.

[0079] The Adam optimizer is used to adjust the network parameters based on gradient information. The forward propagation, loss calculation, back propagation and parameter update steps are repeated until the preset maximum number of iterations is reached.

[0080] Furthermore, by combining long-term historical monitoring data from the bolt axial force risk prediction model, a long short-term memory neural network is applied to achieve temporal health diagnosis and intelligent early warning of bolt axial force. The method used is as follows:

[0081] Long-term historical monitoring data of bolts are collected, including time-series bolt axial force calculation and measurement values ​​and related working condition characteristics. The data is cleaned and normalized to remove outliers. A fixed-length time window sequence is constructed to form a sample set. The health status of bolts is standardized based on empirical labels as a monitoring signal. The health status includes normal, warning and abnormal.

[0082] Design a multi-layer LSTM network, with input being a time step of... The LSTM network effectively captures long-term dependencies in the time series by using a gating mechanism to model the temporal state of the bolt axial force and related feature sequences. The LSTM network outputs the predicted axial force value and the probability of healthy state at future moments, and combines the threshold judgment to identify abnormal states and achieve health diagnosis.

[0083] The prediction results are uploaded to the digital operation and maintenance platform, triggering anomaly warning notifications. The platform automatically generates maintenance suggestions based on the warning results and pushes them to operation and maintenance personnel to achieve full lifecycle management and dynamic risk control of bolt axial force.

[0084] The present invention also provides a bolt axial force analysis device, the analysis device being used to perform the above-described bolt axial force analysis method, comprising:

[0085] The finite element stress analysis module is used to obtain the relevant material and physical and mechanical parameters of the bolts used to equip wind turbines, and to establish a finite element model of the bolts based on the finite element analysis method, to perform axial force distribution simulation analysis on the bolts, and to screen out the weak stress areas of the bolts.

[0086] The sensing and calibration module is used to collect the bolt's acoustic time data, strain signal and ambient temperature in real time at the bolt's weak stress area. Combining Hooke's law and acoustoelastic theory, it analyzes and determines the effective stress area length and longitudinal wave acoustic time difference before and after bolt loading, establishes the bolt preload calibration expression, and performs temperature compensation on the bolt preload calibration expression to obtain the bolt preload.

[0087] The dual-wave acoustic time compensation module establishes the relationship between the transverse and longitudinal dual-wave acoustic time ratio and the bolt axial stress based on the dual-wave measurement method. It introduces a temperature correction coefficient to perform temperature compensation on the transverse and longitudinal dual-wave acoustic time ratio. Based on the temperature-compensated transverse and longitudinal dual-wave acoustic time ratio, it calculates the axial component force of the bolt during operation and combines it with the obtained bolt preload to calculate the bolt axial force.

[0088] The risk prediction modeling module is used to establish a risk prediction model for bolt axial force by taking historically calculated bolt axial force data as input and the bolt risk assessment score as output, and to construct a loss function related to the effective stress zone length of the bolt to optimize the risk prediction model for bolt axial force.

[0089] The health diagnosis and early warning module is used to combine long-term historical monitoring data of the bolt axial force risk prediction model and apply a long short-term memory neural network to realize the time-series health diagnosis and intelligent early warning of bolt axial force. It also links the model's prediction results with the digital operation and maintenance platform to realize automatic early warning of bolt axial force anomalies, push maintenance suggestions, and full life cycle management.

[0090] Compared with the prior art, the beneficial effects of the present invention are:

[0091] This invention effectively solves the problem of online monitoring that is difficult to achieve with the traditional torque method by using finite element analysis to screen weak areas of bolt stress and combining it with real-time monitoring technology using multiple sensors such as ultrasonic sensors and strain gauges. At the same time, it accurately considers the change in the length of the effective stress area caused by bolt axial force, avoiding measurement errors. In addition, it innovatively introduces a dual-wavelength ratio and temperature compensation method. By combining the single-wavelength measurement method with the dual-wavelength measurement method, it overcomes the limitations of the single-wavelength method in the axial force detection of bolts in service, and improves the accuracy and applicability of axial force measurement.

[0092] This invention further constructs a risk prediction model based on neural networks and combines it with expert scoring. The model performance is optimized through a physical consistency loss function to achieve accurate risk assessment of bolt axial force. At the same time, a long short-term memory neural network is used for time-series health diagnosis and intelligent early warning, which enhances the safety of wind turbine bolt operation. Attached Figure Description

[0093] Figure 1 This is a schematic diagram of the overall method flow of the present invention.

[0094] Figure 2 This is a schematic diagram illustrating the analysis of the effective stress area length before and after bolt loading in this invention.

[0095] Figure 3 This is a schematic diagram of the system module flow of the present invention. Detailed Implementation

[0096] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0097] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0098] Example:

[0099] Please see Figure 1 A method for analyzing the axial force of a bolt, the specific steps of which include:

[0100] Step 1: Obtain the relevant material and physical and mechanical parameters of the bolts used to equip the wind turbine, and establish a finite element model of the bolts based on the finite element analysis method. Perform axial force distribution simulation analysis on the bolts and screen out the weak stress areas of the bolts.

