Control method and system of servo press

Abstract

The invention discloses a control method and system of a servo press, and relates to the field of press control, a force displacement curve is obtained in real time from a current punching period, a key feature vector of the force displacement curve is extracted, the key feature vector is compared with an ideal golden feature vector, and a multi-dimensional deviation vector is obtained. Particularly, a deviation decoupling method based on morphological analysis is introduced, and the total deviation can be accurately decomposed into a thermal drift compensation amount caused by the thermal effect of equipment and a material characteristic compensation coefficient caused by plate performance fluctuation by analyzing translation and zooming differences of a current curve and a gold curve in form. Therefore, independent and accurate compensation can be carried out for a physical source, the problems of information aliasing and compensation failure caused by the fact that disturbance sources cannot be distinguished in the prior art are effectively solved, intelligent and feed-forward type cooperative compensation for the next stamping period is achieved, and long-term stability and high precision of the production process are ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of press machine control, and more particularly, to a control method and system of servo press machine. BACKGROUND

[0002] In the field of precision forming such as automobile manufacturing and electronic products, servo press machine is widely used due to its high precision and high flexibility. With the increasingly stringent requirements of product size tolerance in the industry, how to ensure the high consistency of product quality in the continuous high-speed stamping process lasting for several hours or tens of thousands of times has become the core driving force for the development of servo press machine control technology. In order to achieve this goal, it is crucial to build an advanced servo press machine control scheme to cope with various disturbances in the production process.

[0003] However, most of the existing servo press machine control schemes are based on a static or quasi-static process model, that is, they try to make the equipment faithfully reproduce a preset golden force-displacement curve. The essential limitation of this kind of scheme is that they regard the complex stamping process as a black box, mainly monitoring the final results such as the maximum stamping force or the position of the lower dead point at the end of stamping, and making general feedback adjustment for the next cycle accordingly. This compensation logic, which does not address the cause, often fails when faced with actual working conditions where multiple disturbances coexist. For example, when the final stamping force is detected to be decreased, the controller cannot distinguish whether it is caused by insufficient stamping depth due to thermal expansion of the equipment or by softening of the mechanical properties of the incoming material. This information aliasing phenomenon makes the controller unable to make correct compensation decisions. The wrong compensation measures may temporarily cover up the problem, but they will introduce new deviations, causing the system to oscillate between multiple error states and making it difficult to converge to a high-quality production interval. Essentially, the bottleneck of the existing technology is the failure to delve into the internal process of stamping and establish a causal mapping between process characteristic changes and physical causes. In continuous stamping, thermal drift and material property changes are the two most common and prevalent sources of product quality fluctuations, and they exist and affect each other.

[0004] Therefore, to fundamentally improve the stamping precision and stability, the core technical problem becomes how to effectively decouple the two coupled influences of thermal drift and material property changes, and implement coordinated compensation for their respective physical causes. Solving this technical problem is the inevitable requirement for moving from traditional precision control to intelligent control, and it is also the key to be overcome by the present technical scheme. SUMMARY

[0005] In the face of the above-mentioned defects in the prior art, according to an aspect of the present application, a control method of servo press machine is provided, which comprises: obtaining a current stamping force-displacement curve in the Kth stamping cycle; Extract the current stamping feature vector from the current stamping force displacement curve, and calculate the deviation between the current stamping feature vector and the golden stamping feature vector to obtain the stamping feature deviation vector; Based on the stamping golden curve, the stamping characteristic deviation vector and the current stamping force displacement curve are decoupled based on morphological analysis to obtain the thermal drift compensation amount and material property compensation coefficient. Based on thermal drift compensation and material property compensation coefficient, the original target trajectory is compensated to obtain the compensated target trajectory for the next cycle. Based on the compensated target trajectory of the next cycle, control commands for the servo drive are generated.

[0006] According to another aspect of this application, a control system for a servo press is provided, comprising: The current punching force displacement curve acquisition module is used to acquire the current punching force displacement curve in the Kth punching cycle. The feature deviation calculation module is used to extract the current stamping feature vector from the current stamping displacement curve and calculate the deviation between the current stamping feature vector and the golden stamping feature vector to obtain the stamping feature deviation vector. The morphological analysis module is used to perform morphological analysis-based deviation decoupling on the stamping characteristic deviation vector and the current stamping force displacement curve based on the stamping golden curve to obtain the thermal drift compensation amount and material property compensation coefficient. The compensation module is used to compensate the original target trajectory based on the thermal drift compensation amount and the material property compensation coefficient to obtain the compensated target trajectory for the next cycle. The control command generation module is used to generate control commands for the servo drive based on the compensated target trajectory of the next cycle.

[0007] Compared with existing technologies, the control method and system for a servo press provided in this application abandons the traditional black-box control mode that relies on a single endpoint value, and instead utilizes the complete force-displacement curve as the process information carrier. By acquiring the force-displacement curve in real time from the current stamping cycle and extracting its key feature vector, it compares it with the ideal golden feature vector to obtain a multi-dimensional deviation vector. The key innovation lies in its introduction of a deviation decoupling method based on morphological analysis: by analyzing the translation and scaling differences between the current curve and the golden curve in terms of shape, the total deviation can be accurately decomposed into thermal drift compensation caused by the thermal effect of the equipment and material property compensation coefficient caused by the fluctuation of sheet metal properties. In this way, independent and precise compensation can be performed for the physical root cause, effectively solving the problem of information aliasing and compensation failure caused by the inability to distinguish the source of disturbance in existing technologies. This enables intelligent, feedforward collaborative compensation for the next stamping cycle, ensuring long-term stability and high precision of the production process. Attached Figure Description

[0008] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0009] Figure 1 This is a flowchart of a control method for a servo press according to an embodiment of this application.

[0010] Figure 2 This is a schematic diagram of the data flow of a control method for a servo press according to an embodiment of this application.

[0011] Figure 3 This is a flowchart of step S2 in the control method of the servo press according to an embodiment of this application.

[0012] Figure 4 This is a flowchart of step S4 in the control method of the servo press according to an embodiment of this application.

[0013] Figure 5 This is a block diagram of the control system of a servo press according to an embodiment of this application. Detailed Implementation

[0014] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. It should be understood that the drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0015] In view of the problems existing in the application of the aforementioned technologies, this application proposes a control method for a servo press. Figure 1 This is a flowchart of a control method for a servo press according to an embodiment of this application. Figure 2 This is a schematic diagram of the data flow in the control method of a servo press according to an embodiment of this application. Figure 1 and Figure 2As shown, the control method of the servo press according to an embodiment of this application includes: S1, obtaining the current stamping displacement curve in the Kth stamping cycle; S2, extracting the current stamping feature vector from the current stamping displacement curve, and calculating the deviation between the current stamping feature vector and the golden stamping feature vector to obtain the stamping feature deviation vector; S3, based on the golden stamping curve, performing deviation decoupling on the stamping feature deviation vector and the current stamping displacement curve based on morphological analysis to obtain the thermal drift compensation amount and the material property compensation coefficient; S4, based on the thermal drift compensation amount and the material property compensation coefficient, compensating the original target trajectory to obtain the compensated target trajectory for the next cycle; S5, generating control commands for the servo drive based on the compensated target trajectory for the next cycle.

