A method and system for automatic control of a paperboard digital slotter

Abstract

The application relates to the technical field of paperboard processing equipment automatic control, and discloses a paperboard digital slotting machine automatic control method and system. The method comprises the following steps: through synchronous acquisition of a tool vibration signal and cutting force data, a tool wear characteristic vector is obtained through processing and frequency domain analysis; if a wear condition is met, a depth compensation amount is extracted, a servo motor position instruction is generated in combination with a proportional integral differential control algorithm, a Z-axis executing mechanism is driven to dynamically correct a cutting depth; a parameter is optimized through a depth residual feedback, a compensation gain is dynamically calibrated and iteratively corrected until the residual converges to a preset precision. The method realizes real-time perception of a tool wear state and accurate compensation of a depth, solves the problem of unstable slotting depth in traditional control, improves processing precision, product consistency and production efficiency, and provides technical support for high-precision automatic production of paperboard slotting.

Description

Technical Field

[0001] This invention relates to the field of automation control technology, and in particular to an automated control method and system for a digital cardboard slotting machine. Background Technology

[0002] In one existing technology, the depth control of a digital cardboard slotting machine mainly relies on preset fixed parameters or periodic manual calibration. A single cutting depth command is set through the industrial control system to drive the Z-axis actuator. Precision changes caused by tool wear are only adjusted manually in stages, without establishing a linkage mechanism between real-time tool status perception and dynamic cutting depth compensation. While some solutions introduce vibration monitoring, cutting force acquisition, or grating ruler position feedback, they only perform basic processing and independent control on single signals, lacking in-depth analysis of tool wear conditions and failing to achieve coordinated linkage between wear perception, feedforward compensation, and closed-loop fine-tuning, thus failing to form a complete precision control closed loop.

[0003] However, due to the progressive wear of cutting tools during continuous machining, resulting in changes such as edge dulling and micro-chipping, not only does the actual grooving depth deviate from the target value, but it also alters the dynamic characteristics of the cutting process. Existing technologies cannot capture these continuous dynamic changes in real time and lack a comprehensive, targeted control strategy encompassing "prediction-compensation-correction-adaptation." Conventional periodic calibration suffers from significant lag, and simple feedback control can only correct deviations after the fact, failing to prevent deviations from occurring at their source. As tool wear accumulates, it can lead to uneven grooving depth, increased burrs on the groove edge, and substandard folding accuracy, even causing entire batches of products to fail, severely impacting production efficiency and the smooth operation of subsequent processes.

[0004] Therefore, existing technologies suffer from insufficient accuracy in tool depth compensation, making it difficult to meet the demands of high-precision, continuous production. Summary of the Invention

[0005] This invention provides an automated control method and system for a digital cardboard slotting machine to address the shortcomings of insufficient accuracy in tool depth compensation in existing technologies.

[0006] In a first aspect, the present invention provides an automated control method for a digital slotting machine for cardboard, comprising:

[0007] Simultaneously collect the vibration simulation signal and cutting force simulation signal of the tool during the cardboard slotting process, and calculate the initial state index after preprocessing;

[0008] Frequency domain analysis is performed on the initial state index to obtain the harmonic amplitude sequence. A spectral feature matrix is ​​constructed based on the harmonic amplitude sequence as the wear feature vector.

[0009] The energy percentage is calculated based on the wear feature vector. If the energy percentage meets the preset wear feature threshold, a corresponding status identifier is generated. The historical depth deviation set is obtained by searching the preset wear depth mapping library based on the status identifier and the benchmark compensation amount is calculated. The preliminary depth adjustment parameters are calculated based on the benchmark compensation amount.

[0010] Based on the preliminary depth adjustment parameters, a proportional-integral-derivative control algorithm is used to calculate and obtain the motor adjustment command;

[0011] The motor adjustment command is analyzed, electromagnetic torque is generated and executed, linear axial displacement is generated, and the actual cutting depth is calculated by combining the real-time feedback signal of the grating ruler. The actual cutting depth is compared with the target grooving depth to obtain the real-time depth correction result.

[0012] The depth residual is calculated based on the real-time depth correction result, and the depth residual is fed back to the proportional-integral-derivative control algorithm. Iterative compensation is performed in combination with the historical error accumulation to obtain the optimized depth adjustment parameters.

[0013] The dynamic compensation gain is determined based on the wear feature vector, and the depth adjustment parameters are modulated to generate and execute Z-axis servo drive commands. The depth residual is iteratively corrected until it converges to a preset accuracy range.

