Adaptive Control Method and System for MLCC Presses Based on Material Dynamic Identification

By sensing changes in material stiffness and stack uniformity in real time, and combining this with an adaptive control model, the problem of insufficient control precision caused by dynamic changes in material state during the pressing process of multilayer ceramic capacitors is solved, thus achieving efficient and precise control of the pressing process.

CN123113330APending Publication Date: 2026-11-17GUANZHOU TIANHE ELECTRONIC TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611243073.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-17
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing methods for controlling the pressing and molding of multilayer ceramic capacitors cannot adapt to the dynamic changes in material state, leading to problems such as uneven stacking density, deformation of internal electrodes, or delamination and cracking.

Method used

By continuously collecting the servo motor torque current and press head position of the MLCC press, the dynamic stiffness sequence is calculated. Combined with the stack compaction uniformity index and the pressing stage identification model, adaptive control of the pressing stage is achieved, including dynamic stiffness sensing, stack uniformity quantification, and torque slope optimization.

Benefits of technology

It improves the control precision and efficiency in the pressing and molding process, ensures the density and uniformity of the layers, and enhances the consistency and yield of the products.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123113330A_ABST
    Figure CN123113330A_ABST
Patent Text Reader

Abstract

This invention provides an adaptive control method and system for MLCC presses based on dynamic material identification, belonging to the field of process control technology. The method includes: continuously acquiring the torque current of the servo motor and the position of the press head of the MLCC press; analyzing and obtaining pressure and displacement sequences; calculating the ratio of pressure change to displacement change within each sampling interval as a dynamic stiffness sequence; whenever the cumulative displacement of the press head reaches a preset displacement window length, extracting the dynamic stiffness sequence corresponding to all sampling moments within the displacement window length, and analyzing and obtaining the stack compaction uniformity index; identifying the pressing stage; dynamically optimizing the torque rise slope limit value after determining that the final compaction stage has been entered; calculating the deformation pre-compensation position and, combined with the original target position of the press head, obtaining the press head displacement control position. This invention solves the problem of insufficient control accuracy caused by dynamic changes in the material state during the pressing and molding process of multilayer ceramic capacitors.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of process control technology, and specifically to an adaptive control method and system for MLCC presses based on dynamic material identification. Background Technology

[0002] In the manufacturing process of multilayer ceramic capacitors, pressing is a key process that determines the consistency of the product's layered structure. This process applies pressure to a mold filled with ceramic powder and metal internal electrode paste through a press head, which densifies the material and forms the desired layered structure.

[0003] Existing compression molding control methods typically rely on preset fixed pressure or displacement curves, sequentially completing the degassing, plastic flow, and final compaction processes at fixed stage switching points. However, material properties vary between batches, such as powder particle size distribution, slurry viscosity, and filling volume fluctuations, and the material's stiffness, density, and other physical properties continuously change during the same compression cycle. Fixed control curves cannot adapt to these dynamic shifts in material state, easily leading to premature or delayed compression stage switching, resulting in problems such as uneven laminate density, internal electrode deformation, or delamination cracking.

[0004] In summary, existing technologies lack the ability to effectively identify the pressing stage and adaptively adjust control parameters when facing changes in the dynamic properties of materials during the pressing process, ultimately affecting product consistency and yield. Summary of the Invention

[0005] This invention provides an adaptive control method and system for MLCC presses based on dynamic material identification, which is used to solve the problem of insufficient control accuracy caused by dynamic changes in material state during the pressing and molding process of multilayer ceramic capacitors.

[0006] In a first aspect, the present invention provides an adaptive control method for MLCC presses based on dynamic material identification, comprising: The torque current and pressure head position of the servo motor of the MLCC press are continuously collected. The pressure sequence and displacement sequence are analyzed and obtained. The ratio of the pressure change to the displacement change in each sampling interval is calculated as the dynamic stiffness sequence. The displacement sequence includes the displacement change at the next sampling time and the previous sampling time. During the MLCC pressing process, whenever the cumulative displacement of the indenter reaches the preset displacement window length, the dynamic stiffness sequence corresponding to all sampling moments within the displacement window length is extracted, and the stack compactness uniformity index is analyzed and obtained. The pressing stage is identified by combining the current displacement position of the MLCC press head, the current pressing material type, and the stack compaction uniformity index. The pressing stage identification results include the venting stage, the plastic flow stage, and the final compaction stage. After determining that the MLCC press has entered the final compaction stage, the torque rise slope limit value is dynamically optimized based on the stack compaction uniformity index. By combining the current value of the elastic modulus of the transmission chain and the density uniformity index of the stack, the deformation pre-compensation position is calculated, and by combining the original target position of the indenter, the displacement control position of the indenter is obtained.

[0007] Secondly, the present invention provides an adaptive control system for MLCC presses based on dynamic material identification, comprising: The data acquisition and preprocessing module is used to continuously acquire the torque current and pressure head position of the servo motor of the MLCC press, analyze and obtain the pressure sequence and displacement sequence, and calculate the ratio of the pressure change to the displacement change in each sampling interval as the dynamic stiffness sequence. The displacement sequence includes the displacement change at the next sampling time and the previous sampling time. The stack compaction uniformity analysis module is used to extract the dynamic stiffness sequence corresponding to all sampling times within the preset displacement window length during the MLCC pressing process, and analyze and obtain the stack compaction uniformity index. The pressing stage identification module is used to identify the pressing stage by combining the current displacement position of the MLCC press head, the current pressing material type, and the stack compaction uniformity index. The pressing stage identification results include the venting stage, the plastic flow stage, and the compaction final pressing stage. The torque slope dynamic optimization module is used to dynamically optimize the torque rise slope limit value of the MLCC press based on the stack compaction uniformity index after determining that it has entered the compaction final pressing section. The deformation pre-compensation and position control module is used to calculate the deformation pre-compensation position by combining the current value of the elastic modulus of the transmission chain and the density uniformity index of the stack, and to obtain the displacement control position of the pressure head by combining the original target position of the pressure head.

[0008] One or more technical solutions provided in this invention have at least the following technical effects or advantages: The technical solution of this invention first continuously collects the torque current and pressure head position of the servo motor, converts the torque current into pressure value to form a pressure sequence, obtains the displacement change sequence by differentiating the pressure head displacement at adjacent time points, and calculates the ratio of pressure change to displacement change within each sampling interval to form a dynamic stiffness sequence, thereby realizing the quantitative expression of the real-time stiffness change of the material.

[0009] Secondly, during the pressing process, whenever the cumulative displacement of the pressure head reaches the preset displacement window length, the dynamic stiffness sequence within that window is extracted, and the ratio of the standard deviation to the average value of the stiffness values ​​within the window is calculated to obtain the stacking density uniformity index, which reflects the fluctuation of the stacking density within each displacement segment in real time with dimensionless values.

[0010] Next, the current displacement position of the pressure head, the current material type encoding, and the stacking density uniformity index are input into the trained pressing stage identification model. The model outputs the identification results of the exhaust section, plastic flow section, or compaction final pressing section. Through multi-dimensional feature fusion, the pressing stage is accurately identified, avoiding the stage misjudgment caused by material batch differences due to the fixed displacement threshold switching method.

[0011] Furthermore, after entering the final compaction stage, the change in the uniformity index of the laminated compaction is calculated for adjacent displacement windows. When uniformity improves rapidly, the current torque rise slope limit remains unchanged; when the improvement in uniformity slows down, the slope attenuation coefficient is calculated based on the deviation between the current pressure and the target pressure and the current value of the uniformity index, and the torque rise slope limit is dynamically adjusted to ensure uniformity while taking into account compaction efficiency.

[0012] Finally, the elastic deformation is calculated based on the elastic modulus of the transmission chain and the current pressure value. The dynamic compensation coefficient is obtained by combining the current value of the stack compactness uniformity index. The two are multiplied to obtain the deformation pre-compensation position. The original target position of the pressure head is subtracted from the compensation amount to obtain the pressure head displacement control position, so that the thickness of the blank after unloading and rebound is restored to the target value, thus improving the control accuracy of the pressing size.

[0013] In summary, the technical solution of this invention forms a complete closed-loop control from state perception and stage determination to parameter optimization by real-time sensing of material dynamic stiffness, continuous quantification of stack uniformity, intelligent identification of pressing stage, and adaptive adjustment of torque slope and deformation compensation. This ensures both the density and uniformity of the stack while taking into account pressing efficiency and product dimensional accuracy, thereby solving the problem of insufficient control accuracy caused by dynamic changes in material state during the pressing and molding process of multilayer ceramic capacitors. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating the adaptive control method for MLCC presses based on dynamic material identification provided by this invention.

[0015] Figure 2 This is a schematic diagram of the identification process in the pressing stage of the adaptive control method for MLCC press based on material dynamic identification provided by the present invention.

[0016] Figure 3This is a schematic diagram of the structure of the adaptive control system for MLCC press based on material dynamic identification provided by the present invention.

[0017] In the attached diagram, the labels representing each component are as follows: Data acquisition and preprocessing module 11, stacked compaction uniformity analysis module 12, pressing stage identification module 13, torque slope dynamic optimization module 14, deformation pre-compensation and position control module 15. Detailed Implementation

[0018] This invention provides an adaptive control method and system for MLCC presses based on dynamic material identification, which solves the problem of insufficient control accuracy caused by dynamic changes in the material state during the pressing and molding process of multilayer ceramic capacitors.

[0019] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to these processes, methods, products, or devices.

[0020] Example 1, as Figure 1 As shown, this invention provides an adaptive control method for MLCC presses based on dynamic material identification, the method comprising: S100: Continuously acquire the torque current and pressure head position of the servo motor of the MLCC press, analyze and obtain the pressure sequence and displacement sequence, and calculate the ratio of the pressure change to the displacement change in each sampling interval as the dynamic stiffness sequence. The displacement sequence includes the displacement change at the next sampling time and the previous sampling time.

[0021] This step involves synchronously acquiring the torque current and pressure head position of the servo motor, converting the torque current into pressure values ​​to form a pressure sequence, obtaining a displacement change sequence by differentiating the pressure head displacements at adjacent times, and then calculating the ratio of pressure change to displacement change within each sampling interval to form a dynamic stiffness sequence.

