Intelligent identification method for precursor signals of fatigue failure of forgings in complex stress environment

Abstract

This invention relates to the field of industrial intelligent nondestructive testing, and more specifically, to an intelligent identification method for fatigue failure precursor signals of forgings under complex stress environments. The method includes: first, constructing a physical space mapping; extracting streamline vector fields based on finite element simulation data of the forging; calculating sensor path coupling factors; and quantifying the anisotropic constraints of complex stress structures on signal propagation. Next, dynamically setting the signal decomposition algorithm penalty factor according to the streamline geometric curvature; constructing a reference signal by combining the load spectrum and coupling factors; and extracting characteristic responses from stress environment monitoring data. Subsequently, calculating the real-time signal-to-noise ratio (SNR), constructing an adaptive index, and obtaining a blocking index by combining the coupling factor and directional transfer entropy. Finally, setting a dual statistical window based on the load period; and completing the identification and judgment of failure precursor signals according to their statistical characteristics. Through streamline field physical constraints and adaptive SNR modulation, the accuracy and timeliness of fatigue failure early warning are significantly improved.

Description

Technical Field

[0001] This invention relates to the field of industrial intelligent nondestructive testing. More specifically, this invention relates to an intelligent identification method for precursor signals of fatigue failure in forgings under complex stress environments. Background Technology

[0002] Large load-bearing forgings, such as tie rods, typically operate in extremely complex stress environments, bearing alternating loads with multiple axial axes and varying amplitudes. Under these conditions, the initiation and propagation of microcracks within the forgings are often extremely subtle, generating weak signal energy, which is known as precursory failure signals. Accurately capturing and identifying these precursory signals before catastrophic fracture occurs is crucial for ensuring equipment safety.

[0003] However, existing identification methods have significant limitations when facing complex stress environments. First, the intense plastic deformation experienced by forgings during manufacturing creates complex forging streamlines or fibrous structures within the material, resulting in strong anisotropy. Under complex stress, the attenuation characteristics of precursor signals propagating along different streamline directions are drastically different. Traditional methods that treat materials as isotropic ignore the physical constraints of streamline structures on signal propagation paths, easily leading to signal misjudgment. Second, complex stress environments are usually accompanied by high-intensity background noise and non-stationary load fluctuations. These interferences can mask weak failure precursor signals. Traditional fixed-threshold detection strategies cannot distinguish between environmental noise fluctuations and true precursor signals, resulting in frequent false alarms at low signal-to-noise ratios and insufficient sensitivity at high signal-to-noise ratios.

[0004] Therefore, how to combine the internal physical structure characteristics of forgings with the real-time state of the complex external stress environment to construct an identification method that can intelligently sense and extract weak precursor features is a technical problem that urgently needs to be solved. Summary of the Invention

[0005] To address the problem of inaccurate precursor signal identification under complex stress environments due to neglecting material anisotropy and lack of adaptive capability, this invention proposes an intelligent identification method for precursor signals of fatigue failure in forgings under complex stress environments. This method includes the following steps:

[0006] The finite element simulation data of the straight tie rod arm is obtained, a physical space mapping is established, the unit streamline vector field is extracted, and the coupling factor between any two sensors is calculated. The coupling factor characterizes the energy coupling efficiency of the signal propagation in the streamline structure formed by complex forging stress.

[0007] The average geometric curvature of the unit streamline vector field on the calculation path is calculated. The penalty factor of the variational mode decomposition algorithm is dynamically set according to the average geometric curvature. A reference signal is constructed using the load spectrum and coupling factor. Based on the reference signal, the characteristic response signal is extracted from the vibration data of the complex stress environment collected in real time.

[0008] Calculate the real-time signal-to-noise ratio (SNR) of the characteristic response signal, construct an adaptive index based on the real-time SNR, the adaptive index being positively correlated with the real-time SNR; calculate the blocking index by combining the coupling factor, the directional transfer entropy under the reference state, and the directional transfer entropy at the current moment.

[0009] Set short-term and long-term windows based on load cycles, and statistically analyze the mean and standard deviation of the blocking index within the two windows. Based on the statistical results, intelligently identify whether there are fatigue failure precursor signals in the forging.