[0014] In a second aspect, the present invention provides an automated control system for a digital cardboard slotting machine, comprising:

[0015] The data processing module is used to synchronously collect the vibration simulation signal and cutting force simulation signal of the tool during the cardboard slotting process, and calculate the initial state index after preprocessing.

[0016] The wear analysis module is used to perform frequency domain analysis on the initial state index to obtain the octave amplitude sequence, and construct a spectral feature matrix based on the octave amplitude sequence as the wear feature vector;

[0017] The parameter preliminary module is used to calculate the energy ratio based on the wear feature vector. If the energy ratio meets the preset wear feature threshold, a corresponding status identifier is generated. The module retrieves the historical depth deviation set from the preset wear depth mapping library based on the status identifier and calculates the benchmark compensation amount. The module then calculates the preliminary depth adjustment parameters based on the benchmark compensation amount.

[0018] The integral calculation module is used to calculate the motor adjustment command by using the proportional-integral-derivative control algorithm based on the preliminary depth adjustment parameters.

[0019] The dynamic correction module is used to parse the motor adjustment command, generate electromagnetic torque and execute it, generate linear axial displacement, and calculate the actual cutting depth in combination with the real-time feedback signal of the grating ruler. The actual cutting depth is compared with the target grooving depth to obtain the real-time depth correction result.

[0020] The parameter optimization module is used to calculate the depth residual based on the real-time depth correction result, and feed the depth residual back to the proportional-integral-derivative control algorithm. Iterative compensation is performed in combination with the historical error accumulation to obtain the optimized depth adjustment parameters.

[0021] The iterative correction module is used to determine the dynamic compensation gain based on the wear feature vector, modulate the depth adjustment parameters, generate and execute Z-axis servo drive commands, and iteratively correct the depth residual until it converges to a preset accuracy range.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] (1) This invention constructs a full-process collaborative control system of wear perception, feedforward compensation, closed-loop fine adjustment and adaptive optimization, which breaks through the core limitation of the disconnect between perception and control in the existing technology. It is not a simple accumulation of known technologies. Through the accurate identification of tool status, the depth deviation is pre-adjusted in advance, reducing the machining deviation from the root and solving the inherent defects of the conventional scheme of delayed correction.

[0024] (2) The present invention realizes the deep coordination of feedforward compensation and feedback closed loop. The adjustment parameters generated based on wear characteristics provide a precise benchmark for closed loop control, while the actual processing depth feedback can reversely calibrate the compensation accuracy, forming a two-way linkage control mechanism, which greatly improves the control accuracy and stability of slotting depth.

[0025] (3) The adaptive parameter adjustment of the present invention works in synergy with the preceding and following control links. It can match the optimal control parameters according to the real-time cutting state, adapt to the characteristic changes of the tool throughout its entire life cycle, and eliminate the need for frequent manual shutdown for calibration. While ensuring machining accuracy, it effectively improves continuous production efficiency and reduces product scrap rate. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of an automated control method for a digital cardboard slotting machine provided in the first embodiment of the present invention;

[0027] Figure 2 This is a schematic diagram of the automated control system for a digital cardboard slotting machine provided in the second embodiment of the present invention. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] Reference Figure 1 The first embodiment of the present invention provides an automated control method for a digital cardboard slotting machine, comprising the following steps:

[0030] S1, synchronously collect the vibration simulation signal and cutting force simulation signal of the tool during the cardboard slotting process, and calculate the initial state index after preprocessing;

[0031] S2, Simultaneously perform frequency domain analysis on the initial state index to obtain the harmonic amplitude sequence, and construct a spectral feature matrix based on the harmonic amplitude sequence as the wear feature vector;

[0032] S3, calculate the energy percentage based on the wear feature vector. If the energy percentage meets the preset wear feature threshold, generate a corresponding status identifier. Retrieve the historical depth deviation set from the preset wear depth mapping library based on the status identifier and calculate the benchmark compensation amount. Calculate the preliminary depth adjustment parameters based on the benchmark compensation amount.

[0033] S4. Based on the preliminary depth adjustment parameters, a proportional-integral-derivative control algorithm is used to calculate and obtain the motor adjustment command;

[0034] S5, parse the motor adjustment command, generate electromagnetic torque and execute it, generate linear axial displacement, and calculate the actual cutting depth by combining the real-time feedback signal of the grating ruler. Compare the actual cutting depth with the target grooving depth to obtain the real-time depth correction result.