[0022] Step S100 in the method provided by the present invention includes: The torque current and pressure head position of the servo motor of the MLCC press are continuously collected at a preset sampling period, and the timestamp of each sampling moment is recorded. For each sampling moment, the torque current is multiplied by the torque constant and transmission ratio of the servo motor to obtain the pressure value at that sampling moment, and the pressure values ​​at all sampling moments are arranged in order of timestamp to form the pressure sequence; For each sampling time, based on the pressure head position, the pressure head displacement value at that sampling time is obtained, and the pressure head displacement values ​​at all sampling times are arranged in order of timestamp to form the original displacement sequence; For every two adjacent sampling times in the original displacement sequence, the displacement value of the indenter at the later sampling time is subtracted from the displacement value of the indenter at the previous sampling time to obtain the displacement change within the sampling interval. The displacement changes of all the sampling intervals are arranged in order of timestamp to form the displacement sequence. For each sampling interval, the difference between the pressure value corresponding to the next sampling moment and the pressure value corresponding to the previous sampling moment in the pressure sequence is taken as the pressure change within the sampling interval. The pressure change is then divided by the displacement change in the displacement sequence corresponding to the sampling interval to obtain the dynamic stiffness value corresponding to the sampling interval. The dynamic stiffness values ​​of all sampling intervals are arranged in order of timestamp to form the dynamic stiffness sequence.

[0023] Specifically, a preset sampling period is first set in the controller of the MLCC press. Using this sampling period as the interval, the torque current value of the servo motor and the pressure head position value are continuously collected, and a corresponding timestamp is recorded for each sampling point. The timestamp can be the absolute time of the system clock or a relative time calculated from the start of pressing; the accuracy must match the sampling period.

[0024] The controller of the MLCC press is the main control unit of the MLCC press, usually a PLC or industrial computer. The sampling period is configured through timer parameters in the controller software. The value of this sampling period must be able to capture the changes in the material state during the pressing process, and it can usually be set to a fixed value in the range of 1 millisecond to 10 milliseconds, such as 2 milliseconds.

[0025] For example, the sampling period is set to 2 milliseconds. During the pressing process of a certain MLCC press, at the 10.000th millisecond after the start of pressing, the torque current is collected as 2.35A and the pressure head position is 5.000mm; at the 12.000th millisecond, the torque current is collected as 2.41A and the pressure head position is 4.982mm; at the 14.000th millisecond, the torque current is collected as 2.48A and the pressure head position is 4.961mm. And so on, continuously collecting data until the end of pressing.

[0026] Secondly, for each sampling moment, the torque current value collected at that moment is obtained, and multiplied by the torque constant and transmission ratio of the servo motor to obtain the pressure value corresponding to that sampling moment. The pressure values ​​calculated at each sampling moment are arranged in chronological order according to their timestamps to form a pressure sequence.

[0027] The torque constant of a servo motor is an inherent parameter of the motor, measured in N·m / A, and can be obtained from the motor nameplate or technical manual. The transmission ratio is the reduction ratio or force amplification ratio between the motor output shaft and the press head drive mechanism, determined by the design parameters of the press transmission mechanism. For example, a ball screw transmission mechanism converts the motor's rotary motion into the press head's linear motion; its transmission ratio can be expressed as a conversion factor between the output force and the input torque, with a value of 200 N / N·m. An example torque constant value of 0.5 N·m / A is provided.

[0028] For example, taking a torque constant of 0.5 N·m / A and a transmission ratio of 200 N / N·m, the pressure value at the 10th,000th millisecond is 2.35A × 0.5 N·m / A × 200 N / N·m = 235 N; the pressure value at the 12th,000th millisecond is 2.41 × 0.5 × 200 = 241 N; and the pressure value at the 14th,000th millisecond is 2.48 × 0.5 × 200 = 248 N. Arranging these three pressure values ​​in chronological order, the first three terms of the pressure sequence are [235, 241, 248], in N.

[0029] Next, for each sampling moment, the corresponding indenter displacement value is obtained based on the collected indenter position value. The indenter displacement value is obtained by subtracting the indenter position value at the current sampling moment from the initial indenter position value at the start of compression; the difference is the cumulative indenter displacement from the start of compression to the current moment. The indenter displacement values ​​at each sampling moment are arranged in chronological order according to their timestamps to form the original displacement sequence.

[0030] The initial position of the indenter is the position value recorded when the indenter contacts the upper surface of the MLCC preform. It can be automatically recorded before pressing begins by slowly pressing the indenter down to contact the preform surface and detecting a change in the pressure threshold. For example, let the initial position of the indenter be 10.000 mm at the start of pressing.

[0031] For example, suppose the initial position of the indenter is 10.000 mm at the start of pressing. At the 10.000th millisecond, the indenter position is 5.000 mm, and the indenter displacement is 10.000 - 5.000 = 5.000 mm; at the 12.000th millisecond, the indenter position is 4.982 mm, and the indenter displacement is 10.000 - 4.982 = 5.018 mm; at the 14.000th millisecond, the indenter position is 4.961 mm, and the indenter displacement is 10.000 - 4.961 = 5.039 mm. Therefore, the first three terms of the original displacement sequence are [5.000, 5.018, 5.039], in mm.

[0032] Next, for every two adjacent sampling times in the original displacement sequence, the indenter displacement value corresponding to the sampling time with the later timestamp is subtracted from the indenter displacement value corresponding to the sampling time with the earlier timestamp. The difference is the displacement change within that sampling interval. The calculated displacement changes for all sampling intervals are arranged in chronological order of their timestamps to form a displacement sequence.

[0033] For example, if the first sampling interval is from 10.000 ms to 12.000 ms, the displacement change is 5.018 - 5.000 = 0.018 mm; if the second sampling interval is from 12.000 ms to 14.000 ms, the displacement change is 5.039 - 5.018 = 0.021 mm. Then the first two terms of the displacement sequence are [0.018, 0.021], in mm.

[0034] Furthermore, for each sampling interval, the pressure values ​​at the two sampling times corresponding to that interval are taken from the pressure sequence. The pressure value at the later sampling time is subtracted from the pressure value at the earlier sampling time to obtain the pressure change within that sampling interval. This pressure change is then divided by the displacement change corresponding to the same sampling interval in the displacement sequence. The quotient is the dynamic stiffness value corresponding to that sampling interval. The physical meaning of the dynamic stiffness value is the pressure increment required for the material near the current pressing position to resist a unit displacement change, with units of N / mm.

[0035] For example, in the first sampling interval, from the 10th to the 12th millisecond, the pressure change is 241 - 235 = 6 N, the displacement change is 0.018 mm, and the dynamic stiffness value is 6 / 0.018 = 333.33 N / mm; in the second sampling interval, from the 12th to the 14th millisecond, the pressure change is 248 - 241 = 7 N, the displacement change is 0.021 mm, and the dynamic stiffness value is 7 / 0.021 = 333.33 N / mm.

[0036] Finally, the dynamic stiffness values ​​of each sampling interval calculated above are arranged in chronological order according to the timestamps of their corresponding sampling intervals to form a time series of dynamic stiffness values, i.e., a dynamic stiffness sequence.

[0037] For example, the dynamic stiffness values ​​for the first two sampling intervals are both 333.33 N / mm. Arranged in chronological order, the first two terms of the dynamic stiffness sequence are [333.33, 333.33], with units of N / mm. As the pressing process continues, the dynamic stiffness values ​​for subsequent sampling intervals will be calculated sequentially and added to the sequence.

[0038] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0039] In summary, this step quantifies the stiffness change of the material as displacement increases during the pressing process in the form of a time series, providing a direct, continuous and real-time updatable data foundation for subsequent pressing stage identification and adaptive control based on stiffness characteristics, thereby improving the control process's ability to perceive the dynamic changes in the material's state.

[0040] S200: During the MLCC pressing process, whenever the cumulative displacement of the indenter reaches the preset displacement window length, the dynamic stiffness sequence corresponding to all sampling times within the displacement window length is extracted, and the stack density uniformity index is obtained through analysis.

[0041] This step involves extracting the dynamic stiffness sequence according to a fixed displacement window during the pressing process, calculating the ratio of the standard deviation to the average value of the stiffness values ​​within the window, and obtaining the stacked compactness uniformity index.

[0042] Step S200 in the method provided by the present invention includes: Starting from the moment of suppression, the displacement change in the displacement sequence is accumulated. When the accumulated value is greater than or equal to the preset displacement window length for the first time, the current moment is recorded as the window truncation moment. Taking the window truncation time as the end point, backtracking to the starting time when the accumulated value first exceeds or equals the preset displacement window length, the dynamic stiffness values ​​corresponding to all sampling times between the starting time and the window truncation time constitute the window dynamic stiffness sequence. Calculate the arithmetic mean of all dynamic stiffness values ​​in the window dynamic stiffness sequence as the window average stiffness value, and calculate the standard deviation of all dynamic stiffness values ​​in the window dynamic stiffness sequence as the window stiffness standard deviation. Divide the standard deviation of the window stiffness by the average stiffness value of the window to obtain the stack density uniformity index corresponding to the current displacement window.

[0043] Specifically, starting from the initial compression moment, the displacement changes in the displacement sequence are accumulated sequentially over time. Each accumulation is compared with a preset displacement window length. When the accumulated value is greater than or equal to the preset displacement window length for the first time, the sampling moment corresponding to the last displacement change currently being accumulated is recorded as the window cutoff moment.

[0044] The preset displacement window length is an empirical value that can be set according to the pressing stroke and material characteristics. It is typically between 5% and 15% of the total pressing stroke. It should be noted that, for ease of demonstration of the cumulative window capture process, the example uses a smaller stroke scenario: a total pressing stroke of 2mm and a preset displacement window length of 0.1mm. In a real scenario with an 8mm total pressing stroke, the displacement window length can be proportionally set to 1.0mm.