[0010] This invention introduces streamline vector fields to construct physical constraints and mathematically aligns the sensor layout with the anisotropic structure inside the forging, effectively eliminating signal attenuation interference caused by structural factors under complex stress environments. At the same time, by adaptively modulating the recognition sensitivity through the signal-to-noise ratio, it achieves intelligent locking and extraction of weak failure precursor signals under strong noise interference.

[0011] Preferably, the method for calculating the coupling factor between any two sensors specifically includes:

[0012]

[0013] In the formula, Indicates sensor and sensors Coupling factors between them; Represents the characteristic scale constant; Indicates sensor and sensors A straight path between them; Represents the positional variable on the path; Indicates the path at position Tangential unit vector at the location; Represents the unit streamline vector field at position Unit streamline vector at; This represents the vector dot product operation; This represents the absolute value operation; This represents an exponential function with the natural constant as its base.

[0014] This invention utilizes integral operations to accurately quantify the degree to which the signal propagation path is obstructed by the internal streamline structure, providing accurate physical weights for distinguishing between physical structural attenuation and damage-related blockage under complex stress environments.

[0015] Preferably, another method for calculating the coupling factor between any two sensors includes:

[0016] Discretize the path into several sampling points, and calculate the absolute value of the dot product of the tangential unit vector and the unit streamline vector at each sampling point;

[0017] Calculate the arithmetic mean of the absolute values ​​of the dot products of all sampling points, and use the arithmetic mean as the coupling factor.

[0018] This invention uses weighted summation of discrete nodes to replace continuous integration, which reduces computational complexity and improves the real-time performance of intelligent recognition algorithms on edge computing devices.

[0019] Preferably, the step of dynamically setting the penalty factor of the variational mode decomposition algorithm based on the average geometric curvature includes:

[0020] Obtain the baseline bandwidth parameters of the variational mode decomposition algorithm;

[0021] The product of the average geometric curvature of the unit streamline vector field on the calculation path and the preset sensitivity coefficient is added to the value 1 to obtain the adjustment denominator;

[0022] The penalty factor is obtained by dividing the reference bandwidth parameter by the adjustment denominator.

[0023] This invention takes into account that streamline bending caused by complex stress can lead to waveguide effects and signal dispersion. By dynamically adjusting the penalty factor of the VMD algorithm based on curvature, it automatically adapts to the dispersion characteristics and effectively prevents the missed detection of broadband failure precursor signals.

[0024] Preferably, the construction of the reference signal using the load spectrum and coupling factor includes:

[0025] Construct a delay pulse function, wherein the delay of the delay pulse function is the theoretical speed of sound delay;

[0026] Multiplying the delayed impulse function by the coupling factor yields the weighted impulse response;

[0027] The reference signal is obtained by convolving the load spectrum with the weighted impulse response.

[0028] This invention combines actual working condition loads with physical coupling characteristics to construct an ideal reference signal, providing a precise guiding template for extracting weak precursor features from strong background noise in complex stress environments.

[0029] Preferably, the calculation of the real-time signal-to-noise ratio of the characteristic response signal includes:

[0030] Calculate the root mean square value of the characteristic response signal;

[0031] Calculate the ratio of the root mean square value to the pre-stored background noise constant;

[0032] Calculate the common logarithm of the ratio and multiply it by 20 to obtain the real-time signal-to-noise ratio.

[0033] Preferably, the step of constructing the adaptive exponent based on the real-time signal-to-noise ratio includes:

[0034] A sigmoid function is used to map the real-time signal-to-noise ratio to a preset sensitivity range;

[0035] The input variable of the Sigmoid function is the difference between the real-time signal-to-noise ratio and the signal-to-noise ratio threshold. The output increases monotonically between the lower and upper bounds of the sensitivity, thus obtaining the adaptive exponent.