[0035] S6. Calculate the depth residual based on the real-time depth correction result, and feed the depth residual back to the proportional-integral-derivative control algorithm. Combine the historical error accumulation for iterative compensation to obtain the optimized depth adjustment parameters.

[0036] S7. Determine the dynamic compensation gain based on the wear feature vector, modulate the depth adjustment parameter, generate and execute the Z-axis servo drive command, and iteratively correct the depth residual until it converges to the preset accuracy range.

[0037] In step S1, the vibration simulation signal and cutting force simulation signal of the tool during the cardboard slotting process are collected synchronously. After preprocessing, the initial state index is calculated, including:

[0038] S11, synchronously acquires vibration simulation signals and real-time cutting force simulation signals, and generates the original multi-dimensional synchronization signal sequence by aligning them with a unified clock pulse;

[0039] S12, the original multidimensional synchronous signal sequence is denoised to separate pure vibration waveform data and pure cutting force change curve;

[0040] S13, calculate the root mean square amplitude in the time domain based on the pure vibration waveform data, identify the cutting force peak value from the pure cutting force change curve, and generate a dynamic feature dataset;

[0041] S14, calculate the amplitude change rate and force fluctuation deviation based on the dynamic feature dataset to generate the initial state index.

[0042] In step S11, vibration simulation signals and real-time cutting force simulation signals are acquired synchronously and generated into an original multidimensional synchronization signal sequence by aligning them with a unified clock pulse.

[0043] It should be noted that the vibration simulation signal is acquired through an accelerometer installed near the tool holder or machine tool spindle, while the real-time cutting force simulation signal is acquired through a force sensor installed at the tool clamping location. Both types of sensors are installed along the critical path of cutting force and vibration transmission, accurately capturing the dynamic response of the tool during operation. The sampling frequency of the unified clock pulse is set to 10,000 times per second. This frequency is based on the fact that the cardboard slotting cycle is typically in the millisecond range, and high-frequency sampling can completely preserve the instantaneous impact characteristics during the cutting process, avoiding signal distortion. The clock pulse alignment process involves adding a precise timestamp to the acquired data of each signal. Timestamp matching is used to map the vibration signal to the cutting force signal, generating an original multidimensional synchronous signal sequence containing both vibration acceleration and cutting force components. This ensures the consistency of the two types of data in the time dimension, laying the foundation for subsequent joint analysis.

[0044] In step S12, the original multidimensional synchronization signal sequence is denoised to separate pure vibration waveform data and pure cutting force change curve.

[0045] It should be noted that the denoising process employs a combination of wavelet threshold denoising and low-pass filtering. This combination is chosen because wavelet threshold denoising effectively removes high-frequency random noise, while the low-pass filter eliminates background vibration interference generated by the machine tool's own operation. Together, they preserve the effective signal characteristics to the greatest extent possible. The wavelet threshold is set based on the statistical noise intensity of the original signal, using 1.5 times the noise standard deviation as the threshold. Noise components below the threshold are directly set to zero, while signal components above the threshold are retained and reduced. The cutoff frequency of the low-pass filter is set to 1000Hz. This frequency is chosen because the vibration and cutting force signals generated by tool cutting are mainly concentrated in the low-frequency range, and 1000Hz can effectively isolate high-frequency machine tool noise. After denoising, only periodic cutting impact characteristics are retained in the vibration waveform data, and the cutting force variation curve clearly shows the loading-unloading process of each grooving cycle, providing a clean data foundation for subsequent feature extraction.

[0046] In step S13, the root mean square amplitude in the time domain is calculated based on the pure vibration waveform data, the peak value of the cutting force is identified from the pure cutting force variation curve, and a dynamic feature dataset is generated.

[0047] It should be noted that the root mean square (RMS) amplitude in the time domain is calculated by selecting vibration waveform data from 100 consecutive grooving cycles, calculating the square average of the vibration acceleration within each cycle, and then taking the square root of this average to obtain the RMS amplitude for each cycle. This calculation method is based on the fact that the RMS amplitude comprehensively reflects the strength of vibration energy and is suitable as a characterizing indicator of tool condition. Under normal tool conditions, this amplitude is stable in the range of 0.18g to 0.22g. When the tool experiences minor chipping, the amplitude will rapidly rise to above 0.35g. The peak cutting force is identified by traversing the cutting force variation curve and extracting the maximum force value within each grooving cycle. Under normal circumstances, the peak force for a single grooving operation is approximately 420N to 460N. When the tool becomes dull or the coating peels off, the peak force will increase dramatically to the range of 580N to 620N, accompanied by a significant secondary impact peak. The RMS amplitudes and peak cutting forces of all cycles are arranged chronologically to generate a dynamic feature dataset, visually presenting the dynamic changes during tool operation.

