Wire drawing die abrasion loss online judgment method and wire drawing machine

Abstract

The invention is suitable for the technical field of data processing for judging the abrasion loss of a die, and particularly relates to a wire-drawing die abrasion loss online judgment method and a wire drawing machine. The method comprises the steps that the current drawing force, the current die temperature, the current drawing speed and the area reduction rate are obtained; according to the current drawing speed and the area reduction rate, the ideal drawing force and the ideal mold temperature of the current period are calculated through digital twin bodies in the wire drawing process; determining a current equivalent friction coefficient based on the current drawing force, the current mold temperature, the ideal drawing force and the ideal mold temperature; based on the current equivalent friction coefficient, the current abrasion loss is determined, and the current hole diameter of the wire drawing die is calculated according to the current abrasion loss; and based on the current equivalent friction coefficient, predicting the service life of the wire-drawing die to obtain the residual service life of the wire-drawing die. According to the method, the abnormal wear state of the mold is sensed in advance through the change of the equivalent friction coefficient, and early recognition and prediction and early warning of wear are realized.

Description

Technical Field

[0001] This application belongs to the field of data processing technology for determining the wear of molds, and particularly relates to an online method for determining the wear of wire drawing dies and a wire drawing machine. Background Technology

[0002] In modern high-speed precision wire drawing production, die wear directly affects the dimensional stability and surface quality of the product. Therefore, real-time monitoring of the die wear status is crucial to ensuring production efficiency and quality. Traditional manual die disassembly and inspection methods suffer from long cycles, large errors, and easy die damage, making them unsuitable for continuous production. To address this, online inspection technology has emerged. By using high-precision sensing devices such as laser diameter gauges to monitor the wire diameter in real time, the wear trend of the die holes can be indirectly determined.

[0003] In existing technologies, online inspection methods rely on product quality inspection results, such as monitoring defects like wire diameter deviations and surface scratches to determine if there are problems with the mold. When wire diameter deviations or obvious surface scratches can be consistently detected, the mold wear is already quite severe, and a certain number of defective products may have already been produced.

[0004] In summary, when inspecting the wear condition of wire drawing dies, there is a problem that the wear condition may not be detected in a timely manner due to the lag in detection methods. Summary of the Invention

[0005] This application provides an online method for determining the wear of wire drawing dies and a wire drawing machine, which can solve the problem in related technologies where the wear condition of wire drawing dies is not detected in a timely manner due to the lag in detection methods.

[0006] In a first aspect, embodiments of this application provide an online method for determining the wear amount of a wire drawing die, including: The current drawing force, current die temperature, current drawing speed, and reduction rate are obtained; wherein, the current drawing force, current die temperature, and current drawing speed are all actual values ​​for the current cycle, and the reduction rate is used to characterize the ratio of the reduction in drawing area to the die area; Based on the current drawing speed and the reduction rate, the ideal drawing force and ideal die temperature for the current cycle are calculated using a digital twin of the drawing process; wherein, the digital twin includes a physical model, which includes a drawing mechanics model and a thermodynamic model; The current equivalent coefficient of friction is determined based on the current drawing force and the current die temperature, as well as the ideal drawing force and the ideal die temperature. Based on the current equivalent friction coefficient, the current wear amount is determined, and based on the current wear amount, the current aperture of the wire drawing die is calculated; wherein, the current wear amount is the aperture increment of the wire drawing die in the current cycle; Based on the current equivalent friction coefficient, the lifespan of the wire drawing die is predicted to obtain the remaining lifespan of the wire drawing die.

[0007] The technical solutions described in this application embodiment have at least the following technical effects: The online method for determining the wear of wire drawing dies provided in this application first obtains the current drawing force, current die temperature, current drawing speed, and reduction ratio (used to characterize the ratio of the reduction in drawing area to the die area) to determine the actual drawing force, actual die temperature, actual drawing speed, and reduction ratio for the current cycle. Then, based on the current drawing speed and reduction ratio, the ideal drawing force and ideal die temperature for the current cycle are calculated using the drawing mechanics and thermodynamics models of the digital twin of the wire drawing process. Next, based on the current drawing force and current die temperature, as well as the ideal drawing force and ideal die temperature, the current equivalent friction coefficient is determined. Then, based on the current equivalent friction coefficient, the current wear amount (used to characterize the aperture increment of the wire drawing die in the current cycle) is determined, and based on the current wear amount, the current aperture of the wire drawing die is calculated. Finally, based on the current equivalent friction coefficient, the life of the wire drawing die is predicted to obtain the remaining life of the wire drawing die. This method compares and analyzes actual data with digital twin simulation results, enabling early detection of abnormal mold wear conditions through changes in the equivalent friction coefficient. This allows for early identification and prediction of wear, overcoming the limitations of traditional post-detection methods and transforming them into proactive prediction and real-time sensing. This effectively avoids quality losses and downtime risks caused by untimely detection of wear conditions. Furthermore, this method dynamically assesses the remaining mold lifespan using the equivalent friction coefficient, allowing mold maintenance or replacement to be based on remaining lifespan rather than a fixed cycle. This optimizes mold utilization efficiency and reduces maintenance costs and downtime risks.

[0008] Secondly, embodiments of this application provide an online device for determining the wear amount of a wire drawing die, comprising: The acquisition unit is used to acquire the current drawing force, current die temperature, current drawing speed, and reduction rate; wherein the current drawing force, current die temperature, and current drawing speed are all actual values ​​for the current cycle, and the reduction rate is used to characterize the ratio of the reduction in drawing area to the die area. An ideal value calculation unit is used to calculate the ideal drawing force and ideal die temperature for the current cycle based on the current drawing speed and the reduction rate, using a digital twin of the drawing process; wherein, the digital twin includes a physical model, and the physical model includes a drawing mechanics model and a thermodynamic model; An equivalent friction coefficient determination unit is used to determine the current equivalent friction coefficient based on the current drawing force and the current die temperature, as well as the ideal drawing force and the ideal die temperature. The wear calculation unit is used to determine the current wear amount based on the current equivalent friction coefficient, and to calculate the current aperture of the wire drawing die based on the current wear amount; wherein, the current wear amount is the aperture increment of the wire drawing die in the current cycle; The life prediction unit is used to predict the life of the wire drawing die based on the current equivalent friction coefficient, and obtain the remaining life of the wire drawing die.

[0009] Thirdly, embodiments of this application provide a wire drawing machine, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method described in any of the embodiments of the first aspect.

[0010] It is understood that the beneficial effects of the second and third aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a schematic flowchart of an embodiment of the online determination method for wire drawing die wear provided in this application; Figure 2 This is a schematic diagram illustrating the implementation process of calculating the equivalent friction coefficient for each cycle in the online determination method for wire drawing die wear provided in this application embodiment; Figure 3 This is an example diagram showing typical values ​​of the strength coefficient K and strain hardening index n in the online determination method for the wear amount of wire drawing dies provided in the embodiments of this application. Detailed Implementation

[0013] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0014] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0015] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0016] In related technologies, the wear of the die during wire drawing is a gradual process. In the early stages of wear, it does not immediately cause obvious product quality problems, such as wire diameter deviation or surface defects, so the wear is not detected in time before it has a significant impact. Traditional detection methods, such as relying on the quality inspection of later products (e.g., surface defects or dimensional deviations of finished wires), usually only discover problems when the wear has become quite severe.

[0017] Traditional online inspection methods rely on product quality inspection results, such as monitoring defects like wire diameter deviations and surface scratches to determine if there are problems with the mold. However, this method is more about post-event confirmation than pre-event prediction. By the time obvious defects appear, the problem has already occurred, and timely corrective measures cannot be taken. This post-event confirmation approach cannot provide early warnings and adjustments, thus losing the opportunity to detect and resolve problems early, leading to higher production costs, quality losses, and potential production downtime for maintenance.

[0018] When severe wear or mold failure occurs during the wire drawing process, machine shutdown for repair or mold replacement is necessary. Unexpected maintenance downtime can disrupt production plans and delay delivery times. Furthermore, if mold wear is not detected in time, it means that molds that are no longer in optimal condition are being used during production, which will affect production efficiency and product quality, and in the long run, significantly increase overall production costs.