[0036] Preferably, the blocking index is calculated by combining the coupling factor, the directional transfer entropy under the reference state, and the directional transfer entropy at the current moment, satisfying the following relationship:

[0037]

[0038] In the formula, Indicates the blocking index; Indicates the coupling factor; This represents the directional transfer entropy under the reference state; This represents the directional propagation entropy at the current moment; Indicates zero or minute quantity; This represents the adaptive index.

[0039] Preferably, the method for intelligently identifying whether a forging has fatigue failure precursor signals based on statistical results includes:

[0040] Calculate the mean and standard deviation of the blocking index within the long-term window;

[0041] Calculate the mean blocking index within the short-term window;

[0042] If the mean of the blocking index within the short-term window is greater than the sum of the mean of the blocking index within the long-term window and three times the standard deviation of the long-term window, then fatigue failure precursor signals are identified.

[0043] Preferably, another method for intelligently identifying whether a forging has fatigue failure precursor signals based on statistical results includes:

[0044] The cumulative log-likelihood ratio is calculated using the sequential probability ratio test method.

[0045] The decision threshold is calculated based on the preset tolerance for false alarm rate and tolerance for missed alarm rate;

[0046] If the cumulative log-likelihood ratio is greater than the decision threshold, then a fatigue failure precursor signal is determined to exist.

[0047] The present invention has the following beneficial effects:

[0048] This invention addresses the challenge of weak and easily interfered precursor signals under complex stress environments. For the first time, it integrates anisotropic physical models of streamline fields into the signal processing flow, distinguishing between structural attenuation and failure-related blocking from a mechanistic perspective, thereby significantly improving the accuracy of identification.

[0049] Furthermore, the proposed signal-to-noise ratio-driven adaptive recognition mechanism enables the system to automatically adjust its sensitivity in complex stress environments, achieving both high accuracy under high signal-to-noise ratio and high stability under low signal-to-noise ratio. Attached Figure Description

[0050] Figure 1 This is a flowchart of the steps of the intelligent identification method for fatigue failure precursor signals of forgings in complex stress environments provided in the embodiments of the present invention;

[0051] Figure 2 This is the original grayscale image provided in the embodiments of the present invention;

[0052] Figure 3 This is a defect detection diagram provided by a conventional method in an embodiment of the present invention;

[0053] Figure 4 The residual diagram provided in the embodiments of the present invention;

[0054] Figure 5 The final defect detection diagram provided for an embodiment of the present invention. Detailed Implementation

[0055] Please see Figure 1 The diagram illustrates the steps of the intelligent identification method for fatigue failure precursor signals of forgings in complex stress environments provided in Example 1. The method includes the following steps:

[0056] S1: Obtain finite element simulation data of the straight tie rod arm, establish physical space mapping, extract unit streamline vector field, and calculate the coupling factor between any two sensors.

[0057] S10: Obtain finite element simulation data of the straight tie rod arm, establish physical space mapping, and extract the unit streamline vector field.

[0058] Specifically, the finite element simulation data of the tie rod arm is obtained, a physical space mapping is established, and the unit streamline vector field is extracted, including:

[0059] The forging process of the straight tie rod arm was simulated using finite element analysis software to obtain finite element simulation data in the final forging state. The finite element simulation data specifically includes: spatial coordinate information of mesh nodes and metal rheological velocity vector information at each mesh node.

[0060] Based on the metal rheological velocity vector information, the flow direction of metal fibers at each grid node is extracted and defined as the unit streamline vector at that point. The unit streamline vectors of all grid nodes are normalized to obtain the unit streamline vector, thereby constructing a unit streamline vector field covering the entire internal space of the forging.

[0061] A unified three-dimensional Cartesian coordinate system is constructed to align the physical dimensions of the straight tie rod arm with the mesh dimensions of the finite element simulation model, thus completing the physical space mapping. Under this coordinate system, the coordinates of the transmitting sensor and the receiving sensor are determined, and the straight path connecting the two sensors is obtained.

[0062] Calculate the direction vector from the coordinates of the transmitting sensor to the coordinates of the receiving sensor, and normalize the direction vector to obtain the constant tangential unit vector of the straight path between the two sensors at any position.