[0048] In step S14, the amplitude change rate and force fluctuation deviation values ​​are calculated based on the dynamic feature dataset to generate initial state indicators.

[0049] It should be noted that the amplitude change rate is calculated by taking the root mean square amplitude of 10 consecutive cycles, calculating the difference between the next cycle and the previous cycle, and then dividing by the number of cycles to obtain the average change rate. This indicator reflects the cumulative trend of vibration energy. When the amplitude change rate exceeds 0.08g / cycle, it indicates rapid accumulation of vibration energy, and the tool may experience early wear. The force fluctuation deviation is calculated by taking the difference between the peak cutting force of the current cycle and the average peak force of the previous 20 cycles, and then dividing by the average peak force to obtain the deviation percentage. When the deviation is consistently greater than 18%, it indicates a significant deterioration in cutting stability, which may indicate tool dulling or fluctuations in the cardboard material. Combining the amplitude change rate and the force fluctuation deviation forms an abnormal state feature vector as an initial state indicator. This indicator can mutually verify the two dimensions of vibration energy accumulation and cutting force mutation, improving the sensitivity and reliability of tool condition monitoring. For example, when the amplitude change rate exceeds the threshold and the force fluctuation deviation is simultaneously greater than 15%, it can be determined in advance that the tool has entered the early wear stage, avoiding subsequent out-of-tolerance grooving dimensions.

[0050] In step S2, frequency domain analysis is performed on the initial state index to obtain an octave amplitude sequence. A spectral feature matrix is ​​constructed based on the octave amplitude sequence as a wear feature vector, including:

[0051] S21, extract the vibration signal segment and cutting force signal segment associated with the initial state index, and perform frequency domain transformation to generate vibration frequency domain signal and cutting force frequency domain spectrum;

[0052] S22, extract energy values ​​from the vibration frequency domain signal and the cutting force frequency domain spectrum to generate a harmonic amplitude sequence;

[0053] S23, calculate the frequency doubling amplitude and harmonic distortion coefficient of the main shaft rotation frequency based on the frequency doubling amplitude sequence;

[0054] S24, the frequency doubling growth amplitude and harmonic distortion coefficient are combined to construct a spectral feature matrix and determine the wear feature vector.

[0055] In step S21, the vibration signal segment and cutting force signal segment associated with the initial state index are extracted and frequency domain transformation is performed to generate the vibration frequency domain signal and the cutting force frequency domain spectrum.

[0056] It should be noted that the signal segment extraction method involves selecting continuous data segments containing multiple complete grooving cycles. The data segment length is set to 50 grooving cycles. This length is chosen to fully cover the dynamic characteristics of a single continuous tool operation, avoiding frequency domain analysis bias caused by insufficient data volume. Fast Fourier Transform (FFT) is used for signal conversion. This transform is chosen because it efficiently converts time-domain signals to frequency-domain signals, clearly presenting the frequency distribution characteristics of the signal. In the vibration frequency domain signal, the spindle rotation frequency and its harmonic components are most significant, while the cutting force frequency domain spectrum prominently displays the harmonic distribution related to tool cutting impact. Through frequency domain conversion, tool state changes can be captured from the perspective of energy distribution, providing complementary verification with time-domain indicators.

[0057] In step S22, energy values ​​are extracted from the vibration frequency domain signal and the cutting force frequency domain spectrum to generate an octave amplitude sequence.

