A method for predicting the breaking of a stainless steel cable tie

Abstract

The application discloses a kind of stainless steel cable ties fracture prediction method, specifically related to material forming processing technical field, including in cable tie punching forming process, collection multi-source original data, construct time series associated multidimensional production behavior data group, and extract spectrum, deformation, thermal imaging and acoustic emission feature constitute fluctuation judgment basis;When meeting risk threshold condition, calculate crack potential length index and structure stability index, input trained model, output fracture influence grade, realize the risk prediction and grading processing of stainless steel cable ties, ensure that it adapts to different application scenarios or rejects high-risk products;The application realizes the real-time acquisition and analysis of multi-source data in the punching process of cable ties, dynamically judges the risk by constructing fluctuation judgment basis, learns the model prediction by combining crack potential length index and structure stability index, outputs fracture grade, realizes cable tie grading processing and precise application, improves reliability and intelligent level.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of material forming processing, and more particularly to a stainless steel cable tie fracture prediction method. BACKGROUND

[0002] As a key basic connecting component, the stainless steel cable tie is widely used in aerospace, military equipment, high-voltage power transmission, rail transit, deep sea exploration and other key systems with high requirements for structural safety and operation stability under extreme conditions such as high temperature, high vibration and high electromagnetic interference. Its stability directly affects the integrity of internal wiring, the safety of energy transmission and the synergy of structural components, and is an irreplaceable component for achieving high reliability manufacturing.

[0003] However, in the actual production process of the stainless steel cable tie, due to factors such as stamping load fluctuation, mold structure difference, material micro-defects or thermal-mechanical coupling, potential structural risks such as edge initial crack, internal strain accumulation or local tension overload are easily induced. Such risks are often difficult to detect in traditional process inspection, showing strong concealment, sudden consequences and difficult traceability. Once the cable tie breaks during service, it may cause wiring to fall off, signal interruption, or even machine failure and major safety accidents. The fracture risk control has become a key technical bottleneck restricting the widespread application of cable ties in high reliability scenarios.

[0004] At present, the quality control method in the industry still mainly relies on product sampling detection, static mechanical test or visual appearance inspection, and lacks structured modeling and real-time prediction mechanism for the dynamic physical behavior of the cable tie forming process. It cannot identify early signs of micro-crack evolution and structural stability weakening process, especially lacking a prediction-driven grading strategy matching the risk level of the actual application scenario. Some research attempts to introduce thermal imaging, acoustic emission or stress response into the fracture judgment model, but due to isolated feature dimension, lack of information fusion mechanism and weakened physical interpretation, it is difficult to meet the dual requirements of high-precision prediction and scene adaptation. Therefore, the present application proposes a stainless steel cable tie fracture prediction method to solve the above problems. SUMMARY

[0005] To achieve the above purpose, the present application provides the following technical scheme:

[0006] A stainless steel cable tie fracture prediction method, comprising the following steps:

[0007] During the cable tie punching forming process, collect multi-source original data associated with fracture risk in continuous production batches, and build a multi-dimensional production behavior data set with time sequence correlation;

[0008] The multi-dimensional production behavior data set is subjected to feature extraction to form a basis for determining whether to start the fracture risk assessment, wherein the basis is a multi-dimensional feature fluctuation basis, specifically including a frequency spectrum fluctuation rate in a stamping stroke, a deformation profile edge curvature change amount, a thermal distribution abnormal area pixel ratio, and an acoustic emission time domain energy jump value;

[0009] When the basis meets a preset risk threshold condition, a fracture risk assessment process is started, and two groups of feature data sets are constructed, wherein the first group is a crack trend feature data set composed of micro-crack boundary continuous curvature values, strain gradient distribution values, and residual thermal tension distribution values; and the second group is a structure stability feature data set composed of acoustic emission signal spectrum concentration, thermal area centroid offset values, and pressure release response delay time;

[0010] The crack trend feature data set is used to calculate a crack potential length index, and the structure stability feature data set is used to calculate a structure stability index, and the two indexes are used as input variables of the fracture risk assessment;

[0011] The crack potential length index and the structure stability index are input into a pre-trained learning model to output a fracture influence degree grade, and the stainless steel strap is classified and processed according to the fracture influence degree grade to determine whether it is used in a preset different application scenario or an elimination instruction is executed.

[0012] The original data includes stamping frequency spectrum data reflecting the stamping mechanical process, micro-deformation profile data representing the micro-deformation state of the strap surface, local thermal imaging data identifying thermal abnormal areas, and acoustic emission response data sensing internal crack activities of the material.

[0013] In a preferred embodiment, in the multi-dimensional feature fluctuation basis, the frequency spectrum fluctuation rate in the stamping stroke is obtained by dividing the difference between the maximum frequency amplitude and the minimum frequency amplitude in the same stroke segment of the stamping frequency spectrum data by the average frequency amplitude; the deformation profile edge curvature change amount is calculated by the difference between the second derivative values of multiple equidistant points at the strap edge in the micro-deformation profile data; the thermal distribution abnormal area pixel ratio is calculated by the ratio of the number of pixels in the region with a temperature higher than three times the standard deviation of the ambient temperature to the total number of pixels in the local thermal imaging data; and the acoustic emission time domain energy jump value is calculated by the ratio of the energy peak value difference per unit time to the average energy value in the reference smooth period in the acoustic emission response data.

[0014] In a preferred embodiment, the process of determining whether the basis meets the preset risk threshold condition includes selecting a corresponding determination mode according to the stamping rhythm stability feature in the multi-dimensional production behavior data set; when the standard deviation of the frequency spectrum main peak amplitude in the continuous three stroke periods in the stamping frequency spectrum data is less than a first stability threshold, a first determination mode is adopted; and when the standard deviation is greater than or equal to the first stability threshold, a second determination mode is adopted.

[0015] In the first determination mode, if the spectral fluctuation rate within the stamping stroke is greater than a first threshold value, and the pixel proportion of the thermal distribution abnormal area is greater than a second threshold value, it is determined that the preset risk threshold condition is met; in the second determination mode, if the acoustic emission time domain energy jump value is greater than a third threshold value, or the deformation profile edge curvature change amount is greater than a fourth threshold value, it is determined that the preset risk threshold condition is met, and a fracture risk evaluation process is started.

[0016] In a preferred embodiment, the crack tendency feature data set is constructed by the following method:

[0017] The microcrack boundary continuous curvature value is obtained by equally sampling the edge pixel point sequence in the micro deformation profile data, calculating the local curvature value of each point based on the three-point difference method, and taking the first derivative mean square deviation of all curvature values as the continuous curvature value;

[0018] The strain gradient distribution value is obtained by constructing a difference matrix for each stress amplitude based on the stamping stroke length as a reference, and taking the maximum change gradient as the distribution value, from the multi-point stamping stress response curve extracted from the stamping spectrum data;

[0019] The residual thermal tension distribution value is obtained by converting the temperature difference between the thermal anomaly area and its boundary neighborhood in the local thermal imaging data into a thermal tension value, fitting the tension distribution gradient in the form of Laplace transform on a two-dimensional pixel matrix, and taking the value of the maximum gradient point as the residual thermal tension distribution value.

[0020] In a preferred embodiment, the structure stability feature data set is constructed by the following method:

[0021] The acoustic emission signal spectrum concentration is obtained by calculating the proportion of the energy integral value in the preset target frequency band to the total energy in the full frequency domain after the acoustic emission response data is subjected to Fourier transform in the preset target frequency band;

[0022] The thermal zone centroid offset value is obtained by performing centroid calculation on the pixel set in the region where the temperature is greater than twice the standard deviation of the statistical average value in the local thermal imaging data, and taking the Euclidean distance between the centroid position and the center axis of the strap as the offset value;

[0023] The pressure release response delay time is directly calculated in milliseconds from the time difference between the end time of the stroke in the stamping spectrum data and the time of the first abnormal peak value in the acoustic emission response data.