[0019] To address the aforementioned issues, this application provides an online method for determining the wear of a wire drawing die and a wire drawing machine. The method first obtains the current drawing force, current die temperature, current drawing speed, and reduction ratio (characterized by the ratio of the reduction in drawing area to the die area) to determine the actual drawing force, actual die temperature, actual drawing speed, and reduction ratio for the current cycle. Then, based on the current drawing speed and reduction ratio, the ideal drawing force and ideal die temperature for the current cycle are calculated using a digital twin model of the wire drawing process's drawing mechanics and thermodynamics. Next, based on the current drawing force and current die temperature, as well as the ideal drawing force and ideal die temperature, the current equivalent friction coefficient is determined. Then, based on the current equivalent friction coefficient, the current wear amount (characterized by the increase in the die's aperture in the current cycle) is determined, and based on the current wear amount, the current aperture of the wire drawing die is calculated. Finally, based on the current equivalent friction coefficient, the lifespan of the wire drawing die is predicted to obtain the remaining lifespan of the wire drawing die. This method compares and analyzes actual data with digital twin simulation results, enabling early detection of abnormal die wear by monitoring changes in the equivalent friction coefficient. This allows for early identification and prediction of wear, overcoming the limitations of traditional post-detection methods and shifting to pre-detection and real-time sensing. This effectively avoids quality losses and downtime risks caused by untimely detection of wear. The method dynamically assesses the remaining die life using the equivalent friction coefficient, allowing die maintenance or replacement based on remaining life rather than fixed cycles. This optimizes die utilization efficiency and reduces maintenance costs and downtime risks. Furthermore, by establishing ideal models of drawing force and die temperature, the method allows for comparative analysis of actual parameters with ideal conditions in each cycle. This not only improves the control precision of the wire drawing process but also enables process optimization and adjustment through understanding the interrelationships of parameters, ensuring long-term consistency and stability of key indicators such as wire diameter and surface quality.

[0020] The online determination method for the wear of wire drawing dies provided in this application embodiment can be applied to a wire drawing machine. In this case, the wire drawing machine is the executing entity of the online determination method for the wear of wire drawing dies provided in this application embodiment. This application embodiment does not impose any restrictions on the specific type of wire drawing machine.

[0021] For example, a wire drawing machine may include a drawing force acquisition device, a temperature acquisition device, a drawing speed acquisition device, and a control device communicatively connected to the drawing force acquisition device, the temperature acquisition device, and the drawing speed acquisition device. The drawing force acquisition device is capable of acquiring the actual drawing force for each cycle, and may be a force sensor, tension sensor, etc.; the temperature acquisition device is capable of acquiring the actual temperature of the area where the wire drawing die sizing strip is located for each cycle, and may be a non-contact infrared thermometer, embedded thermocouple, etc.; the drawing speed acquisition device is capable of acquiring the drawing speed for each cycle, and may be an encoder, motor drive feedback (servo drive feedback), etc.; the control device is capable of data processing and controlling the drawing force acquisition device, the temperature acquisition device, and the drawing speed acquisition device, and may be an industrial control computer (IPC), a PLC and host computer (such as a SCADA system, industrial PC, etc.), an edge computing device (Edge AI gateway), etc.

[0022] To better understand the online determination method for wire drawing die wear provided in the embodiments of this application, the specific implementation process of the online determination method for wire drawing die wear provided in the embodiments of this application will be described by way of example below.

[0023] Figure 1 This paper presents a schematic flowchart of an online method for determining the wear of wire drawing dies according to an embodiment of this application. The online method for determining the wear of wire drawing dies includes: S100 obtains the current drawing force, current die temperature, current drawing speed, and reduction ratio. Among them, the current drawing force, current die temperature, and current drawing speed are all actual values ​​for the current cycle, and the reduction ratio is used to characterize the ratio of the reduction in drawing area to the die area.

[0024] As can be understood, drawing force refers to the pulling force experienced by the metal wire as it passes through the die during the drawing process, reflecting the magnitude of processing resistance. Die temperature reflects the friction and heat conduction between the die and the wire, and is a parameter for judging die wear and temperature rise effects. Drawing speed refers to the speed at which the wire passes through the die, and is closely related to motor output and take-up speed; it is a dynamic parameter characterizing the process rhythm. Reduction rate is a calculated value used to reflect the degree of reduction in cross-sectional area of ​​the wire as it passes through the die; it is an indicator for judging the degree and intensity of deformation.

[0025] For example, the current pulling force on the wire can be detected in real time by a pulling force acquisition device (such as a force sensor or tension sensor), and the mechanical pulling force signal can be converted into an electrical signal (analog or digital) and transmitted to the control system. The sensor can be installed at the inlet or outlet of the drawing die.

[0026] Temperature acquisition devices (such as non-contact infrared thermometers, embedded thermocouples, etc.) can be used to collect the current mold temperature, and the control system can read the actual temperature value in each cycle at regular intervals. A non-contact infrared thermometer can be used to collect the mold temperature at the mold exit or 2-3 cm after the wire exits the mold; a K-type or T-type thermocouple head can be embedded in the mold body (near the sizing section) to collect the mold temperature.

[0027] If the drawing speed acquisition device is an encoder, the encoder can be installed on the main motor or the take-up mechanism to output pulse signals. The control system calculates the current drawing speed based on the number of pulses per unit time. If it is a wire drawing machine controlled by a servo motor, the actual speed value can be read directly from the driver.

[0028] The formula for calculating the reduction rate is: ,in, Indicates the reduction rate. Indicates the cross-sectional area of ​​the die. Indicates the initial die diameter. Indicate the cross-sectional area of ​​the mold. This indicates the initial die exit diameter. The initial die entry and initial die exit diameters can be the factory calibration values ​​of the wire drawing die or obtained through manual measurement. The subsequent half-cone angle and sizing section length can also be obtained using the same method.

[0029] All data is sampled synchronously, allowing for a unified timestamp to ensure periodic consistency across data points. The acquired raw signal can be denoised (e.g., using Kalman filtering or moving average) to remove outliers, thus ensuring reliable data quality.

[0030] This step enables the real-time and high-precision acquisition and calculation of key parameters without disrupting production, providing a solid data foundation for subsequent digital twin analysis, friction condition assessment, wear calculation, and mold life prediction.

[0031] S200 calculates the ideal drawing force and ideal die temperature for the current cycle using a digital twin of the wire drawing process, based on the current drawing speed and reduction rate. The digital twin includes a physical model, which comprises a drawing mechanics model and a thermodynamic model.

[0032] As can be understood, a digital twin of the wire drawing process is a high-fidelity model of the physical wire drawing process in virtual space. Its core consists of two coupled physical models: a drawing mechanics model that predicts deformation and frictional resistance under different reduction rates; and a thermodynamic model that estimates the temperature rise caused by plastic deformation and friction. Before actual operation, the digital twin is trained and calibrated based on extensive experimental data and theoretical calculations, exhibiting strong generalization capabilities.

[0033] For example, a digital twin can be deployed in a control device (such as a PLC, edge server, or industrial computer) to continuously receive real-time input. The digital twin can pre-load a physical model and, based on the current cycle's drawing speed (i.e., the current drawing speed) and reduction rate, as well as the equivalent friction coefficient of the previous cycle, complete the deduction within milliseconds, returning the ideal drawing force and ideal die temperature for the current cycle, providing a benchmark for subsequent comparative analysis. When using the physical model for the first time, an initial equivalent friction coefficient is used.

[0034] The drawing mechanics model construction process is as follows: It can be assumed that the wire follows ideal plastic flow laws during the drawing process; within a single cycle, the drawing speed and friction state remain stable; the die geometry is an axisymmetric conical shape, including a cone segment in the deformation zone (half-cone angle is...). ) and sizing section (sizing section length is The friction model can adopt Coulomb's friction law, where frictional stress is proportional to normal force; the material deformation is constant strain flow, and the flow stress can be represented by the average value (i.e., average flow stress).

[0035] In the metal drawing process, the drawing force mainly overcomes the material deformation resistance and frictional resistance. The resultant force of the material deformation resistance and frictional resistance is the ideal drawing force. The frictional resistance includes the frictional resistance of the tapered section of the die and the frictional resistance of the sizing section.

[0036] Material deformation resistance: We can assume there is no frictional resistance during the drawing process; then, the material deformation resistance at this point is the drawing force. Metal materials deform under the action of the drawing force, but due to the incompressibility of metals, volume conservation is satisfied, i.e. ,in, Indicates the wire feed length into the die. This indicates the wire's exit length from the die. The definition of true strain is... ,in, Let L represent the original length and L represent the final length. During the drawing process, the wire is stretched, and its length increases. According to the law of conservation of volume, the equivalent strain can be expressed as: ,in, It indicates equivalent change.

[0037] Plastic deformation work refers to the integral of the work done by all the tiny units in the entire plastic volume, i.e. ,in, It represents the work done by plastic deformation (i.e., the work done by the resistance to material deformation during the drawing process). Let V represent the average flow stress and V represent the volume of the deformation zone. The volume of the deformation zone can be approximated as the product of the outlet area and the deformation length; therefore, the work of plastic deformation can be expressed as... ,in, This indicates the length of the deformation zone. Since the work done by the resistance to material deformation is the same as the work done by the pulling force, i.e. Where F represents the pull-out force, Indicates the resistance to material deformation, due to Then the resistance to material deformation .

[0038] According to Hollomon's constitutive equation ( Where K represents the strength coefficient. The strain hardening index is represented as (0~1), and the relationship between the mean flow stress and the equivalent strain is as follows: The values ​​of K and n can be found in literature, GB standards, or MatWeb and material databases, depending on the type of material used (e.g., Q195, Q235, T8, etc.). For example, typical values ​​of K and n are as follows: Figure 3 As shown.