[0058] It should be noted that the spindle rotation frequency is determined based on the machine tool spindle speed. For example, when the spindle speed is 1500 rpm, the spindle rotation frequency is 25 Hz. The range for extracting the harmonic amplitude is from the 1st to the 6th harmonic. This range is based on the fact that tool wear leads to a significant increase in high-harmonic energy, and harmonics 1-6 can fully cover the main frequency components generated by tool cutting. The extraction method involves extracting the amplitude at each harmonic position from the vibration frequency domain signal and the cutting force frequency domain spectrum, respectively, to obtain the vibration harmonic amplitude sequence and the cutting force harmonic amplitude sequence. Then, the amplitudes at corresponding positions in the two sets of sequences are weighted and averaged. The weights are set to 0.6 for vibration amplitude and 0.4 for cutting force amplitude. The basis for this weighting is that the vibration signal is more sensitive to tool wear, while the cutting force signal can reflect changes in cutting load, and the weighted average can combine the advantages of both types of signals. Under normal tool conditions, the first harmonic amplitude accounts for about 65% of the total energy, the second harmonic amplitude accounts for about 18%, and the higher harmonic amplitudes decay rapidly to below 5%. After tool wear, the proportions of the second and third harmonic amplitudes will increase significantly, reaching about 28% and 15% respectively.

[0059] In step S23, the frequency doubling amplitude and harmonic distortion coefficient of the spindle rotation frequency are calculated based on the frequency doubling amplitude sequence.

[0060] It should be noted that the frequency doubling amplitude is calculated as the ratio of the current period's second harmonic amplitude to the average second harmonic amplitude of the previous 30 periods. This ratio reflects the growth trend of high-harmonic energy. When the ratio exceeds 1.45, it indicates that tool wear has entered a perceptible stage. This threshold is set based on statistical analysis of a large amount of tool wear experimental data, determining 1.45 as the critical value between normal and worn states. The harmonic distortion coefficient is calculated by dividing the sum of all harmonic amplitudes above the second harmonic by the first harmonic amplitude. This coefficient reflects the degree of nonlinear distortion in the cutting process. Under normal tool conditions, the harmonic distortion coefficient remains in the range of 0.35 to 0.42. After tool dulling, this coefficient rapidly rises to above 0.68 because tool wear leads to uneven cutting impact and exacerbates nonlinear distortion.

[0061] In step S24, the frequency doubling growth amplitude and harmonic distortion coefficient are fused to construct a spectral feature matrix and determine the wear feature vector.

[0062] It should be noted that the fusion method involves multiplying the frequency doubling amplitude by a weight of 0.6 and the harmonic distortion coefficient by a weight of 0.4, and then summing them to obtain the feature value for a single detection moment. The basis for these weights is that the frequency doubling amplitude is more sensitive to early wear, while the harmonic distortion coefficient is more sensitive to later wear stages. Weighted fusion can comprehensively cover the wear characteristics throughout the entire tool lifecycle. After continuous acquisition over multiple cycles, the feature values ​​for each detection moment are arranged chronologically to form a two-dimensional spectral feature matrix. Each row of the matrix corresponds to a detection moment, and each column corresponds to a feature index. This matrix directly serves as the wear feature vector. For example, when multiple rows in the matrix show frequency doubling amplitudes consistently higher than 1.3 and harmonic distortion coefficients exceeding 0.55, it can be determined that the tool has reached moderate wear, providing a basis for subsequent depth compensation.

[0063] In step S3, if the energy percentage is calculated based on the wear feature vector, and if the energy percentage meets a preset wear feature threshold, a corresponding status identifier is generated. A preset wear depth mapping library is retrieved based on the status identifier to obtain a historical depth deviation set, and a baseline compensation amount is calculated. Preliminary depth adjustment parameters are calculated based on the baseline compensation amount, including:

[0064] S31, parse the wear feature vector, separate the high-frequency signal components, and calculate the ratio of the high-frequency signal components to the total signal energy;

[0065] S32, if the ratio exceeds the preset wear characteristic threshold, the tool is determined to be in a moderate wear state, and a moderate wear state identifier is generated;

[0066] S33, based on the moderate wear state identifier, retrieve the preset wear depth mapping library, obtain the historical depth deviation set, and perform weighted average calculation to obtain the benchmark compensation amount of the depth deviation;

[0067] S34, the reference compensation amount is superimposed on the standard cutting depth command to obtain the preliminary depth adjustment parameters.

[0068] In step S31, the wear feature vector is parsed, the high-frequency signal components are separated, and the ratio of the high-frequency signal components to the total signal energy is calculated.

[0069] It should be noted that signal component separation employs wavelet transform or bandpass filtering. The high-frequency band is defined as the portion with frequencies exceeding five times the spindle rotation frequency. This definition is based on the fact that the minute impacts, frictional noise, and transient distortions generated after tool edge dulling are primarily concentrated in the high-frequency band, while the low-frequency band mainly reflects the energy distribution of normal cutting. The energy ratio is calculated by integrating the power spectral density, dividing the energy integral value of the high-frequency signal by the total energy integral value of the entire signal to obtain the high-frequency energy ratio. This value quantitatively reflects the intensity of high-frequency noise caused by tool wear.