[0063] Figure 2 This diagram visualizes the streamline blocking effect. Blue streamlines represent the unit streamline vector field, showing the natural orientation of the material's fibrous structure. Black dots indicate the locations of monitoring sensors arranged on the surface of the tie rod arm, with the lines connecting them forming the monitoring path. Thick red lines represent the microcrack blocking zone that initiates during fatigue testing, laterally severing the originally continuous streamlines. Green arrows indicate that in undamaged areas, signal energy is transmitted losslessly along the streamline direction with high efficiency. Red arrows indicate that in cracked areas, the signal transmission efficiency drops sharply in that direction due to the physical severance of the streamlines. This diagram visually demonstrates the streamline structure within the forging.

[0064] S11: Calculate the coupling factor between any two sensors.

[0065] It is important to note that under complex stress environments, the streamline structure within forgings leads to strong anisotropy in the material. This anisotropy means that the natural attenuation of the signal is no longer a uniform constant, but rather strictly depends on the geometric angle between the propagation path and the streamline direction. If this structurally determined directional difference is not analyzed, and only the signal amplitude is directly analyzed, it is easy to misjudge the normal structurally high attenuation caused by cross-streamline propagation as abnormal damage-related attenuation caused by microcrack blockage. Therefore, it is essential to first construct a streamline model to describe the differentiated constraints of anisotropy on different paths, thereby establishing a theoretical energy transfer benchmark under non-destructive conditions and providing a model foundation for subsequently stripping away structural influences and identifying actual damage.

[0066] Preferably, as an example, the coupling factor between any two sensors is calculated to satisfy the following relationship:

[0067]

[0068] In the formula, Indicates sensor and sensors Coupling factors between them; This represents the characteristic scale constant, which is usually taken as the diameter of the neck of the straight tie rod arm; Indicates the straight path between the two sensors; Represents the positional variable on the path; Indicates the path at position Tangential unit vector at the location; Represents the unit streamline vector field at position Unit streamline vector at; This represents the vector dot product operation; This represents the absolute value operation; This represents an exponential function with the natural constant as its base.

[0069] Understandably, in the formula The term precisely describes the direction of signal propagation on each micro-element segment ( ) and the direction of internal streamlines of the material ( ) Directional deviation. When When the directional deviation is 0, the integral result is minimized, and the coupling factor is minimized. Approaching This indicates that the physical channel is unobstructed and signal transmission is not hindered by structure. At this point, the directional deviation is at its maximum, the integral result increases rapidly, and the coupling factor... This indicates that the physical channel is blocked and signal transmission is obstructed by the structure.

[0070] Furthermore, through the aforementioned integral mapping mechanism, anisotropy was successfully transformed into path-dependent transmission reference weights, thereby eliminating signal attenuation caused by propagation across streamlines. This ensures that subsequent detected signal changes truly reflect material damage rather than the influence of the structure itself, solving the problem of misjudgment caused by anisotropy and providing a key basis for achieving accurate identification.

[0071] In addition, considering the need for real-time computing, the coupling factor can also be calculated in the following ways.

[0072] Optionally, as another example, the coupling factor can be calculated using a weighted summation method based on discrete nodes. The specific calculation method is as follows:

[0073] First, the path Discretized One sampling point.

[0074] Then, the coupling factor is calculated, and the coupling factor satisfies the following relationship:

[0075]

[0076] In the formula, Indicates sensor and sensors Coupling factors between them; Indicates the total number of sampling points; Indicates the first Tangential unit vector at each sampling point; Indicates the first Unit streamline vector at each sampling point.

[0077] In this way, the above operations can establish a physical transmission reference in a lossless state, effectively eliminating the interference of structural anisotropy on signal analysis.

[0078] S2: Calculate the average geometric curvature of the unit streamline vector field on the path, dynamically set the penalty factor of the variational mode decomposition algorithm based on the average geometric curvature, construct a reference signal using the load spectrum and coupling factor, and extract the characteristic response signal from the vibration data of the complex stress environment acquired in real time based on the reference signal.

[0079] It should be noted that in order to perform fatigue failure analysis, it is necessary to extract signals that reflect the characteristics of fatigue failure from complex signals.