[0039] The stress-strain curve can be measured by tensile testing, and K and strain can be fitted. The value of can be used to perform a standard tensile test on the material to be stretched to obtain a series of stress-strain data. In a program (such as Excel, MATLAB, Python), input all stress-strain data and apply it through the linear form of the Hollomon constitutive equation: A linear regression was performed on the stress-strain data, and the slope of the fitted regression line was... The intercept is K is obtained by taking the exponent.

[0040] Frictional resistance: Conical segment frictional resistance: In the Avitzur model, the work per unit volume includes the work of ideal plastic deformation, frictional work, and redundant work, i.e. The Avitzur model incorporates two sources of power consumption: ideal plastic deformation work (from the uniform plastic flow of the material in the cone segment) and cone segment friction work (from the friction of the material on the surface of the deformation zone, introduced through projection). The Avitzur model does not include friction in the sizing zone because it assumes an ideal die, and the material deforms in the conical section of the deformation zone. The frictional work can be extracted from the Avitzur model as the work done by the frictional resistance per unit volume of the conical section. After deformation through the die, the volume of the material is... Then the total frictional work of the cone segment in the deformation zone is ,in, This represents the coefficient of friction. Since the total frictional work of the cone segment in the deformation zone is the product of the frictional resistance of the cone segment and the deformation length, i.e. Then the frictional resistance of the cone segment is .

[0041] Frictional resistance in the sizing section: according to Coulomb's law of friction (in, (where p represents the frictional force per unit area and p represents the normal pressure), the frictional resistance of the sizing section can be calculated, i.e. ,in, , This represents the cross-sectional area of ​​the fixed-diameter section. This indicates the frictional resistance of the fixed-sizing section.

[0042] When the material is in contact with the mold (i.e., pressed against the mold wall), under axisymmetric conditions, the material is subjected to normal compressive force and tangential frictional force on the outer wall. According to plasticity analysis (von Mises yield criterion, stress tensor transformation, etc.), the normal pressure can be expressed as the average stress of the material: Then the frictional resistance of the sizing section can be expressed as: .

[0043] By summing the above-mentioned material deformation resistance, tapered section friction resistance, and sizing section friction resistance, we obtain the drawing mechanics model: ,in, This represents the ideal pull-out force. This is a theoretical input, where the mold and wire satisfy Coulomb's law of friction. However, in reality, friction is affected not only by the material but also by surface roughness, lubricant, contact pressure, temperature, surface hardening / wear, etc. These factors cannot be written into the model item by item, so an equivalent friction coefficient can be used. replace This makes the model more closely resemble the real system.

[0044] Thermodynamic model construction process: The total heat source is the heat of plastic deformation and frictional heat. The heat of plastic deformation is the conversion of work done during plastic deformation into heat, i.e. ,in, Represents the heat of plastic deformation. This indicates the plastic work-heat conversion efficiency (e.g., 0.9~0.95). Indicates the pulling speed.

[0045] Frictional heat is the conversion of frictional work into heat, that is... ,in, Indicates frictional heat. Represents normal force, , use Replacement.

[0046] The mold transfers heat to the coolant (water or oil) through contact cooling. ,in, Indicates the amount of heat dissipated during cooling. Indicates the heat transfer coefficient between the mold and the coolant. Indicates the cooling contact area. Indicates the ideal mold temperature. This indicates the temperature of the coolant.

[0047] When the mold reaches thermal stability (the temperature no longer rises), then Then the thermodynamic model can be expressed as ,because , , These are unmeasurable parameters; unmeasurable parameters in a thermodynamic model can be used as parameters to be estimated. ,in, express , express .

[0048] It can collect multiple sets (N sets, i=1,2,...,N) of experimental data, including actual pull-out force. Actual mold temperature Pulling speed The formula for calculating the ideal mold temperature for each set of experimental data is as follows: ,in, .

[0049] The sum of squared errors between the ideal mold temperature and the actual mold temperature can be used as the objective function, i.e. The goal is to simultaneously minimize the objective function J on all experimental data, thereby obtaining... , , The optimal value.

[0050] During the first fitting process, for By setting a fixed value, the result can be calculated for each set of experimental data. We can assume that the linear model is... Based on the experimental data of each group and By fitting a linear model using the linear least squares method, we obtain... , .

[0051] In the present , , Next, calculate the ideal mold temperature for each set of experimental data. The function value is calculated using the objective function J, and continuously updated through scanning or one-dimensional optimization. The optimal equivalent friction coefficient (i.e., the initial equivalent friction coefficient) is obtained by minimizing the function value.

[0052] Scanning method: can be preset Reasonable range of values For example, [0.02, 0.25] sets a scan step size. (e.g., 0.005 or 0.01), generating a series of candidate values ​​within the range. k=1,2,...,n Calculate each corresponding .

[0053] For each Based on the experimental data of each group and By fitting a linear model using the linear least squares method, each... corresponding , According to each and its corresponding , Calculate each Ideal mold temperature for each set of experimental data .according to and Each The corresponding function value.

[0054] Compare each The corresponding function value, select the value corresponding to the minimum function value. The optimal equivalent friction coefficient, and the corresponding value of the optimal equivalent friction coefficient. , For optimal , .

[0055] One-dimensional optimization algorithm: It can utilize any one-dimensional unconstrained optimization method to minimize the objective function J, updating it each time. Internally, it performs the aforementioned linear least squares method to fit the linear model and calculate the function value of the objective function J, and returns the function value. The optimization algorithm updates the algorithm based on the function value. .

[0056] The optimal , , Substitution This allows us to obtain the final thermodynamic model. Similarly, the optimal... Substituting into the above pull-out mechanics model, we obtain the final pull-out mechanics model.

[0057] The physical model, based on the theories of plastic deformation of materials, tribology, and the principle of energy conservation, uses analytical mathematical expressions to uniformly describe the drawing force, deformation work, friction work, and heat exchange process between the die and the cooling system. This establishes a clear mechanistic relationship between the main physical quantities in the drawing process. With the help of the physical model, ideal drawing force and ideal die temperature can be predicted directly based on parameters without requiring extensive experiments or empirical parameters, thus significantly improving the accuracy and generalization ability of process analysis and prediction.

[0058] In actual production, the measured drawing force or die temperature may deviate from the theoretical value of the physical model due to lubrication degradation, die wear, or changes in material condition. By comparing the ideal and actual values, the current equivalent friction coefficient can be deduced. Therefore, the physical model can transform the equivalent friction coefficient, which is difficult to measure directly, into a calculable and traceable state variable, providing a key foundation for intelligent drawing, fault diagnosis, and remaining life prediction.

[0059] S300 determines the current equivalent friction coefficient based on the current drawing force and current die temperature, as well as the ideal drawing force and ideal die temperature.

[0060] It is understandable that the equivalent friction coefficient is a comprehensive parameter used to characterize the overall friction effect of the interface between the die and the metal material during the wire drawing process. It not only reflects the actual friction state, but also indirectly includes the influence of various factors such as lubrication effect, die wear and temperature change on friction behavior.

[0061] For example, the error between the current pull-out force and the ideal pull-out force can be calculated, i.e. ,in, Indicates the current pulling force. This represents the ideal drawing force for the current cycle; it can be used to calculate the error between the current die temperature and the ideal die temperature. ,in, Indicates the current mold temperature. This indicates the ideal mold temperature for the current cycle.

[0062] We can assume that at the current equivalent friction coefficient The ideal drawing force and ideal die temperature are respectively , , Equivalent friction coefficient compared to the previous cycle The difference between them is It is possible to... Nearby, first-order approximations are made to the pull-out mechanics model and the thermodynamic model, respectively. , ,in, This represents the sensitivity of the pull-out force to the equivalent coefficient of friction, which can be obtained through analytical differentiation or numerical differentiation. This indicates the sensitivity of mold temperature to the equivalent coefficient of friction.

[0063] The error is mainly caused by Caused, therefore , , can Minimize the objective function: , to the objective function Taking the derivative and setting it to zero, we can obtain Analytical solution: Calculate After that, calculations can be performed. : ,in, It is an optional coefficient. =1, directly using the complete correction value, resulting in fast convergence; 0< <1, only a portion of the correction is needed each time, resulting in smoother convergence.

[0064] This step enables real-time identification and dynamic updating of the friction state during the drawing process. It uses the drawing force error and temperature error to correct the equivalent friction coefficient without relying on additional tests, thereby achieving rapid perception and accurate quantification of lubrication degradation, mold wear and abnormal operating conditions.

[0065] In one possible implementation, please refer to Figure 2 S300, based on the current drawing force and current die temperature, as well as the ideal drawing force and ideal die temperature, determines the current equivalent friction coefficient, including: S310, calculate the error between the current drawing force and the ideal drawing force to obtain the drawing error, and calculate the error between the current mold temperature and the ideal mold temperature to obtain the temperature error.

[0066] For example, the formula can be used: ,in, This represents the drawing error in the current period n. This represents the current pulling force in the current period n. This represents the ideal drawing force for the current period n, used to calculate the drawing error; the formula is: ,in, The current period n represents the temperature error. This represents the current mold temperature in the current period n. This represents the ideal mold temperature for the current cycle n, and the temperature error is calculated. and Together, they describe the deviation between the current operating conditions and the ideal state.