[0070] In step S32, if the ratio exceeds the preset wear characteristic threshold, the tool is determined to be in a moderate wear state, and a moderate wear state identifier is generated.

[0071] It should be noted that the preset wear characteristic threshold is set at 18%. This threshold is based on statistical data on the high-frequency energy percentage of a large number of sharp and moderately worn cutting tools. The high-frequency energy percentage of sharp tools is usually maintained between 8% and 12%, mainly contributed by slight vibrations during normal cutting. However, after the tool enters moderate wear, the high-frequency energy percentage will rise significantly. When it exceeds 18%, tearing or burrs are likely to appear at the grooving edge, requiring depth compensation to be activated. The determination method is as follows: if the high-frequency energy percentage value exceeds 18% for three consecutive grooving cycles, the system will automatically generate a moderate wear status indicator to avoid misjudgment caused by a single fluctuation and ensure the reliability of wear status determination.

[0072] In step S33, the preset wear depth mapping library is retrieved according to the moderate wear state identifier to obtain the historical depth deviation set, and a weighted average calculation is performed to obtain the benchmark compensation amount of the depth deviation.

[0073] It should be noted that the preset wear depth mapping library was initially constructed through offline calibration experiments and supports online self-optimization updates during mass production. During the offline calibration phase, under standard operating conditions consistent with actual mass production, a full lifecycle processing experiment was conducted using brand-new standard grooving tools from the same batch. A fixed number of cardboard pieces were processed at each interval, and wear characteristic data such as the proportion of high-frequency energy corresponding to the tool vibration and cutting force signals at that node were collected synchronously. Simultaneously, a depth measurement device with an accuracy of 0.001mm was used to collect the deviation data between the actual grooving depth and the target depth of the corresponding cardboard. The experiment covered common mass production conditions with different cardboard materials, target grooving depths, and processing parameters. After completing multiple sets of repeated experiments to obtain full and valid data, outliers were removed from the data and the data was categorized by wear. The system groups wear feature intervals and correlates them with corresponding depth deviation values ​​to establish a mapping relationship with wear feature intervals as search keys and baseline depth deviation values ​​as storage values, forming an initial mapping library. During online operation, the system continuously collects real-time wear feature data and actual depth residual data from grating ruler feedback during mass production. After classifying the wear feature intervals, the system uses a sliding weighted average method to iteratively update the baseline deviation values ​​of the corresponding intervals, while automatically filtering out abnormal data with deviations exceeding limits to ensure the accuracy of the mapping library and its adaptability to operating conditions. During the retrieval process, the system matches the historical depth deviation set of the corresponding interval with the current wear feature, assigns weights according to the timeliness of the data collection time, and performs a weighted average calculation to finally obtain the baseline compensation amount for the depth deviation.

[0074] Specifically, new cutting tools are used to perform full lifecycle processing under typical working conditions, such as standard cardboard and fixed process parameters. The high-frequency energy percentage at each processing node is collected simultaneously as wear characteristics, and the corresponding actual grooving depth deviation is used as target values ​​to form an initial training dataset. As production data from different cutting tools and batches accumulates, this mapping library is continuously optimized and updated using machine learning algorithms. In this embodiment, the preferred machine learning algorithm is a multiple linear regression model, whose expression is: ,in, This is the depth deviation compensation amount. For the proportion of high-frequency energy, The hardness of the cardboard material can be quantified based on the cardboard density or compressive strength. , , The regression coefficients are used for the training dataset. The training dataset's feature dimensions include high-frequency energy percentage (continuous values) and cardboard material stiffness (categorical or continuous values). The sample size accumulates with production, and at least 200 valid data sets are included in the initial modeling. The training and validation sets are determined by sorting historical data chronologically, using the first 70% as the training set and the last 30% as the validation set. Five-fold cross-validation is used to evaluate the model's prediction accuracy, with the root mean square error (RMSE) as the metric. For linear regression models, the least squares method is used to solve for the regression coefficients, with no additional hyperparameters. If Support Vector Regression (SVR) is used as an alternative, the radial basis function (RBF) is chosen, with a penalty coefficient of 1.0 and a kernel parameter of 0.1, optimized through grid search. For every 100 newly added valid data sets, the model is retrained using all the data, or the regression coefficients are updated each time new samples are acquired using online gradient descent, ensuring that the mapping relationship adapts to changes in working conditions.