[0101] In a specific embodiment of this invention, by obtaining the material and mechanical parameters of the bolt and combining them with finite element analysis to establish an accurate bolt model, we can deeply simulate the axial force distribution characteristics of the bolt under actual loads and accurately identify the weak stress areas. The specific steps for constructing the finite element model of the bolt are as follows: Create a new project using finite element analysis software, enter the modeling module, and draw the two-dimensional cross-sectional outline of the bolt using the software's sketching tools according to the actual size parameters of the bolt. Through modeling commands such as extrusion, rotation, and sweep, the two-dimensional sketch is converted into a three-dimensional geometric model. For example, extrusion generates the bolt shank, rotation generates the head, and, as needed, the software's thread function or custom thread parameters can be used to add a thread structure to the bolt to enhance the realism of the model. Check whether the dimensions and structure of the model are consistent with the actual bolt. After confirming that there are no errors, save the geometric model of the bolt and import the geometric model into the finite element analysis module to prepare for mesh generation and mechanical analysis, finally obtaining the finite element model of the bolt.

[0102] Furthermore, a finite element model of the bolt was established, and the weak stress areas of the bolt were identified. The method used was as follows:

[0103] Collect relevant material and physical and mechanical parameters of the bolt, including: bolt material composition and grade, bolt density, elastic modulus, Poisson's ratio, Lamé elastic constant and fatigue performance parameters. The relevant material and physical and mechanical parameters of the bolt are obtained from experiments or supplier technical data.

[0104] Based on the specifications of actual wind turbine bolts, including bolt diameter, length, thread shape and clamping length, a three-dimensional geometric model of the bolt is created using CAD software, and constraints are applied to fix one end of the bolt to simulate the bolt installation clamping state, and apply a preload force equal to the axial stress of the bolt to the other end to simulate the actual preload condition of the bolt.

[0105] The three-dimensional geometric model of the bolt is divided into mesh elements, and an axial tensile force equal to the nominal preload of the bolt is applied to the top of the bolt. The nominal preload of the bolt is determined by multiplying the yield strength of the bolt material by a safety factor. The stress value in each mesh element of the three-dimensional geometric model is calculated, and 70% of the bolt yield strength is taken as the stress threshold. Mesh elements in the three-dimensional geometric model that exceed the stress threshold are counted. Spatial topology analysis is performed on the mesh elements that exceed the stress threshold using the adjacency method. Mesh elements with common edges or common nodes are grouped into the same category, forming several continuous stress concentration areas. Each continuous stress concentration area is defined as a weak stress region, and the number of meshes contained in it, its spatial location, and its maximum stress value are counted.

[0106] Step 2: Real-time acquisition of bolt acoustic time data, strain signal and ambient temperature at the weak stress area of ​​the bolt. Combining Hooke's law and acoustic elasticity theory, the effective stress area length and longitudinal wave acoustic time difference before and after bolt loading are determined. A bolt preload calibration expression is established, and temperature compensation is applied to the bolt preload calibration expression to obtain the bolt preload.

[0107] In a specific embodiment of this invention, by real-time acquisition of key physical signals in the weak stress area of ​​the bolt, combined with Hooke's law and acoustoelasticity theory, the state changes of the bolt before and after loading are accurately analyzed. This allows for the determination of the effective stress area length and longitudinal wave transit time, establishing a scientifically reasonable preload calibration expression. Compared to traditional methods that often treat the effective stress area length of the bolt as a fixed value, we consider that the effective stress area length of the bolt will change due to stress variations during actual operation. Therefore, based on Hooke's law and the axial stress state of the bolt, we dynamically calculate the change in the effective stress area length before and after loading, achieving a more realistic physical model. Furthermore, we incorporate the ultrasonic transit time caused by temperature into the preload calculation, proposing a temperature-compensated bolt preload expression. This significantly improves the robustness of the measurement results to changes in ambient temperature and solves the error problem caused by temperature disturbances in ultrasonic measurements.

[0108] Furthermore, the effective stress zone length and longitudinal wave acoustic time difference before and after bolt loading were determined, and a calibration expression for the bolt preload was established. Temperature compensation was then applied to this calibration expression to obtain the bolt preload. The method used was as follows:

[0109] Based on Hooke's Law, the effective stress zone length of a bolt is analyzed. When the bolt has no axial stress, its effective stress zone, according to empirical formulas in mechanics of materials, is equal to the sum of the bolt clamping length and the bolt diameter. The formula is as follows:

[0110]

[0111] in, Indicates the effective stress area of ​​the bolt. This refers to the bolt clamping length. The diameter of the bolt;

[0112] When a bolt is under axial stress, the axial strain of the bolt under axial stress is obtained according to the generalized Hooke's law. Based on this axial strain, the increment of the bolt's effective length is calculated to obtain the overall effective stress-bearing region length of the bolt under axial stress. The formula used is as follows:

[0113]

[0114] in, This indicates the length of the effective stress-bearing area of ​​the bolt under load. It is the axial stress applied to the cross-sectional area of ​​the bolt, and It is known. The elastic modulus of the bolt;

[0115] Based on the acoustoelastic theory, the longitudinal wave time difference before and after bolt loading is analyzed. Since the axial stress on the bolt causes changes in ultrasonic wave velocity and bolt length, the ultrasonic wave velocity and bolt length before and after loading are obtained respectively to calculate the longitudinal wave time difference before and after bolt loading. The formula used is as follows:

[0116]

[0117]

[0118]

[0119] in, When representing the longitudinal wave sound before bolt loading, When representing the longitudinal wave sound after bolt loading, It is the length of the effective stress zone before bolt loading. It is the longitudinal wave velocity before bolt loading. It is the longitudinal wave velocity after bolt loading. It is the longitudinal wave acoustic time difference before and after bolt loading;

[0120] Based on the fundamental assumptions of the single-wave method combined with acoustoelastic theory, when the ultrasonic longitudinal wave propagates along the stress direction, the relationship between the longitudinal wave velocity and the stress is expressed as:

[0121]

[0122] in, This indicates the material density of the bolt. Let the wave velocity be the longitudinal wave. Let Lame be the elastic constant. For fatigue performance parameters, , All elastic constant, Let be the acoustic elastic calibration coefficient, and ;

[0123] When the stress is zero, that is The expression for the propagation speed of longitudinal waves in a solid medium is: Substituting this into the above equation relating longitudinal wave velocity and stress, we obtain the following simultaneous formula:

[0124]

[0125] The above equation is processed using the first-order Taylor approximation to obtain... and The relation is:

[0126]

[0127]

[0128] in, It is the ultrasonic acoustoelastic coefficient, which is the ultrasonic velocity of a material under stress-free conditions.

[0129] Obtain the cross-sectional area of ​​the bolt and calculate the bolt preload using empirical formulas from mechanics of materials. The formulas used are:

[0130]

[0131] in, Indicates the preload of the bolt. This represents the cross-sectional area of ​​the bolt.

[0132] The expression for the longitudinal wave acoustic time difference before and after bolt loading is compared with... and The relationships are combined, and the expression for the longitudinal wave acoustic time difference before and after bolt loading is further written as:

[0133]

[0134] because The order of magnitude is ,and The order of magnitude is ,therefore Therefore, the above formula can be simplified to:

[0135]

[0136] Based on the simplified expression for the longitudinal wave acoustic time difference before and after bolt loading, the expression for bolt preload is substituted into it to obtain the bolt preload calibration expression related to the longitudinal wave acoustic time difference:

[0137]

[0138]

[0139] in, The calibration factor representing the bolt preload;

[0140] To correct for the influence of temperature, based on the temperature compensation method, the ultrasonic transit time caused by temperature is substituted into the bolt preload calibration expression. The resulting formula for calculating the bolt preload after temperature compensation is as follows:

[0141]

[0142] in, This indicates the time difference of ultrasonic waves caused by temperature. This is the bolt preload after temperature compensation.

[0143] Step 3: Establish the relationship between the transverse and longitudinal dual-wave acoustic time ratio and the bolt axial stress based on the dual-wave measurement method. Introduce a temperature correction coefficient to compensate for the temperature of the transverse and longitudinal dual-wave acoustic time ratio. Calculate the axial component of the bolt during operation based on the temperature-compensated transverse and longitudinal dual-wave acoustic time ratio, and combine it with the calculated bolt preload to obtain the bolt axial force.

[0144] In a specific embodiment of the present invention, we use a dual-wave measurement method to obtain the axial component force of the bolt under service conditions. The advantage of the dual-wave measurement method is that it can simultaneously obtain the acoustic time ratio of the transverse wave and the longitudinal wave, which can make full use of the different propagation characteristics of the two ultrasonic waves in the material and the difference in sensitivity to force, improve the detection accuracy and reliability of the bolt axial stress, overcome the limitations of the single-wave measurement method under complex working conditions, and also introduce temperature compensation here to effectively eliminate the influence of ambient temperature changes on the ultrasonic wave propagation speed and acoustic time ratio.

[0145] It should be noted that the temperature correction factor The time-to-sound ratio (TST) is usually obtained through experimental calibration. The specific steps are as follows: Under a controlled temperature environment, a known axial stress is applied to the bolt, and the transverse and longitudinal dual-wave acoustic time-to-sound data at different temperatures are collected simultaneously. The TST is then measured at multiple different temperature points. In terms of temperature change The independent variable is , and the relative change in the transverse and longitudinal acoustic time-to-sound ratio is the dependent variable. A linear or nonlinear regression method is used to fit the functional relationship between the acoustic time-to-sound ratio and temperature. The temperature sensitivity parameter, i.e., the temperature correction coefficient, is extracted from the slope of the fitted curve or related parameters. .