[0016] In step S1, the current stamping force-displacement curve for the Kth stamping cycle is obtained. It should be understood that in the precision stamping process performed by a servo press, any minute disturbance can lead to dimensional deviations in the final product. Existing control methods often rely on a single scalar value of the final state, such as the maximum stamping force or the bottom dead center position, to provide feedback adjustment to the production process. The fundamental limitation of this approach lies in the one-sidedness of its information; it cannot effectively distinguish between multiple physical disturbance sources such as thermal drift and changes in material properties, resulting in a lack of targeted compensation measures and making it difficult to fundamentally guarantee quality consistency in continuous production. Therefore, to achieve accurate identification and decoupling compensation of disturbance sources, an information carrier that can comprehensively and dynamically reflect the inherent physical characteristics of the stamping process is needed, and a complete force-displacement curve can provide this overall process information. Obtaining the current stamping force-displacement curve for the Kth stamping cycle is precisely to capture this complete process fingerprint information, providing the necessary and sufficient data foundation for subsequent feature extraction, deviation decoupling, and collaborative compensation.

[0017] In one possible implementation, step S1, obtaining the current stamping force displacement curve in the Kth stamping cycle, includes: S11, obtaining the original force sensor timing reading and the original slider position timing reading of the Kth stamping cycle; S12, performing digital low-pass filtering and curve construction on the original force sensor timing reading and the original slider position timing reading of the Kth stamping cycle to obtain the current stamping force displacement curve.

[0018] In the above embodiment, the operation flow of step S1 is as follows: In a high-strength steel sheet stamping application for automobiles, the servo press is set to work continuously at a stroke rate of 60 strokes per minute. Before formally starting the steps of this embodiment, a complete stamping cycle needs to be predefined. This cycle is determined by the motion trajectory of the servo motor. For example, it is defined as the process of the slider moving down from the top dead center (e.g., position coordinate 400 mm), passing through the bottom dead center (e.g., position coordinate 150 mm), and then returning to the top dead center.

[0019] S11 is executed at the start of the Kth (e.g., the 1000th) stamping cycle. In this embodiment, the force sensor is a high-precision piezoelectric sensor mounted on the fuselage column, whose output charge signal is converted into a voltage signal proportional to the force via a charge amplifier. The slider position is measured by an optical linear encoder with a resolution of 1 micrometer, which is directly mounted on the side of the slider. A high-speed data acquisition module performs analog-to-digital conversion on the analog signals from the two sensors at a synchronous sampling frequency of 10 kHz, thereby obtaining two sets of discrete time-series data. The raw force sensor timing readings of the Kth stamping cycle are a set of noisy force value sequences, such as [0.1, 0.2, 0.5, ..., 250.8, 498.5, 501.2, 499.1, ..., 0.3] kN, with a time interval of 0.1 ms between data points. The sequence contains high-frequency noise caused by motor drive, mechanical vibration, and electromagnetic environment. Meanwhile, the original slider position timing readings obtained for the Kth stamping cycle are slider position sequences corresponding to the same time point, such as [400.000,399.985,...,155.321,150.002,150.001,150.003,...,399.998] mm.

[0020] Subsequently, step S12 is executed. The purpose of filtering is to filter out measurement noise to the greatest extent possible without damaging the actual stamping process signal. Specifically, the original force sensor timing readings and the original slider position timing readings are respectively input into a pre-configured third-order Butterworth digital low-pass filter. The transfer function structure of this filter is fixed, and its key parameter is the cutoff frequency, which is set according to the dynamic characteristics of the stamping process. For the stamping speed of 60 times per minute in this embodiment, the main force-displacement changes occur within a time scale of approximately 100 milliseconds. Therefore, the cutoff frequency of the filter is set to 500 Hz, which can effectively suppress noise above 1 kHz while allowing low-frequency signals containing the main information of the stamping process to pass through without distortion. After filtering, smoothed force data sequences are obtained, such as [0.1,0.2,0.4,...,250.5,499.0,500.0,499.5,...,0.2] kN, and smoothed position data sequences, such as [400.000,399.985,...,155.320,150.001,150.001,150.002,...,399.998] mm. It can be seen that the sharp fluctuations between adjacent data points in the original data are effectively smoothed. Next, curve construction is performed, a process based on the principle of timestamp alignment. Since the two sets of smoothed data sequences were obtained under the same sampling clock, the i-th force data point and the i-th position data point correspond to the same time. By pairing these two sets of sequences point by point, a set of discrete (position, force) data points can be formed, such as {(400.000,0.1),(399.985,0.2),...,(150.001,500.0),...,(399.998,0.2)}. This set of data points, plotted in a two-dimensional coordinate system with displacement as the abscissa and force as the ordinate, constitutes a current stamping force-displacement curve that accurately reflects the dynamic relationship between force and slider displacement in the Kth stamping cycle.

[0021] In step S2, the current stamping feature vector is extracted from the current stamping displacement curve, and the deviation between the current stamping feature vector and the golden stamping feature vector is calculated to obtain the stamping feature deviation vector. Correspondingly, after obtaining the force-displacement curve containing rich physical information about the stamping process, the original curve data itself is high-dimensional and unstructured, making it difficult to directly use for quantitative closed-loop control. To extract core indicators that accurately characterize the current process state from this complex process fingerprint and quantitatively compare it with the ideal state, the curve information needs to be reduced in dimensionality and characterized. Therefore, extracting the current stamping feature vector from the current stamping displacement curve and calculating its deviation from the golden standard is to transform the abstract curve shape difference into a structured input composed of specific physical quantity deviations, which can be used for precise decoupling and control.

[0022] Figure 3 This is a flowchart of step S2 in the control method of the servo press according to an embodiment of this application. In one possible implementation, such as... Figure 3 As shown, step S2, extracting the current stamping feature vector from the current stamping displacement curve and calculating the deviation between the current stamping feature vector and the golden stamping feature vector to obtain the stamping feature deviation vector, includes: S21, performing key physical feature calculations based on the current stamping displacement curve to obtain the maximum stamping force, actual bottom dead center position, elastic segment stiffness, and total stamping work, wherein the maximum stamping force, actual bottom dead center position, elastic segment stiffness, and total stamping work constitute the current stamping feature vector; S22, calculating the element-by-element difference between the current stamping feature vector and the golden stamping feature vector to obtain the stamping feature deviation vector.