[0024] In a preferred embodiment, the crack potentiality index is calculated by the following steps:

[0025] The microcrack boundary continuous curvature value K, the strain gradient distribution value E, and the residual thermal tension distribution value T obtained in the crack tendency feature data set are standardized to obtain the corresponding standardized values , according to the image edge energy calculation principle, the local nonlinear amplification function is introduced to enhance the sensitivity of high risk value, and three enhanced functions are defined in turn 、 、 :

[0026] ;

[0027] The synergistic cross terms are calculated in turn 、 、 , representing the covariant synergistic strength between different risk factors:

[0028] The weighted comprehensive crack potential function value S is constructed, defined as:

[0029] ; All are non-zero amplification coefficients set in advance according to experience;

[0030] The crack potential function value S is input into the Logit function for normalized mapping to form the final crack potential index CGPI: .

[0031] In a preferred embodiment, the calculation steps of the structure stability index are as follows:

[0032] The obtained acoustic emission signal spectrum concentration A, thermal zone centroid offset value H and pressure release response delay time D in the structure stability feature data set are standardized, respectively denoted as:

[0033] ; The historical mean and standard deviation of the corresponding index X are represented, and then the disturbance synergism matrix M is constructed to obtain the following two-dimensional disturbance tensor: ;

[0034] Where each element represents the synergistic disturbance strength between different characteristics, the diagonal elements represent the single factor self disturbance variance, and the non diagonal elements represent the cross disturbance effect. The matrix Frobenius norm is used to measure the overall disturbance energy , and the calculation formula is: ;

[0035] The risk perception mapping function is constructed to output the structure stability index SIFI: ; is a preset amplification parameter to control the overall response rate, is a preset risk response nonlinear index.

[0036] In a preferred embodiment, the learning model is a machine learning classification model that has completed pre-training, the input of which is a two-dimensional feature vector composed of the crack potential index and the structure stability index, the model is trained through a training sample set constructed based on historical stainless steel strap fracture data, and the training sample set contains label information of actual fracture and corresponding crack potential index and structure stability index values;

[0037] The machine learning classification model adopts a light gradient boosting tree model or a deep neural network model based on residual structure, and establishes the coupling relationship between the crack development trend and the structure stability through supervised learning. The model output is the fracture influence degree grade prediction result, and the prediction result is a continuous numerical score index, denoted as the fracture influence grade score value, the numerical range is 0 to 100, indicating the comprehensive score of the possibility and consequence severity of the strap fracture;

[0038] The fracture influence grade score value output by the model adopts a skew probability density modeling method in the training process, specifically uses a lognormal distribution or a Weibull distribution function to fit the risk level interval of the historical fracture label sample, and constructs an asymmetric mapping relationship between the score interval and the fracture possibility. Finally, it is used according to the preset standard in different application scenarios or executes the rejection instruction.

[0039] The technical effects and advantages of the present application are as follows:

[0040] The present application realizes the whole process perception and information integration of the strap production state by collecting multi-source original data in the strap punching forming process and constructing a multi-dimensional production behavior data set with time sequence correlation. Compared with the traditional method of relying on sampling detection or visual inspection, this method can extract signals from stamping vibration, surface micro-deformation, heat distribution and acoustic emission in each production batch in real time, realize comprehensive coverage and dynamic capture of abnormal behavior, improve the timeliness and accuracy of fracture prediction, and is especially suitable for high rhythm and continuous automatic production scenes.

[0041] The present application uses multi-dimensional feature fluctuation as the basis for determining whether to start fracture risk assessment, breaking through the problem of excessive dependence on single index threshold in the past. By extracting statistical fluctuation characteristics in four features such as stamping frequency spectrum volatility, contour curvature change, thermal imaging anomaly and acoustic emission energy jump, and combining with the stability of stamping rhythm, the judgment mode is dynamically switched, so that the judgment basis is more adaptive and has higher fault tolerance, which can effectively deal with atypical structure risk signals caused by working condition fluctuation, material batch difference or die wear in the production process, and reduce the misjudgment rate and the omission rate.

[0042] The application realizes the quantitative prediction path from physical behavior data to risk level by constructing crack trend characteristic data set and structure stability characteristic data set, respectively calculating crack potential index and structure stability index, and introducing the two as input variables into the learning model to output the fracture influence degree grade. The model combines the supervised learning mechanism of actual fracture data and the multi-dimensional feature modeling capability, can grade the strap according to the risk level result or reject operation, thereby establishing a set of application adaptation mechanism driven by prediction score, and significantly improving the use safety and screening efficiency of the strap product in different scenes such as aviation, electric power and military industry. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to facilitate the understanding of those skilled in the art, the application will be further described below in conjunction with the drawings;

[0044] Figure 1 The principle diagram of the fracture prediction method of the stainless steel strap in the application. DETAILED DESCRIPTION

[0045] The technical solutions in the embodiments of the application will be clearly and completely described below in conjunction with the drawings of the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0046] REFERENCE Figure 1 The following embodiments are obtained:

[0047] Embodiment 1: The application relates to a method for fracture prediction of a stainless steel strap, and the core is that based on original data collected in a punching forming process of the strap, a time sequence associated multi-dimensional production behavior data set is constructed, and by extracting characteristic fluctuations in dimensions such as frequency spectrum, deformation, thermal imaging and acoustic emission, a priori judgment basis for judging whether there is a fracture risk is formed. When the system identifies that the frequency spectrum stability deviates from a preset threshold, a fracture risk assessment process is automatically triggered, so that the conversion of strap quality control from post-failure screening to real-time prediction is realized, and the adaptation capability of the product in a high reliability scene is improved.

[0048] After triggering the assessment process, by finely analyzing the microstructure changes and energy responses in the strap forming process, two groups of risk characteristic data sets are constructed, one group describes the crack development trend, and the other group describes the structure stability state. The crack trend characteristics include boundary curvature, strain gradient and thermal tension distribution, and the structure stability characteristics include frequency spectrum energy concentration, thermal centroid change and response delay time. Crack potential index and structure stability index are respectively constructed according to the two groups of characteristics, and the two indexes reflect different physical mechanisms of the strap fracture risk and are used as input basis for fracture influence assessment.

[0049] The application adopts the trained learning model to cooperatively fuse the two indexes, outputs the breaking impact degree rating of the cable tie, and combines the risk score distribution structure constructed by the model based on historical breaking data to divide into multiple levels and correspondingly allocate to high reliability use scenarios, general industrial use or perform rejection processing. Through deep integration of multi-physical source data behavior and machine learning model, the whole process prediction and application adaptation of the stainless steel cable tie breaking risk are realized, and the quality evaluation ability and production intelligent level of the key components are improved. Specifically, the following steps are included:

[0050] In the cable tie punching forming process, multi-source original data associated with the breaking risk in the continuous production batch are collected, and a multi-dimensional production behavior data group with time sequence correlation is constructed. The significance of this step is to obtain various signals from mechanical stamping response, material deformation state, temperature distribution anomaly and micro-crack acoustic emission in the source stage of the production process, realize comprehensive perception of the cable tie forming state, and organize the data structure in a time sequence manner, thereby providing a data basis for subsequent dynamic feature analysis and trend judgment, so as to replace the traditional method of relying only on factory sampling, and have higher real-time and traceability.