[0080] It should be further explained that the physical structure of streamlined curvature is similar to a waveguide. The greater the curvature, the more significant the dispersion effect of elastic waves during propagation. This causes the energy of the failure precursor signal to be dispersed over a wider frequency band. Traditional fixed-parameter decomposition methods cannot adapt to this bandwidth variation caused by changes in structural curvature, and are prone to missing critical broadband signals due to an excessively narrow capture range. Therefore, to better adapt to the bandwidth variation caused by changes in structural curvature, we consider analyzing the physical characteristics of streamlined curvature to adaptively adjust key parameters affecting bandwidth, thereby achieving accurate extraction of signals that reflect fatigue failure characteristics.

[0081] Preferably, as an example, the average geometric curvature of the unit streamline vector field along the path is calculated. Based on this average geometric curvature, the penalty factor for the variational mode decomposition algorithm is dynamically set. A reference signal is constructed using the load spectrum and coupling factor. Based on this reference signal, characteristic response signals are extracted from the real-time acquired vibration data of complex stress environments. The specific implementation method is as follows:

[0082] First, the penalty factor for the variational mode decomposition (VMD) algorithm is dynamically set based on the average geometric curvature, specifically including:

[0083]

[0084] In the formula, This represents the penalty factor for the variational mode decomposition algorithm; Indicates the reference bandwidth parameter; This indicates the preset sensitivity coefficient; Representing a path The average geometric curvature of the unit streamline vector field.

[0085] Understandably, in the VMD algorithm, the penalty factor... It is a key parameter for controlling modal bandwidth. The larger the bandwidth, the narrower the bandwidth. The smaller the value, the wider the bandwidth.

[0086] The above relation is expressed through With curvature The negative correlation mapping mechanism can achieve:

[0087] When streamlines along a path are severely curved, it physically leads to a strong dispersion effect, broadening the signal energy. This causes the denominator in the formula to increase, forcibly pulling the signal lower. Value, smaller This allows the VMD algorithm to automatically relax the bandwidth limit, thus enabling it to completely encompass the failure signal that has broadened due to dispersion.

[0088] For example, assuming a reference bandwidth If the curvature is 0, then If the increase in curvature results in a denominator of 2, then (broadband).

[0089] This intuitively verifies that by utilizing streamlined geometric features, the algorithm effectively solves the problem of narrowband missed detection caused by dispersion effects, thereby enabling adaptive capture of broadband failure precursor signals.

[0090] Then, load data under working conditions are collected in real time by load sensors installed at the stress points of the forging to obtain the load spectrum. A reference signal is constructed by convolving the load spectrum with the ideal transfer function, where the ideal transfer function is obtained by weighting the delay pulse function and the coupling factor.

[0091] The system acquires real-time vibration data collected by the receiving sensor, performs VMD decomposition on the data, calculates the cross-correlation coefficient between each modal component and the reference signal, and extracts the modal component with the highest cross-correlation coefficient as the characteristic response signal.

[0092] Thus, the above processing can overcome the influence of dispersion effect and accurately locate and extract characteristic response signals that reflect fatigue failure characteristics from complex vibration data.

[0093] Figure 3 The image is a schematic diagram of vibration time series data. As can be seen from the image, the waveform is chaotic and the main component is the variable amplitude load noise that fluctuates drastically over time, which completely masks the weak damage characteristics.

[0094] Figure 4 The image shows a schematic diagram of the characteristic response signal. As can be seen from the image, three damage impact characteristic peaks can be clearly seen in the signal, thus proving that load noise interference has been eliminated.

[0095] S3: Calculate the real-time signal-to-noise ratio of the characteristic response signal, construct an adaptive index based on the real-time signal-to-noise ratio, and calculate the blocking index by combining the coupling factor, the directional transfer entropy under the reference state, and the directional transfer entropy at the current moment.

[0096] It should be noted that in order to determine fatigue failure, the extracted characteristic response signals that reflect the characteristics of fatigue failure need to be processed and analyzed in order to construct an index that reflects the characteristics of fatigue failure.