[0023] In the above implementation, the operation flow of step S2 is as follows: using this curve as input, step S21 is executed, that is, the calculation of key physical characteristics is performed based on this curve. The selection of these physical characteristics closely revolves around the core physical laws of the stamping process, comprehensively characterizing the state of the current stamping cycle from four dimensions: peak force, endpoint of position, material response characteristics, and energy consumption. First, the first characteristic is calculated: maximum stamping force. This calculation process is achieved by traversing all force values ​​in the current stamping force displacement curve data set. In the aforementioned embodiment, the filtered force data sequence is [0.1,0.2,0.4,...,250.5,499.0,500.0,499.5,...,0.2] kN. Through a simple peak-finding algorithm, the maximum value in this sequence can be determined to be 500.0 kN. This value directly reflects the maximum normal force exerted by the die on the workpiece in this stamping cycle, and is the most direct indicator for measuring the stamping load. Next, the second characteristic is calculated: actual bottom dead center position. This process is similar to the calculation of the maximum stamping force, but it traverses the position data. According to the aforementioned embodiment, the slider position coordinate system is defined as a decrease in position value during downward movement; therefore, the actual bottom dead center position corresponds to the minimum value in the position data sequence. For the smoothed position sequence [400.000, 399.985, ..., 155.320, 150.001, 150.001, 150.002, ..., 399.998] mm, its minimum value is calculated to be 150.001 mm. This value accurately represents the deepest position reached by the slider movement in this cycle, directly related to the height dimension of the finished part. Then, the third feature is calculated: elastic segment stiffness. This feature reflects the elastic deformation resistance of the entire system consisting of mold-workpiece-press during the initial return phase of stamping loading, and its variation is mainly related to the elastic modulus and thickness of the workpiece material. The calculation of this value is divided into two steps. The first step is to define a purely elastic loading interval on the current stamping displacement curve. To ensure the linearity of the selected interval and avoid the nonlinear region of initial contact and plastic deformation, this interval is defined by a force value range. Specifically, using the calculated maximum punching force of 500.0 kN as a benchmark, a force range of 10% to 40% (i.e., from 50.0 kN to 200.0 kN) is selected. This force range is determined to avoid the nonlinear instability stage during the initial contact between the punch and the workpiece, and to ensure that the stiffness calculation is entirely within the elastic loading segment where the force-displacement relationship is most linear, before significant plastic deformation of the material occurs. Subsequently, from the complete set of current punching force-displacement curve data points, all data points with force values ​​within this range are selected. The second step involves performing least squares linear regression analysis on the selected subset of data points to find a best-fitting straight line, in the form of Force = Slope × Position + Intercept. The slope of this straight line is defined as the elastic segment stiffness of that punching cycle.For example, the slope of the best-fit line for this interval was calculated to be 300.0 kN / mm. Finally, the fourth feature is calculated: total stamping work. This feature represents the total mechanical work done by the press on the workpiece during the entire forming process, from the moment the punch contacts the workpiece to the moment the slide reaches the bottom dead center. Its magnitude is closely related to the material's yield strength, work hardening index, and deformation. During calculation, the starting point of the stamping work needs to be accurately determined on the current stamping pressure-displacement curve, i.e., the moment when the punch and workpiece just make effective contact. This can be achieved by setting a small force threshold, such as 0.5% of the maximum stamping pressure, or 2.5 kN. This force threshold is an engineering parameter determined through experimental optimization during the process calibration phase, designed to ensure that it can sensitively capture the true starting point of the material's deformation under stress, while effectively ignoring the inherent background noise signal of the sensor. Starting from the beginning of the curve, the first data point with a force value exceeding 2.5 kN is identified as the contact point. Subsequently, numerical integration is used to calculate the position from this contact point. to the actual bottom dead center position Within the interval (150.001 mm), the area enclosed by the current impact force displacement curve and the position axis. In the case of discrete data points, this is achieved using the trapezoidal rule, and its calculation formula is: In this formula, the summation symbol Σ indicates that each pair of adjacent data points (the i-th and the (i+1)-th) between the contact point and the bottom dead center is traversed. and These are the force values ​​at these two points, This represents the average force within this tiny interval. and This corresponds to the position value. Since the position decreases downwards in this embodiment, This represents the displacement increment within this tiny interval, and its value is positive. By calculating and summing the areas of each tiny trapezoid, the total stamping power is obtained. For example, after accumulating all relevant data points, the total stamping power of this cycle is 12.5 kJ. Finally, they are combined in a preset order [maximum stamping force, actual bottom dead center position, elastic segment stiffness, total stamping power] to form the output of step S21, namely the current stamping feature vector [500.0, 150.001, 300.0, 12.5].

[0024] After feature extraction is completed, proceed to step S22. The gold stamping feature vector represents the standard stamping characteristics that should be achieved under ideal production conditions, i.e., the equipment is in a thermally stable state and raw materials that fully meet specifications are used. This gold vector is obtained through a teaching or calibration process before the production task begins. Specifically, the operator selects a batch of sheet metal of the highest quality and performs several trial stampings after the press is cold-started and runs to thermal equilibrium. In particular, the gold stamping feature vector includes the gold elastic segment stiffness, the gold total stamping power, the gold maximum stamping force, and the gold bottom dead center position. The force-displacement curves of these near-perfect stamping cycles are collected, and their feature vectors are extracted using the same method as in step S21. Finally, the average value of each component of these vectors is taken to obtain the gold stamping feature vector for this production task. For example, after calibration, the obtained gold vector is [Gold maximum stamping force 510.0 kN, gold bottom dead center position 150.000 mm, gold elastic segment stiffness 305.0 kN / mm, gold total stamping power 12.8 kJ]. The process of calculating the deviation is a simple element-by-element vector subtraction: Stamping characteristic deviation vector = Current stamping characteristic vector - Golden stamping characteristic vector. Substituting the values ​​in this embodiment, we get: Stamping characteristic deviation vector = [500.0-510.0, 150.001-150.000, 300.0-305.0, 12.5-12.8] = [-10.0, +0.001, -5.0, -0.3]. In this vector, the maximum stamping force is 10.0 kN lower than the ideal value; the actual bottom dead center position is 0.001 mm higher than the ideal value (i.e., the stamping depth is slightly insufficient); the stiffness of the system's elastic section is reduced by 5.0 kN / mm; and the total stamping energy is reduced by 0.3 kJ. This deviation vector, in a highly condensed and quantitative way, precisely indicates the direction and magnitude of the deviation between the current stamping cycle and the ideal state.

[0025] In step S3, based on the stamping golden curve, the stamping characteristic deviation vector and the current stamping force-displacement curve are decoupled based on morphological analysis to obtain the thermal drift compensation amount and the material property compensation coefficient. It is understandable that after obtaining the deviation vector characterizing the current stamping cycle state, this vector itself is only a comprehensive representation of the final result; it indicates the deviation from the ideal state but fails to reveal the fundamental physical causes of these deviations. During continuous stamping, the thermal expansion effect of the equipment causes the entire force-displacement curve to shift along the position axis, while fluctuations in the mechanical properties of the raw materials change the slope and shape of the curve. The influence of these two physical phenomena is coupled together, jointly leading to the final deviation. Without indiscriminate compensation, precise control cannot be achieved. Therefore, in order to extract independent control inputs directly corresponding to the physical root cause from this coupled deviation information, this application, based on the stamping golden curve, performs deviation decoupling based on morphological analysis on the stamping characteristic deviation vector and the current stamping force-displacement curve to decompose the overall, fuzzy deviation into a pure position offset caused by thermal effects and a shape adjustment coefficient caused by changes in material properties.

[0026] In one possible implementation, step S3, based on the stamping golden curve, decouples the stamping characteristic deviation vector and the current stamping displacement curve based on morphological analysis to obtain the thermal drift compensation amount and material property compensation coefficient, including: S31, extracting high-pressure segment curve slices from the stamping golden curve and the current stamping displacement curve to obtain the stamping golden high-pressure segment curve and the current stamping displacement high-pressure segment curve; S32, solving for the optimal position offset of the stamping golden high-pressure segment curve and the current stamping displacement high-pressure segment curve to obtain the thermal drift compensation amount.

[0027] In the above implementation, the operation flow of step S3 is as follows: The stamping gold curve is generated in the same calibration stage as the acquisition of the gold stamping feature vector. Specifically, when using the optimal sheet metal and ensuring the press reaches a thermally stable state for trial stamping, the nearly perfect complete force-displacement curve that can represent the ideal process is collected and stored as the stamping gold curve. It serves as the benchmark fingerprint for morphological comparison in all subsequent stamping cycles.