[0051] The multi-dimensional production behavior data group is subjected to feature extraction to form the basis for judging whether to start the breaking risk assessment, wherein the basis is a multi-dimensional feature fluctuation basis, specifically including frequency spectrum fluctuation rate within the stamping stroke, deformation profile edge curvature change amount, thermal distribution anomaly area pixel ratio and acoustic emission time domain energy jump value. This step converts the physical meanings of various original data through engineering conversion to extract key numerical features that can reflect the potential instability, abnormal energy release and micro-crack propagation trend in the cable tie forming process, thereby establishing a prior trigger mechanism for initially judging whether the cable tie has a breaking risk, and realizing an efficient evaluation mode from "full monitoring" to "precise response".

[0052] When the judgment basis meets the preset risk threshold condition, the breaking risk assessment process is started, and two groups of feature data groups are constructed. The first group is a crack trend feature data group, which is composed of micro-crack boundary continuous curvature value, strain gradient distribution value and residual thermal tension distribution value; the second group is a structure stability feature data group, which is composed of acoustic emission signal spectrum concentration, thermal zone centroid offset value and pressure release response delay time. The significance of this step is that once the cable tie is identified to have structural risk signs, the system no longer relies on single signal judgment, but enters the in-depth evaluation stage, models through two paths of crack development trend and structure stability state in parallel, separates two independent risk sources of stress evolution and state disturbance, and helps to realize risk source classification, mechanism tracing and index representation.

[0053] The crack potential index is calculated by using the crack tendency characteristic data set, and the structure stability index is calculated by using the structure stability characteristic data set, and the two indexes are used as input variables for fracture risk assessment, this step is through the establishment of crack potential index and structure stability index, the multi-dimensional engineering characteristics are nonlinearly modeled and fused and mapped, so that the complex crack evolution and instability mode reflected by the original signal can be quantified, and the cross-batch comparison, trend sorting and risk scoring ability is obtained, which is a key step for converting perception data into an intelligent model that can be understood and processed.

[0054] The crack potential index and the structure stability index are input into the pre-trained learning model, and the fracture influence degree level is output, and the stainless steel strap is graded according to the fracture influence degree level, and it is decided to be used in different application scenarios or to execute the rejection instruction, this step is through the introduction of the intelligent learning model trained based on the historical fracture sample, the foregoing physical index is mapped to the fracture risk level, and the automatic closed loop from prediction judgment to grading decision is realized. By delimiting a plurality of risk response intervals, the system can dynamically decide the subsequent flow path according to the actual fracture possibility and potential impact result of the strap, thereby significantly improving the adaptation efficiency and risk control ability of the stainless steel strap in the high reliability field such as military industry, electric power and aviation.

[0055] In the fracture prediction method of the stainless steel strap, the original data includes: stamping frequency spectrum data for reflecting the stamping mechanical process, micro-deformation contour data for characterizing the micro-deformation state of the strap surface, local thermal imaging data for identifying the thermal abnormal area, and acoustic emission response data for sensing the internal fission activity of the material. The above-mentioned original data is used as the basis for constructing a multi-dimensional production behavior data set, and its acquisition process is synchronized with the strap punching forming process, and is organized in time series form, ensuring the real-time and continuity of the data, facilitating subsequent feature extraction and risk assessment operations.

[0056] The stamping frequency spectrum data refers to the vibration signal generated between the processing equipment and the strap material during the contact and deformation process during the punching processing of the strap, which is obtained by a high-sensitivity frequency sensor, and the signal is subjected to Fourier transform to obtain a frequency domain expression, which is used to reflect the energy distribution, rhythm fluctuation and instantaneous response state of the material during the impact process. In the present application, the stamping frequency spectrum data is used to calculate the frequency spectrum fluctuation rate in the stamping stroke, which is used as one of the important bases for judging whether there is structural instability in the strap forming process, and further participates in the construction of the strain gradient distribution value and the pressure release response delay time and other key characteristics.

[0057] Micro-deformation profile data refers to the micro-deformation profile image of the strap surface after forming, which is obtained by a micro-displacement laser measuring instrument or a high-resolution profile imaging device. The image reflects the micro-scale deformation traces of the strap caused by stress concentration or die wear during the stamping process. The edge curvature, local collapse and linear discontinuous features can be obtained by pixel-level profile extraction algorithm. In the present application, the micro-deformation profile data is used to calculate the deformation profile edge curvature change and the micro-crack boundary continuous curvature value, so as to capture the initial nucleation phenomenon of potential cracks and provide basic input data for crack trend modeling.

[0058] Local thermal imaging data refers to the thermal distribution image obtained by a medium-wave or long-wave infrared thermal imaging sensor during the forming process of the strap. The data can be used to detect temperature abnormal areas caused by local plastic deformation, friction heat or material thermal sensitivity. The high-temperature area, temperature gradient boundary and thermal zone spatial distribution information can be extracted by image recognition algorithm. In the present application, the local thermal imaging data is used to calculate the pixel proportion of thermal distribution abnormal area, residual thermal tension distribution value and thermal zone centroid offset value, which serve as the basis for risk trigger judgment, crack trend modeling and structure stability modeling, and play a multiple supporting role.

[0059] Acoustic emission response data refers to the instantaneous elastic wave signal released by the internal structure when micro-crack propagation, structure damage or material friction slip occurs in the strap material. The signal is captured by an acoustic emission sensing device and processed by time domain analysis and frequency spectrum analysis to form a response data set reflecting the internal fission activity of the strap. In the present application, the acoustic emission response data is used to calculate the acoustic emission time domain energy jump value, acoustic emission signal spectrum concentration and pressure release response delay time, which can sensitively reflect the hidden damage behavior of the strap material during impact forming process, and is an important index source for evaluating structure integrity and fracture risk.

[0060] The basis for judging whether to start fracture risk assessment in the present application is multi-dimensional feature fluctuation judgment basis, which is constructed based on targeted feature extraction of original data during the stamping and cutting forming process of the strap, to reflect the dynamic response fluctuation of the strap material during the processing process, aiming to find the signs of potential crack formation and structure instability signals induced by abnormal manufacturing behavior. The multi-dimensional feature fluctuation judgment basis includes four core indexes: frequency spectrum fluctuation rate within stamping stroke, deformation profile edge curvature change, thermal distribution abnormal area pixel proportion and acoustic emission time domain energy jump value. The above four indexes describe the processing state of the strap from four dimensions of frequency response, deformation structure, thermal force feature and internal fission behavior, forming a multi-dimensional information joint judgment mechanism.

[0061] The frequency spectrum fluctuation rate in a punch stroke refers to a parameter reflecting the stability of the processing rhythm extracted based on the punch spectrum data. The punch spectrum data is the vibration response signal generated during the contact impact between the strap and the punch. After frequency spectrum processing, the frequency amplitude distribution map can be obtained. In specific implementation, the frequency amplitude data corresponding to all sampling points in the same stroke segment are selected to determine the maximum frequency amplitude and the minimum frequency amplitude. The difference value is divided by the average value of the frequency amplitude in the segment to obtain the frequency spectrum fluctuation rate in the punch stroke. The larger the index, the more uneven the impact energy distribution in the processing process, indicating the abnormal risks such as die impact deviation and local deformation of the material.

[0062] The curvature variation of the deformed profile edge refers to a parameter calculated based on the curvature variation trend of the edge of the strap in the micro-deformed profile data. The micro-deformed profile data is collected by a non-contact profile scanning method, and records the micro-topography of the edge of the strap after forming. In specific operation, a plurality of equidistant points are extracted from the edge of the strap, the local second derivative of each point is calculated by the three-point difference method, representing the curvature value of the corresponding position, and then the maximum difference value in the curvature values is obtained, which is the curvature variation of the deformed profile edge. The index reflects whether there is an abnormal deformation region in the edge of the strap, and reveals the potential stress concentration point of the micro-crack.