[0045] For example, suppose the displacement changes of the first five terms in the displacement sequence are 0.018mm, 0.021mm, 0.019mm, 0.022mm, and 0.020mm, respectively, and the preset displacement window length is 0.100mm. Starting from the beginning of the compression, the displacement changes are accumulated sequentially over time: after the first term is added, the accumulated value is 0.018mm, which is less than 0.100mm; after adding the second term, the accumulated value is 0.039mm, which is less than 0.100mm; after adding the third term, the accumulated value is 0.058mm, which is less than 0.100mm; after adding the fourth term, the accumulated value is 0.080mm, which is less than 0.100mm; after adding the fifth term, the accumulated value is 0.100mm, which is equal to the window length of 0.100mm for the first time. At this point, the end time of the sampling interval corresponding to the fifth displacement change is the window truncation time, denoted as the 20.000th millisecond.

[0046] Secondly, using the determined window cutoff time as the endpoint, we trace back along the time axis to find the starting moment during the accumulation process when the accumulated value first exceeds or equals the preset displacement window length. That is, we begin accumulating the displacement change from this starting moment, ensuring that the accumulated value reaches or exceeds the displacement window length at the window cutoff time. We then extract the dynamic stiffness values ​​corresponding to all sampling intervals between this starting moment and the window cutoff time, arranging them in chronological order to form a window dynamic stiffness sequence. The actual head displacement covered by this sequence is the preset displacement window length.

[0047] For example, by backtracking, it can be seen that starting from the beginning of the 5th sampling interval, i.e., 18.000 ms, after accumulating displacement over 5 sampling intervals, the accumulated value reaches 0.100 mm at the end of the 5th sampling interval, i.e., 20.000 ms. Therefore, the starting time is 18.000 ms. The window dynamic stiffness sequence consists of the dynamic stiffness values ​​corresponding to the 5 sampling intervals between 18.000 ms and 20.000 ms, which are [333.33 N / mm, 350.00 N / mm, 366.67 N / mm, 381.82 N / mm, 400.00 N / mm].

[0048] Next, the arithmetic mean of all dynamic stiffness values ​​in the window dynamic stiffness sequence is calculated, and the result is the window average stiffness value corresponding to the current displacement window. The window average stiffness value reflects the overall average resistance to deformation of the material within that displacement interval.

[0049] For example, if the window dynamic stiffness sequence is [333.33, 350.00, 366.67, 381.82, 400.00], with a total of 5 values, then the average window stiffness value = (333.33 + 350.00 + 366.67 + 381.82 + 400.00) / 5 = 366.36 N / mm.

[0050] Next, the standard deviation of all dynamic stiffness values ​​in the window dynamic stiffness sequence is calculated. The result is the standard deviation of the window stiffness corresponding to the current displacement window. The standard deviation can be calculated using either the population standard deviation or the sample standard deviation. In the calculation of the sample standard deviation, the degrees of freedom are n-1, where n is the number of dynamic stiffness values ​​within the window. The standard deviation of the window stiffness characterizes the degree of dispersion of the material stiffness values ​​relative to the average stiffness within this displacement interval.

[0051] For example, first calculate the squared deviations of each value from the average stiffness value: (333.33-366.36)²=1091.09, (350.00-366.36)²=267.65, (366.67-366.36)²=0.10, (381.82-366.36)²=238.98, (400.00-366.36)²=1131.89. The sum of squared deviations is 1091.09+267.65+0.10+238.98+1131.89=2729.71. Using the sample standard deviation, n=5, then dividing by (n-1)=4, we get the variance = 2729.71 / 4 = 682.43, and the standard deviation = ≈26.12N / mm.

[0052] Finally, the obtained standard deviation of window stiffness is divided by the average stiffness value of the window, and the resulting ratio is the stack compactness uniformity index corresponding to the current displacement window. The stack compactness uniformity index is a dimensionless value. The larger the value, the more uneven the compressive deformation between the electrode layer and the ceramic dielectric layer inside the MLCC stack, and the more dispersed the distribution of residual air gaps between layers; the smaller the value and the more stable it is, the more uniform and dense the layers inside the MLCC stack tend to be.

[0053] For example, if the standard deviation of window stiffness is 26.12 N / mm and the average stiffness of window is 366.36 N / mm, then the uniformity index of laminated density is approximately 26.12 / 366.36 ≈ 0.0713.

[0054] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0055] In summary, this step reflects the fluctuation of material density in each pressing displacement segment in real time in the form of dimensionless numerical values. The smaller the index value, the more uniform the layering, which provides a quantitative basis for accurate identification and dynamic adjustment of process parameters in the subsequent pressing stage.

[0056] S300: Combine the current displacement position of the MLCC press head, the current pressing material type, and the stack compaction uniformity index to identify the pressing stage, wherein the pressing stage identification results include the venting stage, the plastic flow stage, and the compaction final pressing stage.

[0057] like Figure 2 As shown, in this step, the current displacement position of the pressure head, the current material type encoding, and the stack compaction uniformity index are input into the trained identification model. The model outputs the determination result of the exhaust section, plastic flow section, or compaction final compression section.

[0058] Step S300 in the method provided by the present invention includes: Obtain the current displacement position of the MLCC press head, where the current displacement position of the press head is the displacement value of the press head at the current sampling time; Obtain the current pressing material type, wherein the current pressing material type includes the stacked structure information of the ceramic dielectric film and the inner electrode layer; A pressing stage identification model is constructed. The current displacement position of the pressing head, the current pressing material type, and the stack compaction uniformity index are input into the pressing stage identification model. The pressing stage identification model outputs the pressing stage identification result at the current moment. The pressing stage identification result includes the venting stage, the plastic flow stage, and the compaction final pressing stage.

[0059] Specifically, at the current sampling time, the pressure head displacement value corresponding to that time is read from the displacement sequence as the current displacement position of the pressure head, representing the cumulative distance moved by the pressure head from the start of the pressing to the current time.

[0060] For example, if the current sampling time is 120.000 milliseconds, and the displacement value of the indenter at that time is 4.520 mm, then the current displacement position of the indenter is 4.520 mm.

[0061] Furthermore, the material type information of the currently being pressed MLCC preforms is retrieved from the production management system. The production management system is an information management platform on the MLCC production line, storing process parameters for each batch of preforms, including the ceramic dielectric film formulation code, single-layer thickness of the dielectric film, number of internal electrode layers, total number of layers, and other material type information, as well as the corresponding material type code. The currently pressed material type includes the laminated structure information of the ceramic dielectric film and the internal electrode layers, represented in coded form.

[0062] For example, the currently pressed MLCC preform uses a ceramic dielectric film with an X7R formulation, a single film thickness of 2.5 micrometers, 200 internal electrode layers, and a total of 200 layers. According to the preset material type coding table, the code corresponding to this material type is 3.

[0063] In this step, the construction of the suppression stage identification model includes: Acquire historical pressing data, which includes the indenter displacement value, current pressing material type, and stack compaction uniformity index of multiple MLCC preforms at each sampling moment during the complete pressing process; Each sampling moment in the historical compression data is labeled with a stage to obtain an output supervision set, wherein the output supervision set includes the exhaust section label, the plastic flow section label, and the compaction final pressure section label corresponding to the historical compression data; A suppression stage identification model is constructed. The input layer of the suppression stage identification model contains three input nodes, and the output node outputs the suppression stage labels obtained by identification. The suppression stage identification model is trained under supervision using the historical suppression data and the output supervision set. The suppression stage labels in the output supervision set are converted into one-hot encoded vectors. The training objective is to minimize the cross-entropy loss between the three confidence scores output by the suppression stage identification model and the one-hot encoded vectors. The weight parameters of the suppression stage identification model are iteratively updated until the cross-entropy loss converges, thus obtaining the trained suppression stage identification model.

[0064] Specifically, historical pressing data was first collected from multiple MLCC preforms during the complete pressing process. This historical pressing data included indenter displacement values ​​at each sampling time, the current pressing material type, and stack compaction uniformity indices, covering the entire process from the start to the end of pressing. Material types were stored in coded form, and the sample size should cover pressing batches under different material types and combinations of process parameters to ensure the representativeness of the training data.

[0065] For example, data from 500 completed MLCC preforms were collected. The pressing process for each preform included approximately 5,000 sampling moments. At each sampling moment, the indenter displacement value, material type code, and stack compaction uniformity index were recorded. Among the collected historical pressing data, the material type codes included code 3, corresponding to the X7R-200 layer, and code 5, corresponding to the C0G-150 layer, covering the pressing processes of various material types.

[0066] In this step, the step of marking each sampling time in the historical suppression data includes: For each sampling time in the historical suppression data, the dynamic stiffness sequence of the displacement window where the sampling time is located is obtained, and the arithmetic mean of all dynamic stiffness values ​​in the dynamic stiffness sequence is calculated as the window average stiffness value of the sampling time. For each sampling moment in the historical suppression data, the pressure value at the sampling moment is obtained. When the pressure value is less than the preset upper limit threshold of the low pressure range, it is determined that the pressure sequence at the sampling moment is in the low value range. When the pressure sequence is in a low value range, and the growth rate of the average stiffness value of the window of multiple consecutive displacement windows is less than the preset gradual growth threshold, the sampling time covered by the multiple consecutive displacement windows is marked as the exhaust segment. When the average stiffness value of the window is positively increasing compared to the previous adjacent displacement window, and multiple consecutive displacement windows maintain a positive increasing trend, it is determined that the dynamic stiffness sequence is in a continuously rising state, and the stack compaction uniformity index corresponding to the multiple consecutive displacement windows is obtained. When the stack compaction uniformity index first increases and then decreases continuously, the sampling time covered by the displacement window that satisfies the dynamic stiffness sequence being in a continuously rising state and the stack compaction uniformity index showing an upward and then downward trend is marked as the plastic flow section. When the average stiffness value of the window exceeds the preset stiffness limit corresponding to the current pressing material type, and the stack compaction uniformity index is less than the preset upper limit threshold of the stable range, it is determined that the stack compaction uniformity index has entered the low-level stable range, and the sampling time covered by the displacement window that satisfies the condition that the average stiffness value of the window exceeds the preset stiffness limit and the stack compaction uniformity index has entered the low-level stable range is marked as the final compaction section.