[0028] At the start of the decoupling process, step S31 is executed. The physical significance of this step lies in the fact that the force-displacement relationship during the high-pressure stage of the stamping process—the stage where the punch and workpiece are deeply engaged and the material undergoes significant plastic flow—is primarily determined by the rigid profile of the die and the constitutive relationship of the material. The curve shape of this section is most sensitive to minute positional shifts caused by thermal drift and is least affected by noise from the initial contact stage. Therefore, by focusing the analysis on this key section, displacement information related to thermal drift can be extracted most effectively. Specifically, the maximum stamping force of gold is first read from the gold stamping feature vector [510.0, 150.000, 305.0, 12.8], which has a value of 510.0 kN. Based on this value, a lower limit threshold for the force value of a high-pressure segment is set. For example, it can be set to 80% of the maximum stamping force of gold, i.e., 0.8 × 510.0 = 408.0 kN. The determination of this force value threshold aims to capture the stamping stage where the material has entered deep plastic deformation and the die and workpiece have formed a stable rigid contact. This is because the curve shape at this stage is most sensitive to the positional translation caused by thermal effects and can minimize the nonlinear interference of the initial contact and elastic deformation stages. Subsequently, all data points with force values ​​greater than 408.0 kN are selected from both the gold stamping curve and the current stamping force-displacement curve. For the gold stamping curve, the set of selected data points constitutes the high-pressure segment curve of the gold stamping, whose shape accurately represents the force-displacement response when the material is fully deformed under ideal conditions. For the current stamping force-displacement curve (from the 1000th cycle), all data points with force values ​​greater than 408.0 kN are also selected to constitute the current high-pressure segment curve of the stamping force-displacement curve.

[0029] Next, step S32 is executed. Thermal drift, as reflected in the force-displacement curve, is essentially a rigid translation. Due to thermal expansion of the machine body and mold, the actual mold clearance of the slider changes under the same encoder reading. This change causes the current curve to shift relative to the golden curve on the position axis. Solving for this shift is equivalent to translating the current high-pressure segment curve of the stamping displacement on the position axis until it achieves an optimal match with the shape of the stamping golden high-pressure segment curve. This optimal translation is the quantification result of the thermal drift. In this embodiment, this optimal position shift is solved by calculating the cross-correlation function of the two curve slices. In one possible implementation, step S32, solving for the optimal position shift of the stamping golden high-pressure segment curve and the current high-pressure segment curve to obtain the thermal drift compensation amount, includes: solving for the optimal position shift of the stamping golden high-pressure segment curve and the current high-pressure segment curve using the following formula: ; ;in, To press the high-pressure section curve of gold, This is the current pressure-displacement high-pressure section curve. The position of the curve. This is a positional offset, a tentative offset, meaning it assumes the current curve has shifted relative to the golden curve. , This is the thermal drift compensation amount. The first formula defines the correlation function. The calculation method. Its physical meaning is to measure when... The curve was shifted After that, with The degree of similarity between curves. During calculation, for a given... traverse position coordinates In each Point, read the force value corresponding to the golden curve. Then read the current curve's position after the offset. Force value at the point Then multiply these two force values. Finally, sum the products at all position points to obtain Σ. If the two curves are at this offset... The shape matches very well, so in most cases At a given point, higher force values ​​are multiplied together, and lower force values ​​are multiplied together, resulting in a final sum. A large positive value is obtained. Conversely, if the shapes do not match, the sum of multiplications will be smaller. The second formula indicates the method for solving the thermal drift compensation. By systematically changing the trial offset in very small steps within a preset, reasonable range. The value, and for each Calculate the corresponding relevant function values Ultimately, this will enable the relevant functions... The one that reaches the maximum value The position offset that is determined to be optimal is denoted as the thermal drift compensation amount. In practice, the first step is to determine... The search range and step size are determined. Considering that the thermal drift of a servo press is typically at the micrometer level, the search range can be set between -0.050 mm and +0.050 mm, with a step size of 0.0001 mm to ensure sufficient solution accuracy. The calculation process is as follows: 1. Let τ = -0.0500 mm, calculate R(-0.0500). 2. Let τ = -0.0499 mm, calculate R(-0.0499). 3.... and so on, until τ = +0.0500 mm, calculate R(+0.0500). In this series of calculations, for example, when calculating R(+0.0120), it means uniformly subtracting 0.0120 mm from the position coordinates of all data points on the current punch displacement high-pressure section curve, i.e., shifting to the left, and then multiplying and accumulating with the fixed golden high-pressure section curve. After calculating all 1001 τ values, a result will be obtained. about The function curve. By finding the peak point of this curve, the optimal offset can be determined. In the aforementioned embodiment, the total position deviation calculated in step S2 is +0.001 mm. However, by matching the curve shape of the high-pressure section, it was found that... The function reaches its maximum value at τ = +0.0120 mm. Therefore, the final output thermal drift compensation is... =+0.012 mm. That is, although the final measured bottom dead center position deviation is only +0.001 mm, in-depth analysis of the curve shape reveals a systematic positional offset of +0.012 mm caused by thermal effects. This means that the thermal expansion of the machine tool resulted in an equivalent stamping depth loss of 12 micrometers.

[0030] In one possible implementation, step S3, which involves decoupling the stamping characteristic deviation vector and the current stamping displacement curve based on morphological analysis to obtain the thermal drift compensation amount and the material property compensation coefficient, further includes: S33, extracting the elastic segment stiffness deviation and the total stamping work deviation from the stamping characteristic deviation vector; S34, extracting the gold elastic segment stiffness and the gold total stamping work from the gold stamping characteristic vector; and S35, determining the material property compensation coefficient based on the elastic segment stiffness deviation, the total stamping work deviation, the gold elastic segment stiffness, and the gold total stamping work.

[0031] In the second stage of this decoupling process, step S33 is executed first. Based on the calculation results of the previous embodiment, the obtained stamping characteristic deviation vector is [-10.0, +0.001, -5.0, -0.3], whose four components correspond to the maximum stamping force deviation, the actual bottom dead center position deviation, the elastic segment stiffness deviation, and the total stamping work deviation, respectively. The elastic segment stiffness is directly related to the elastic modulus of the material, while the total stamping work comprehensively reflects the yield strength and work hardening characteristics of the material. Therefore, these two deviation components are key indicators characterizing changes in material properties. This step extracts the third and fourth elements from the vector through a simple analytical operation. Thus, the elastic segment stiffness deviation = -5.0 kN / mm and the total stamping work deviation = -0.3 kJ are obtained.

[0032] Next, step S34 is executed. The gold stamping eigenvector = [510.0, 150.000, 305.0, 12.8] represents the ideal process state. Similarly, through analytical operations, the third and fourth elements are extracted. Thus, the gold elastic segment stiffness = 305.0 kN / mm, and the total gold stamping work = 12.8 kJ are obtained. These two values ​​constitute a stable benchmark for calculating the relative deviation.