[0146] Furthermore, a relationship between the transverse and longitudinal double-wave acoustic time ratio and the bolt axial stress is established, and temperature compensation is applied to the transverse and longitudinal double-wave acoustic time ratio. Based on the temperature-compensated transverse and longitudinal double-wave acoustic time ratio, the axial component of the bolt during operation is calculated, and the bolt axial force is calculated in conjunction with the obtained bolt preload. The method used is as follows:

[0147] Based on the acoustoelastic coefficients of the bolt's transverse and longitudinal waves obtained using the dual-wave detection method, the acoustic time of the transverse wave and the acoustic time of the longitudinal wave are calculated separately when the bolt is under non-zero stress. The ratio of the transverse wave acoustic time to the longitudinal wave acoustic time is the transverse-longitudinal dual-wave acoustic time ratio, based on the following formula:

[0148]

[0149]

[0150]

[0151] in, , These represent the acoustic times of the longitudinal and transverse waves of the bolt under non-zero stress conditions, respectively. , These represent the acoustic times of the longitudinal and transverse waves of the bolt under stress-free conditions, respectively. The transverse and longitudinal dual-wave acoustic time ratio under non-zero stress conditions. , These are the acoustic elastic coefficients of the longitudinal and transverse waves, respectively. This represents the axial component of the working load acting on the bolt cross-section;

[0152] The acoustic time of the transverse wave and the acoustic time of the longitudinal wave of the bolt under stress-free conditions are obtained using the same method described above, so as to obtain the acoustic time ratio of the bolt under stress-free conditions. Furthermore, based on the obtained acoustoelastic coefficients of the bolt's longitudinal and transverse waves, the difference between the two is the bolt's acoustoelastic anisotropy coefficient. ;

[0153] Based on the analysis of the effective region length of the bolt under axial stress and the effective region length without axial stress, the ratio of these two lengths is calculated and denoted as the effective region length deformation ratio of the bolt. The axial force sensitivity coefficient of the bolt is then calculated comprehensively based on the acoustic time ratio of the bolt under stress-free conditions, the acoustic elastic anisotropy coefficient of the bolt, the effective region length deformation ratio of the bolt, and the cross-sectional area of ​​the bolt thread. The formula used is as follows:

[0154]

[0155]

[0156] in, This represents the sensitivity coefficient of the bolt's axial force component. The effective area length deformation ratio of the bolt;

[0157] Based on the bolt's axial force sensitivity coefficient, bolt preload, and the acoustic time ratio under stress-free conditions, a relationship is established between the transverse and longitudinal dual-wave acoustic time ratio under non-zero stress conditions and the axial force of the bolt under test. The formula used is as follows:

[0158]

[0159] in, This represents the axial component of the force on the bolt to be measured.

[0160] Temperature compensation is applied to the relationship between the transverse and longitudinal dual-wave acoustic time ratio and the axial component force of the bolt under test under non-zero stress conditions. Based on the temperature compensation, the calibration relationship between the transverse and longitudinal dual-wave acoustic time ratio and the axial component force of the bolt under test under non-zero stress conditions is obtained:

[0161]

[0162] in, This is the temperature correction factor. This refers to the change in temperature.

[0163] Based on the calibration relationship between the transverse and longitudinal dual-wave acoustic time ratio under non-zero stress and the axial component of the bolt under test, the axial component of the bolt under test, i.e., the axial component of the bolt during operation, is obtained. Combined with the obtained temperature-compensated preload of the bolt, the axial force of the bolt is calculated using the following formula:

[0164]

[0165] in, This indicates the axial force of the bolt.

[0166] Step 4: Using historically calculated bolt axial force data as input and bolt risk assessment score as output, establish a risk prediction model for bolt axial force, and construct a loss function related to the effective stress zone length of the bolt to optimize the risk prediction model for bolt axial force.

[0167] In a specific embodiment of the present invention, bolts are an important component of the connection structure in key equipment such as wind turbines. Their axial force state is directly related to the safety of the overall structure. By deeply mining and analyzing historical axial force data and using a neural network model to build a risk prediction system, it is possible to accurately assess the current and future working state of bolts.

[0168] It should be noted that this prediction model not only introduces expert scoring and combines domain expertise to effectively label historical data, but also innovatively integrates traditional error terms and physical consistency error terms into the loss function. This ensures that the model can predict bolt axial force while taking into account the physical constraints of the effective stress-bearing area length of the bolt, thereby enhancing the model's physical rationality and generalization ability.