[0067] Specifically, for each sampling moment in the historical suppression data, the displacement window in which that sampling moment falls is determined, and the dynamic stiffness sequence within that displacement window is extracted. The arithmetic mean of all dynamic stiffness values ​​in the dynamic stiffness sequence is calculated as the window average stiffness value corresponding to that sampling moment. If a sampling moment belongs to multiple displacement windows simultaneously, such as in the case of overlapping sliding windows, the displacement window first included at that moment is used for calculation to ensure that each sampling moment corresponds to a unique window average stiffness value.

[0068] For example, the dynamic stiffness sequence of the displacement window at the 120.000th millisecond sampling time in the historical compression data is [280N / mm, 310N / mm, 340N / mm, 370N / mm, 400N / mm], a total of 5 values. The average stiffness value of the window at this sampling time = (280+310+340+370+400) / 5 = 340N / mm.

[0069] Secondly, for each sampling time in the historical suppression data, the pressure value corresponding to that sampling time is obtained. The pressure value is compared with a preset upper threshold of the low-pressure range. When the pressure value is less than the upper threshold of the low-pressure range, the pressure sequence at that sampling time is determined to be in the low-value range.

[0070] The preset upper limit threshold of the low-pressure range can be set according to the type of material being pressed and the specifications of the press. For example, for X7R-200 layer material, under the condition of a rated pressing force of 2000N, the upper limit threshold of the low-pressure range can be set to 600N.

[0071] For example, if the pressure value at the 120.000th millisecond sampling time is 238N, and the preset upper threshold for the low-pressure range is 600N, then 238N < 600N, and the pressure sequence at that sampling time is determined to be in the low-value range. Furthermore, when the pressure sequence at the sampling time is in the low value range, and the growth rate of the average stiffness value of multiple consecutive displacement windows ending at that sampling time is less than the preset gradual growth threshold, the sampling time covered by the multiple consecutive displacement windows is marked as the exhaust segment.

[0072] The specific number of consecutive displacement windows can be preset to 3, and the gradual growth threshold can be preset to 5%. The number of windows is set to 3 to achieve a balance between suppressing sampling noise interference and maintaining the sensitivity of the phase switching response. The gradual growth threshold of 5% is based on the fact that the exhaust section is mainly characterized by gas discharge and particle rearrangement, with a slow stiffness increase, typically not exceeding 5%, while the growth rate generally exceeds 10% after entering the plastic flow section. 5% effectively distinguishes between the two.

[0073] The above values ​​can be calibrated and adjusted based on the distribution of exhaust section stiffness growth rate in historical data: increase to 5 sampling times when the sampling frequency is high to smooth noise, and decrease to 2 sampling times when the sampling frequency is low to avoid response lag; decrease to 2 sampling times when the total pressing stroke is large, and increase to 4 to 5 sampling times when the total pressing stroke is small. Gradual growth threshold: decrease to 3% for high viscosity binder systems and increase to 8% for low viscosity systems; decrease to 4% when the number of internal electrode layers exceeds 300 layers, and increase to 6% when the number of internal electrode layers is less than 100 layers.

[0074] For example, within three consecutive displacement windows ending at the 120.000th millisecond sampling time, the average stiffness values ​​are [320 N / mm, 332 N / mm, 340 N / mm]. The growth rates of adjacent windows are (332-320) / 320=3.75% and (340-332) / 332=2.41%, respectively, both less than the gradual growth threshold of 5%. Furthermore, the pressure values ​​at the sampling times within these three windows are all less than 600 N, falling within the low-value range. Therefore, all sampling times covered by these three displacement windows are labeled as the exhaust section.

[0075] Next, when the average stiffness value of the sampling window shows a positive increase compared to the previous adjacent displacement window, and multiple consecutive displacement windows ending at that sampling time all maintain a positive growth trend, the dynamic stiffness sequence is determined to be in a continuously rising state. The specific number of multiple consecutive displacement windows can be preset to 3, consistent with the principle of determining the number of windows with a gradual increase in stiffness in the exhaust section labeling rules. At the same time, the stack compaction uniformity index corresponding to these multiple consecutive displacement windows is obtained. When it shows a change characteristic of first continuously rising and then continuously falling, the sampling time covered by these multiple consecutive displacement windows is labeled as the plastic flow section. The first continuous rise can be preset to at least 2 consecutive windows of rise, and the subsequent continuous fall can be preset to at least 2 consecutive windows of fall, which can effectively filter out random fluctuation interference and ensure that what is captured is the real trend reversal caused by the change in the material's plastic flow mechanism.

[0076] For example, the average stiffness values ​​of three consecutive displacement windows ending at the 200.000th millisecond sampling time are [520 N / mm, 610 N / mm, 650 N / mm], respectively. The average stiffness value of the preceding adjacent displacement window is 450 N / mm. 520 > 450, and 610 > 520, 650 > 610. Each window shows a positive increase compared to the previous window, satisfying the condition of continuous increase. The stacking density uniformity index of the above three displacement windows and the two windows before them are [0.10, 0.16, 0.22, 0.18, 0.13], respectively. The first two windows correspond to uniformity indices of 0.10 and 0.16. The first three values ​​(0.10, 0.16, 0.22) show a continuous increase, while the last two values ​​(0.18, 0.13) show a continuous decrease, exhibiting a characteristic of first increasing and then decreasing. Therefore, all sampling times covered by the above three windows are labeled as the plastic flow section.

[0077] Furthermore, when the average stiffness value of the window at the sampling time exceeds the preset stiffness limit corresponding to the current pressing material type, and the stack compaction uniformity index at that sampling time is less than the preset upper limit threshold of the stable interval, it is determined that the stack compaction uniformity index has entered the low-level stable interval. The sampling time covered by the displacement window that satisfies the above two conditions is marked as the final compaction section.

[0078] The preset stiffness limit is set according to the material type. For example, the preset stiffness limit for the X7R-200 layer is 800 N / mm, which is the average distribution value of the average stiffness value at the starting point of the compacted final compression section in the historical pressing data of the X7R-200 layer material. Exceeding this value indicates that the gas inside the laminate has been basically expelled and close contact has been formed between the particles. The upper limit threshold of the stable range can be preset to 0.05. When the uniformity index is lower than 0.05, the relative dispersion of interlayer stiffness is less than 5%, and the distribution of residual air gaps has been basically eliminated.

[0079] For example, the average stiffness value of the window at sampling time 350.000 ms is 850 N / mm, exceeding the preset stiffness limit of 800 N / mm for the X7R-200 layer. The stack compaction uniformity index at this sampling time is 0.03, which is less than the preset upper threshold of the stable interval of 0.05, indicating that it has entered the low-level stable interval. Therefore, all sampling times covered by the displacement window at this sampling time are labeled as the final compaction section.

[0080] Furthermore, all sampling moments with completed stage labeling in the above steps and their corresponding stage labels are summarized to form an output supervision set. Each sampling moment in the output supervision set includes its timestamp, indenter displacement value, pressure value, window average stiffness value, stack compaction uniformity index, and labeled stage label, which are used for subsequent supervised training of the compression stage identification model.

[0081] Furthermore, a fully connected neural network is constructed as the model for identifying the compression stage. The input layer of the model contains three input nodes, corresponding to the current displacement position of the indenter, the current compression material type encoding, and the stack compaction uniformity index, respectively. Two hidden layers are set after the input layer: the first hidden layer contains 16 nodes, and the second hidden layer contains 8 nodes. Each hidden layer uses the ReLU function as the activation function. The output layer contains 3 nodes, using the Softmax activation function, and outputs the confidence scores for three categories: the venting stage, the plastic flow stage, and the compaction final compression stage. The sum of the three confidence scores is 1. The category corresponding to the highest confidence score is taken as the identified compression stage label.

[0082] The weight matrix and bias vector between each layer of the model are iteratively updated during training using the gradient descent algorithm. The initial weights can be randomly initialized, for example, using the Xavier initialization method, to keep the variance of the weights in each layer consistent, thereby accelerating the training convergence speed.

[0083] For example, the input feature vector is [4.520, 3, 0.0713]. It is passed from the input layer to the first hidden layer. The first hidden layer output is obtained by calculating the weight matrix W1 (3×16 dimension) and the bias vector b1 (16 dimension) and then performing ReLU activation. The second hidden layer output is obtained by calculating the weight matrix W2 (16×8 dimension) and the bias vector b2 (8 dimension) and then performing ReLU activation. Finally, the output layer output is obtained by calculating the weight matrix W3 (8×3 dimension) and the bias vector b3 (3 dimension) and then performing Softmax activation. Three confidence values ​​are output, and the category corresponding to the maximum value is taken as the recognition result.

[0084] Furthermore, the suppression stage identification model is trained under supervision using historical suppression data and the output supervision set. The suppression stage labels in the output supervision set are converted into one-hot encoded vectors, for example, [1,0,0] for the exhaust stage, [0,1,0] for the plastic flow stage, and [0,0,1] for the compaction final compression stage. The cross-entropy loss between the three confidence scores of the model output and the one-hot encoded vector is used as the loss function. With minimizing the loss as the training objective, the weight parameters of the model are iteratively updated using the gradient descent algorithm until the cross-entropy loss converges, thus obtaining the trained suppression stage identification model.

[0085] The training termination condition is set as follows: when the cross-entropy loss value on the validation set no longer decreases in N consecutive training rounds (i.e., the difference between the validation set loss value in the current round and the lowest validation set loss value in the previous N rounds is less than a preset loss convergence threshold, for example, a loss convergence threshold of 0.001 and N = 10), then the loss is considered converged and training stops; or when the number of training rounds reaches a preset upper limit, for example, 500 rounds, training is forcibly terminated even if the loss has not fully converged, to prevent excessively long and ineffective iterations. The model parameters corresponding to the lowest validation set loss are taken as the final model parameters to obtain the trained suppression phase recognition model.

[0086] For example, in a certain training round, the model outputs three confidence scores of [0.10, 0.75, 0.15] at a certain sampling time, with corresponding one-hot encoded labels of [0, 1, 0]. The cross-entropy loss is calculated. During training, the validation set cross-entropy loss value gradually decreases from above 1.0 initially, reaching 0.048 by the 200th round. For the next 10 consecutive rounds, the validation set loss remains at its lowest value of 0.048. The difference between the current round's loss value and the lowest value is consistently less than 0.001, satisfying the convergence condition. Training terminates at the 210th round, and the model parameters from the 200th round are used as the final model, completing the training.