[0063] The pixel ratio of the thermal distribution abnormal area refers to the proportion of the high-temperature abnormal area identified from the local thermal imaging data. The local thermal imaging data is extracted from the thermal distribution image obtained by the infrared thermal imager at the moment of strap forming, and is processed by combining statistical temperature analysis. In specific implementation, first, the average value and the standard deviation of the pixel temperature in the entire image are calculated, then all pixel points with a temperature value higher than three times the standard deviation of the ambient temperature are identified, the total number of pixels in the area is counted, and the ratio calculation is performed with the total number of pixels in the entire image to obtain the pixel ratio of the thermal distribution abnormal area. The higher the ratio, the more serious the local stress release abnormality or friction heat concentration phenomenon in the forming process, which may be accompanied by local tissue damage or plastic deformation residue.

[0064] The time domain energy jump value of acoustic emission refers to an energy fluctuation parameter reflecting internal crack or microstructure mutation obtained by analyzing acoustic emission response data. The acoustic emission response data is derived from the elastic strain wave rapidly released by the internal structure of the strap at the moment of punching, and the energy change in unit time is captured by high-frequency sampling. In actual operation, a fixed length time window before and after a punch stroke is selected, the difference between the maximum peak value and the minimum valley value of the energy in the time segment is extracted, the average value of the energy in the reference stable period (i.e. the period without punching) is determined, and the ratio of the two values is calculated, which is the time domain energy jump value of acoustic emission. The larger the value, the more likely there are crack propagation, micro-pore collapse or other irreversible instability phenomena in the material during the punching process, which is an important reference basis for identifying initial damage signals of the structure.

[0065] In the present application, the process of determining whether the condition of meeting the preset risk threshold is satisfied comprises selecting a corresponding determination mode according to the stamping rhythm stability feature in the multi-dimensional production behavior data set. The so-called stamping rhythm stability feature refers to the fluctuation law of the main peak frequency amplitude in the stamping frequency spectrum data collected in different stroke periods in the strapping cutting forming process. This feature reflects the consistency of impact energy in multiple continuous processing cycles. When the standard deviation of the main peak amplitude in the frequency spectrum in three consecutive stroke periods in the stamping frequency spectrum data is less than a first stability threshold, it indicates that the processing rhythm is stable, the vibration consistency is good, and there are few abnormal signals. At this time, the first determination mode is adopted. When the standard deviation is greater than or equal to the first stability threshold, it indicates that the stamping rhythm fluctuation is enhanced, which may be affected by the change of die gap, the slight difference of material thickness or the inconsistent transient response of equipment, and has shown signs of instability. Therefore, the second determination mode is adopted.

[0066] The first determination mode is suitable for processing batches with stable rhythm and is mainly used to identify potential abnormal heat aggregation and uneven energy distribution in the forming process. In this mode, if the frequency spectrum fluctuation rate in the stamping stroke is greater than a first threshold, and the pixel ratio of the abnormal heat distribution area is greater than a second threshold, it is determined that the condition of meeting the preset risk threshold is satisfied. Here, the frequency spectrum fluctuation rate in the stamping stroke is obtained by dividing the difference between the maximum and minimum frequency amplitudes in the same stroke segment by the average value, which reflects the fluctuation range of impact energy in one processing cycle. The pixel ratio of the abnormal heat distribution area is identified by taking three times the standard deviation as the threshold to identify the proportion of the high-temperature area in the image. The joint judgment of these two features helps to discover potential risk situations that exist local forming stress abnormalities or heat abnormal aggregation although the overall rhythm is stable. For example, in a certain batch of strapping production, although the device running rhythm has no obvious fluctuation, the local pressure is uneven due to the slight wear of the die edge, causing local temperature rise. This mode can identify such risks in time and trigger the subsequent process.

[0067] The second determination mode is suitable for batches with unstable rhythm and focuses on identifying deep structural changes caused by impact rhythm disorder. In this mode, if the acoustic emission time domain energy jump value is greater than a third threshold, or the deformation profile edge curvature change amount is greater than a fourth threshold, it is determined that the condition of meeting the preset risk threshold is satisfied, and the fracture risk assessment process is started. The acoustic emission time domain energy jump value is constructed based on the ratio of the energy peak value per unit time to the average value in the smooth period, which is used to detect the material fission response induced by rhythm abnormalities in the forming process. The deformation profile edge curvature change amount reveals the dramatic change of the strapping deformation boundary by calculating the difference of the local second-order derivatives of multiple edge points. Taking a batch as an example, if there is mechanical interference or unstable material feeding rhythm in the impact process, it may cause abnormal stress release and trigger the formation of micro-cracks. At this time, the acoustic emission signal and the edge curvature are abnormal synchronously, and the second determination mode can effectively respond and trigger the subsequent fracture risk assessment.

[0068] Through the division and setting of the two determination modes, not only the adaptation ability of the risk identification mechanism to different processing states is realized, but also the differentiated judgment logic of the risk triggering condition is enhanced, avoiding false positives or missed judgments caused by a single mode. In actual application, the mechanism can dynamically select the judgment basis path according to different stamping conditions, reflecting the engineering flexibility and accurate judgment of the present application in the fracture prediction logic, ensuring that the fracture risk assessment has high reliability and intelligent judgment ability under various manufacturing conditions.

[0069] The crack trend feature data set is constructed in the following manner: the micro-crack boundary continuous curvature value is obtained by equally sampling the edge pixel point sequence in the micro-deformation contour data, calculating the local curvature value of each point based on the three-point difference method, and obtaining the first derivative mean square deviation of all curvature values as the continuous curvature value; the strain gradient distribution value is obtained by constructing a difference matrix for each stress amplitude based on the multi-point stamping stress response curve extracted from the stamping frequency spectrum data, and taking the maximum change gradient as the distribution value; the residual thermal tension distribution value is obtained by converting the temperature difference between the thermal anomaly area and its boundary neighborhood in the local thermal imaging data into thermal tension value, fitting the tension distribution gradient in the form of Laplace transform on the two-dimensional pixel matrix, and taking the value of the maximum gradient point as the residual thermal tension distribution value.

[0070] Specifically, the extraction process of the micro-crack boundary continuous curvature value first relies on the micro-deformation contour data, which is collected by a non-contact high-precision surface scanning device after the edge contour image of the strap forming. The pixel point sequence represents the discrete representation of the edge curve of the strap. In actual processing, the edge curve is sampled at a fixed interval to form a set of discrete point columns with spatial continuity, and the three-point difference method is used to calculate the local curvature value by the coordinate difference value of each three adjacent points, reflecting the bending degree of the edge shape at this place. To measure the stability of the curvature change on the entire boundary, the curvature values of all points are calculated by the first derivative, and the micro-crack boundary continuous curvature value is established based on the mean square deviation. The larger the index is, the more frequent the edge fluctuation is, and there may be multiple micro-cracks or initial crack nucleation points.

[0071] The strain gradient distribution value is calculated based on the stamping spectrum data, which is the stress response of the strap material under impact load during the punching process, and is converted by spectrum to obtain the stress amplitude sequence corresponding to the processing time sequence. In constructing the strain gradient distribution value, first, the stamping stroke length is taken as the time or position dimension reference, and the entire stamping process is equally divided into several sampling nodes, and the stress response amplitude of each node is extracted; then a difference matrix is constructed, that is, the adjacent difference values of each point amplitude are arranged as a group of sequences, and the maximum gradient value is taken to quantify the mutation degree of stress response on the forming path. This index is used to identify whether there is strong strain concentration in the local area of the material, and is an important basis for judging whether the material structure enters the nonlinear response or is about to lose stability.