[0067] S320 determines the current equivalent friction coefficient based on the drawing error, temperature error, and the equivalent friction coefficient of the previous cycle.

[0068] For example, inversion methods such as gradient descent, least squares, and Kalman filtering can be used to calculate the current equivalent friction coefficient. Since the relationship between the equivalent friction coefficient and the drawing force and die temperature is non-linear (described by the physical model), an extended Kalman filter (EKF) can be used, which provides the optimal estimate in a noisy environment. The Kalman gain is calculated based on the drawing error, temperature error, and the equivalent friction coefficient of the previous cycle, and the current equivalent friction coefficient is then calculated using the Kalman gain.

[0069] Optionally, S320, based on the drawing error, temperature error, and the equivalent friction coefficient of the previous cycle, determines the current equivalent friction coefficient, including: S321 determines the equivalent friction coefficient of the previous cycle as the predicted value of the current state, and constructs the drawing error and temperature error as a residual vector.

[0070] For example, since the friction coefficient exhibits a continuous and slow evolution with changing operating conditions, it is reasonable to assume that the equivalent friction coefficient of the previous cycle is used as a priori estimate (i.e., a predicted value for the current state) for the current cycle. ,in, This represents the predicted value of the current state. This reflects the system's best prior judgment of the current periodic friction state based on historical states before any new observation information is introduced.

[0071] The drawing error and temperature error can be combined into a two-dimensional residual vector in a fixed order: The residual vector contains all the deviation information and is the core input used to correct the state prediction value. By constructing the residual vector and retaining the friction coefficient from the previous cycle as the state prediction value, the foundation is laid for subsequent dynamic optimization of the friction coefficient through model linearization and Kalman gain calculation.

[0072] S322, take the partial derivative of the physical model in the digital twin at the current state prediction value to obtain the first Jacobian matrix, and calculate the Kalman gain based on the first Jacobian matrix.

[0073] For example, under the condition that the current drawing speed and reduction rate are fixed, the partial derivatives of the drawing mechanics model and the thermodynamic model can be calculated at the current state prediction value to obtain the first Jacobian matrix. This is helpful in understanding the extent to which small changes in the equivalent friction coefficient affect the drawing force and temperature, i.e. The Jacobian matrix (a 2×1 matrix in the single-state case) can be obtained directly by differentiating the analytical expression of the physical model, or by numerical differentiation. The approximate derivative is calculated in the vicinity. The first Jacobian matrix is ​​the sensitivity coefficient that maps the observation residuals back to the state correction.

[0074] The Kalman gain for the current period n can be calculated using the Kalman filter formula: ,in, This represents the covariance after the state update in the previous cycle. This represents the measurement noise covariance matrix. The Kalman gain is used to automatically adjust the correction magnitude of force and temperature errors on the equivalent friction coefficient based on the magnitude of measurement noise and model uncertainty. When the measurement is more reliable, the Kalman gain is larger, and the correction amount increases; when the model prediction is more reliable or the measurement noise is large, the Kalman gain is smaller, and the correction amount decreases accordingly.

[0075] The initial value of the covariance P can be set based on the possible range of variation of the equivalent friction coefficient. For example, if the equivalent friction coefficient varies from approximately 0.02 to 0.20 in production, and the variation within each cycle is very small (e.g., 0.01), then the initial value of P can be... If a historical equivalent friction coefficient is available, the variance of the historical equivalent friction coefficient can be calculated as the initial value of the covariance P.

[0076] The expression is ,in, This indicates the variance of the noise measured by the force sensor. This represents the variance of the noise measured by the temperature sensor. and This can be determined through the sensor's instruction manual. For example, if the force sensor has an accuracy of ±0.5% and the temperature sensor has a noise level of ±1℃, then... , ,in, This indicates the maximum tensile force that the force sensor can measure.

[0077] S323, based on the Kalman gain and residual vector, corrects the current state prediction value to obtain the current equivalent friction coefficient.

[0078] For example, the product of the residual vector and the Kalman gain can be calculated to obtain the correction amount for the current period n, i.e. The correction amount can be added to the current state prediction value to obtain the current equivalent friction coefficient, i.e. .

[0079] The covariance can be updated based on the Kalman gain, the covariance after the state update in the previous cycle, and the first Jacobian matrix to obtain the covariance of the current cycle n, which is then used to calculate the equivalent friction coefficient for the next cycle. , where I represents the identity matrix, and since the state only has the equivalent friction coefficient, I=[1].

[0080] By taking these steps, the deviations in force and temperature during each cycle can be utilized to achieve real-time, adaptive updates to the equivalent friction coefficient. This ensures that the estimation of the friction state is both mechanistically accurate and capable of responding quickly to changes in actual working conditions, providing accurate state input for subsequent wear prediction, aperture evolution calculation, and mold life assessment.

[0081] S400: Based on the current equivalent friction coefficient, determine the current wear amount, and calculate the current aperture of the wire drawing die based on the current wear amount. The current wear amount is the aperture increment of the wire drawing die in the current cycle.

[0082] For example, a wear model can be constructed based on Archard's wear law, using the equivalent coefficient of friction as the lubrication condition; good lubrication... Smaller size means less wear; poorer lubrication. Larger areas result in greater wear. The wear model can be represented as... ,in, The wear coefficient is represented by t, the time of one control cycle is represented by t, and the hardness of the mold material is represented by H (PCD, TC, ND, etc. can be found).

[0083] The wear coefficient can be obtained through experimental calibration: Samples identical to the drawing material and drawing die material can be selected, and the samples can be drawn for a period of time using a fixed reduction rate and speed. The drawing force is recorded, and the die aperture is periodically measured and the aperture increment (i.e., wear) is calculated. Multiple sets of experimental data can be obtained, including speed, drawing force for each cycle, and wear amount. Using the initial equivalent friction coefficient and all experimental data, a wear model can be fitted using the least squares method to obtain... .

[0084] The current equivalent friction coefficient can be Substituting these values ​​into the wear model, the current wear amount (wear amount for one cycle) can be calculated. Since the current wear amount is the increase in the die's aperture in the current cycle, the sum of the die's aperture in the previous cycle and the current wear amount can be calculated to obtain the current die aperture. The initial die aperture can be obtained using the initial die exit diameter. .

[0085] This step eliminates the need for disassembly and inspection or manual estimation to estimate mold wear. Instead, the growth of the mold aperture can be calculated in real time by observing changes in the equivalent friction coefficient during operation, providing a quantitative basis for subsequent dimensional compensation, mold replacement decisions, and quality control.

[0086] In one possible implementation, please refer to Figure 2 S400, based on the current equivalent friction coefficient, determines the current wear amount, and calculates the current bore diameter of the wire drawing die based on the current wear amount, including: S410, calculate the current wear amount using the wear mapping model based on the current equivalent friction coefficient and the current thread feed. The wear mapping model is pre-built and is used to characterize the difference between the equivalent friction coefficient and the initial equivalent friction coefficient, as well as the relationship between the thread feed and the wear amount.

[0087] It can be understood that the current wire feed is the cumulative wire feed from the start of drawing to the current cycle n, and the current wear is the cumulative wear from the start of drawing to the current cycle n.

[0088] For example, the current equivalent friction coefficient can be... and current fiber content The current wear level is calculated by inputting the data into a pre-built wear mapping model. The process of building the wear mapping model is described in detail below.

[0089] If the wire drawing machine can provide the current drawing speed in real time, the wire feed rate in one cycle can be obtained by integrating the speed. For discrete sampling conditions, the wire feed rate in one cycle can be approximated as the product of the current drawing speed and the time length of one cycle. The cumulative wire feed rate can be updated recursively, so each cycle has a real-time updated current wire feed rate.

[0090] If the wire drawing machine has a meter counter, the current wire passing amount can be obtained by directly reading the value from the meter counter.

[0091] S420, calculate the sum of the current wear and the initial aperture to obtain the current aperture of the wire drawing die.

[0092] For example, unlike the calculation of wear over a cycle in step S400, this step uses a wear mapping model to calculate the cumulative wear, so the initial bore diameter (i.e., the initial die diameter) can be used. The current wear amount obtained in step S410 is superimposed to obtain the wire drawing die diameter for the current period n.

[0093] In one possible implementation, the online method for determining the wear of the wire drawing die also includes: S401: Obtain the preset equivalent friction coefficient, initial drawing speed, multiple initial drawing forces, and initial die temperature; and calculate the average drawing force of the multiple initial drawing forces and the average die temperature of the multiple initial die temperatures. The multiple initial drawing forces and initial die temperatures are continuously collected under the conditions of initial drawing speed and reduction ratio.

[0094] For example, an initial guess of the equivalent friction coefficient (preset equivalent friction coefficient) can be preset. This preset equivalent friction coefficient can come from empirical estimation, literature reference, or the initial equivalent friction coefficient of a physical model. The preset equivalent friction coefficient is just an initial guess, which is then iteratively corrected.