[0169] Furthermore, a risk prediction model for bolt axial force is established, and a loss function related to the effective stress-bearing region length of the bolt is constructed to optimize the risk prediction model for bolt axial force. The method used is as follows:

[0170] A risk prediction model for bolt axial force is constructed based on a neural network. The model structure includes an input layer, a hidden layer, and an output layer. The input layer receives bolt axial force data calculated from historical bolt data. By statistically analyzing the axial force measurement results of bolts under different service conditions, representative feature values ​​are extracted. An expert scoring method is introduced, and experts in relevant fields are organized to comprehensively evaluate and score the historical axial force data to form an expert score label for bolt axial force, which is output by the output layer as the output target of the model training. The hidden layer is responsible for nonlinear feature extraction and transformation of the input historical bolt axial force data.

[0171] The loss function consists of a traditional error term and a physical consistency error term. The traditional error term is the mean square error between the historical calculated axial force data sequence and the actual axial force data sequence. The physical consistency error term is the mean square error between the calculated value and the actual value of the effective stress zone length of the bolt. The true axial force value and the true length value of the effective stress area of ​​the bolt sample are used to calculate the traditional error term and the physical consistency error term. A linear weighted summation method is then used to assign weights to each term, which together form the loss function. The formula upon which the loss function is based is:

[0172]

[0173] in, This represents the construction of a loss function, the value of which is used to measure the overall error between the model's predictions and the actual results. This is the weighting coefficient for the mean square error of axial force. This is the weighting coefficient for the mean square error of the effective force-bearing length, and , This is the total number of bolt samples. Indicates the first The true axial force value of a sample bolt. Indicates the first The axial force prediction values ​​for each bolt sample are obtained from the model prediction. Indicates the first The actual length of the effective stress zone for each bolt sample Indicates the first The effective stress zone length of each bolt sample is predicted by the model. For the index of the bolt sample, and ;

[0174] The risk prediction model for bolt axial force is optimized based on the constructed loss function. The gradient of the loss function relative to the network output is calculated, the error signal is propagated in the back, and the gradient of the loss function with respect to the weights and biases of each layer of the network is calculated using the chain rule. The gradient of the physical consistency error term is propagated through the mapping link from the predicted axial force to the effective force-bearing area length to ensure the influence of physical constraints on parameter updates.

[0175] The Adam optimizer is used to adjust the network parameters based on gradient information. The forward propagation, loss calculation, back propagation and parameter update steps are repeated until the preset maximum number of iterations is reached.

[0176] Step 5: Combining long-term historical monitoring data from the risk prediction model, apply a long short-term memory neural network to achieve time-series health diagnosis and intelligent early warning of bolt axial force, and link the model's prediction results with the digital operation and maintenance platform to achieve automatic early warning of abnormal bolt axial force, push maintenance suggestions, and full life cycle management.

[0177] In a specific embodiment of this invention, by combining long-term historical monitoring data from a risk prediction model, a Long Short-Term Memory (LSTM) neural network is applied to perform time-series health diagnosis and intelligent early warning of bolt axial force. This achieves dynamic, continuous monitoring and accurate prediction of bolt status. The use of LSTM for bolt axial force monitoring is primarily due to its excellent time-series modeling capabilities, which effectively capture the dynamic trends and long-term dependence characteristics of bolt axial force over time. Compared to traditional neural networks, LSTM overcomes the long-term dependence problem, accurately memorizes and utilizes key historical information, and adapts to the complex and nonlinear changes in bolt axial force in actual working conditions.

[0178] Furthermore, by combining long-term historical monitoring data from the bolt axial force risk prediction model, a long short-term memory neural network is applied to achieve temporal health diagnosis and intelligent early warning of bolt axial force. The method used is as follows:

[0179] Long-term historical monitoring data of bolts are collected, including time-series bolt axial force calculation and measurement values ​​and related working condition characteristics. The data is cleaned and normalized to remove outliers. A fixed-length time window sequence is constructed to form a sample set. The health status of bolts is standardized based on empirical labels as a monitoring signal. The health status includes normal, warning and abnormal.

[0180] Design a multi-layer LSTM network, with input being a time step of... The LSTM network effectively captures long-term dependencies in the time series by using a gating mechanism to model the temporal state of the bolt axial force and related feature sequences. The LSTM network outputs the predicted axial force value and the probability of healthy state at future moments, and combines the threshold judgment to identify abnormal states and achieve health diagnosis.

[0181] The prediction results are uploaded to the digital operation and maintenance platform, triggering anomaly warning notifications. The platform automatically generates maintenance suggestions based on the warning results and pushes them to operation and maintenance personnel to achieve full lifecycle management and dynamic risk control of bolt axial force.

[0182] Please see Figure 2 As can be seen from the schematic diagram analyzing the effective stress zone length before and after bolt loading, This indicates the length of the effective stress zone of the bolt in the unloaded state. It is the effective stress-bearing area length of the bolt, defined as the sum of the clamping length and the bolt diameter. It is the bolt clamping length, that is, the distance between two or more bolted components when they are clamped together. Where is the diameter of the bolt. This refers to the length deformation of the bolt due to axial stress. This is the total length after the bolt is loaded. The effective stress zone length of the bolt under load, i.e., the effective length after considering strain, is the sensor position of the probe device on the bolt head used for ultrasonic transmission and reception.