[0097] It should be further explained that noise information under complex stress environments is inherently dynamic. This dynamic characteristic means that the signal-to-noise ratio (SNR) of fatigue failure signals fluctuates drastically between high and low levels. In this situation, using a single fixed threshold or fixed sensitivity easily leads to a dilemma: at low SNR, the signal fluctuates violently, and a fixed high sensitivity can easily trigger false alarms; while at high SNR, even small changes may indicate damage, and insufficient sensitivity can lead to missed alarms. Therefore, it is necessary to construct a nonlinear modulation mechanism that can automatically adjust the decision scale according to the real-time SNR, and to build an index that reflects fatigue failure characteristics applicable to different SNRs, thereby maintaining optimal detection performance under various noise environments. This invention uses the blocking index as an index of fatigue failure characteristics.

[0098] Preferably, as an example, the real-time signal-to-noise ratio of the characteristic response signal is calculated, an adaptive exponent is constructed based on the real-time signal-to-noise ratio, and a blocking exponent is calculated by combining the coupling factor, the directional transfer entropy under the reference state, and the directional transfer entropy at the current moment, including:

[0099] First, the real-time signal-to-noise ratio of the characteristic response signal is calculated, specifically including:

[0100]

[0101] In the formula, Indicates the real-time signal-to-noise ratio; This represents the root mean square (RMS) calculation function; Indicates the characteristic response signal; It represents the background noise constant that is measured and stored in advance when the equipment is stopped or unloaded.

[0102] Then, an adaptive exponent is constructed based on the real-time signal-to-noise ratio:

[0103]

[0104] In the formula, Indicates the adaptive index; These represent the lower bounds of sensitivity, for example. Take 1 and 3 respectively; For example, the adjustment coefficient is... Take 0.5; Indicates the real-time signal-to-noise ratio; This represents the signal-to-noise ratio threshold.

[0105] Understandably, this relationship achieves an S-shaped mapping through the Sigmoid function; when the signal-to-noise ratio is high, Tend to This improves detection sensitivity.

[0106] Next, acquire the data from the transmitting sensor. signal sequence and receiving sensor signal sequence Under the initial health condition of the forging during its service life, calculations were performed from... arrive The directional transfer entropy, denoted as the directional transfer entropy under the reference state. At the current monitoring moment, real-time calculations are performed from... arrive The directional transfer entropy, denoted as the directional transfer entropy at the current moment. .

[0107] The blocking exponent is calculated by combining the coupling factor, the directional transfer entropy under the reference state, and the directional transfer entropy at the current moment, satisfying the following relationship:

[0108]

[0109] In the formula, Indicates the blocking index; Indicates the coupling factor; This represents the directional transfer entropy under the reference state; This represents the directional propagation entropy at the current moment; Indicates zero-value (e.g.) ); This represents the adaptive index.

[0110] Understandably, fatigue microcracks can block information transmission, leading to... Decrease, ratio The number of these factors increases dramatically, resulting in a larger calculated blocking index when fatigue microcracks are present.

[0111] Furthermore, the exponential term It is dynamically adjusted; when the environment is harsh, Automatic reduction weakens the amplification capability of the power operation comparison value fluctuation, thereby suppressing noise interference and resulting in a smaller calculated blocking exponent; when the environment is favorable, The automatic increase utilizes the nonlinear amplification characteristics of the power function to significantly amplify the weak crack precursor signal, thereby resulting in a larger calculated blocking index.

[0112] Furthermore, the coefficient It can forcibly block out the obstruction caused by the obstruction of the streamline structure itself, and the calculated obstruction index can only reflect the obstruction caused by the actual damage.

[0113] S4: Set short-term and long-term windows based on the load cycle, calculate the mean and standard deviation of the blocking index within the two windows, and intelligently identify whether there are fatigue failure precursor signals in the forging based on the statistical results.

[0114] It should be noted that, in order to identify fatigue failure, the fatigue failure status of the forging must be judged based on the blocking index calculated above.