[0087] Finally, the current displacement position of the indenter, the current material type code, and the stack compaction uniformity index calculated by the current displacement window are input into the trained pressing stage identification model. The model calculates and outputs three confidence values ​​through forward propagation, and takes the stage label corresponding to the maximum confidence value as the pressing stage identification result at the current moment.

[0088] For example, inputting the current displacement position of the indenter (4.520 mm), material type code (3), and laminate compaction uniformity index (0.0713) into the trained model, the model output confidence scores are [0.05, 0.85, 0.10]. The confidence score for the plastic flow section is the highest at 0.85, therefore the current identification result is the plastic flow section. At this moment, the laminate compaction uniformity index (0.0713) is in a decreasing phase after falling from its peak value of 0.22, consistent with the characteristic of the uniformity index in the plastic flow section first increasing and then decreasing.

[0089] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0090] In summary, this step integrates the dynamic stiffness fluctuation characteristics of materials with displacement processes and material properties in a multi-dimensional manner, enabling accurate identification of the exhaust section, plastic flow section, and compaction final pressure section, thus avoiding misjudgment of stages caused by batch differences in materials due to fixed displacement threshold switching methods.

[0091] S400: After determining that the MLCC press has entered the final compaction stage, the torque rise slope limit value of the MLCC press is dynamically optimized based on the stack compaction uniformity index.

[0092] In this step, the torque rise slope limit is dynamically adjusted according to the changing trend of the uniformity index of the laminated compaction in the final compaction section. When the uniformity improves rapidly, it remains unchanged; when the improvement slows down, the slope attenuation coefficient is calculated in combination with the pressure deviation and adjusted to balance uniformity and pressing efficiency.

[0093] Step S400 in the method provided by the present invention includes: After determining that the compaction has entered the final compaction section, the current value of the stack compaction uniformity index corresponding to the current displacement window and the previous window value of the stack compaction uniformity index corresponding to the previous adjacent displacement window are obtained. The difference between the current value of the stack compaction uniformity index and the previous window value of the stack compaction uniformity index is calculated as the uniformity change. When the uniformity change is negative and its absolute value is greater than the preset fast convergence threshold, the current proportional gain parameter and torque rise slope limit value of the MLCC press are kept unchanged. When the uniformity change is negative and its absolute value is less than or equal to the preset fast convergence threshold, the pressure deviation value between the pressure value at the current sampling time and the preset target pressure value is obtained, and the torque rise slope limit value is adjusted according to the pressure deviation value.

[0094] Specifically, after determining that the compaction has entered the final compaction stage, the current displacement window's corresponding compaction uniformity index is read from the calculated and stored sequence of stacked compaction uniformity indices as the current value. Simultaneously, the compaction uniformity index corresponding to the displacement window immediately preceding the current displacement window is read as the previous window's value.

[0095] For example, suppose the current displacement window in the final compaction section is the 35th displacement window, and its current value of the stack compaction uniformity index is 0.032. The previous adjacent displacement window is the 34th displacement window, and its previous window value of the stack compaction uniformity index is 0.038.

[0096] Furthermore, the difference between the current value of the stack compaction uniformity index and the previous window value is the uniformity change. When the uniformity change is negative, it indicates that the uniformity index is decreasing and the stack compaction is becoming more uniform; when the uniformity change is positive, it indicates that the uniformity index is rising and the stack compaction uniformity is deteriorating. The absolute value of the change reflects the rate of improvement or deterioration of uniformity.

[0097] For example, the change in uniformity is 0.032 - 0.038 = -0.006. The negative value indicates that the uniformity index is decreasing, and the layers are becoming more uniform. Its absolute value is 0.006.

[0098] Furthermore, a fast convergence threshold is preset to measure whether the improvement in uniformity is fast enough. The fast convergence threshold is preset to 0.005, based on the fact that this value can distinguish whether the decrease in uniformity index between adjacent displacement windows is significant enough: a decrease greater than 0.005 in a single window indicates rapid improvement in uniformity, requiring no adjustment of control parameters; a decrease less than or equal to 0.005 indicates that the improvement is slowing down, requiring adjustment in conjunction with pressure deviation. This threshold can be adjusted according to process requirements; it can be relaxed to 0.008 when rapid convergence is required, and tightened to 0.003 when stable convergence is required.

[0099] When the uniformity change is negative and its absolute value is greater than the fast convergence threshold, it indicates that the uniformity of the stack compaction is improving rapidly, the current control parameter settings are reasonable and no adjustment is needed, and the current proportional gain parameter and torque rise slope limit of the MLCC press should be kept unchanged.

[0100] For example, the preset fast convergence threshold is 0.005. The current absolute value of the uniformity change is 0.006, which is greater than 0.005, indicating that the uniformity is improving rapidly. The current proportional gain parameter and torque rise slope limit value are maintained unchanged.

[0101] In this step, when the uniformity change is negative and its absolute value is less than or equal to the preset fast convergence threshold, the pressure deviation between the pressure value at the current sampling time and the preset target pressure value is obtained, and the torque rise slope limit is adjusted according to the pressure deviation value, including: When the uniformity change is negative and its absolute value is less than or equal to the preset fast convergence threshold, the pressure value at the current sampling time and the preset target pressure value are obtained, and the pressure deviation value is obtained by subtracting the pressure value at the current sampling time from the preset target pressure value. Obtain the current value of the stack compaction uniformity index corresponding to the current displacement window, and calculate the slope attenuation coefficient based on the current value of the stack compaction uniformity index and the absolute value of the pressure deviation value. Obtain a reference value for the torque rise slope limit value, multiply the reference value for the torque rise slope limit value by the slope attenuation coefficient to obtain the adjusted torque rise slope limit value, and update the torque rise slope limit value to the adjusted torque rise slope limit value.

[0102] Specifically, when the uniformity change is negative and its absolute value is less than or equal to a preset fast convergence threshold, the pressure value at the current sampling moment and the preset target pressure value are obtained. The pressure deviation value is obtained by subtracting the pressure value at the current sampling moment from the preset target pressure value. When the pressure deviation value is positive, it indicates that the current pressure has not yet reached the target pressure and the pressure needs to be increased further; when the pressure deviation value approaches zero, it indicates that the current pressure is close to the target pressure and the pressure increase rate needs to be slowed down to prevent overshoot.

[0103] For each type of pressed material and laminated structure combination, a preset target pressure value can be obtained through statistical analysis of the pressure values ​​at the end of the final compaction stage in historical pressing data. For example, the pressure values ​​of multiple qualified preforms of this material type at the end of the final compaction stage, i.e., when the indenter displacement reaches the total pressing stroke, can be collected, and their average value can be taken as the target pressure value corresponding to this material type. Alternatively, the upper and lower limits of the pressing pressure specified in the process specification can be used as constraints to ensure that the target pressure value is within a reasonable range that the material can withstand.

[0104] For example, if the preset target pressure value is 2000N and the pressure value at the current sampling time is 1850N, then the pressure deviation value = 2000 - 1850 = 150N.

[0105] Furthermore, the current value of the stack compaction uniformity index corresponding to the current displacement window is obtained. Based on this current value and the absolute value of the obtained pressure deviation, the slope attenuation coefficient is calculated. The slope attenuation coefficient is used to moderately attenuate the torque rise slope limit value when the uniformity improvement slows down and pressure deviation exists, so as to prevent the uniformity index from rebounding due to excessively rapid pressure rise.

[0106] The slope attenuation coefficient is calculated as follows: the absolute value of the pressure deviation is divided by a preset reference deviation value, then multiplied by the ratio of the current value of the laminated compaction uniformity index to the preset reference uniformity index. This ratio is then subtracted from 1, and the result is limited to a preset lower limit. The slope attenuation coefficient is calculated as follows: Slope attenuation coefficient = 1 - (absolute value of pressure deviation / preset reference deviation value) × (current value of laminated compaction uniformity index / preset reference uniformity index). The slope attenuation coefficient ranges from 0 to 1; a smaller value indicates a greater degree of attenuation. The preset lower limit can be set to 0.3 to prevent excessive attenuation that could prevent the pressure from reaching the target value within a specified time.

[0107] The preset reference deviation value is determined based on the difference between the target pressure value corresponding to the current pressing material type and the starting pressure value of the final compaction stage. Specifically, the distribution of pressure difference between the pressure value at the beginning of the final compaction stage and the target pressure value at the end of the pressing stage is statistically analyzed in the historical pressing data of this material type. The average value of this pressure difference is taken as the reference deviation value, which represents the typical span of pressure increase within the final compaction stage. This value is used to normalize the actual pressure deviation value to the order of 0 to 1, so that it can be used in the calculation of the slope attenuation coefficient.

[0108] For example, for X7R-200 layer material, the average pressure value of 500 qualified preforms at the beginning of the compaction final pressing section is 1400N, the target pressure value at the end of pressing is 2000N, and the average pressure difference is 600N. Then the preset reference deviation value can be set to 600N.

[0109] The preset reference uniformity index is determined based on the distribution of the laminated compaction uniformity index at the beginning of the final compaction stage in the historical pressing data of this material type. Specifically, the laminated compaction uniformity index of multiple qualified preforms of this material type at the moment of entering the final compaction stage is statistically analyzed, and the average value is taken as the reference uniformity index. This index characterizes the typical level of laminated uniformity at the beginning of the final compaction stage and is used to normalize the current uniformity index so that it can be used in conjunction with the pressure deviation value in the calculation of the adjustment factor.

[0110] For example, for X7R-200 layer material, if the average value of the stack compaction uniformity index of 500 qualified preforms at the beginning of the compaction final compression section is 0.05, then the preset reference uniformity index can be set to 0.05.

[0111] For example, suppose the current value of the layer compaction uniformity index is 0.034, and the absolute value of the pressure deviation is 150N. The preset reference deviation value is 600N, the preset reference uniformity index is 0.05, and the preset lower limit value is 0.3. Calculate the slope attenuation coefficient = 1 - (150 / 600) × (0.034 / 0.05) = 0.83. 0.83 is greater than the lower limit value of 0.3, therefore the slope attenuation coefficient is taken as 0.83.