[0095] The experiment can be conducted using combinations of materials and drawing dies identical or equivalent to those used in actual production. Initial drawing speeds and reduction rates can be set to match actual production conditions. The drawing machine is started, and the drawing speed is stabilized at the initial speed. Drawing force and die temperature are continuously collected over a period of time, yielding multiple sets of initial drawing force and initial die temperature. The average values ​​of all initial drawing forces and all initial die temperatures are calculated to obtain the average drawing force and average die temperature. The average drawing force effectively filters out instantaneous fluctuations, reflecting the true drawing resistance under the initial drawing speed and reduction rate. The average die temperature eliminates short-term fluctuations and noise interference during temperature measurement.

[0096] The initial drawing force and initial die temperature were obtained by conducting a short-term stable drawing test before production started. This established the baseline friction characteristics under healthy conditions, laying the data foundation for subsequent calculation of the initial equivalent friction coefficient, and providing a reliable reference point for subsequent friction coefficient inversion and wear prediction.

[0097] S402, Ideal value calculation: Based on the initial drawing speed, reduction ratio and preset equivalent friction coefficient, the initial ideal drawing force and initial ideal die temperature are calculated through a physical model.

[0098] For example, the initial drawing speed, reduction ratio, and preset equivalent friction coefficient can be input into the drawing mechanics model and thermodynamic model in step S200 to obtain the initial ideal drawing force and the initial ideal die temperature.

[0099] S403, Error Calculation: Calculate the first error between the average drawing force and the initial ideal drawing force, and calculate the second error between the average die temperature and the initial ideal die temperature.

[0100] For example, the difference between the average pull-out force and the initial ideal pull-out force can be calculated, which is the first error.

[0101] The difference between the average mold temperature and the initial ideal mold temperature can be calculated, which is the second error.

[0102] This step allows us to obtain error information in both the mechanical and temperature dimensions, providing the necessary basis for subsequently using the Jacobian matrix and iterative correction methods to obtain an accurate initial equivalent friction coefficient.

[0103] S404, Jacobian matrix calculation: Take the partial derivative of the physical model at the preset equivalent friction coefficient to obtain the second Jacobian matrix.

[0104] For example, under the condition that the initial drawing speed and the reduction ratio are fixed, the physical model can be regarded as a nonlinear mapping model with the equivalent friction coefficient as input and the ideal drawing force and the ideal die temperature as output: It can be done by and For the preset equivalent friction coefficient Taking the derivative, we obtain the second Jacobian matrix: .

[0105] The second Jacobian matrix quantitatively reflects the influence of small changes in the equivalent friction coefficient on the ideal drawing force and ideal die temperature under the current working conditions. In subsequent iterations, the force error and temperature error are mapped to the correction amount of the equivalent friction coefficient through the second Jacobian matrix, thereby realizing the gradual optimization and convergence of the preset equivalent friction coefficient, and finally determining the initial equivalent friction coefficient.

[0106] S405, Update: Based on the second Jacobian matrix, the first error, and the second error, solve for the correction amount, and update the preset equivalent friction coefficient based on the correction amount.

[0107] For example, the first error and the second error can be combined into an error vector: ,in, Indicates the first error. This represents the second error. It can be corrected by fine-tuning the equivalent friction coefficient (i.e., adding a correction factor). This allows the physical model's output of drawing and die temperatures to more closely approximate the measured average drawing force and average die temperature. The aforementioned nonlinear mapping model can be used to achieve this. The vicinity is represented by a first-order linear approximation. ,in, , This represents the initial ideal pull-out force. This represents the initial ideal mold temperature. To make... To get as close as possible to the average, so ,in, Indicates the average pull-out force. This represents the average mold temperature. and The difference between them is the error vector, i.e. Since e cannot be exactly equal to Therefore, least squares is used here. That is, to solve , The value is .

[0108] The preset equivalent friction coefficient can be added to the correction amount to obtain a new equivalent friction coefficient, which is the updated preset equivalent friction coefficient.

[0109] This step utilizes the physical model's sensitivity to the equivalent friction coefficient to accurately map force and temperature errors into the correction range of the equivalent friction coefficient, thereby avoiding the instability and inefficiency caused by blind searching or fixed step size adjustment.

[0110] S406, iterate through the steps of ideal value calculation, error calculation, Jacobian matrix calculation, and update until the first error and the second error are less than the corresponding thresholds, and then determine the preset equivalent friction coefficient of the last iteration as the initial equivalent friction coefficient.

[0111] For example, thresholds can be pre-set for the first error and the second error, respectively. and ,like and or and Setting the thresholds for the two errors to multiples of the standard deviation of the corresponding sensor measurement noise demonstrates that when the two errors have been reduced to near the level of the sensor measurement noise, further iterations can no longer significantly improve the performance; the remaining difference is mainly due to noise and model simplification.

[0112] The steps S401, S402, S403, S404, and S405 are executed repeatedly. As the number of iterations increases, the first error and the second error will gradually decrease. When the absolute values ​​of the first error and the second error are detected to be less than [a certain value], [the process continues]. and When the convergence condition is met, the iteration terminates. The preset equivalent friction coefficient obtained in the last iteration can be used as the initial equivalent friction coefficient under the initial drawing speed and reduction rate.

[0113] Through an iterative update mechanism, the equivalent friction coefficient can significantly approximate the actual friction state within a finite number of iterations, enabling the mechanical and thermal outputs of the physical model to simultaneously satisfy the experimental data. This establishes a friction benchmark that accurately reflects the initial working conditions, providing high-precision initial parameters for subsequent online friction identification, wear prediction, and mold life modeling.

[0114] In one possible implementation, a wear mapping model is constructed, including: S10, Obtain the sample dataset and divide it into a training set and a test set. The sample dataset includes multiple sample data, each of which includes the equivalent friction coefficient, wire feed amount, and actual wear amount. The actual wear amount is obtained by measuring the offline diameter of the drawing die at different time points.

[0115] For example, the testing method in step S401 can be used to conduct a drawing test over multiple time periods. Multiple initial drawing forces and initial die temperatures are collected for each time period, and the average drawing force and average die temperature for each time period are calculated. Since the drawing speed and reduction rate are known, the equivalent friction coefficient for each time period can be derived from the average drawing force and average die temperature using a physical model, and the difference between the equivalent friction coefficient for each time period and the initial equivalent friction coefficient can be calculated. The amount of wire passing L in each time period can be obtained through speed integration or a meter counter. At the end of each stage, the machine is stopped, and the die exit diameter for that time period is measured offline. The die exit diameter for each time period is then calculated. The difference between them is the actual wear and tear in each time period. Through the above process, a set of three data points can be obtained for each time period ( Wire feed amount L and actual wear amount (i.e., sample data).

[0116] All sample data can be divided into a training set (e.g., 70%) and a test set (e.g., 30%).

[0117] S20: Using an expression tree structure, multiple random expression trees are initialized as an initial population. Each expression tree represents a wear mapping model.

[0118] For example, a set of basic operators and non-linear functions can be defined as a set of function nodes (such as addition "+", subtraction "-", multiplication "+"). The code snippet mentions operators like ' / ', power functions (such as squares and cubes), exponents, and logarithms, and protects division, logarithms, and root extraction operators (e.g., replacing zero divisors with very small constants, truncating negative numbers when taking their logarithms) to prevent numerical divergence. L, along with several random constants (such as real constants), form the set of terminal nodes. Internal nodes are selected from the function node set, and leaf nodes are selected from the terminal node set.

[0119] For each expression tree: an integer can be randomly selected from a preset tree depth range (e.g., maximum tree depth of 3-6 levels, minimum tree depth of 2-3 levels) as the maximum depth of the tree. For internal nodes that have not reached the maximum depth, a function can be randomly selected from the function node set as the current node, based on the function's atom count (e.g., +, -, ...). (where ' / ' is a binary function, and exponential and logarithmic functions are unary functions). Create the corresponding number of child nodes for each function and continue recursively generating new nodes. If the maximum depth is reached, a terminal node can be randomly selected from the set of terminal nodes as a leaf node.

[0120] If the expression tree does not contain The variables L and L can be used to force the replacement of certain leaf nodes with one of two variables, which is beneficial because the wear mapping model is based on the mapping of the equivalent friction coefficient and the amount of yarn passing through, rather than a pure constant. Through the above process, a set of mathematical expressions with different structures and forms can be obtained, such as linear combinations, product forms, and composite forms containing nonlinear functions. This expression tree corresponds to a candidate wear mapping model. ,in, , This represents the initial equivalent friction coefficient.

[0121] To ensure the genetic algorithm has sufficient search space, N different random expression trees can be generated, such as N=50 or 100. This process is repeated N times to obtain an initial population consisting of N random expression trees.

[0122] S30 calculates the predicted wear amount corresponding to each expression tree based on the equivalent friction coefficient and wire feed amount in the training set.

[0123] For example, for each expression tree in the initial population: each sample data from the training set can be used. of and Substitute the values ​​sequentially into the expression tree for evaluation: starting from the leaf nodes, you can substitute the variable symbols in the leaf nodes... L uses sample data The value ( , The substitution process proceeds from the bottom of the tree upwards, performing the operations corresponding to each internal node until the root node is reached. The result of the calculation at the root node represents the expression tree in the sample data. Predicted wear amount .