[0075] It is worth noting that the regression coefficients are not fixed values ​​set in advance, but are obtained by fitting historical experimental data using the least squares method. Specifically, during the offline calibration stage, actual depth deviation data under different high-frequency energy ratios and different cardboard material hardness are collected to construct the aforementioned multiple linear regression equation. The least squares method is used to minimize the sum of squared residuals between the predicted and actual values, thereby calculating the regression coefficients. , , As new data accumulates during mass production, the system can dynamically update the regression coefficients using recursive least squares or batch retraining to ensure that the mapping relationship always adapts to the current working conditions. Therefore, the regression coefficients are set based on the statistical fitting of the measured data. Their specific values ​​vary with the data samples, but the determination method (least squares) is public and clear, and those skilled in the art can fully implement it according to the description in the manual.

[0076] In step S34, the reference compensation amount is superimposed on the standard cutting depth command to obtain the preliminary depth adjustment parameters.

[0077] It should be noted that the standard cutting depth command is set according to the requirements of the cardboard processing technology. For example, the standard slotting depth of a certain batch of cardboard is 2.50mm. The superposition method is to directly add the depth deviation benchmark compensation amount to the standard cutting depth command. For example, when the compensation amount is 0.019mm, the initial depth adjustment parameter is 2.50mm + 0.019mm = 2.519mm. Through this superposition, the insufficient cutting depth caused by tool wear can be offset, making the actual processing depth closer to the target size. When continuously producing the same batch of cardboard, the system repeats the above high-frequency energy ratio monitoring and depth compensation calculation every 10 slotting cycles, gradually accumulating and adjusting the parameters to ensure that the slotting depth remains stable and consistent throughout the entire tool life cycle.

[0078] In step S4, based on the preliminary depth adjustment parameters, a proportional-integral-derivative (PID) control algorithm is used to calculate and obtain motor adjustment commands, including:

[0079] S41, Collect the current position feedback value output by the servo motor encoder, calculate the difference between the current position feedback value and the preliminary depth adjustment parameter, and obtain the position error signal;

[0080] S42, Perform discrete-time integration on the position error signal according to the preset integral time constant, and combine proportional and differential operations to obtain the integral cumulative compensation, proportional adjustment component and differential prediction correction.

[0081] S43, the integral cumulative compensation, proportional adjustment component and derivative prediction correction are linearly superimposed to generate the total control output;

[0082] S44, perform pulse sequence format conversion on the total control output to obtain a motor adjustment command that includes integral term cumulative compensation.

[0083] In step S41, the current position feedback value output by the servo motor encoder is acquired, and the difference between the current position feedback value and the preliminary depth adjustment parameter is calculated to obtain the position error signal.

[0084] It should be noted that the servo motor encoder resolution is set to 0.001mm. This resolution is based on the fact that the accuracy requirement for cardboard slotting depth is typically within ±0.005mm, and a high-resolution encoder can accurately reflect the actual position of the cutter. The acquisition frequency of the current position feedback value is synchronized with the slotting cycle to ensure that real-time position data can be acquired in each slotting cycle. The position error signal is calculated by subtracting the current position feedback value from the initial depth adjustment parameter. A positive difference indicates that the actual depth is too shallow, and a negative difference indicates that the actual depth is too deep. This error signal essentially represents the depth deviation residual caused by accumulated cutter wear or other mechanical factors, providing input for subsequent control calculations.

[0085] In step S42, discrete-time integration is performed on the position error signal according to the preset integral time constant, and the integral cumulative compensation, proportional adjustment component and differential prediction correction are obtained by combining proportional and differential operations.

[0086] It should be noted that the preset integral time constant is set to a cumulative time window corresponding to 5 to 10 slotting cycles. This time window is set to balance the steady-state error elimination capability and dynamic response speed of the integral component. Too short a time will result in insufficient integral compensation, while too long a time will lead to response lag. The discrete-time integral operation involves summing the position error signals of each cycle to obtain the cumulative integral compensation. This compensation reflects the system's memory effect on long-term deviations, effectively eliminating steady-state errors. For example, when the position error signal is consistently positive, the integral term will gradually increase, continuously compensating for insufficient cutting depth. The proportional adjustment component is calculated by multiplying the position error signal by a preset proportional gain (usually set to 0.8). This gain is based on the principle of rapid response to instantaneous error changes. For example, when a slotting operation is affected by unforeseen factors causing a sudden increase in position error, the proportional term directly amplifies the error by the gain, providing immediate correction. The differential prediction correction is calculated by multiplying the difference between the position error signal of the current period and the previous period by a preset differential gain (usually set to 0.2). The basis of this gain is that it can predict the trend in advance based on the rate of change of the error. If the error is increasing from small to large, the differential term outputs a negative value to suppress overshoot. Conversely, it helps to accelerate convergence and improve the quality of dynamic response.