[0115] It is important to further explain that during long-term service, the operating baseline of equipment will slowly drift with changes in operating conditions. This drift means that the statistical characteristics of the signal are no longer stationary, but rather exhibit a long-term trend. In this situation, if an absolute fixed threshold is used for judgment, this normal baseline drift can easily be misjudged as an abnormal failure mutation. Therefore, this step introduces a dual-window relative statistical judgment logic, using a long-term window to lock the dynamic baseline and a short-term window to capture immediate mutations, thereby eliminating the influence of baseline drift while accurately identifying the true fatigue failure.

[0116] Preferably, as an example, a short-term window and a long-term window based on the load cycle are set, and the mean and standard deviation of the blocking index within the two windows are statistically analyzed. Based on the statistical results, the presence of fatigue failure precursor signals in the forging is intelligently identified, including:

[0117] Calculate the blocking index at each time point, arrange the blocking indices in time sequence to obtain the blocking index sequence, and set a short-term window. and long-term window Exemplary Take 5 load cycles, Use 50 load cycles. Slide the blocking index sequence using both short-term and long-term windows. After each slide, calculate the mean blocking index using the blocking index data within the window. and standard deviation Similarly, calculate the mean of the blocking index within the short-term window. .

[0118] If it satisfies 3 times consecutively If this continues, an early warning will be triggered, indicating the presence of a precursor signal to fatigue failure.

[0119] Furthermore, considering the higher requirements for response speed, the following methods can also be used to identify fatigue failure precursor signals:

[0120] Optionally, as another example, the cumulative log-likelihood ratio can be calculated using a sequential probability ratio test based on the blocking exponent sequence. .like Then an alarm will be triggered, among which These are the set tolerance levels for false alarm rate and missed alarm rate, respectively.

[0121] Understandably, this method triggers an alarm as soon as the accumulated failure probability exceeds a set threshold, resulting in a faster response.

[0122] In this way, by performing the above operations, potential fatigue failure risks can be quickly identified while ensuring an extremely low false alarm rate, thus gaining valuable time for timely maintenance.

[0123] Figure 5 This graph compares the sensitivity of monitoring indicators with their early warning capabilities. The horizontal axis represents the number of loading cycles in the fatigue test, and the vertical axis represents the normalized monitoring indicator values. The gray dashed line represents the monitoring curve calculated using the traditional vibration energy method, which fluctuates significantly and rises slowly. The blue solid line represents the blocking index curve calculated by this invention, which is stable and rises sharply. In the background area, the green area on the left represents the healthy stage, and the red area on the right represents the damage propagation stage; the dividing line is the moment of microcrack initiation. The arrows indicate that this invention triggers an early warning at the initial stage of microcrack initiation, while traditional methods exhibit a significant lag.

[0124] This image demonstrates that, compared to traditional methods, the present invention can achieve more accurate fatigue failure identification.

[0125] This concludes the embodiment.

[0126] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intelligent identification of precursor signals of fatigue failure in forgings under complex stress environments, characterized in that, include: The finite element simulation data of the straight tie rod arm is obtained, a physical space mapping is established, the unit streamline vector field is extracted, and the coupling factor between any two sensors is calculated. The coupling factor characterizes the energy coupling efficiency of the signal propagation in the streamline structure formed by complex forging stress. The average geometric curvature of the unit streamline vector field on the calculation path is calculated. The penalty factor of the variational mode decomposition algorithm is dynamically set according to the average geometric curvature. A reference signal is constructed using the load spectrum and coupling factor. Based on the reference signal, the characteristic response signal is extracted from the vibration data of the complex stress environment collected in real time. Calculate the real-time signal-to-noise ratio of the characteristic response signal, and construct an adaptive index based on the real-time signal-to-noise ratio, wherein the adaptive index is positively correlated with the real-time signal-to-noise ratio; An adaptive exponent is constructed based on the real-time signal-to-noise ratio, including: A sigmoid function is used to map the real-time signal-to-noise ratio to a preset sensitivity range; The input variable of the Sigmoid function is the difference between the real-time signal-to-noise ratio and the signal-to-noise ratio threshold. The output increases monotonically between the lower and upper bounds of the sensitivity, thus obtaining the adaptive exponent. The blocking exponent is calculated by combining the coupling factor, the directional transfer entropy under the reference state, and the directional transfer entropy at the current moment, satisfying: Indicates the blocking index; Indicates the coupling factor; This represents the directional transfer entropy under the reference state; This represents the directional propagation entropy at the current moment; Indicates zero or minute quantity; Indicates the adaptive index; Set short-term and long-term windows based on load cycles, and statistically analyze the mean and standard deviation of the blocking index within the two windows. Based on the statistical results, intelligently identify whether there are fatigue failure precursor signals in the forging.