[0124] By following the steps above, we can obtain the predicted wear sequence for each expression tree.

[0125] S40 calculates the error of each expression tree based on the actual wear amount in the training set and the predicted wear amount corresponding to each expression tree.

[0126] For example, for each expression tree: each sample data can be computed. Actual wear Compared with predicted wear The difference between them, i.e. ,in, express and The difference between them. The mean squared error (MSE) or mean absolute error (MAE) can be used as the error function to calculate the overall fitting effect of the expression tree. The mean squared error can be expressed as... Where n represents the number of sample data; the mean absolute error can be expressed as .

[0127] The smaller the error, the more accurately the wear mapping model corresponding to the expression tree fits the actual wear amount on the training set. In the subsequent genetic algorithm evolution process, this error can be used as the basis for fitness evaluation to select, cross over and mutate the expression tree, thereby gradually evolving a better wear mapping model structure.

[0128] S50: Based on the error of each expression tree, a genetic algorithm is used to perform multiple rounds of evolutionary iterations on the initial population to obtain a set of candidate mapping models.

[0129] For example, a genetic algorithm can be employed using a genetic programming (GP) algorithm based on expression trees. The error of each expression tree can be inverted or normalized to serve as the fitness value for that tree. The genetic algorithm performs a selection operation: based on the fitness value of each expression tree, it selects a set of parents from the population for crossover and mutation. The selection operator can be either tournament selection or roulette wheel selection. Tournament selection randomly selects a number of individuals (e.g., 3-5) from the population and selects the individual with the highest fitness as the parent; roulette wheel selection samples individuals based on their fitness with probability, where higher fitness increases the probability of being selected as a parent.

[0130] Crossover operation: For the selected parent expression tree, new child expression trees can be generated by exchanging its local subtree structures. For example, randomly selecting a subtree node from each of the two parent expression trees and swapping the two subtrees will generate two new expression trees. The crossover operation can recombine structural fragments of different individuals while maintaining the validity of the original function nodes and terminal nodes. This allows structures that originally only had good fitting ability in certain regions to be combined into a better overall expression, thus enriching the search space. The crossover probability can be set to 0.7~0.9.

[0131] Mutation operation: Applying a low-probability random structural perturbation to some offspring or parent individuals to prevent the population from getting trapped in local optima. This involves randomly changing the function type of a node, adjusting the constant value of a terminal node, or replacing an existing subtree with a new one, introducing new structures and parameter diversity. Although the mutation operation has a low probability of occurrence, it can introduce new expressions into the structural space, thereby improving the algorithm's global search capability. The mutation probability can be set to 0.05~0.2.

[0132] By performing selection, crossover, and mutation operations, a new generation of the population can be obtained. For each expression tree in this new generation, the predicted wear and corresponding error can be recalculated on the training set, and its fitness updated. The selection, crossover, and mutation operations are then repeated iteratively for several generations (e.g., 50-200 generations). When the number of iterations reaches a preset upper limit, or when the overall error of the population no longer decreases significantly within several generations, the population can be considered converged. A batch of expression trees with smaller errors can be selected from the final population as a candidate mapping model set.

[0133] This step, through multiple rounds of evolution using a genetic algorithm based on the error of each expression tree, can automatically search for the optimal or near-optimal wear mapping model from a large number of candidate structures without pre-setting a specific function form. This enables autonomous learning and structure mining of the complex nonlinear relationship between the equivalent friction coefficient and the amount of yarn passing through. It provides a high-quality candidate space for subsequent selection of the optimal wear mapping model based on a test set, thereby significantly improving the accuracy and robustness of the wear prediction model.

[0134] S60. Based on the test set, calculate the error of each candidate mapping model in the candidate mapping model set, and select the candidate mapping model with the smallest error as the wear mapping model.

[0135] For example, for each candidate mapping model: the equivalent friction coefficient and yarn passage amount of each sample data in the test set can be substituted into the candidate mapping model to calculate the predicted wear amount of each sample data in the test set. The error calculation process of the candidate mapping model is the same as step S40. The candidate mapping model with the smallest error can be selected as the wear mapping model from all the errors of the candidate mapping models.

[0136] This step helps the selected wear mapping model to have good generalization ability, avoiding the structural rigidity and overfitting problems of traditional empirical models. The wear mapping model can accurately predict the wear amount under different friction states and wire passage amounts, while maintaining the interpretability of the model form and computational efficiency, thus providing a reliable data-driven basic model for online wear assessment, aperture evolution prediction, and die life management.

[0137] In one possible implementation, the online method for determining the wear of the wire drawing die also includes: S101, based on the set of candidate mapping models, calculate the complexity of each candidate mapping model.

[0138] For example, each candidate mapping model in the candidate mapping model collection corresponds to an expression tree, and each expression tree includes internal nodes (operators / functions) and leaf nodes (…). (L, constants). For each expression tree in the candidate mapping model set: the entire tree can be traversed starting from the root node (e.g., using preorder or postorder traversal), and the total number of nodes in the expression tree can be counted for each node visited (whether it is a function node or a terminal node). This total number of nodes is used as the complexity value of the candidate mapping model.

[0139] S102, map the error and complexity values ​​of each candidate mapping model to a two-dimensional coordinate space to generate a model performance distribution map. The horizontal axis of the model performance distribution map represents complexity, and the vertical axis represents error.

[0140] For example, we can assume that the complexity of the j-th candidate mapping model is O(n). The error is In a coordinate system with complexity on the horizontal axis and error on the vertical axis, with ( , The coordinates of the candidate mapping model are plotted. By repeating the above operation on all models in the candidate mapping model set, a model performance distribution map consisting of multiple scattered points is obtained. This model performance distribution map provides an intuitive basis for subsequent identification of the trade-off between error and complexity and for screening the Pareto optimal frontier.

[0141] S103, based on the model performance distribution map, selects candidate mapping models from the candidate mapping model set that are not simultaneously dominated by other candidate mapping models in terms of error and complexity, and obtains the Pareto optimal frontier.

[0142] For example, for candidate mapping models j and k, if the following conditions are met... , And at least one of them is satisfied. or This indicates that candidate mapping model k dominates candidate mapping model j in terms of error and complexity. That is, while candidate mapping model k has higher prediction accuracy or simpler structure, it is not worse in another indicator, and therefore outperforms candidate mapping model j.

[0143] We can perform pairwise comparisons on all models in the candidate mapping model set, eliminating models that are simultaneously dominated by other models in terms of both error and complexity. The remaining set of models constitutes the Pareto optimal front. The Pareto optimal front does not require a single optimal model, but rather provides a performance boundary composed of multiple non-dominated models, offering a reliable candidate space for selecting the most suitable wear mapping model.

[0144] S104, based on the model performance distribution map, selects the wear mapping model from the Pareto optimal frontier.

[0145] For example, the minimum and maximum values ​​of complexity and error on the Pareto optimal front can be found from the model performance distribution plot. Based on the minimum and maximum values ​​of complexity and error, the complexity and error of each candidate mapping model on the Pareto optimal front are normalized, i.e. , ,in, and These represent the minimum and maximum values ​​of the complexity, respectively. and These are the minimum and maximum values ​​of the error, respectively. This represents the normalized complexity of the j-th candidate mapping model. This represents the normalization error of the j-th candidate mapping model.

[0146] The weighted sum of the normalized complexity and normalization error of each candidate mapping model can be calculated as a comprehensive indicator, i.e. ,in, ∈[0,1] is used to adjust the relative preference between model simplicity and prediction accuracy. A smaller value can be chosen when accuracy is prioritized. When the focus is more on the simplicity and ease of interpretation of the model, the size can be appropriately increased. .

[0147] The candidate mapping model with the smallest comprehensive index among all candidate mapping models is selected as the wear mapping model. This step provides a clear trade-off between error and complexity while ensuring that the model is at the Pareto optimal frontier, thus uniquely determining the wear mapping model from the Pareto optimal frontier.

[0148] S500, based on the current equivalent friction coefficient, predicts the life of the wire drawing die and obtains the remaining life of the wire drawing die.

[0149] For example, based on process quality requirements and mold design specifications, the rejection criteria for wire drawing dies can be preset, such as setting the die aperture to a certain limit value. When the cumulative wear reaches a certain threshold, the die can be considered unable to guarantee product dimensions and surface quality, and needs to be replaced. For any period n, we can assume the current aperture of the wire drawing die is... Then the remaining allowable wear is The increment of the die aperture in the nth period can be expressed as: This formula can be normalized with respect to time to obtain the instantaneous wear rate in the nth cycle, i.e. The ratio of the remaining allowable wear to the instantaneous wear rate can be calculated, which is the predicted remaining service life (remaining life) of the wire drawing die. .