[0183] By measuring the change in the effective stress region length and the longitudinal wave acoustic time difference before and after bolt loading, and combining the mechanical and acoustic performance parameters of the bolt material, a calibration expression for bolt preload was established. This expression considers both the length change caused by the axial strain of the bolt and the influence of stress on the stress wave velocity, realizing a direct correlation between ultrasonic measurement and the bolt stress state. The advantage of this approach is that the preload of the bolt under test can be directly obtained based on the established calibration expression for bolt preload, thus providing assistance for obtaining the bolt axial force later.

[0184] Please see Figure 3 The present invention also provides a bolt axial force analysis device, the analysis device being used to perform the above-described bolt axial force analysis method, comprising:

[0185] The finite element stress analysis module is used to obtain the relevant material and physical and mechanical parameters of the bolts used to equip wind turbines, and to establish a finite element model of the bolts based on the finite element analysis method, to perform axial force distribution simulation analysis on the bolts, and to screen out the weak stress areas of the bolts.

[0186] The sensing and calibration module is used to collect the bolt's acoustic time data, strain signal and ambient temperature in real time at the bolt's weak stress area. Combining Hooke's law and acoustoelastic theory, it analyzes and determines the effective stress area length and longitudinal wave acoustic time difference before and after bolt loading, establishes the bolt preload calibration expression, and performs temperature compensation on the bolt preload calibration expression to obtain the bolt preload.

[0187] The dual-wave acoustic time compensation module establishes the relationship between the transverse and longitudinal dual-wave acoustic time ratio and the bolt axial stress based on the dual-wave measurement method. It introduces a temperature correction coefficient to perform temperature compensation on the transverse and longitudinal dual-wave acoustic time ratio. Based on the temperature-compensated transverse and longitudinal dual-wave acoustic time ratio, it calculates the axial component force of the bolt during operation and combines it with the obtained bolt preload to calculate the bolt axial force.

[0188] The risk prediction modeling module is used to establish a risk prediction model for bolt axial force by taking historically calculated bolt axial force data as input and the bolt risk assessment score as output, and to construct a loss function related to the effective stress zone length of the bolt to optimize the risk prediction model for bolt axial force.

[0189] The health diagnosis and early warning module is used to combine long-term historical monitoring data of the bolt axial force risk prediction model and apply a long short-term memory neural network to realize the time-series health diagnosis and intelligent early warning of bolt axial force. It also links the model's prediction results with the digital operation and maintenance platform to realize automatic early warning of bolt axial force anomalies, push maintenance suggestions, and full life cycle management.