[0072] The residual thermal tension distribution value is obtained by converting the temperature difference between the thermal anomaly area identified in the local thermal imaging data and its boundary neighborhood. The local thermal imaging data is collected by an infrared thermal imaging device to obtain the thermal distribution image of the strap surface at the forming moment. The hot spot area in the image is usually related to local plastic strain or material friction. After identifying the thermal anomaly area, the average temperature difference between the area and its adjacent boundary area is extracted, and the temperature difference value is converted into a thermal tension distribution graph in tension units according to the thermodynamic model of the material. Subsequently, in the constructed two-dimensional pixel matrix, the Laplace transform method is applied to the tension image to perform second-order derivative enhancement, and the points with the most significant tension gradient change are extracted as the residual thermal tension distribution value, which reflects the strength of the hidden damage that the material may produce under the action of thermal force coupling. In some actual production scenarios, for example, when the mold and the material are not uniformly in contact, leading to the localization of edge heat input, the thermal tension distribution mutation point calculated by the Laplace transform often has a high degree of coincidence with the position where micro-cracks occur.

[0073] The three indicators of the crack tendency feature data set describe the geometric anomaly, mechanical strain aggregation and thermal damage behavior of the strap material in the punching forming process from the three dimensions of curvature topography, strain distribution and thermal tension characteristics, which together constitute a quantitative feature set reflecting the crack generation and expansion tendency, providing a reliable basis for subsequent crack potential index calculation and fracture risk assessment.

[0074] The structural stability feature data set is constructed by the following methods: the acoustic emission signal spectrum concentration is calculated by Fourier transform of the acoustic emission response data in the preset target frequency band, and the proportion of the energy integral value in this frequency band to the total energy in the full frequency domain; the thermal area centroid offset value is calculated by the centroid of the pixel set in the area where the temperature is greater than twice the standard deviation of the statistical average value in the local thermal imaging data, and the Euclidean distance between this centroid position and the center axis of the strap is taken as the offset value; the pressure release response delay time is directly calculated in milliseconds from the time difference between the end time of the stroke and the time of the first abnormal peak in the acoustic emission response data.

[0075] Specifically, the acoustic emission signal spectrum concentration degree refers to a parameter for measuring the concentration degree of acoustic emission energy in the frequency spectrum space. The acoustic emission response data is a high-frequency acoustic signal released when the instantaneous behavior of micro-crack propagation, interface slip or structure disintegration occurs inside the material during the strap cutting process. The signal is collected by an acoustic sensitive device and processed by frequency domain transformation. In specific implementation, the collected acoustic emission signal is input into the Fourier transform algorithm to obtain its frequency energy spectrum diagram, and a target frequency band (for example, thirty to sixty kilohertz) related to the micro-crack behavior of the material is selected as the energy attention interval. Then, the energy integral total value in the frequency band is calculated, and the ratio operation is performed with the energy integral total value in the full frequency domain (for example, zero to one hundred kilohertz) range. The obtained ratio value is the acoustic emission signal spectrum concentration degree. The value reflects whether the acoustic energy presents concentration enhancement, and usually shows obvious energy focusing phenomenon at the initial stage of structure instability or before the propagation of potential cracks. It is an important feature for identifying micro-scale discontinuity.

[0076] The thermal zone centroid offset value refers to the geometric distance between the geometric center of the high-temperature anomaly region identified in the local thermal imaging data and the axis of the strap structure. The local thermal imaging data is collected by an infrared thermal imaging device, and each pixel in the pixel matrix represents a temperature measurement point. In the process of constructing the offset value, first, the high-temperature identification threshold is set based on the statistical temperature parameters (mean value and standard deviation) of the image, and the region pixels with temperature greater than twice the average value of the standard deviation are usually taken as the thermal anomaly area. Second, the two-dimensional coordinates of all pixels in the region are weighted and averaged according to the heat distribution to obtain the position of the thermal zone geometric centroid. Finally, the Euclidean distance between the centroid point and the strap center axis (usually represented by the strap geometric center line or the thermal image symmetry axis) is calculated, and the distance is the thermal zone centroid offset value. The greater the value, the more significant the thermal anomaly deviates from the structure symmetry, which may be caused by local pressure unevenness, mold heating asymmetry, etc., reflecting the weakening trend of the structure thermal stability.

[0077] The pressure release response delay time refers to the time lag degree of the internal structure of the strap after the stress is conducted in the stamping forming process. The stamping spectrum data records the impact start and end time and energy release characteristics corresponding to each stroke in the processing process; the acoustic emission response data captures the fluctuation behavior when the internal structure adjustment or energy release occurs. In specific implementation, first, the end signal of the stroke in the stamping spectrum data is taken as the starting reference time; then the signal sequence after the stroke is extracted from the acoustic emission response data, and the time point corresponding to the first abnormal peak value exceeding the reference energy threshold is identified; the time difference between the two is the pressure release response delay time, which is in milliseconds. The time interval reflects the reaction time required for the internal structure to reach a non-steady state after the external impact stops. If the delay time is too long, it usually means that there are potential risk phenomena such as structural loosening and slow crack propagation in the material, which is an important delay response index for judging the stability of the structure.

[0078] The structure stability feature data set models the structural integrity and stable state of the strap in the die cutting forming process through three dimensions of spectrum energy distribution, thermal space offset and response time lag, can cover the energy concentration state, geometric offset state and response lag state before structural variation, constitutes the core risk factor group in the strap fracture prediction, and provides high reliability and quantifiable data support foundation for the calculation of the subsequent structure stability index.

[0079] The calculation steps of the crack potential length index are as follows:

[0080] The obtained micro-crack boundary continuous curvature value K, strain gradient distribution value E and residual thermal tension distribution value T in the crack trend feature data set are standardized to obtain the corresponding standardized values Standardization processing is used to eliminate the influence of physical dimension, scale difference and sample dispersion, so that different sources of features can be combined and modeled in a unified numerical space. The specific standardization method can be zero mean unit standard deviation processing, or using the maximum and minimum normalization method, introducing a local nonlinear amplification function to enhance the sensitivity of high risk values according to the image edge energy calculation principle, and defining three enhancement functions in turn , , :

[0081] The introduction of these cross terms is derived from the coupling mechanism of multi-physical field characteristics found in engineering practice, for example: thermal tension anomalies are usually accompanied by stress disturbance enhancement, edge curvature anomalies are more likely to trigger rapid expansion of cracks under local thermal field, therefore the introduction of cross terms helps to build a more realistic behavior prediction score model. The strain gradient distribution value adopts a logarithmic exponential mixed function, which is suitable for a wide dynamic range of strain gradient response. The square root enhances the sensitivity of small values and maintains the positive number characteristics, which adapts to the local dramatic change behavior of thermal tension gradient.

[0082] Sequentially calculate the synergistic cross terms 、 、 , which represents the covariant synergistic strength between different risk factors:

[0083] The introduction of these cross terms is derived from the coupling mechanism of multi-physical field characteristics found in engineering practice, for example: thermal tension anomalies are usually accompanied by stress disturbance enhancement, edge curvature anomalies are more likely to trigger rapid expansion of cracks under local thermal field, therefore the introduction of cross terms helps to build a more realistic behavior prediction score model.

[0084] A weighted comprehensive crack potential function value S is constructed, defined as: ; All are non-zero amplification coefficients set in advance according to experience, which can be adjusted according to the weight proportion of the actual application scene. The above function S aggregates single factor response and double factor coupling relationship, which is a unified expression score of crack potential risk behavior.

[0085] The crack potential function value S is input into the Logit function for normalized mapping to form the final crack potential index CGPI: .