2. The intelligent identification method for precursor signals of fatigue failure in forgings under complex stress environments according to claim 1, characterized in that, The method for calculating the coupling factor between any two sensors specifically includes: In the formula, Indicates sensor and sensors Coupling factors between them; Represents the characteristic scale constant; Indicates sensor and sensors A straight path between them; Represents the positional variable on the path; Indicates the path at position Tangential unit vector at the location; Represents the unit streamline vector field at position Unit streamline vector at; This represents the vector dot product operation; This represents the absolute value operation; This represents an exponential function with the natural constant as its base.

3. The intelligent identification method for precursor signals of fatigue failure in forgings under complex stress environments according to claim 1, characterized in that, Another method for calculating the coupling factor between any two sensors includes: Discretize the path into several sampling points, and calculate the absolute value of the dot product of the tangential unit vector and the unit streamline vector at each sampling point; Calculate the arithmetic mean of the absolute values ​​of the dot products of all sampling points, and use the arithmetic mean as the coupling factor.

4. The intelligent identification method for precursor signals of fatigue failure in forgings under complex stress environments according to claim 1, characterized in that, The step of dynamically setting the penalty factor for the variational mode decomposition algorithm based on the average geometric curvature includes: Obtain the baseline bandwidth parameters of the variational mode decomposition algorithm; The product of the average geometric curvature of the unit streamline vector field on the calculation path and the preset sensitivity coefficient is added to the value 1 to obtain the adjustment denominator; The penalty factor is obtained by dividing the reference bandwidth parameter by the adjustment denominator.

5. The intelligent identification method for precursor signals of fatigue failure in forgings under complex stress environments according to claim 1, characterized in that, The construction of the reference signal using the load spectrum and coupling factor includes: Construct a delay pulse function, wherein the delay of the delay pulse function is the theoretical speed of sound delay; Multiplying the delayed impulse function by the coupling factor yields the weighted impulse response; The reference signal is obtained by convolving the load spectrum with the weighted impulse response.

6. The intelligent identification method for precursor signals of fatigue failure in forgings under complex stress environments according to claim 1, characterized in that, The calculation of the real-time signal-to-noise ratio of the characteristic response signal includes: Calculate the root mean square value of the characteristic response signal; Calculate the ratio of the root mean square value to the pre-stored background noise constant; Calculate the common logarithm of the ratio and multiply it by 20 to obtain the real-time signal-to-noise ratio.

7. The intelligent identification method for precursor signals of fatigue failure in forgings under complex stress environments according to claim 1, characterized in that, The method for intelligently identifying whether fatigue failure precursor signals exist in forgings based on statistical results includes: Calculate the mean and standard deviation of the blocking index within the long-term window; Calculate the mean blocking index within the short-term window; If the mean of the blocking index within the short-term window is greater than the sum of the mean of the blocking index within the long-term window and three times the standard deviation of the long-term window, then fatigue failure precursor signals are identified.

8. The intelligent identification method for precursor signals of fatigue failure in forgings under complex stress environments according to claim 1, characterized in that, Another method for intelligently identifying the presence of fatigue failure precursor signals in forgings based on statistical results includes: The cumulative log-likelihood ratio is calculated using the sequential probability ratio test method. The decision threshold is calculated based on the preset tolerance for false alarm rate and tolerance for missed alarm rate; If the cumulative log-likelihood ratio is greater than the decision threshold, then a fatigue failure precursor signal is determined to exist.

Images