[0190] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0191] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0192] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0193] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method of analyzing an axial force of a bolt, characterized by, The specific steps include: Step 1: Obtain the material parameters and physical and mechanical parameters of the bolt used to equip the wind turbine, and establish a finite element model of the bolt based on the finite element analysis method, simulate and analyze the axial force distribution of the bolt, and screen out the stress weak area of the bolt; Step 2: Real-time acquisition of acoustic time data, strain signal and environmental temperature of the bolt at the stress weak area of the bolt, combined with Hook's law and acoustic elastic theory analysis to determine the effective stress area length and longitudinal wave time difference before and after the bolt loading, establish the bolt pretightening force calibration expression, and temperature compensation for the bolt pretightening force calibration expression to obtain the bolt pretightening force; Step 3: Establish the relationship between the transverse and longitudinal double wave time ratio and the axial stress of the bolt based on the double wave measurement method, introduce the temperature correction coefficient, temperature compensation for the transverse and longitudinal double wave time ratio, calculate the axial force of the bolt in the working state based on the transverse and longitudinal double wave time ratio after temperature compensation, and calculate the axial force of the bolt combined with the obtained bolt pretightening force; Step 4: Take the historical calculated bolt axial force data as input, and take the bolt risk evaluation score as output, establish a risk prediction model of the bolt axial force, and construct a loss function related to the effective stress area length of the bolt to optimize the risk prediction model of the bolt axial force; Step 5: Combined with the long-term historical monitoring data of the risk prediction model, the long and short term memory neural network is applied to realize the time sequence health diagnosis and intelligent early warning of the bolt axial force, and the prediction results of the model are linked with the digital operation and maintenance platform to realize the automatic early warning, maintenance suggestion pushing and whole life cycle management of the bolt axial force; The finite element model of the bolt is established, and the stress weak area of the bolt is screened out, and the method is: Collect the relevant material parameters and physical and mechanical parameters of the bolt, including: the material composition and grade of the bolt, the density, elastic modulus, Poisson's ratio, Lame elastic constant and fatigue performance parameters of the bolt, the relevant material parameters and physical and mechanical parameters of the bolt are obtained by experiment or supplier technical data; Based on the specification parameters of the actual wind turbine bolt, including the diameter, length, thread shape and clamping length of the bolt, a three-dimensional geometric model of the bolt is established by using CAD software, and constraint conditions are applied, one end of the bolt is fixed to simulate the bolt installation clamping state, and the other end is applied with a pretightening force equal to the axial stress of the bolt to simulate the actual pretightening working condition of the bolt; The three-dimensional geometric model of the bolt is divided into grid elements, and an axial tension equal to the nominal pretightening force of the bolt is applied at the top of the bolt, the nominal pretightening force of the bolt is determined according to the yield strength of the bolt material multiplied by the safety factor, and the stress value in each grid element of the three-dimensional geometric model is calculated, the stress threshold is taken as 70% of the yield strength of the bolt, the grid elements exceeding the stress threshold in the three-dimensional geometric model are counted, the space topology analysis is carried out on the grid elements exceeding the stress threshold by adjacency method, the grid elements with common edges or common nodes are classified into the same class, forming several continuous stress concentration zones, each continuous stress concentration zone is defined as a stress weak area, and the number of grid elements contained, the spatial position and the maximum stress value are counted; Determine the effective stress area length before and after the bolt loading and the longitudinal wave acoustic time difference, and establish the bolt pre-tightening force calibration expression, and temperature compensation is carried out on the bolt pre-tightening force calibration expression to obtain the bolt pre-tightening force, and the method is as follows: Based on Hooke's law, the effective stress area length of the bolt is analyzed, when the bolt has no axial stress, the effective stress area length is equal to the sum of the bolt clamping length and the bolt diameter according to the empirical formula in material mechanics; when the bolt has axial stress, the axial strain of the bolt under axial stress is obtained according to the generalized Hooke's law, and the increment of the effective length of the bolt is calculated based on the axial strain to obtain the effective stress area length of the bolt under axial stress; According to the acoustic elasticity theory, the longitudinal wave acoustic time before and after the bolt stress is analyzed, because the axial stress of the bolt will cause the change of the ultrasonic wave velocity and the bolt length, so the ultrasonic wave velocity before and after the bolt loading is obtained, and the bolt length is obtained, so as to calculate the longitudinal wave acoustic time difference before and after the bolt loading; The cross-sectional area of the bolt is obtained, and according to the calculation method of the bolt pre-tightening force in material mechanics: the bolt pre-tightening force is equal to the product of the axial stress of the bolt and the cross-sectional area of the thread, the expression of the longitudinal wave acoustic time difference before and after the bolt loading is combined with the expression of the bolt pre-tightening force to obtain the bolt pre-tightening force calibration expression related to the longitudinal wave acoustic time difference, and the temperature compensation is carried out on the expression, the temperature caused acoustic time difference is substituted into the bolt pre-tightening force calibration expression to obtain the bolt pre-tightening force in the real situation.

2. The bolt axial force analysis method according to claim 1, characterized by, The relationship between the transverse and longitudinal double wave acoustic time ratio and the bolt axial force is established, and the transverse and longitudinal double wave acoustic time ratio is temperature compensated, and the axial force of the bolt in the working state is calculated based on the transverse and longitudinal double wave acoustic time ratio after temperature compensation, and the axial force of the bolt is calculated combined with the obtained bolt pre-tightening force, and the method is as follows: Based on the double wave detection method combined with the obtained acoustic elastic coefficients of the bolt transverse wave and longitudinal wave, the acoustic time of the transverse wave and the acoustic time of the longitudinal wave of the bolt in the non-zero stress state are calculated respectively, and the ratio of the acoustic time of the transverse wave to the acoustic time of the longitudinal wave is the transverse and longitudinal double wave acoustic time ratio; According to the same method, the acoustic time of the transverse wave and the acoustic time of the longitudinal wave of the bolt in the non-stress state are obtained to obtain the acoustic time ratio of the bolt in the non-stress state, and according to the obtained acoustic elastic coefficients of the bolt longitudinal wave and transverse wave, the difference between the two is the acoustic elastic anisotropy coefficient of the bolt; According to the effective area length of the bolt under axial stress and the effective area length of the bolt without axial stress obtained by analysis, the ratio of the two lengths is calculated, which is the effective area length deformation ratio of the bolt, and the axial force sensitivity coefficient of the bolt is calculated based on the acoustic time ratio of the bolt in the non-stress state, the acoustic elastic anisotropy coefficient of the bolt, the effective area length deformation ratio of the bolt and the cross-sectional area of the bolt thread; Based on the axial force sensitivity coefficient of the bolt, the bolt pre-tightening force and the acoustic time ratio of the bolt in the non-stress state, the relationship between the transverse and longitudinal double wave acoustic time ratio and the axial force of the bolt to be measured under the non-zero stress state is established. A temperature correction coefficient is introduced and combined with a temperature change amount to compensate the transverse and longitudinal double wave acoustic time ratio for temperature, so as to obtain a calibrated formula of the transverse and longitudinal double wave acoustic time ratio and the bolt axial force, and the bolt pre-tightening force is calculated to obtain the bolt axial force.