[0086] The calculation of crack potential index adopts a comprehensive scoring method based on enhancement function and cross synergistic factor. The index is used to measure the potential crack development trend of the strap in the punching forming process due to the combined action of edge deformation anomaly, strain response mutation or thermal tension concentration and other factors. The core calculation idea is to highlight the response of high-risk samples through nonlinear enhancement mapping, and to introduce multi-factor synergistic effect expression model to express the implicit coupling behavior, and then output a risk score value with normalized characteristics as an important input basis for fracture risk assessment.

[0087] Firstly, three indicators that constitute the crack tendency feature dataset are extracted: micro-crack boundary continuous curvature value, strain gradient distribution value, and residual thermal tension distribution value. These three data are derived from micro deformation profile data, stamping frequency spectrum data, and local thermal imaging data, which have been extracted in the previous step. In order to facilitate the fusion processing between different physical dimension data, the three indicators are standardized respectively to unify their numerical scales, eliminate dimensional differences and extreme value interference, and make them suitable for combination in the same function model.

[0088] Next, the three standardized indicators are processed by enhancement functions. The design of the enhancement function is based on making the sample response more sensitive in the high-risk value area, thereby improving the model's ability to identify dangerous samples. The hyperbolic tangent square function is used for the enhancement function of the micro-crack boundary continuous curvature value, which quickly saturates and stabilizes the output when there is an edge mutation. The logarithmic exponential mixed function is used for the enhancement of the strain gradient distribution value, which is suitable for the characteristic of large fluctuation range in the medium and high interval. The absolute value square root function is used for the enhancement function of the residual thermal tension distribution value, which still has amplification response ability in the low amplitude stage. The introduction of these three types of enhancement functions constitutes the basis factor of the crack potential index.

[0089] On this basis, three groups of cross-collaboration factors are constructed to express the collaborative action characteristics between different risk sources. Specifically, the collaborative factor between the micro-crack boundary and the residual thermal tension is used to express the coupling effect of crack initial expansion and local thermal field; the collaborative factor between the micro-crack boundary and the strain gradient is used to capture the coupling effect of edge geometric change on material mechanical response; and the collaborative factor between the strain gradient and the residual thermal tension is used to describe the stress fluctuation and thermal concentration resonance phenomenon. The three collaborative factors are the product of two enhanced functions in calculation, reflecting the joint strength of nonlinear response between variables.

[0090] Subsequently, the three enhanced function values and the three collaborative factors are respectively weighted and combined to form a comprehensive score. In terms of weight setting, a set of non-zero coefficients is preset according to experimental verification and actual experience, corresponding to the influence strength of different cross factors. These weight coefficients do not require accurate learning, but are used to reflect the relative importance of risk transmission mechanisms between different physical paths. The comprehensive score value after weighting is defined as the crack potential function value, which is a one-time summary of the three types of crack inducements and their coupling relationships.

[0091] Finally, the crack potential function value is input into the normalized mapping function for output transformation. A normalized formula with monotonicity and boundary is adopted, so that the final crack potential index value is strictly limited between zero and one. The normalization process not only improves the interpretability of the score value, but also makes it more suitable as an input feature in a machine learning model or as a quantitative indicator in risk grading. Through the above construction method, the crack potential index can sensitively reflect the comprehensive influence of multiple physical factors on the development trend of microcracks, especially when multiple abnormal signals coexist, showing higher prediction accuracy, effectively improving the foresight, accuracy and practical value of the invention in the stainless steel strap fracture risk prediction task.

[0092] The calculation steps of the structural stability index are as follows:

[0093] The obtained acoustic emission signal spectrum concentration A, hot zone centroid offset H and pressure release response delay time D in the structural stability feature data set are standardized, respectively denoted as ; , which represents the historical mean and standard deviation of the corresponding index X, and then a disturbance synergy matrix M is constructed to obtain the following two-dimensional disturbance tensor: ;

[0094] Where each element represents the cooperative disturbance intensity between different features, the diagonal elements represent the self-disturbance variance of a single factor, and the non-diagonal elements represent cross-disturbance effects. The matrix Frobenius norm is used to measure the overall disturbance energy , and the calculation formula is: ;

[0095] A risk perception mapping function is constructed to output the structural stability index SIFI: ; is a preset amplification parameter to control the overall response rate, is a preset risk response nonlinear index.

[0096] The three indicators obtained in the structural stability feature data set are respectively the acoustic emission signal spectrum concentration denoted by the letter A, the hot zone centroid offset denoted by the letter H, and the delay time of the pressure release response denoted by the letter D. Standardization is performed in sequence. The purpose of standardization is to convert original data with different physical dimensions, scales and fluctuation ranges into data with comparable values in numerical space, so that abnormal values can be uniformly processed and eliminated in the subsequent modeling process. The standardization adopts the mean standard deviation normalization formula, that is, each index is subtracted by the historical sample mean and divided by the historical sample standard deviation, respectively denoted as the standardized , and .

[0097] where the historical mean represents the average level of the indicator in the historical normal or typical samples, and the standard deviation represents the fluctuation degree of the indicator in the sample population. The normalized values obtained by normalization , and reflect the relative abnormality degree of the current to-be-evaluated strap in each structural dimension.

[0098] Next, the three normalized values are constructed into a two-dimensional disturbance synergy matrix, denoted as M. The matrix is a three-by-three square tensor structure, where each element represents the product result between indicators to measure the synergy disturbance strength between them. The elements on the diagonal square, square, square respectively represent the self-disturbance strength of each indicator, i.e., the univariate self-disturbance variance; the elements on the non-diagonal line, such as and , and , and , respectively represent the cooperative change strength between different indicators, which is used to capture the cross-coupling effect between structural disturbance factors.

[0099] In actual scenarios, for example, when a batch of straps has a significant centroid shift in the hot zone due to asymmetric mold pressure, and the acoustic emission energy is highly concentrated during the impact response, the and product items in the matrix will show larger values, indicating that the batch of straps has a local thermal-acoustic abnormal coupling phenomenon, and the potential structural stability risk is significant.

[0100] After constructing the matrix M, to further extract the overall disturbance strength information, the matrix overall norm is introduced as a comprehensive measurement index, and the Frobenius norm is used for processing. The Frobenius norm is essentially the square root of the sum of squares of all elements in the matrix, which is used to measure the overall strength of the information contained in the entire matrix. In this calculation, the elements of the constructed matrix M are squared item by item, and the combination of all multiplication items and cross items is summed up to take the square root, and the result obtained is the structural disturbance energy Q.

[0101] The physical meaning of the disturbance energy Q is that it comprehensively reflects the structural stability disturbance strength of the current strap in three dimensions. The larger the value, the more unstable and coupled the structure is, and the higher the risk of crack evolution.

[0102] Finally, to map the perturbation energy Q into a risk index with interpretability and risk level, a risk perception mapping function is introduced to construct a structural stability index, denoted as SIFI. This function adopts an asymmetric amplification model, i.e., one minus an exponential function, to simulate the nonlinear trend of dramatic score fluctuations caused by small perturbations in high-risk areas. The function contains two preset parameters: the amplification parameter α, which controls the overall response rate, and the nonlinear index β, which regulates the steepness of the risk influence curve. Both parameters can be set according to specific production scenarios and risk sensitivity levels.

[0103] The final output of the structural stability index SIFI is a real number between zero and one. The closer the value is to one, the worse the structural stability, and the closer it is to zero, the more stable the structure of the strap. This index can be directly used as one of the input variables in the learning model or as a basis for risk classification to determine whether to reject the strap or use it in secondary application scenarios.

[0104] The learning model is a pre-trained machine learning classification model whose input is a two-dimensional feature vector composed of the crack potential index and the structural stability index. The model is trained using a training sample set based on historical stainless steel strap fracture data, which includes label information of actual fracture and corresponding crack potential index and structural stability index values.