[0033] After obtaining the absolute value of the deviation and the corresponding reference value, proceed to step S35. This step aims to determine a single, dimensionless material property compensation coefficient based on the above information. In one possible implementation, step S35, based on the elastic segment stiffness deviation, the total stamping work deviation, the golden elastic segment stiffness, and the golden total stamping work, determines the material property compensation coefficient, including: based on the elastic segment stiffness deviation, the total stamping work deviation, the golden elastic segment stiffness, and the golden total stamping work, the material property compensation coefficient is determined using the following formula, which is: ;in, For the stiffness deviation of the elastic segment, The stiffness is the golden elastic segment. For the total stamping power deviation, This represents the total work done in gold stamping. and For preset weighting coefficients, This is the material property compensation coefficient. This process involves two main calculation steps: calculating the relative rate of change of deviation and weighted fusion calculation. The first calculation step is calculating the relative rate of change. Directly comparing a stiffness deviation of -5.0 kN / mm and a total work deviation of -0.3 kJ is physically meaningless because their dimensions and numerical scales are completely different. To eliminate this difference and allow for comparison and fusion within a unified framework, their absolute deviations need to be converted into dimensionless relative rates of change. The specific calculation is as follows: First, calculate the relative rate of change of stiffness = elastic segment stiffness deviation / golden elastic segment stiffness. Substituting the values ​​from the example, we get: relative rate of change of stiffness = -5.0 / 305.0 ≈ -0.01639. This result indicates that the system elastic segment stiffness of the current stamping cycle has decreased by approximately 1.64% compared to the ideal state. Then, calculate the relative rate of change of total work = total stamping work deviation / golden stamping total work. Substituting the values ​​from the example, we get: the relative change rate of total work = -0.3 / 12.8 ≈ -0.02344. This result indicates that the total work consumed in the current stamping cycle is reduced by approximately 2.34% compared to the ideal state. The second calculation step is a weighted fusion calculation. In order to obtain a final, comprehensive compensation coefficient, a weighted summation method is used to integrate the two indicators. and These are preset weighting coefficients, and their setting is crucial, reflecting the degree of importance attached to stiffness and total work changes when evaluating the overall performance of materials. The sum of these two weighting coefficients is set to 1 (i.e., + =1), making the final result a weighted average of the two relative rates of change. Weighting coefficient and The determination is based on a deep understanding of the specific stamping process and the analysis of experimental data. For example, for some precision stamping parts with extremely high requirements for springback control, the finished dimensions are very sensitive to the elastic modulus of the material (i.e., the stiffness of the elastic segment), which may require... A larger value (e.g., 0.7). However, for some deep-drawn parts characterized by large deformation, the forming quality depends more on the material's flow and hardening properties (reflected in the total stamping work), which may impart a higher value. A relatively large value (e.g., 0.8). In this embodiment, for the stamping of high-strength steel sheets for automobiles, the initial stiffness of the material and the energy absorption during the entire plastic deformation process are equally important. However, considering that the total stamping work integrates more comprehensive plastic performance information such as yielding and hardening, a weight is set. =0.4, =0.6, indicating a relatively greater emphasis on the impact of total work change on material properties. Substituting all values ​​into the weighted fusion formula above, the material property compensation coefficient can be calculated: =0.4*(-0.01639)+0.6*(-0.02344)=-0.02062, which is a dimensionless scalar. Its negative sign indicates that the overall mechanical properties of the currently used sheet metal are softer or have lower deformation resistance than the ideal gold standard material. Its value, approximately -2.06%, precisely quantifies the degree of this deviation. This coefficient, together with the previously decoupled thermal drift compensation of +0.012 mm, constitutes a complete, decoupled quantitative description of the current stamping cycle state deviation.

[0034] Specifically, in the aforementioned disturbance decoupling scheme, the calculation of the material property compensation coefficient mainly relies on the deviations in two dimensions: the stiffness of the elastic segment and the total stamping work. However, this method treats the two physical quantities as independent indicators, ignoring the inherent physical coupling relationship between the ultimately measured deviations. Preferably, to obtain higher decoupling accuracy and estimation robustness, the approach of independently estimating the optimal position offset and material property compensation coefficient is improved to coupling effect deconstruction and multi-factor fusion estimation. The core is to recognize that the final observed total deviation of the bottom dead center position is the result of the combined effect of equipment thermal drift and changes in material springback characteristics. Therefore, using the most reliable preliminary estimate of thermal drift obtained through cross-correlation analysis of the high-pressure segment curves, the position deviation component caused by changes in material properties, i.e., the equivalent springback deviation, can be accurately extracted from the total position deviation. This new information, containing the elastic recovery characteristics of the material, along with another robust characteristic—the maximum stamping force deviation—is then incorporated into the calculation model of the material property compensation coefficient. This allows for the construction of a more accurate and stable quantitative index of material properties that comprehensively reflects the four dimensions of material strength, elasticity, toughness, and elastic recovery, thereby enabling higher-order and more reliable intelligent compensation for the stamping process.

[0035] Based on this, in a possible preferred embodiment, step S35, based on the elastic segment stiffness deviation, the total stamping power deviation, the golden elastic segment stiffness, and the golden stamping power, determines the material property compensation coefficient, including: performing servo press disturbance decoupling based on closed-loop feedback and multi-feature fusion on the elastic segment stiffness deviation, the total stamping power deviation, the golden elastic segment stiffness, and the golden stamping power to obtain the material property compensation coefficient.

[0036] More specifically, the material property compensation coefficients are obtained by decoupling the servo press disturbance based on closed-loop feedback and multi-feature fusion for the stiffness deviation of the elastic segment, the total stamping power deviation, the stiffness of the golden elastic segment, and the total stamping power, including: The equivalent springback deviation of material properties is calculated based on the bottom dead center position deviation and thermal drift compensation in the stamping characteristic deviation vector. It should be understood that the total bottom dead center position deviation measured in the preceding steps is a comprehensive reflection of the combined effects of two different physical mechanisms: one is the physical positional shift caused by the thermal expansion and contraction of the machine tool, i.e. Secondly, the difference in post-forming position is due to the difference between the current sheet metal's springback characteristics and the ideal sheet metal. Without differentiation, targeted compensation is impossible. Therefore, this application separates these two coupled effects through a deconstruction operation, obtaining a physical quantity that can purely and directly quantify the change in the material's elastic recovery capability: the equivalent springback deviation. This provides a crucial new dimension for subsequent multi-factor fusion evaluation. The specific implementation process involves performing a vector subtraction. It utilizes the physical relationship that the total deviation equals the sum of the deviations of each component. Therefore, subtracting the known thermal drift component from the total deviation leaves the component caused by material properties. The calculation formula is: ,in, It is the material property equivalent springback deviation. It's a bottom dead center position deviation. This is the thermal drift compensation amount. In the above embodiment, the stamping characteristic deviation vector = [-10.0, +0.001, -5.0, -0.3] has been obtained, from which the total deviation of the bottom dead center position can be determined. =+0.001 mm. Meanwhile, the thermal drift compensation amount has also been calculated through morphological analysis of the high-pressure section curve. =+0.012 mm. Substitute these values ​​into the above formula: =0.001-0.012=-0.011 mm. The calculated equivalent springback deviation of the material properties is -0.011 mm, which means that the actual springback of the material is 11 micrometers larger than the ideal state. This is logically consistent with the previous conclusion that the material is relatively soft and provides quantitative data support. This step utilizes the final and most realistic physical observation (total deviation of the bottom dead center position). This achieves a closed-loop calibration. It breaks the reliance on thermal drift estimation ( ) and material property estimation ( The assumption is that the two compensation values ​​are considered as two independent parallel lines. By calculating the equivalent springback deviation, any small errors in the initial estimate of thermal drift (e.g., those coupled with the springback effect) are not discarded, but are fed back into the subsequent calculation of the material property compensation coefficient. In this way, the estimates of the two compensation values ​​are no longer isolated guesses, but form a mechanism of mutual verification and synergistic approximation of the true physical state, thereby significantly improving the final accuracy of decoupling.