3. The bolt axial force analysis method according to claim 2, characterized by, A risk prediction model of the bolt axial force is established, and a loss function related to the length of the effective force area of the bolt is constructed to optimize the risk prediction model of the bolt axial force, and the method is as follows: The risk prediction model of the bolt axial force is constructed based on a neural network, the model structure includes an input layer, a hidden layer and an output layer, the bolt axial force data calculated from the bolt history is input as the input of the input layer, the representative characteristic values are extracted by statistical analysis on the bolt axial force measurement results in the past service state, the expert scoring method is introduced, the related field experts are organized to comprehensively evaluate and score the past history axial force data, the expert scoring label of the bolt axial force is formed as the output target of the model training and is output by the output layer, and the hidden layer is responsible for nonlinear feature extraction and conversion of the input bolt historical axial force data; The loss function is composed of a traditional error term and a physical consistency error term, the traditional error term is the mean square error between the historical calculated axial force data sequence and the real axial force data sequence, the physical consistency error term is the mean square error between the calculated value of the effective stress area length of the bolt and the real value, collect The real axial force value and the real length value of the effective stress area of the bolt sample are calculated, the traditional error term and the physical consistency error term are calculated, and each term is given a weight by using a linear weighted sum method to jointly constitute the loss function. The risk prediction model of the bolt axial force is optimized based on the constructed loss function, the gradient of the loss function with respect to the network output is calculated, the error signal is back propagated, and the gradient of the loss function to the weight and bias of each layer network is calculated by using the chain rule, and the gradient of the physical consistency error term is transmitted through the mapping link from the predicted axial force to the effective force area length, so as to ensure the influence of physical constraints on parameter updating; The Adam optimizer is used to adjust the network parameters according to the gradient information, and the steps of forward propagation, loss calculation, back propagation and parameter updating are repeatedly executed until the preset maximum iteration number is reached.

4. The bolt axial force analysis method according to claim 3, characterized by, The long-term historical monitoring data of the risk prediction model of the bolt axial force is combined, and the long short-term memory neural network is applied to realize the time sequence health diagnosis and intelligent early warning of the bolt axial force, and the method is as follows: The long-term historical monitoring data of the bolt is collected, including the time sequence bolt axial force calculation measurement value and related working condition characteristics, the data is cleaned and normalized, the abnormal points are removed, the fixed length time window sequence is constructed, the sample set is formed, the health status of the standard bolt is based on the experience label as the supervision signal, and the health status includes normal, warning and abnormal; A multi-layer LSTM network is designed, and the input is the bolt axial force and related feature sequence with a time step of The LSTM effectively captures the long-term dependence in the time series through the gating mechanism, realizes the time series state modeling, and outputs the axial force prediction value and health state probability at the future time. In combination with the threshold judgment, the abnormal state is identified, and the health diagnosis is realized. The prediction result is uploaded to the digital operation and maintenance platform, the abnormal early warning notification is triggered, the platform automatically generates maintenance suggestions based on the early warning result, and pushes the operation and maintenance personnel, so as to realize the whole life cycle management and dynamic risk control of the bolt axial force.

5. A device for analyzing an axial force of a bolt, characterized by The analysis device is used to execute the bolt axial force analysis method in any one of claims 1-4, comprising: A finite element stress analysis module is used to obtain the material parameters and physical and mechanical parameters of the bolt for equipping the wind turbine, and a finite element model of the bolt is established based on the finite element analysis method, the bolt axial force distribution simulation analysis is performed, and the weak force area of the bolt is screened out. The sensing monitoring and calibration module is configured to collect acoustic time data, strain signals and environmental temperature of the bolt at a weak force area of the bolt in real time, determine effective force area length and longitudinal wave acoustic time difference before and after loading of the bolt by combining Hook's law and acoustic elasticity theory analysis, establish a bolt pretightening force calibration expression, and perform temperature compensation on the bolt pretightening force calibration expression to obtain the bolt pretightening force. The double-wave acoustic time compensation module is configured to establish a relationship between transverse-longitudinal double-wave acoustic time ratio and bolt axial stress based on a double-wave measurement method, introduce a temperature correction coefficient to compensate the transverse-longitudinal double-wave acoustic time ratio for temperature, calculate the axial force of the bolt in operation based on the temperature-compensated transverse-longitudinal double-wave acoustic time ratio, and calculate the axial force of the bolt in combination with the obtained bolt pretightening force. The risk prediction modeling module is configured to take bolt axial force data calculated in history as input, take a risk evaluation score of the bolt as output, establish a risk prediction model of the bolt axial force, and construct a loss function related to the effective force area length of the bolt to optimize the risk prediction model of the bolt axial force. The health diagnosis and early warning module is configured to combine long-term historical monitoring data of the risk prediction model of the bolt axial force, apply a long short-term memory neural network to realize time sequence health diagnosis and intelligent early warning of the bolt axial force, link the prediction result of the model with a digital operation and maintenance platform, and realize automatic early warning, maintenance suggestion pushing and full life cycle management of the bolt axial force.

Images