[0105] The machine learning classification model uses a light gradient boosting tree model or a deep neural network model based on residual structure to establish the coupling relationship between crack development trend and structural stability through supervised learning. The model output is the fracture impact level prediction result, which is a continuous numerical score indicator denoted as fracture impact level score value, with a value range of 0 to 100, representing the comprehensive score of the possibility and severity of strap fracture.

[0106] The fracture impact level score value output by the model uses a skew probability density modeling method during training. Specifically, a lognormal distribution or Weibull distribution function is used to fit the risk level interval of historical fracture label samples, establishing an asymmetric mapping relationship between the score interval and the fracture possibility. Finally, it is used in different application scenarios or executes the rejection instruction according to the preset standard.

[0107] The learning model is a pre-trained machine learning classification model whose input is a two-dimensional feature vector composed of the crack potential index and the structural stability index. The crack potential index is used to express the potential trend strength of crack formation in the strap driven by multiple physical factors during the cutting and forming process, and the structural stability index is used to measure the concentration of perturbation in the structural response of the strap during the forming stage. Both of them together constitute a multi-dimensional numerical expression of the fracture risk.

[0108] The model is trained by a training sample set constructed based on historical stainless steel cable tie fracture data. The so-called training sample set refers to the actual cable tie fracture data collected in existing production batches, which contains label information of whether fracture occurs, and crack potential index and structure stability index values corresponding to each sample. By establishing a corresponding relationship between actual fracture occurrence and model input, the model can identify and fit the mapping path between input features and fracture risk.

[0109] The machine learning classification model adopts a light gradient boosting tree model or a deep neural network model based on residual structure. The light gradient boosting tree model is a high-efficiency classification algorithm based on gradient boosting decision tree mechanism, which is suitable for tasks with low feature dimension but complex discrimination boundary. The deep neural network model based on residual structure belongs to multi-layer neural structure, which can capture high-order nonlinear coupling relationship between input variables and has stronger feature extraction capability. In the present application, different model architectures can be selected for implementation according to data size, real-time requirement and deployment platform difference. The model establishes the coupling relationship between crack development trend and structure stability through supervised learning, i.e. learning the nonlinear function mapping between input and output under the guidance of existing labels.

[0110] The model output is the fracture influence degree grade prediction result, and the prediction result is a continuous numerical score index, denoted as fracture influence grade score value. The score value ranges from zero to one hundred, representing the comprehensive score of the possibility of cable tie fracture and the severity of consequences, and the higher the value, the higher the risk level. The score index has both probability interpretation and risk ordering ability, which is convenient for engineers to make quality judgment and use strategy decision for different batches of cable ties.

[0111] The fracture influence grade score value output by the model adopts a skew probability density modeling method in the training process, specifically using a lognormal distribution or Weibull distribution function to fit the risk level interval of historical fracture label samples. Lognormal distribution is suitable for reflecting the skewness characteristic that risk level is concentrated in low value area in most sample sets and a few high value samples have extreme risk; Weibull distribution is widely used in reliability analysis to model the distribution relationship between failure time, fracture probability and grade intensity, which can more accurately reflect the natural distribution characteristics of "high risk low frequency but high danger" samples in actual production.

[0112] By fitting the skewness density as described above, an asymmetric mapping relationship between the score interval and the fracture possibility is established, so that the model can not only output a single numerical value, but also provide a probabilistic understanding of the risk level corresponding to different score values. Finally, the model segments the score results according to the preset standard, and applies the cable to the preset different application scenarios or executes the rejection instruction. The preset standard can be set according to the industry safety requirements, such as classifying the cable with a score value below thirty into a low-risk level, which can be used in high-reliability situations such as aviation and military industry; the cable with a score value between thirty and seventy is considered to be in a medium-risk level, which can be used in general industrial bundling situations; and the cable with a score value higher than seventy is identified as a high-risk level, which should execute the rejection instruction and not enter the actual application link.

[0113] For example, in a certain military batch cable production task, through the fracture risk assessment model analysis, it is found that the fracture influence level score value of a batch of cables is concentrated in the interval of eighty-five to ninety, which is determined as a high-risk level. According to the model output result, this batch of cables is classified as a batch unsuitable for high-reliability tasks and is rejected, avoiding the potential risk of structural failure. This mechanism not only improves the intelligent level of risk identification, but also enhances the accuracy of cable classification and use in different application scenarios.

[0114] In the present application, the fracture influence degree level score value is taken as the quantitative output result of the cable fracture risk, with a value range of zero to one hundred. According to the severity of the risk level, the score interval is divided into three preset levels, each corresponding to a different application scenario or processing method, as follows:

[0115] The first type is the low-risk level (score value between zero and thirty), which indicates that the crack potential index and the structural stability index of the cable during the forming process are at a low level, and the model judges that the fracture risk is extremely low, with high stability and structural integrity. This type of cable is preferentially used in key fields with high reliability requirements, such as aviation equipment cable bundling, high-speed train sensor wiring fixation, satellite or rocket fairing accessory fixation, high-voltage transmission tower structural component pre-tightening device, and detachable cable group fastener in national defense equipment. Once the cable fracture occurs in such scenarios, the consequences may involve aircraft signal loss, electrical fire, or component shedding, so the lowest risk level cable must be selected to ensure overall operational safety.

[0116] The second type: medium risk level (score value between thirty and seventy), which indicates that the cable tie has certain use stability, but it shows slight abnormalities in some characteristic indicators, such as local heat concentration deviation or impact response slightly delayed, and there is a weak risk of fracture. This type of cable tie is suitable for regular industrial use or non-critical structure binding, such as urban bridge cable layout, industrial plant weak electricity construction binding, energy storage battery integrated structure boundary fixation, civilian vehicle cable auxiliary binding, commercial building pipeline auxiliary hanging structure, in these scenes, even if the cable tie breaks, it usually does not cause structural disaster, and only needs to be checked and replaced periodically, so it can accept slight risk.

[0117] The third type: high risk level (score value higher than seventy), which indicates that the cable tie has significant potential crack propagation ability or structural instability, such as micro-crack boundary curvature mutation, heat tension concentration extreme or frequent acoustic emission abnormalities, and the model assesses that the fracture risk is extremely high. This type of cable tie is not suitable for any practical application scene, and needs to be rejected at the factory stage, including: refusing to enter the warehouse stage, marking as unqualified products for scrap or repair, prohibiting the user delivery link, and being included in the quality traceability and improvement record system.

[0118] For example, in a batch production of industrial cable ties, if the model evaluation finds that the unstable stamping rhythm of a batch of cable ties causes multiple characteristic parameter abnormalities, and the final score value reaches eighty-two, according to the preset rules, the batch of cable ties is identified as high-risk products, and is rejected in whole before packaging, effectively avoiding the fracture accident during the service period caused by "hidden cracks".

[0119] The above formulas are dimensionless values, and the formulas are obtained by software simulation of a large amount of data to obtain a formula closest to the actual situation. The preset parameters in the formula are set by those skilled in the art according to the actual situation.

[0120] It should be understood that the size of the sequence number of the above processes in various embodiments of the present application does not mean the order of execution, and the execution order of the processes should be determined according to their functions and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0121] Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0122] Those skilled in the art can clearly understand the specific working process of the foregoing described device and unit by referring to the corresponding process in the foregoing method embodiments, and thus will not be described here again.