[0037] This paper describes a multi-dimensional feature deviation normalization and weighted fusion method for the deviations of stiffness and total stamping power in the elastic segment, stiffness and total stamping power in the elastic segment, maximum stamping force deviation and maximum stamping force, as well as the equivalent springback deviation and material thickness, to obtain a material property compensation coefficient. In other words, relying solely on stiffness and total power to evaluate material properties results in a relatively simplistic evaluation model. When any feature measurement is affected by noise, the compensation coefficient is prone to significant fluctuations. To significantly improve the robustness and noise resistance of the evaluation results, this application expands the evaluation dimensions from two to four. It introduces two new, high-quality feature dimensions: maximum stamping force, which directly reflects the material's strength and has a high and stable signal strength; and the equivalent springback deviation decoupled in the previous step, which accurately characterizes the material's elastic recovery properties. By normalizing and weightedly fusing the information from these four dimensions (strength, elasticity, toughness, and elastic recovery), a compensation coefficient that comprehensively reflects the material's overall mechanical properties without relying on any single feature is obtained. This ensures that even with minor fluctuations in a particular feature, the final compensation decision remains smooth and reliable. The specific implementation involves applying an extended weighted summation formula. First, the deviations of the four dimensions are normalized to eliminate the influence of units and numerical ranges. Then, a weighted fusion is performed based on their respective importance. The calculation formula is as follows: Specifically, each deviation comes from the calculation results of the previous steps: elastic segment stiffness deviation. =-5.0 kN / mm; Total stamping power deviation =-0.3 kJ; Maximum impact force deviation =-10.0 kN; and the equivalent springback deviation calculated in the previous step. =-0.011 mm. The various gold reference values ​​also originate from previous steps: gold elastic segment stiffness. =305.0 kN / mm; Total power of gold stamping =12.8 kJ; maximum resistance level for gold =510.0 kN. This is the nominal thickness of the workpiece, a preset process parameter. In this application of high-strength steel sheet stamping for automobiles, the sheet thickness is set. =1.5 mm. This parameter is used to convert springback deviation, which has the dimension of length, into a dimensionless relative value. , , , Four preset weighting coefficients are used, summing to 1. Their settings are based on process experience and the assessment of the importance of each feature. In this embodiment, maximum impact force is considered the strongest indicator of material strength, springback and total work are also crucial, while stiffness has a relatively smaller impact. Therefore, the weights can be set as follows: =0.15 (corresponding to stiffness), =0.25 (corresponding to total power) =0.35 (corresponding to the maximum impact force), =0.25 (corresponding to springback). Substitute all values ​​into the formula for calculation: First, calculate the four normalized relative rates of change: Stiffness term: -5.0 / 305.0≈-0.01639; Total work term: -0.3 / 12.8≈-0.02344; Maximum impact force term: -10.0 / 510.0≈-0.01961; Springback term: -0.011 / 1.5≈-0.00733. Then perform a weighted summation: =0.15×(-0.01639)+0.25×(-0.02344)+0.35×(-0.01961)+0.25×(-0.00733)≈-0.01701. Therefore, the material property compensation coefficient calculated through this preferred embodiment is -0.01701. Compared to the -2.06% obtained by the basic method, the superiority of the -1.70% compensation coefficient obtained by this preferred method lies in the fact that the judgment of material properties no longer relies solely on the stiffness (elasticity) and area (toughness) of the force-displacement curve, but integrates complete information from four dimensions: peak force (strength), stiffness (elasticity), work done (toughness), and rebound (elastic recovery). This multi-factor fusion mechanism makes the estimation results more robust and resistant to noise. Specifically, even if a noise-sensitive feature (such as total stamping power) experiences slight fluctuations due to measurement noise or local process variations, the presence of the other three features (especially the maximum stamping force as a strong signal) can play a significant stabilizing role, ensuring that the final calculated material property compensation coefficient will not undergo drastic changes, thereby making the overall control compensation performance smoother and more reliable.

[0038] In step S4, the original target trajectory is compensated based on the thermal drift compensation amount and the material property compensation coefficient to obtain the compensated target trajectory for the next cycle. It should be understood that in the preceding steps, through in-depth insight and morphological analysis of the stamping process, the complex and coupled process deviations have been successfully decomposed into two independent quantitative indicators with clear physical meaning, corresponding to equipment thermal effects and material property fluctuations: the thermal drift compensation amount and the material property compensation coefficient. However, these two indicators are merely diagnostic results. To truly achieve closed-loop control of stamping quality, this diagnostic information needs to be transformed into precise correction instructions for the future behavior of the press. The motion of the servo press follows a preset target trajectory (containing the target profile of position and force). Therefore, in order to put the decoupled compensation information into practice and proactively eliminate potential deviations in the next cycle in a feedforward manner, this original target trajectory needs to be dynamically and specifically modified to generate a completely new compensated target trajectory that adapts to the current equipment state and material properties.

[0039] Figure 4 This is a flowchart of step S4 in the control method of the servo press according to an embodiment of this application. In one possible implementation, such as Figure 4 As shown, step S4, based on the thermal drift compensation amount and the material property compensation coefficient, compensates the original target trajectory to obtain the compensated target trajectory for the next cycle, including: S41, applying thermal drift compensation to the original target trajectory to obtain the thermal drift compensated target trajectory; S42, applying material property compensation to the thermal drift compensated target trajectory to obtain the compensated target trajectory for the next cycle.

[0040] In the above implementation, the operation flow of step S4 is as follows: The original target trajectory is calibrated and generated together with the gold stamping curve and the gold feature vector before the production task begins. It is an ideal curve that defines the change of slider position and stamping force with time or position under ideal working conditions, and is the reference command for the servo drive to perform trajectory tracking. One of its key parameters, namely the original target bottom dead center position, is consistent with the gold bottom dead center position, which is 150.000 mm.

[0041] After obtaining two independent compensation values, the compensation process is performed in two steps. First, step S41 is executed to accurately offset the equivalent stamping depth loss caused by the thermal expansion of the equipment. Based on the decoupling results of the previous steps, the thermal drift compensation value is +0.012 mm. This positive value indicates a 12-micron stamping depth loss due to thermal expansion. To compensate for this loss, the servo motor is instructed to drive the slider to move an additional 12 microns downward in the next cycle. Since downward movement of the coordinate system in this embodiment represents a decrease in coordinate value, the compensation operation is a complete and pure translation of the position profile of the original target trajectory. The compensation formula is: Target position after thermal drift compensation = Original target position - Thermal drift compensation value. This formula is applied to every time point t on the original target trajectory. For example, if the bottom dead center position defined in the original target trajectory is 150.000 mm, then after thermal drift compensation, the new target bottom dead center position becomes: 150.000 - 0.012 = 149.988 mm. Similarly, all other points on the trajectory are also subtracted by 0.012 mm. This generates an intermediate trajectory, namely the thermal drift compensation target trajectory. This new trajectory is identical in shape to the original trajectory, but its overall position is shifted downwards by 0.012 mm on the position coordinate axis.

[0042] After compensating for changes in equipment status, step S42 is executed to finely adjust the stamping depth based on the current material's hardness to ensure the final forming force or dimensions reach the ideal state. According to the calculation results from the previous steps, the material property compensation coefficient is -0.02062, a negative value indicating that the material's overall mechanical properties are relatively soft. For softer materials, stamping at the standard depth may result in insufficient deformation resistance leading to a lower final stamping force, or changes in springback characteristics causing dimensional deviations in the finished product. Appropriately increasing the stamping depth is necessary for compensation. Material property compensation is achieved by adjusting the target bottom dead center after thermal compensation. The adjustment amount is determined by a function f(), which converts the dimensionless material property compensation coefficient into a position adjustment amount with physical units (millimeters). In this embodiment, a simple linear proportional function is used: Position adjustment amount = K × Material characteristic compensation coefficient. The final calculation formula for the target bottom dead center position after compensation is: Target bottom dead center position after compensation = Target bottom dead center position after thermal drift compensation + Position adjustment amount = Target bottom dead center position after thermal drift compensation + K × Material characteristic compensation coefficient. In this formula, K is a preset compensation gain coefficient, and its unit is millimeters. The physical meaning of this coefficient is how much position adjustment is needed to compensate for a unit change in the material characteristic compensation coefficient. The determination of the K value needs to be based on process knowledge and experimental data. For example, two plates with known mechanical property differences (e.g., a difference of 5%) can be selected for trial stamping, and the difference in bottom dead center position required to obtain the same qualified parts can be recorded, thereby calculating the K value. For the high-strength steel plate stamping in this embodiment, after experimental calibration, the compensation gain coefficient K is determined to be 0.5 mm. This value means that when the material characteristic compensation coefficient is -1 (i.e., -100%), in the extreme case, the stamping depth needs to be increased by 0.5 mm to compensate. Substituting the values ​​from the example into the calculation: Position adjustment amount = 0.5 × (-0.02062) = -0.01031 mm, Final compensated target bottom dead center = 149.988 + (-0.01031) = 149.97769 mm. To match the controller's resolution, this value can be taken as 149.978 mm. This calculation result clearly shows that, for the current material that is 2.06% softer, an additional stamping depth of approximately 10 micrometers is needed on top of the already compensated 12 micrometers of thermal drift.