[0150] In practical applications, the aforementioned remaining life prediction process can be repeated in each cycle. Whenever the equivalent friction coefficient is updated using the physical model and measured data, the system simultaneously updates the current bore diameter, remaining allowable wear, and wear rate, thereby updating the remaining life estimate of the wire drawing die in real time. Because this method uniformly reflects factors such as changes in lubrication status, load fluctuations, and speed changes in the equivalent friction coefficient, it can promptly adjust the remaining life prediction results when operating conditions change, achieving dynamic assessment and early warning of the wire drawing die's life, providing a reliable basis for die replacement decisions and production plan optimization.

[0151] In one possible implementation, S500, based on the current equivalent friction coefficient, predicts the life of the wire drawing die to obtain the remaining life of the wire drawing die, including: S510, obtains the equivalent friction coefficient and wire feed amount for N cycles.

[0152] For example, during the drawing process, the equivalent friction coefficient and the statistical wire passage amount can be calculated for each cycle. The equivalent friction coefficient and the wire passage amount can be continuously stored in the control system or host computer in the form of a time series, such as in a database or local file. When life prediction is required, the equivalent friction coefficient and the wire passage amount of the most recent N cycles can be directly read from the storage medium.

[0153] S520, the equivalent friction coefficient and wire feed amount are fitted over N cycles to obtain the first model. The first model characterizes the relationship between the equivalent friction coefficient and the wire feed amount.

[0154] For example, a suitable function can be selected as the fitting model form based on the actual characteristics of frictional changes. This could be a polynomial function (such as a first- or second-order polynomial), an exponential decay / growth function, or a piecewise linear function. Taking a polynomial function as an example, the fitting model can be expressed as follows: The equivalent friction coefficient and over-filament amount for N cycles are constructed as a set of discrete sample points. , sample points Substitute each parameter into the fitted model and solve for the model parameters using least squares regression. That is, minimize and Sum of squared errors between: Through the above fitting process, the first model can be obtained. The first model can describe the overall evolution of the equivalent friction coefficient caused by factors such as the deterioration of lubrication and wear of the mold surface as the cumulative wire feed increases.

[0155] S530, based on the wear mapping model and the first model, a prediction function is constructed. This prediction function characterizes the relationship between wear amount and thread passage amount.

[0156] For example, the wear mapping model is as follows: ,in, The first model is Substituting the first model into the wear mapping model yields a wear prediction function with only the amount of yarn passing as the independent variable. It is used to characterize the evolution of aperture wear of the wire drawing die with the amount of wire passing throughout the entire service process.

[0157] S540, based on the initial aperture and the scrap threshold, the prediction function is solved to obtain the total wire feed. The scrap threshold is used to characterize the preset scrap aperture of the drawing die.

[0158] For example, the rejection threshold for the aperture can be set according to process requirements and dimensional accuracy standards. It can be based on the initial aperture and scrapping threshold Calculate the maximum allowable cumulative wear of the mold: It can be Combined with the prediction function: The total amount of silk can be obtained by directly solving the formula. If the prediction function has an explicit analytical form and the above formula can be solved directly, the total amount of silk can be obtained by algebraic methods; if the prediction function is a more complex nonlinear function, it can be obtained by... Treating it as a univariate nonlinear equation, it is solved numerically within a reasonable range of wire passage amounts using methods such as the bisection method and Newton's iteration method, to find the total wire passage amount that makes the equation valid. During the solution process, it can be assumed that the prediction function monotonically increases with the wire passage amount, thereby ensuring the uniqueness of the solution and the convergence of the numerical solution.

[0159] S550 calculates the remaining wire feed based on the total wire feed and the current wire feed, thus obtaining the remaining life of the wire drawing die. The remaining life is the remaining wire feed.

[0160] For example, the remaining wire feed can be obtained by subtracting the current wire feed from the total wire feed, which is the remaining life of the wire drawing die. This remaining life is a life in the sense of length.

[0161] In addition to the methods mentioned above, the remaining lifespan can also be a lifespan in the sense of time. The average drawing speed of N cycles can be calculated (again, the drawing speed of each cycle can be stored in a database or a local file), and the ratio between the remaining wire feed and the average drawing speed of N cycles can be calculated to obtain the remaining service time of the drawing die, that is, the remaining lifespan of the drawing die.

[0162] As new data is continuously generated, the remaining life prediction of the wire drawing die can be updated periodically. For example, the remaining life of the wire drawing die can be recalculated every 8 hours.

[0163] Compared to traditional methods that rely on fixed empirical wear rates or simply use cumulative machining volume to estimate lifespan, the above steps explicitly incorporate influencing factors (referring to the equivalent friction coefficient) such as actual friction state changes, lubrication degradation, and accelerated die wear. This allows for adaptive reflection of wear rate changes under different operating conditions, resulting in more accurate predictions of total usable thread throughput and remaining lifespan. It can provide early warnings of lifespan reduction when die wear accelerates abnormally or lubrication performance deteriorates, enabling refined management of die lifespan, improving production stability and economy, and reducing downtime and quality risks caused by sudden die failures.

[0164] In one possible implementation, the online method for determining the wear of the wire drawing die also includes: S501, obtain the wire drawing die orifice diameter and wire feed amount for N cycles.

[0165] For example, similarly, the wire drawing die aperture and wire feed amount for N cycles can be continuously stored in the control system or host computer in the form of a time series, such as in a database or local file. When life prediction is required, the wire drawing die aperture and wire feed amount for the most recent N cycles can be directly read from the storage medium.

[0166] S502, the drawing die aperture and wire feed rate are fitted over N cycles to obtain the second model. The second model characterizes the relationship between the drawing die aperture and the wire feed rate.

[0167] For example, if the wear process is nonlinear: similarly, the wire drawing die aperture and wire feed amount for N cycles can be constructed as a set of discrete sample points. The form of the fitting model can be selected based on actual wear behavior, the shape of the empirical curve, or the performance of the fitting error. Examples include polynomial functions (such as quadratic or cubic polynomials), exponential growth functions, power functions, logistic or saturating models (suitable for cases where the wear rate gradually slows down), etc. The fitting model can be represented as... , sample points Substituting these values ​​into the fitting model, the model parameters b are solved using the nonlinear least squares method. Through the above fitting process, the second model can be obtained. The second model can accurately depict the nonlinear evolution of mold wear during use, avoiding life estimation bias caused by linear assumptions, thereby significantly improving the reliability and accuracy of remaining life prediction.

[0168] If the wear process follows a linear relationship: The wear rate can be calculated based on the die orifice diameter and wire feed rate of the last cycle out of N cycles. Where K represents the wear rate, This represents the wire drawing die orifice diameter in the last cycle of N cycles. This represents the amount of yarn fed in the last cycle out of N cycles. Based on the scrap threshold... Calculate the allowable wear and tear before scrapping: According to And K, calculate the remaining amount of yarn: ,in, This indicates the remaining amount of yarn.

[0169] S503, based on the scrap threshold, solve the second model to obtain the total amount of yarn.

[0170] For example, the scrapping threshold can be... Second Model United: Since this formula has only one independent variable L, solving it yields the total amount of silk thread. If the second model is a univariate linear function or a low-order polynomial with an explicit analytical solution, the total amount of silk thread can be directly calculated using algebraic operations; if the second model is a nonlinear function, it can be... Treating it as a univariate nonlinear equation, numerical solution methods (such as the bisection method, Newton's iteration method, etc.) are used to search within a preset range of wire feed to achieve... and Equal solutions. It can be assumed that the aperture increases monotonically with the wire feed rate, thus ensuring the uniqueness of the solution and the stable convergence of the numerical solution. The total wire feed rate can provide a direct basis for subsequent calculations of the remaining wire feed rate and remaining lifespan.

[0171] S504: Calculate the remaining wire feed based on the total wire feed and the current wire feed, and obtain the remaining life of the wire drawing die.

[0172] For example, the implementation process of this step is the same as that of step S550, and will not be described again here.

[0173] The above steps capture the phased changes in wear rate, enabling a natural representation of accelerated and decelerated wear behavior in the die, thereby improving the accuracy and stability of life prediction. The remaining life results of the wire drawing die have physical interpretability and engineering credibility, providing a reliable basis for die replacement decisions, production line scheduling optimization, and preventive maintenance.

[0174] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0175] Corresponding to the online determination method for wire drawing die wear described in the above embodiments, this application also provides an online determination device for wire drawing die wear, wherein each unit of the device can realize each step of the online determination method for wire drawing die wear.

[0176] The device includes: The acquisition unit is used to acquire the current drawing force, current die temperature, current drawing speed, and reduction ratio. Among them, the current drawing force, current die temperature, and current drawing speed are all actual values ​​for the current cycle, and the reduction ratio is used to characterize the ratio of the reduction in drawing area to the die area.

[0177] The ideal value calculation unit is used to calculate the ideal drawing force and ideal die temperature for the current cycle based on the current drawing speed and reduction rate, using a digital twin of the drawing process. The digital twin includes a physical model, which comprises a drawing mechanics model and a thermodynamic model.

[0178] The equivalent friction coefficient determination unit is used to determine the current equivalent friction coefficient based on the current drawing force and current die temperature, as well as the ideal drawing force and ideal die temperature.