[0123] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for predicting the breakage of stainless steel cable ties, characterized in that, Includes the following steps: During the cable tie punching process, multi-source raw data related to breakage risk were collected from continuous production batches, and a multi-dimensional production behavior data set with time series correlation was constructed. Feature extraction is performed on the multidimensional production behavior data set to form the basis for determining whether to initiate a fracture risk assessment. The basis is the multidimensional feature fluctuation judgment basis, which specifically includes the spectral fluctuation rate within the stamping stroke, the change in curvature of the deformation contour edge, the pixel ratio of the abnormal heat distribution area, and the acoustic emission time-domain energy jump value. When the judgment criteria meet the preset risk threshold conditions, the fracture risk assessment process is initiated, and two sets of characteristic data are constructed. The first set is the crack trend characteristic data set, which consists of the continuous curvature value of the microcrack boundary, the strain gradient distribution value, and the residual thermal tension distribution value. The second set is the structural stability characteristic data set, which consists of the acoustic emission signal spectrum concentration, the thermal zone centroid offset value, and the pressure release response delay time. Crack potential length index was calculated using crack trend characteristic data set, and structural stability index was calculated using structural stability characteristic data set. Both indices were used as input variables for fracture risk assessment. The crack potential index and structural stability index are input into a pre-trained learning model, which outputs the fracture impact level. Based on the fracture impact level, the stainless steel cable ties are graded and processed to determine whether they are used in different preset application scenarios or to execute rejection instructions.

2. The method for predicting the breakage of stainless steel cable ties according to claim 1, characterized in that, The raw data includes: stamping spectrum data reflecting the stamping mechanical process, micro-deformation profile data characterizing the micro-deformation state of the cable tie surface, local thermal imaging data for identifying thermal anomaly regions, and acoustic emission response data for sensing internal fission activity of the material.

3. The method for predicting the breakage of stainless steel cable ties according to claim 2, characterized in that, In the multidimensional characteristic fluctuation judgment criteria, the spectral fluctuation rate within the stamping stroke is obtained by dividing the difference between the maximum and minimum frequency amplitudes within the same stroke segment in the stamping spectral data by the average frequency amplitude; the change in curvature of the deformed contour edge is calculated by the difference in the second derivative values ​​of multiple equidistant points at the edge of the cable tie in the micro-deformed contour data; and the pixel proportion of the thermal distribution anomaly area is calculated by the ratio of the number of pixels in the area where the temperature is three times the standard deviation of the ambient temperature to the total number of pixels in the image in the local thermal imaging data. The time-domain energy jump value of acoustic emission is calculated by the ratio of the energy peak difference per unit time to the average energy value during the reference stationary period in the acoustic emission response data.

4. The method for predicting the breakage of stainless steel cable ties according to claim 3, characterized in that, The process of determining whether the preset risk threshold conditions are met includes selecting the corresponding determination mode based on the stability characteristics of the stamping rhythm in the multi-dimensional production behavior data set. When the standard deviation of the amplitude of the main peak of the spectrum in three consecutive stroke cycles in the stamping spectrum data is less than the first stability threshold, the first determination mode is adopted; when the standard deviation is greater than or equal to the first stability threshold, the second determination mode is adopted. In the first judgment mode, if the spectral fluctuation rate within the stamping stroke is greater than the first threshold and the proportion of pixels in the abnormal heat distribution area is greater than the second threshold, then the preset risk threshold condition is met. In the second judgment mode, if the acoustic emission time-domain energy jump value is greater than the third threshold, or the change in curvature of the deformed contour edge is greater than the fourth threshold, then the preset risk threshold condition is met, and the fracture risk assessment process is initiated.

5. The method for predicting the breakage of stainless steel cable ties according to claim 4, characterized in that, The crack trend feature data set was constructed in the following way: The continuous curvature value of the microcrack boundary is obtained by sampling the edge pixel sequence in the micro deformation profile data at equal intervals, calculating the local curvature value of each point based on the three-point difference method, and obtaining the root mean square error of the first derivative of all curvature values ​​as the continuous curvature value. The strain gradient distribution value is obtained from the multi-point stamping stress response curve extracted from the stamping spectrum data. Based on the stamping stroke length, a difference matrix is ​​constructed for the stress amplitude at each point, and the maximum change gradient is taken as the distribution value. The residual thermal tension distribution value is obtained by converting the temperature difference between the thermal anomaly area identified in the local thermal imaging data and its boundary neighborhood into a thermal tension value, fitting the tension distribution gradient in the form of Laplace transform on the two-dimensional pixel matrix, and taking the value of the maximum gradient point as the residual thermal tension distribution value.

6. The method for predicting the breakage of stainless steel cable ties according to claim 5, characterized in that, The structural stability feature dataset was constructed in the following way: The concentration of the acoustic emission signal spectrum is calculated by performing a Fourier transform on the acoustic emission response data within a preset target frequency band, and then calculating the proportion of the energy integral value within that frequency band to the total energy in the entire frequency domain. The centroid offset value of the hot zone is calculated by taking the centroid of the pixel set in the local thermal imaging data where the temperature is greater than twice the standard deviation of the statistical average, and taking the Euclidean distance between the centroid position and the central axis of the cable tie as the offset value. The pressure release response delay time is directly calculated in milliseconds from the time difference between the end of the stroke in the ramming spectrum data and the first abnormal peak time in the acoustic emission response data.

7. The method for predicting the breakage of stainless steel cable ties according to claim 6, characterized in that, The steps for calculating the crack potential index are as follows: The microcrack boundary continuity curvature value K, strain gradient distribution value E, and residual thermal tension distribution value T obtained from the crack trend characteristic data set are standardized to obtain the corresponding standardized values. Based on the principle of image edge energy calculation, a local nonlinear amplification function is introduced to enhance the sensitivity of high-risk values. Three enhancement functions are defined sequentially. , , : ； Calculate the cooperative cross terms sequentially , , This indicates the strength of covariation synergy among different risk factors: ； The weighted composite crack potential function value S is defined as follows: ; All are non-zero amplification factors pre-set based on experience; The crack potential length function value S is input into the Logit function for normalization mapping to form the final crack potential length index CGPI: .

8. The method for predicting the breakage of stainless steel cable ties according to claim 7, characterized in that, The calculation steps for the structural stability index are as follows: The acoustic emission signal spectral concentration A, the hot zone centroid offset H, and the pressure release response delay time D obtained in the structural stability characteristic data set are standardized and denoted as follows: ; Let X represent the historical mean and standard deviation of the corresponding index X. Then, construct the perturbation coherence matrix M to obtain the following two-dimensional perturbation tensor: ; Each element represents the cooperative perturbation strength among different features. The diagonal elements represent the variance of the individual factor's self-perturbation, and the off-diagonal elements represent the cross-cooperative effects. The Frobenius norm of the matrix is ​​used to measure the overall perturbation energy. The calculation formula is: ； Construct a risk perception mapping function to output the structural stability index SIFI: ; The preset amplification parameters are used to control the overall response rate. This is a preset nonlinear index for risk response.

9. The method for predicting the breakage of stainless steel cable ties according to claim 8, characterized in that, The learning model is a pre-trained machine learning classification model. Its input is a two-dimensional feature vector composed of crack potential length index and structural stability index. The model is trained by a training sample set constructed based on historical stainless steel cable tie fracture data. The training sample set contains the label information of the actual fracture situation and the corresponding crack potential length index and structural stability index values. The machine learning classification model adopts a light gradient boosting tree model or a deep neural network model based on residual structure. The coupling relationship between crack development trend and structural stability is established through supervised learning. The model output is the fracture impact level prediction result, which is a continuous numerical score index, denoted as the fracture impact level score value. The value range is 0 to 100, representing the comprehensive score of the possibility of cable tie breakage and the severity of the consequences. The fracture impact level score output by the model is modeled using a skewed probability density method during training. Specifically, a log-normal distribution or Weibull distribution function is used to fit the risk level range of historical fracture label samples to construct an asymmetric mapping relationship between the score range and the fracture probability. Finally, the score is applied to different preset application scenarios or removed according to preset standards.

Images