[0043] Finally, simply having a new target bottom dead center is insufficient; a smooth and complete motion trajectory leading to this new endpoint also needs to be generated. This is achieved through a trajectory generation module. This module takes key path points (such as top dead center, initial contact point, bottom dead center, etc.) as input and automatically generates a complete temporal trajectory connecting these points, satisfying velocity and acceleration constraints, based on preset kinematic algorithms (such as fifth-order polynomial interpolation or cycloidal motion curves). In this embodiment, the updated bottom dead center position of 149.978 mm is input into the trajectory generation module as the new endpoint, while other key points remain unchanged or are adjusted proportionally. The module then recalculates the motion profile of the entire stamping stage from the contact point to the new bottom dead center, ensuring a smooth transition without impact. Thus, a completely new target trajectory, tailored for the 1001st stamping cycle and incorporating thermal drift compensation and material property compensation, is generated. This compensated target trajectory for the next cycle will be sent to the next step to generate specific control commands for the servo drive.

[0044] In step S5, control commands for the servo drive are generated based on the compensated target trajectory for the next cycle. That is, through a series of precise calculations and compensations, a theoretically perfect compensated target trajectory tailored to the next stamping cycle has been successfully generated. However, this trajectory itself is merely a set of digital sequences existing in the controller's memory; it defines the motion target but does not explain how to drive the physical motors and mechanical structures to precisely achieve this target. As a bridge connecting digital control and physical motion, the servo drive cannot understand abstract position-time curves, but rather concrete, real-time, high-frequency electrical signals or digital commands. Therefore, to ensure that all the complex intelligent diagnostic and compensation strategies are ultimately implemented and transformed into the actual, corrected physical motion of the press slide, this application performs this final transformation step. Based on the compensated target trajectory for the next cycle, control commands for the servo drive are generated, translating this static, high-level motion blueprint into a dynamic, low-level control command stream that the servo drive can directly parse and faithfully execute.

[0045] In one possible implementation, the operation flow of step S5 is as follows: This function is completed collaboratively by a motion controller and a servo driver, which internally adopts the classic architecture of cascaded control, namely a three-loop control system consisting of a position loop, a velocity loop, and a torque loop. This process can be decomposed into two main stages: trajectory interpolation and real-time closed-loop tracking.

[0046] The first stage is trajectory interpolation. The compensated target trajectory generated in the previous embodiment, while defining a critical path including the new bottom dead center (BDC), may simply be a collection of key points or a mathematical function describing the entire motion. To enable the servo controller to execute this, it needs to be converted into a dense sequence of equally timed command points. This process is performed by the interpolator within the motion controller. The interpolator samples the entire compensated target trajectory at a very high time resolution (e.g., consistent with the period of the servo control loop, set to 1 millisecond), generating a list containing thousands of command points. Each command point precisely defines the position the slider should reach at a future millisecond. For example, when the slider approaches the new target BDC of 149.978 mm, the command sequence generated by the interpolator might be: at time t = target BDC - 2 ms, the target position should be 149.982 mm; at time t = target BDC - 1 ms, the target position should be 149.979 mm; at time t = target BDC, the target position should be 149.978 mm. This high-density, time-sensitive stream of position commands is the direct input for real-time closed-loop tracking.

[0047] The second stage is real-time closed-loop tracking, which is the core of putting commands into practice. At the start of the 1001st stamping cycle, the generated position command stream is fed point-by-point into the three-loop control structure of the servo controller. The following uses time t = target bottom dead center - 2ms as an example to illustrate the process occurring within one control cycle: 1. Position Loop Control: The function of this loop is to ensure that the actual position of the slider accurately follows the commanded position. At t = target bottom dead center - 2ms, the position loop receives the current target position command from the interpolator = 149.982 mm. Simultaneously, the controller reads the current actual position of the slider through a high-precision linear encoder, such as a reading of 149.984 mm at this moment. The position loop calculates the position error = 149.982 - 149.984 = -0.002 mm. This error is multiplied by a preset position loop proportional gain Kp1 to generate the speed command for the next-level loop. The gain Kp1 is set during the equipment commissioning phase through system identification and tuning. Its physical meaning is to determine the required speed to correct the positional error based on its magnitude. For example, Kp1 = 150s. -1The speed command generated in this cycle is: -0.002 * 150 = -0.3 mm / s. This negative value indicates that a downward speed is needed to reduce the position error. 2. Speed ​​Loop Control: This loop receives the speed command generated by the position loop and controls the motor to rotate at that speed. The speed loop receives the speed command = -0.3 mm / s. At the same time, it obtains the actual speed of the motor by differentiating the encoder position reading (or directly using the speed sensor built into the motor), such as the actual speed at this moment being -0.25 mm / s. The speed loop calculates the speed error = -0.3 - (-0.25) = -0.05 mm / s. This error is processed by a PI (proportional-integral) controller. The proportional term (multiplied by gain Kp2) provides a fast response to the current error, while the integral term (multiplied by gain Ki) is used to eliminate steady-state error. Both gains are also carefully tuned during equipment commissioning to obtain the best dynamic response. After calculation by the PI controller, a torque command is output. For example, this torque command is 250 N·m. 3. Torque Loop (or Current Loop) Control: This loop is the innermost control loop, operating at a higher frequency within the servo driver. Its task is to ensure that the motor can accurately output the torque required by the upper loop. The torque loop receives a torque command of 250 N·m. Since the output torque of the servo motor is proportional to the current flowing through its coils (T=Kt*I, where Kt is the motor torque constant), the control torque is equivalent to the control current. The driver monitors the actual current flowing through the motor windings in real time through an internal current sensor. The torque loop (current loop) also uses a PI controller to compare the commanded torque (converted commanded current) with the actual current, and based on the error, precisely controls the voltage applied to the motor windings by adjusting the duty cycle of the PWM (Pulse Width Modulation) signal. This final PWM voltage signal is the lowest-level and most direct control command driving the physical rotation of the servo motor. It drives the motor to generate a torque of exactly 250 N·m, which is ultimately converted into a downward thrust on the slider through transmission mechanisms such as a reducer and lead screw.

[0048] The entire calculation and execution process, from position error to PWM voltage signal, is completed within one control cycle (1 millisecond). In the next millisecond, the controller receives a new position command (149.979 mm) and repeats the entire closed-loop tracking process. This cycle continues, allowing the servo press's slider to precisely and precisely trace the newly compensated target trajectory step by step, within milliseconds.