[0112] Finally, a baseline value for the torque rise slope limit is obtained. This baseline value is the default torque rise slope limit for the final compaction stage and can be preset according to the press specifications and material type. For example, the distribution of torque rise slope in the final compaction stage can be statistically analyzed in historical pressing data for this material type, and the average slope of 50 N·m / s that makes the stack compaction uniformity index converge stably is taken as the baseline value. The baseline value is multiplied by the calculated slope attenuation coefficient to obtain the adjusted torque rise slope limit value, and the current torque rise slope limit value is updated to this adjusted value.

[0113] For example, if the baseline value of the torque rise slope limit is 50 N·m / s and the slope decay coefficient is 0.83, then the adjusted torque rise slope limit value = 50 × 0.83 = 41.5 N·m / s. This value is updated to the currently used torque rise slope limit value, and the torque rise rate during subsequent pressing will be constrained by this new limit value.

[0114] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0115] In summary, this step controls the pressure rise rate while ensuring continuous optimization of the compaction uniformity of the laminate. This avoids excessive pressure rise that could damage the laminate structure when the improvement in uniformity slows down, and also avoids excessive torque restriction that could prolong the pressing cycle. This achieves a balanced control of uniformity and pressing efficiency within the final compaction stage.

[0116] S500: Combine the current value of the elastic modulus of the transmission chain and the density uniformity index of the stack to calculate the deformation pre-compensation position, and combine it with the original target position of the pressure head to obtain the pressure head displacement control position.

[0117] In this step, the deformation pre-compensation position is calculated based on the elastic modulus of the transmission chain and the uniformity index of the compaction of the layers within the compaction final compression section. Combined with the original target position of the indenter, the final indenter displacement control position is obtained.

[0118] Step S500 in the method provided by the present invention includes: Obtain the elastic modulus of the drive train of the MLCC press; Obtain the pressure value at the current sampling moment, divide the pressure value by the elastic modulus of the transmission chain, and obtain the elastic deformation under the current pressure value; Obtain the current value of the stack compaction uniformity index corresponding to the current displacement window, and obtain the dynamic compensation coefficient according to the preset mapping relationship between the stack compaction uniformity index and the compensation coefficient. Multiply the elastic deformation by the dynamic compensation coefficient to obtain the deformation pre-compensation position. Obtain the original target position of the indenter, which is the indenter position value corresponding to the preset target thickness when the MLCC preform is pressed; The displacement control position of the indenter is obtained by subtracting the deformation pre-compensation position from the original target position of the indenter.

[0119] Specifically, the elastic modulus of the transmission chain corresponding to the current MLCC press is first read from the equipment parameter database of the MLCC press or from the pre-built transmission chain elastic modulus calibration table. The elastic modulus of the transmission chain is the stiffness coefficient of the transmission chain composed of the lead screw, coupling and bearing of the MLCC press when it undergoes elastic deformation under force, and the unit is N / mm. It represents the pressure required for the transmission chain to produce 1mm of elastic deformation. It has been pre-stored in the equipment parameter database through the pre-built transmission chain elastic modulus calibration table.

[0120] The elastic modulus of the transmission chain is obtained through a pre-calibration experiment: a known pressure is applied to the press under no-load conditions, and the elastic displacement of the transmission chain is measured. The ratio of pressure to displacement is then calculated. For example, in the calibration experiment, an elastic displacement of 0.2 mm was measured when a pressure of 1000 N was applied, and an elastic displacement of 0.4 mm was measured when a pressure of 2000 N was applied, with a ratio of 5000 N / mm. After calibration, the elastic modulus value is stored in the equipment parameter database for direct retrieval during operation.

[0121] For example, the elastic modulus of the drive train of the current MLCC press is read from the equipment parameter database as 5000 N / mm.

[0122] Next, obtain the pressure value at the current sampling moment, divide this pressure value by the elastic modulus of the transmission chain, and the quotient is the elastic deformation of the transmission chain under the current pressure value, in mm. For example, if the pressure value at the current sampling moment is 1850N and the elastic modulus of the transmission chain is 5000N / mm, then the elastic deformation = 1850 / 5000 = 0.37mm.

[0123] Next, obtain the current value of the stack compaction uniformity index corresponding to the current displacement window. Based on the preset mapping relationship between the stack compaction uniformity index and the compensation coefficient, determine the corresponding dynamic compensation coefficient.

[0124] The mapping relationship adopts a piecewise function form: when the current value of the layer compactness uniformity index is greater than 0.05, that is, in the high-level fluctuation range, the dynamic compensation coefficient = 0.7 + 0.25 × (0.1 - current value) / 0.05, with a value range of [0.7, 0.95). The larger the current value, the smaller the compensation coefficient. When the current value of the layer compactness uniformity index is less than or equal to 0.05, that is, in the low-level stable range, the dynamic compensation coefficient = 0.95 + 0.05 × (0.05 - current value) / 0.05, with a value range of [0.95, 1.0]. The smaller the current value, the closer the compensation coefficient is to 1.

[0125] For example, the current value of the stacking density uniformity index of the current displacement window is 0.032, which is less than 0.05, and it has entered the low-level stable range. The dynamic compensation coefficient is 0.95 + 0.05 × (0.05 - 0.032) / 0.05 = 0.968.

[0126] Next, the calculated elastic deformation is multiplied by the dynamic compensation coefficient. The resulting product is the deformation pre-compensation position, in mm. The physical meaning of the deformation pre-compensation position is: during the pressing process, the pressure head's downward stroke is reduced in advance to allow space for the elastic deformation of the transmission chain to rebound after pressure holding and unloading, so that the actual thickness of the MLCC preform rebounds to the target position after unloading.

[0127] For example, if the elastic deformation is 0.37 mm and the dynamic compensation coefficient is 0.968, then the deformation pre-compensation position is approximately 0.37 × 0.968 ≈ 0.358 mm.

[0128] Furthermore, the original target position of the indenter corresponding to the current MLCC preform is read from the process parameter database. The process parameter database stores the preset target thickness and the original target position of the indenter for different product specifications, which are entered by the process personnel according to the product specifications before each batch of production. The original target position of the indenter is the indenter position value corresponding to the preset target thickness when the MLCC preform is pressed, in mm. The original target position of the indenter is set according to the product specifications before each batch of production, and its calculation method is: Original target position of indenter = Initial position of indenter - (Initial thickness of preform - Preset target thickness), where the initial position of the indenter is the position value when the indenter contacts the upper surface of the preform.

[0129] For example, if the current preset target thickness of the MLCC preform is 2.0 mm, the initial thickness of the preform is 10.0 mm, and the initial position of the indenter is 10.000 mm (that is, the position value recorded when the indenter contacts the upper surface of the preform), then the original target position of the indenter = 10.000 - (10.0 - 2.0) = 2.000 mm.

[0130] Finally, subtract the deformation pre-compensation position from the original target position of the indenter; the difference is the indenter displacement control position, in mm. The indenter displacement control position is the displacement command value used to actually stop the indenter.

[0131] For example, if the initial target position of the indenter is 2.000mm and the deformation pre-compensation position is 0.358mm, then the indenter displacement control position = 2.000 - 0.358 = 1.642mm. The control system sets the indenter displacement command to 1.642mm. After the indenter moves to this position, it stops and enters the pressure holding stage. After unloading, the elastic deformation of the transmission chain rebounds, and the actual thickness of the blank recovers to the target thickness of 2.0mm.

[0132] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0133] In summary, this step dynamically adjusts the compensation degree of the elastic deformation of the transmission chain according to the real-time uniformity of the material. When the uniformity is poor, the compensation is reduced to prevent local overpressure, and when the uniformity is good, sufficient compensation is provided to ensure the accurate thickness of the blank after unloading and springback, thereby improving the control accuracy of the final pressing size and the consistency of the product.

[0134] In summary, this invention forms a complete closed-loop control from state perception and stage determination to parameter optimization by real-time sensing of material dynamic stiffness, continuous quantification of stack uniformity, intelligent identification of pressing stage, and adaptive adjustment of torque slope and deformation compensation. This ensures both stack density and uniformity while balancing pressing efficiency and product dimensional accuracy.

[0135] Example 2, as Figure 3 As shown, this invention provides an adaptive control system for MLCC presses based on dynamic material identification, the system comprising: The data acquisition and preprocessing module 11 is used to continuously acquire the torque current and pressure head position of the servo motor of the MLCC press, analyze and obtain the pressure sequence and displacement sequence, and calculate the ratio of the pressure change to the displacement change in each sampling interval as the dynamic stiffness sequence. The displacement sequence includes the displacement change at the next sampling time and the previous sampling time.

[0136] The process involves continuously acquiring the torque current and pressure head position of the servo motor of the MLCC press, analyzing and obtaining pressure and displacement sequences, and calculating the ratio of pressure change to displacement change within each sampling interval as a dynamic stiffness sequence. This includes: The torque current and pressure head position of the servo motor of the MLCC press are continuously collected at a preset sampling period, and the timestamp of each sampling moment is recorded. For each sampling moment, the torque current is multiplied by the torque constant and transmission ratio of the servo motor to obtain the pressure value at that sampling moment, and the pressure values ​​at all sampling moments are arranged in order of timestamp to form the pressure sequence; For each sampling time, based on the pressure head position, the pressure head displacement value at that sampling time is obtained, and the pressure head displacement values ​​at all sampling times are arranged in order of timestamp to form the original displacement sequence; For every two adjacent sampling times in the original displacement sequence, the displacement value of the indenter at the later sampling time is subtracted from the displacement value of the indenter at the previous sampling time to obtain the displacement change within the sampling interval. The displacement changes of all the sampling intervals are arranged in order of timestamp to form the displacement sequence. For each sampling interval, the difference between the pressure value corresponding to the next sampling moment and the pressure value corresponding to the previous sampling moment in the pressure sequence is taken as the pressure change within the sampling interval. The pressure change is then divided by the displacement change in the displacement sequence corresponding to the sampling interval to obtain the dynamic stiffness value corresponding to the sampling interval. The dynamic stiffness values ​​of all sampling intervals are arranged in order of timestamp to form the dynamic stiffness sequence.