[0179] The aperture calculation unit is used to determine the current wear amount based on the current equivalent friction coefficient, and to calculate the current aperture of the wire drawing die based on the current wear amount. The current wear amount is the aperture increment of the wire drawing die in the current cycle.

[0180] The life prediction unit is used to predict the life of the wire drawing die based on the current equivalent friction coefficient, and obtain the remaining life of the wire drawing die.

[0181] It should be noted that the information interaction and execution process between the above-mentioned units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, which will not be repeated here.

[0182] This application also provides a wire drawing machine, which includes a drawing force acquisition device, a temperature acquisition device, a drawing speed acquisition device, and a control device communicatively connected to the drawing force acquisition device, the temperature acquisition device, and the drawing speed acquisition device. The control device of this embodiment includes at least one processor, at least one memory, and a computer program stored in the at least one memory and executable on the at least one processor. When the processor executes the computer program, it causes the wire drawing machine to perform the steps in any of the above embodiments of the online determination method for the wear of drawing dies, or to perform the functions of each unit in the above embodiments of the devices.

[0183] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete this application. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the control device of the wire drawing machine.

[0184] The control device for the wire drawing machine can be an industrial control computer (IPC), a PLC and host computer (such as a SCADA system, industrial PC, etc.), an edge computing device (Edge AI gateway), etc. The wire drawing machine may include, but is not limited to, processors and memory, and may also include input / output devices, network access devices, buses, etc.

[0185] The processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0186] In some embodiments, the memory may be an internal storage unit of the control device of the wire drawing machine, such as the hard drive or memory of the wire drawing machine. In other embodiments, the memory may be an external storage device of the wire drawing machine, such as a plug-in hard drive, smart media card (SMC), secure digital (SD) card, flash card, etc. Furthermore, the memory may include both internal and external storage units of the wire drawing machine. The memory is used to store operating systems, applications, bootloaders, data, and other programs, such as the program code of computer programs. The memory can also be used to temporarily store data that has been output or will be output.

[0187] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0188] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for on-line determination of the wear amount of a wire drawing die, characterized by, The method comprises: obtaining a current drawing force, a current die temperature, a current drawing speed and a reduction ratio; wherein the current drawing force, the current die temperature and the current drawing speed are actual values in a current cycle, and the reduction ratio is used to represent a ratio of a drawing area reduction amount to a die entry area; calculating an ideal drawing force and an ideal die temperature in the current cycle by a digital twin of the wire drawing process according to the current drawing speed and the reduction ratio; wherein the digital twin comprises a physical model, and the physical model comprises a drawing mechanics model and a thermodynamics model; determining a current equivalent friction coefficient based on the current drawing force and the current die temperature and the ideal drawing force and the ideal die temperature; determining a current wear amount based on the current equivalent friction coefficient, and calculating a current aperture of a wire drawing die according to the current wear amount; wherein the current wear amount is an aperture increment of the wire drawing die in the current cycle; predicting a service life of the wire drawing die based on the current equivalent friction coefficient to obtain a remaining service life of the wire drawing die.

2. The on-line wire rod die wear amount judgment method according to claim 1, wherein The method further comprises: calculating an error between the current drawing force and the ideal drawing force to obtain a drawing error, and calculating an error between the current die temperature and the ideal die temperature to obtain a temperature error; determining the current equivalent friction coefficient based on the drawing error, the temperature error and an equivalent friction coefficient in a previous cycle.

3. The on-line wire rod die wear amount judgment method according to claim 2, characterized by, The method further comprises: determining the equivalent friction coefficient in the previous cycle as a current state prediction value, and constructing the drawing error and the temperature error as a residual vector; calculating a first Jacobian matrix by taking a partial derivative of the physical model in the digital twin at the current state prediction value, and calculating a Kalman gain according to the first Jacobian matrix; correcting the current state prediction value according to the Kalman gain and the residual vector to obtain the current equivalent friction coefficient.

4. The on-line wire rod die wear amount judgment method according to claim 3, characterized by The method further comprises: calculating the current wear amount by using a wear mapping model according to the current equivalent friction coefficient and a current wire passing amount; wherein the wear mapping model is pre-constructed, and the wear mapping model is used to represent a relationship between a difference between the equivalent friction coefficient and an initial equivalent friction coefficient and a relationship between the wire passing amount and the wear amount; calculating a sum of the current wear amount and an initial aperture to obtain the current aperture of the wire drawing die.

5. The on-line wire rod die wear amount judgment method according to claim 4, characterized by, The method further comprises: obtaining a preset equivalent friction coefficient, an initial drawing speed, a plurality of initial drawing forces and initial die temperatures, and calculating an average drawing force of the plurality of initial drawing forces and an average die temperature of the plurality of initial die temperatures; wherein the plurality of initial drawing forces and the initial die temperatures are obtained continuously under the condition of the initial drawing speed and the reduction ratio. Ideal value calculation: according to the initial drawing speed, the reduction ratio and the preset equivalent friction coefficient, an initial ideal drawing force and an initial ideal die temperature are calculated by the physical model; Error calculation: a first error between the average drawing force and the initial ideal drawing force is calculated, and a second error between the average die temperature and the initial ideal die temperature is calculated; Jacobi matrix calculation: partial derivatives of the physical model at the preset equivalent friction coefficient are obtained to obtain a second Jacobi matrix; Updating: a correction amount is solved according to the second Jacobi matrix, the first error and the second error, and the preset equivalent friction coefficient is updated according to the correction amount; The steps of iterative ideal value calculation, error calculation, Jacobi matrix calculation and updating are performed until the first error and the second error are less than the corresponding threshold values, and the preset equivalent friction coefficient of the last iteration is determined as the initial equivalent friction coefficient.

6. The on-line wire rod die wear amount judgment method according to claim 4, wherein The construction of the wear mapping model comprises: Obtaining a sample data set, and dividing the sample data set into a training set and a test set; wherein the sample data set comprises a plurality of sample data, each sample data comprising an equivalent friction coefficient, a wire passing amount and an actual wear amount, and the actual wear amount is obtained by measuring the offline diameter of the drawing die at different time points; Using an expression tree structure, a plurality of random expression trees are initialized as an initial population; wherein each expression tree is used to represent a wear mapping model; According to the equivalent friction coefficient and the wire passing amount in the training set, the predicted wear amount corresponding to each expression tree is calculated; Based on the actual wear amount in the training set and the predicted wear amount corresponding to each expression tree, the error of each expression tree is calculated; Based on the error of each expression tree, the initial population is evolved and iterated for multiple rounds using a genetic algorithm to obtain a candidate mapping model set; According to the test set, the error of each candidate mapping model in the candidate mapping model set is calculated, and the candidate mapping model with the smallest error is taken as the wear mapping model.

7. The method of claim 6, wherein the amount of wear of the wire drawing die is calculated based on the difference between the initial diameter of the wire rod and the diameter of the wire rod after the wire drawing process. The method further comprises: Based on the candidate mapping model set, the complexity of each candidate mapping model is calculated; The error and complexity values of each candidate mapping model are mapped to a two-dimensional coordinate space to generate a model performance distribution map; wherein the horizontal axis of the model performance distribution map is the complexity, and the vertical axis is the error; Based on the model performance distribution map, a candidate mapping model that is not simultaneously dominated by other candidate mapping models in terms of error and complexity is screened out from the candidate mapping model set to obtain a Pareto optimal front; Based on the model performance distribution map, the wear mapping model is screened out from the Pareto optimal front.

8. The method of claim 4, wherein the amount of wear of the wire drawing die is determined on-line. The prediction of the service life of the drawing die based on the current equivalent friction coefficient comprises: Obtaining the equivalent friction coefficient and the wire passing amount of N cycles; Fitting the equivalent friction coefficient and the wire passing amount of the N cycles to obtain a first model; wherein the first model is used to represent the relationship between the equivalent friction coefficient and the wire passing amount; constructing a prediction function according to the wear mapping model and the first model, wherein the prediction function is used to represent a relationship between the wear amount and the wire passing amount; solving the prediction function according to the initial aperture and a scrap threshold to obtain a total wire passing amount, wherein the scrap threshold is used to represent a preset scrap aperture of the wire drawing die; calculating a remaining wire passing amount according to the total wire passing amount and the current wire passing amount to obtain a remaining service life of the wire drawing die, wherein the remaining service life is the remaining wire passing amount.

9. The method of claim 8, wherein the amount of wear of the wire drawing die is calculated based on the difference between the initial diameter of the wire rod and the diameter of the wire rod after the wire drawing process. The method further comprises: obtaining wire drawing die apertures and wire passing amounts of N periods; fitting the wire drawing die apertures and wire passing amounts of the N periods to obtain a second model, wherein the second model is used to represent a relationship between the aperture of the wire drawing die and the wire passing amount; solving the second model according to the scrap threshold to obtain the total wire passing amount; calculating the remaining wire passing amount according to the total wire passing amount and the current wire passing amount to obtain the remaining service life of the wire drawing die.

10. A wire drawing machine comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that, The processor implements the method of any one of claims 1 to 9 when executing the computer program.

Images