[0049] In summary, the control method of the servo press based on the embodiments of this application is explained. It abandons the traditional black-box control mode that relies on a single endpoint value, instead utilizing the complete force-displacement curve as the process information carrier. By acquiring the force-displacement curve in real time from the current stamping cycle and extracting its key feature vector, it compares this vector with the ideal golden feature vector to obtain a multi-dimensional deviation vector. The key innovation lies in its introduction of a deviation decoupling method based on morphological analysis: by analyzing the translation and scaling differences between the current curve and the golden curve in terms of shape, the total deviation can be accurately decomposed into thermal drift compensation caused by the equipment's thermal effect and material property compensation coefficient caused by fluctuations in sheet metal properties. This allows for independent and precise compensation targeting the physical root cause, effectively solving the problem of information aliasing and compensation failure caused by the inability to distinguish the source of disturbance in existing technologies. This enables intelligent, feedforward-oriented collaborative compensation for the next stamping cycle, ensuring long-term stability and high precision in the production process.

[0050] Figure 5 This is a block diagram of the control system of a servo press according to an embodiment of this application. Figure 5 As shown, the control system 100 of the servo press according to an embodiment of this application includes: a current stamping displacement curve acquisition module 110, used to acquire the current stamping displacement curve in the Kth stamping cycle; a feature deviation calculation module 120, used to extract the current stamping feature vector from the current stamping displacement curve and calculate the deviation between the current stamping feature vector and the golden stamping feature vector to obtain a stamping feature deviation vector; a morphological analysis module 130, used to perform morphological analysis-based deviation decoupling on the stamping feature deviation vector and the current stamping displacement curve based on the golden stamping curve to obtain thermal drift compensation amount and material property compensation coefficient; a compensation module 140, used to compensate the original target trajectory based on the thermal drift compensation amount and material property compensation coefficient to obtain the compensated target trajectory for the next cycle; and a control command generation module 150, used to generate control commands for the servo drive based on the compensated target trajectory for the next cycle.

[0051] Here, those skilled in the art will understand that the specific operations of each step in the control system of the aforementioned servo press have been referenced above. Figures 1 to 4 The control method of the servo press is described in detail in the description of the servo press, and therefore, its repeated description will be omitted.

Claims

1. A control method for a servo press, characterized in that, include: Obtain the current stamping force displacement curve in the Kth stamping cycle; Extract the current stamping feature vector from the current stamping force displacement curve, and calculate the deviation between the current stamping feature vector and the golden stamping feature vector to obtain the stamping feature deviation vector; Based on the stamping golden curve, the stamping characteristic deviation vector and the current stamping force displacement curve are decoupled based on morphological analysis to obtain the thermal drift compensation amount and material property compensation coefficient. Based on thermal drift compensation and material property compensation coefficient, the original target trajectory is compensated to obtain the compensated target trajectory for the next cycle. Based on the compensated target trajectory of the next cycle, control commands for the servo drive are generated.

2. The control method for the servo press according to claim 1, characterized in that, The gold stamping characteristic vector includes the stiffness of the gold elastic segment, the total work done in gold stamping, the maximum stamping force of gold, and the position of the bottom dead center of gold.

3. The control method for the servo press according to claim 1, characterized in that, Obtain the current stamping force displacement curve in the Kth stamping cycle, including: Acquire the original force sensor timing reading and the original slider position timing reading for the Kth stamping cycle; The original force sensor timing readings and the original slider position timing readings of the Kth stamping cycle are digitally low-pass filtered and curves are constructed to obtain the current stamping force-displacement curve.

4. The control method for the servo press according to claim 3, characterized in that, Extract the current stamping feature vector from the current stamping force displacement curve, and calculate the deviation between the current stamping feature vector and the golden stamping feature vector to obtain the stamping feature deviation vector, including: Based on the current punching force displacement curve, key physical characteristics are calculated to obtain the maximum punching force, actual bottom dead center position, elastic segment stiffness, and total punching power. The maximum punching force, actual bottom dead center position, elastic segment stiffness, and total punching power constitute the current punching feature vector. The element-by-element difference between the current stamping feature vector and the gold stamping feature vector is calculated to obtain the stamping feature deviation vector.

5. The control method for a servo press according to claim 1, characterized in that, Based on the stamping golden curve, morphological analysis-based deviation decoupling is performed on the stamping characteristic deviation vector and the current stamping force displacement curve to obtain the thermal drift compensation amount and material property compensation coefficient, including: High-pressure segment curves are extracted from the stamping gold curve and the current stamping force displacement curve to obtain the stamping gold high-pressure segment curve and the current stamping force displacement high-pressure segment curve. The optimal position offset is calculated for the high-pressure section curve of the stamping gold and the current stamping displacement high-pressure section curve to obtain the thermal drift compensation amount.

6. The control method for a servo press according to claim 5, characterized in that, To obtain the thermal drift compensation, the optimal position offset of the high-pressure section curve of the stamping gold and the current high-pressure section curve of the stamping displacement is calculated. This includes: calculating the optimal position offset of the high-pressure section curve of the stamping gold and the current high-pressure section curve of the stamping displacement using the following formula: ; ;in, To press the high-pressure section curve of gold, This is the current pressure-displacement high-pressure section curve. The position of the curve. This is the position offset. This is the thermal drift compensation amount.

7. The control method for a servo press according to claim 5, characterized in that, Morphological analysis-based deviation decoupling is performed on the stamping characteristic deviation vector and the current stamping force displacement curve to obtain the thermal drift compensation amount and material property compensation coefficient, which also includes: Extract the stiffness deviation of the elastic segment and the total stamping work deviation from the stamping characteristic deviation vector; Extract the stiffness of the elastic segment of gold and the total work done in gold stamping from the eigenvectors of gold stamping; Based on the stiffness deviation of the elastic segment, the total stamping work deviation, the stiffness of the golden elastic segment, and the total stamping work of the golden segment, the material property compensation coefficient is determined.

8. The control method for a servo press according to claim 7, characterized in that, Based on the stiffness deviation of the elastic segment, the total stamping work deviation, the stiffness of the golden elastic segment, and the total stamping work, the material property compensation coefficient is determined, including: based on the stiffness deviation of the elastic segment, the total stamping work deviation, the stiffness of the golden elastic segment, and the total stamping work, the material property compensation coefficient is determined by the following formula, which is: ;in, For the stiffness deviation of the elastic segment, The stiffness is the golden elastic segment. For the total stamping power deviation, This represents the total work done in gold stamping. and For preset weighting coefficients, This is the material property compensation coefficient.

9. The control method for a servo press according to claim 1, characterized in that, Based on thermal drift compensation and material property compensation coefficients, the original target trajectory is compensated to obtain the compensated target trajectory for the next cycle, including: Apply thermal drift compensation to the original target trajectory to obtain a thermal drift compensated target trajectory; Material property compensation is applied to the thermal drift compensation target trajectory to obtain the compensated target trajectory for the next cycle.

10. A control system for a servo press, characterized in that, include: The current punching force displacement curve acquisition module is used to acquire the current punching force displacement curve in the Kth punching cycle. The feature deviation calculation module is used to extract the current stamping feature vector from the current stamping displacement curve and calculate the deviation between the current stamping feature vector and the golden stamping feature vector to obtain the stamping feature deviation vector. The morphological analysis module is used to perform morphological analysis-based deviation decoupling on the stamping characteristic deviation vector and the current stamping force displacement curve based on the stamping golden curve to obtain the thermal drift compensation amount and material property compensation coefficient. The compensation module is used to compensate the original target trajectory based on the thermal drift compensation amount and the material property compensation coefficient to obtain the compensated target trajectory for the next cycle. The control command generation module is used to generate control commands for the servo drive based on the compensated target trajectory of the next cycle.

Images