[0137] The stack compactness uniformity analysis module 12 is used to extract the dynamic stiffness sequence corresponding to all sampling times within the preset displacement window length during the MLCC pressing process, and analyze and obtain the stack compactness uniformity index.

[0138] In the MLCC pressing process, whenever the cumulative displacement of the indenter reaches a preset displacement window length, the dynamic stiffness sequence corresponding to all sampling moments within the displacement window length is extracted, and the stack compaction uniformity index is analyzed and obtained, including: Starting from the moment of suppression, the displacement change in the displacement sequence is accumulated. When the accumulated value is greater than or equal to the preset displacement window length for the first time, the current moment is recorded as the window truncation moment. Taking the window truncation time as the end point, backtracking to the starting time when the accumulated value first exceeds or equals the preset displacement window length, the dynamic stiffness values ​​corresponding to all sampling times between the starting time and the window truncation time constitute the window dynamic stiffness sequence. Calculate the arithmetic mean of all dynamic stiffness values ​​in the window dynamic stiffness sequence as the window average stiffness value, and calculate the standard deviation of all dynamic stiffness values ​​in the window dynamic stiffness sequence as the window stiffness standard deviation. Divide the standard deviation of the window stiffness by the average stiffness value of the window to obtain the stack density uniformity index corresponding to the current displacement window.

[0139] The pressing stage identification module 13 is used to identify the pressing stage by combining the current displacement position of the MLCC press head, the current pressing material type, and the stack compaction uniformity index. The pressing stage identification results include the venting stage, the plastic flow stage, and the compaction final pressing stage.

[0140] The step of identifying the pressing stage by combining the current displacement position of the MLCC press head, the current pressing material type, and the stack compaction uniformity index includes: Obtain the current displacement position of the MLCC press head, where the current displacement position of the press head is the displacement value of the press head at the current sampling time; Obtain the current pressing material type, wherein the current pressing material type includes the stacked structure information of the ceramic dielectric film and the inner electrode layer; A pressing stage identification model is constructed. The current displacement position of the pressing head, the current pressing material type, and the stack compaction uniformity index are input into the pressing stage identification model. The pressing stage identification model outputs the pressing stage identification result at the current moment. The pressing stage identification result includes the venting stage, the plastic flow stage, and the compaction final pressing stage.

[0141] The construction of the suppression stage identification model includes: Acquire historical pressing data, which includes the indenter displacement value, current pressing material type, and stack compaction uniformity index of multiple MLCC preforms at each sampling moment during the complete pressing process; Each sampling moment in the historical compression data is labeled with a stage to obtain an output supervision set, wherein the output supervision set includes the exhaust section label, the plastic flow section label, and the compaction final pressure section label corresponding to the historical compression data; A suppression stage identification model is constructed. The input layer of the suppression stage identification model contains three input nodes, and the output node outputs the suppression stage labels obtained by identification. The suppression stage identification model is trained under supervision using the historical suppression data and the output supervision set. The suppression stage labels in the output supervision set are converted into one-hot encoded vectors. The training objective is to minimize the cross-entropy loss between the three confidence scores output by the suppression stage identification model and the one-hot encoded vectors. The weight parameters of the suppression stage identification model are iteratively updated until the cross-entropy loss converges, thus obtaining the trained suppression stage identification model.

[0142] The step of marking each sampling time in the historical suppression data as a stage includes: For each sampling time in the historical suppression data, the dynamic stiffness sequence of the displacement window where the sampling time is located is obtained, and the arithmetic mean of all dynamic stiffness values ​​in the dynamic stiffness sequence is calculated as the window average stiffness value of the sampling time. For each sampling moment in the historical suppression data, the pressure value at the sampling moment is obtained. When the pressure value is less than the preset upper limit threshold of the low pressure range, it is determined that the pressure sequence at the sampling moment is in the low value range. When the pressure sequence is in a low value range, and the growth rate of the average stiffness value of the window of multiple consecutive displacement windows is less than the preset gradual growth threshold, the sampling time covered by the multiple consecutive displacement windows is marked as the exhaust segment. When the average stiffness value of the window is positively increasing compared to the previous adjacent displacement window, and multiple consecutive displacement windows maintain a positive increasing trend, it is determined that the dynamic stiffness sequence is in a continuously rising state, and the stack compaction uniformity index corresponding to the multiple consecutive displacement windows is obtained. When the stack compaction uniformity index first increases and then decreases continuously, the sampling time covered by the displacement window that satisfies the dynamic stiffness sequence being in a continuously rising state and the stack compaction uniformity index showing an upward and then downward trend is marked as the plastic flow section. When the average stiffness value of the window exceeds the preset stiffness limit corresponding to the current pressing material type, and the stack compaction uniformity index is less than the preset upper limit threshold of the stable range, it is determined that the stack compaction uniformity index has entered the low-level stable range, and the sampling time covered by the displacement window that satisfies the condition that the average stiffness value of the window exceeds the preset stiffness limit and the stack compaction uniformity index has entered the low-level stable range is marked as the final compaction section.

[0143] The torque slope dynamic optimization module 14 is used to dynamically optimize the torque rise slope limit value of the MLCC press based on the stack compaction uniformity index after determining that the press has entered the compaction final pressing section.

[0144] The step of dynamically optimizing the torque rise slope limit of the MLCC press based on the laminate compaction uniformity index after determining that it has entered the final compaction stage includes: After determining that the compaction has entered the final compaction section, the current value of the stack compaction uniformity index corresponding to the current displacement window and the previous window value of the stack compaction uniformity index corresponding to the previous adjacent displacement window are obtained. The difference between the current value of the stack compaction uniformity index and the previous window value of the stack compaction uniformity index is calculated as the uniformity change. When the uniformity change is negative and its absolute value is greater than the preset fast convergence threshold, the current proportional gain parameter and torque rise slope limit value of the MLCC press are kept unchanged. When the uniformity change is negative and its absolute value is less than or equal to the preset fast convergence threshold, the pressure deviation value between the pressure value at the current sampling time and the preset target pressure value is obtained, and the torque rise slope limit value is adjusted according to the pressure deviation value.

[0145] Wherein, when the uniformity change is negative and its absolute value is less than or equal to the preset fast convergence threshold, the pressure deviation value between the pressure value at the current sampling time and the preset target pressure value is obtained, and the torque rise slope limit value is adjusted according to the pressure deviation value, including: When the uniformity change is negative and its absolute value is less than or equal to the preset fast convergence threshold, the pressure value at the current sampling time and the preset target pressure value are obtained, and the pressure deviation value is obtained by subtracting the pressure value at the current sampling time from the preset target pressure value. Obtain the current value of the stack compaction uniformity index corresponding to the current displacement window, and calculate the slope attenuation coefficient based on the current value of the stack compaction uniformity index and the absolute value of the pressure deviation value. Obtain a reference value for the torque rise slope limit value, multiply the reference value for the torque rise slope limit value by the slope attenuation coefficient to obtain the adjusted torque rise slope limit value, and update the torque rise slope limit value to the adjusted torque rise slope limit value.

[0146] The deformation pre-compensation and position control module 15 is used to calculate the deformation pre-compensation position by combining the elastic modulus of the transmission chain and the current value of the stack density uniformity index, and to obtain the pressure head displacement control position by combining the original target position of the pressure head.

[0147] The calculation of the deformation pre-compensation position based on the current value of the elastic modulus of the transmission chain and the density uniformity index of the laminate, and the acquisition of the indenter displacement control position based on the original target position of the indenter, includes: Obtain the elastic modulus of the drive train of the MLCC press; Obtain the pressure value at the current sampling moment, divide the pressure value by the elastic modulus of the transmission chain, and obtain the elastic deformation under the current pressure value; Obtain the current value of the stack compaction uniformity index corresponding to the current displacement window, and obtain the dynamic compensation coefficient according to the preset mapping relationship between the stack compaction uniformity index and the compensation coefficient. Multiply the elastic deformation by the dynamic compensation coefficient to obtain the deformation pre-compensation position. Obtain the original target position of the indenter, which is the indenter position value corresponding to the preset target thickness when the MLCC preform is pressed; The displacement control position of the indenter is obtained by subtracting the deformation pre-compensation position from the original target position of the indenter.

[0148] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0149] This specification and accompanying drawings are merely illustrative examples of the invention and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its scope. Therefore, if such modifications and modifications fall within the scope of the invention and its equivalents, the invention is intended to include these modifications and modifications.

Claims

1. An adaptive control method for MLCC presses based on dynamic material identification, characterized in that, include: The torque current and pressure head position of the servo motor of the MLCC press are continuously collected. The pressure sequence and displacement sequence are analyzed and obtained. The ratio of the pressure change to the displacement change in each sampling interval is calculated as the dynamic stiffness sequence. The displacement sequence includes the displacement change at the next sampling time and the previous sampling time. During the MLCC pressing process, whenever the cumulative displacement of the indenter reaches the preset displacement window length, the dynamic stiffness sequence corresponding to all sampling moments within the displacement window length is extracted, and the stack compactness uniformity index is analyzed and obtained. The pressing stage is identified by combining the current displacement position of the MLCC press head, the current pressing material type, and the stack compaction uniformity index. The pressing stage identification results include the venting stage, the plastic flow stage, and the final compaction stage. After determining that the MLCC press has entered the final compaction stage, the torque rise slope limit value is dynamically optimized based on the stack compaction uniformity index. By combining the current value of the elastic modulus of the transmission chain and the density uniformity index of the stack, the deformation pre-compensation position is calculated, and by combining the original target position of the indenter, the displacement control position of the indenter is obtained.

2. The adaptive control method for MLCC presses based on dynamic material identification according to claim 1, characterized in that, The continuous acquisition of the torque current and pressure head position of the servo motor of the MLCC press, analysis to obtain pressure and displacement sequences, and calculation of the ratio of pressure change to displacement change in each sampling interval as a dynamic stiffness sequence, including: The torque current and pressure head position of the servo motor of the MLCC press are continuously collected at a preset sampling period, and the timestamp of each sampling moment is recorded. For each sampling moment, the torque current is multiplied by the torque constant and transmission ratio of the servo motor to obtain the pressure value at that sampling moment, and the pressure values ​​at all sampling moments are arranged in order of timestamp to form the pressure sequence; For each sampling time, based on the pressure head position, the pressure head displacement value at that sampling time is obtained, and the pressure head displacement values ​​at all sampling times are arranged in order of timestamp to form the original displacement sequence; For every two adjacent sampling times in the original displacement sequence, the displacement value of the indenter at the later sampling time is subtracted from the displacement value of the indenter at the previous sampling time to obtain the displacement change within the sampling interval. The displacement changes of all the sampling intervals are arranged in order of timestamp to form the displacement sequence. For each sampling interval, the difference between the pressure value corresponding to the next sampling moment and the pressure value corresponding to the previous sampling moment in the pressure sequence is taken as the pressure change within the sampling interval. The pressure change is then divided by the displacement change in the displacement sequence corresponding to the sampling interval to obtain the dynamic stiffness value corresponding to the sampling interval. The dynamic stiffness values ​​of all sampling intervals are arranged in order of timestamp to form the dynamic stiffness sequence.

3. The adaptive control method for MLCC presses based on dynamic material identification according to claim 1, characterized in that, During the MLCC pressing process, whenever the cumulative displacement of the indenter reaches a preset displacement window length, the dynamic stiffness sequence corresponding to all sampling moments within the displacement window length is extracted, and the stack compaction uniformity index is analyzed and obtained, including: Starting from the moment of suppression, the displacement change in the displacement sequence is accumulated. When the accumulated value is greater than or equal to the preset displacement window length for the first time, the current moment is recorded as the window truncation moment. Taking the window truncation time as the end point, backtracking to the starting time when the accumulated value first exceeds or equals the preset displacement window length, the dynamic stiffness values ​​corresponding to all sampling times between the starting time and the window truncation time constitute the window dynamic stiffness sequence. Calculate the arithmetic mean of all dynamic stiffness values ​​in the window dynamic stiffness sequence as the window average stiffness value, and calculate the standard deviation of all dynamic stiffness values ​​in the window dynamic stiffness sequence as the window stiffness standard deviation. Divide the standard deviation of the window stiffness by the average stiffness value of the window to obtain the stack density uniformity index corresponding to the current displacement window.

4. The adaptive control method for MLCC presses based on dynamic material identification according to claim 1, characterized in that, The identification of the pressing stage is performed by combining the current displacement position of the MLCC press head, the current pressing material type, and the stack density uniformity index, including: Obtain the current displacement position of the MLCC press head, where the current displacement position of the press head is the displacement value of the press head at the current sampling time; Obtain the current pressing material type, wherein the current pressing material type includes the stacked structure information of the ceramic dielectric film and the inner electrode layer; A pressing stage identification model is constructed. The current displacement position of the pressing head, the current pressing material type, and the stack compaction uniformity index are input into the pressing stage identification model. The pressing stage identification model outputs the pressing stage identification result at the current moment. The pressing stage identification result includes the venting stage, the plastic flow stage, and the compaction final pressing stage.

5. The adaptive control method for MLCC presses based on dynamic material identification according to claim 4, characterized in that, The construction of the suppression stage identification model includes: Acquire historical pressing data, which includes the indenter displacement value, current pressing material type, and stack compaction uniformity index of multiple MLCC preforms at each sampling moment during the complete pressing process; Each sampling moment in the historical compression data is labeled with a stage to obtain an output supervision set, wherein the output supervision set includes the exhaust section label, the plastic flow section label, and the compaction final pressure section label corresponding to the historical compression data; A suppression stage identification model is constructed. The input layer of the suppression stage identification model contains three input nodes, and the output node outputs the suppression stage labels obtained by identification. The suppression stage identification model is trained under supervision using the historical suppression data and the output supervision set. The suppression stage labels in the output supervision set are converted into one-hot encoded vectors. The training objective is to minimize the cross-entropy loss between the three confidence scores output by the suppression stage identification model and the one-hot encoded vectors. The weight parameters of the suppression stage identification model are iteratively updated until the cross-entropy loss converges, thus obtaining the trained suppression stage identification model.

6. The adaptive control method for MLCC presses based on dynamic material identification according to claim 5, characterized in that, The step of marking each sampling time in the historical suppression data includes: For each sampling time in the historical suppression data, the dynamic stiffness sequence of the displacement window where the sampling time is located is obtained, and the arithmetic mean of all dynamic stiffness values ​​in the dynamic stiffness sequence is calculated as the window average stiffness value of the sampling time. For each sampling moment in the historical suppression data, the pressure value at the sampling moment is obtained. When the pressure value is less than the preset upper limit threshold of the low pressure range, it is determined that the pressure sequence at the sampling moment is in the low value range. When the pressure sequence is in a low value range, and the growth rate of the average stiffness value of the window of multiple consecutive displacement windows is less than the preset gradual growth threshold, the sampling time covered by the multiple consecutive displacement windows is marked as the exhaust segment. When the average stiffness value of the window is positively increasing compared to the previous adjacent displacement window, and multiple consecutive displacement windows maintain a positive increasing trend, it is determined that the dynamic stiffness sequence is in a continuously rising state, and the stack compaction uniformity index corresponding to the multiple consecutive displacement windows is obtained. When the stack compaction uniformity index first increases and then decreases continuously, the sampling time covered by the displacement window that satisfies the dynamic stiffness sequence being in a continuously rising state and the stack compaction uniformity index showing an upward and then downward trend is marked as the plastic flow section. When the average stiffness value of the window exceeds the preset stiffness limit corresponding to the current pressing material type, and the stack compaction uniformity index is less than the preset upper limit threshold of the stable range, it is determined that the stack compaction uniformity index has entered the low-level stable range, and the sampling time covered by the displacement window that satisfies the condition that the average stiffness value of the window exceeds the preset stiffness limit and the stack compaction uniformity index has entered the low-level stable range is marked as the final compaction section.

7. The adaptive control method for MLCC presses based on dynamic material identification according to claim 1, characterized in that, After determining that the MLCC press has entered the final compaction stage, the torque rise rate limit is dynamically optimized based on the laminate compaction uniformity index, including: After determining that the compaction has entered the final compaction section, the current value of the stack compaction uniformity index corresponding to the current displacement window and the previous window value of the stack compaction uniformity index corresponding to the previous adjacent displacement window are obtained. The difference between the current value of the stack compaction uniformity index and the previous window value of the stack compaction uniformity index is calculated as the uniformity change. When the uniformity change is negative and its absolute value is greater than the preset fast convergence threshold, the current proportional gain parameter and torque rise slope limit value of the MLCC press are kept unchanged. When the uniformity change is negative and its absolute value is less than or equal to the preset fast convergence threshold, the pressure deviation value between the pressure value at the current sampling time and the preset target pressure value is obtained, and the torque rise slope limit value is adjusted according to the pressure deviation value.

8. The adaptive control method for MLCC presses based on dynamic material identification according to claim 7, characterized in that, When the uniformity change is negative and its absolute value is less than or equal to the preset fast convergence threshold, the pressure deviation between the pressure value at the current sampling time and the preset target pressure value is obtained, and the torque rise slope limit is adjusted according to the pressure deviation value, including: When the uniformity change is negative and its absolute value is less than or equal to the preset fast convergence threshold, the pressure value at the current sampling time and the preset target pressure value are obtained, and the pressure deviation value is obtained by subtracting the pressure value at the current sampling time from the preset target pressure value. Obtain the current value of the stack compaction uniformity index corresponding to the current displacement window, and calculate the slope attenuation coefficient based on the current value of the stack compaction uniformity index and the absolute value of the pressure deviation value. Obtain a reference value for the torque rise slope limit value, multiply the reference value for the torque rise slope limit value by the slope attenuation coefficient to obtain the adjusted torque rise slope limit value, and update the torque rise slope limit value to the adjusted torque rise slope limit value.

9. The adaptive control method for MLCC presses based on dynamic material identification according to claim 1, characterized in that, The calculation of the deformation pre-compensation position based on the current values ​​of the transmission chain elastic modulus and the ply compactness uniformity index, combined with the original target position of the indenter, yields the indenter displacement control position, including: Obtain the elastic modulus of the drive train of the MLCC press; Obtain the pressure value at the current sampling moment, divide the pressure value by the elastic modulus of the transmission chain, and obtain the elastic deformation under the current pressure value; Obtain the current value of the stack compaction uniformity index corresponding to the current displacement window, and obtain the dynamic compensation coefficient according to the preset mapping relationship between the stack compaction uniformity index and the compensation coefficient. Multiply the elastic deformation by the dynamic compensation coefficient to obtain the deformation pre-compensation position. Obtain the original target position of the indenter, which is the indenter position value corresponding to the preset target thickness when the MLCC preform is pressed; The displacement control position of the indenter is obtained by subtracting the deformation pre-compensation position from the original target position of the indenter.

10. An adaptive control system for an MLCC press based on dynamic material identification, characterized in that, The adaptive control method for MLCC presses based on material dynamic identification as described in any one of claims 1 to 9 includes: The data acquisition and preprocessing module is used to continuously acquire the torque current and pressure head position of the servo motor of the MLCC press, analyze and obtain the pressure sequence and displacement sequence, and calculate the ratio of the pressure change to the displacement change in each sampling interval as the dynamic stiffness sequence. The displacement sequence includes the displacement change at the next sampling time and the previous sampling time. The stack compaction uniformity analysis module is used to extract the dynamic stiffness sequence corresponding to all sampling times within the preset displacement window length during the MLCC pressing process, and analyze and obtain the stack compaction uniformity index. The pressing stage identification module is used to identify the pressing stage by combining the current displacement position of the MLCC press head, the current pressing material type, and the stack compaction uniformity index. The pressing stage identification results include the venting stage, the plastic flow stage, and the compaction final pressing stage. The torque slope dynamic optimization module is used to dynamically optimize the torque rise slope limit value of the MLCC press based on the stack compaction uniformity index after determining that the press has entered the compaction final pressing section. The deformation pre-compensation and position control module is used to calculate the deformation pre-compensation position by combining the current value of the elastic modulus of the transmission chain and the density uniformity index of the stack, and to obtain the displacement control position of the pressure head by combining the original target position of the pressure head.