[0087] In step S43, the integral cumulative compensation, proportional adjustment component and derivative prediction correction are linearly superimposed to generate the total control output.

[0088] It should be noted that the linear superposition method involves directly adding the three components without additional weight adjustments. This method is based on the fact that the proportional, integral, and derivative components have already achieved a balance in their intensity through their respective gain settings. For example, if the integral term contributes 0.008mm, the proportional term contributes 0.010mm, and the derivative term contributes -0.003mm in a certain cycle, then the total control output is 0.008mm + 0.010mm + (-0.003mm) = 0.015mm. This output represents the position increment command required by the servo motor, which directly corresponds to the depth value that the tool needs to adjust.

[0089] In step S44, a pulse sequence format conversion is performed on the total control output to obtain a motor adjustment command that includes cumulative compensation of integral terms.

[0090] It should be noted that the purpose of format conversion is to convert the total control output into pulse signals that the servo drive can recognize. The frequency and amplitude of the pulse signals correspond to the speed and distance of position adjustment. For example, a position increment of 0.015mm corresponds to a specific number of pulses (150 pulses when the pulse equivalent is 0.0001mm / pulse). The converted motor adjustment command includes integral term cumulative compensation, which enables the depth of cut command in subsequent machining cycles to continuously carry a correction offset, providing strong anti-disturbance capability. In long-term batch production, when tool wear gradually increases and the initial depth adjustment parameters are no longer sufficient to fully compensate, the position error signal continuously accumulates additional compensation through the integral path, thereby controlling the grooving depth deviation within ±0.005mm.

[0091] In step S5, the motor adjustment command is parsed, an electromagnetic torque is generated and executed, a linear axial displacement is generated, and the actual cutting depth is calculated in conjunction with the real-time feedback signal from the grating ruler. The actual cutting depth is compared with the target grooving depth to obtain the real-time depth correction result, including:

[0092] S51, acquire the real-time feedback signal from the grating ruler, receive and parse the motor adjustment command, and generate electromagnetic torque;

[0093] S52, the ball screw is driven to rotate according to the electromagnetic torque, which drives the tool to generate a linear axial displacement;

[0094] S53, the actual cutting depth is calculated based on the linear axial displacement and the real-time feedback signal from the grating ruler;

[0095] S54, compare the actual cutting depth with the target grooving depth to obtain the real-time depth correction result.

[0096] In step S51, the real-time feedback signal from the grating ruler is acquired, the motor adjustment command is received and analyzed, and electromagnetic torque is generated.

[0097] It should be noted that the grating ruler is mounted on the linear motion axis of the tool, with a resolution set to 0.001mm. It forms a dual position feedback system with the servo motor encoder, ensuring accurate position detection. The system reads the grating ruler feedback value every 2ms, forming high-frequency closed-loop position data. Motor adjustment commands are received via industrial Ethernet, with transmission latency controlled within 10ms to ensure real-time command execution. Command parsing involves converting the pulse sequence format position adjustment command into a corresponding electromagnetic torque setpoint. The conversion is based on the servo motor's torque constant. For example, when the position adjustment command is a positive increment of 0.015mm, the required electromagnetic torque, calculated based on the motor torque constant, is approximately 2.8Nm. This torque value can drive the motor to generate the corresponding rotation, meeting the power requirements for tool depth adjustment.

[0098] In step S52, the ball screw is driven to rotate according to the electromagnetic torque, which in turn causes the tool to generate a linear axial displacement.

[0099] It should be noted that the ball screw pitch is set to 5mm / revolution. This pitch is chosen to create a suitable transmission ratio between motor rotation and tool displacement, balancing adjustment accuracy and speed. For every 1 revolution of the motor, the tool produces a 5mm axial displacement. Precise motor control allows for micron-level depth adjustments. The electromagnetic torque is applied to the ball screw via a coupling connecting the servo motor output shaft. When the motor rotates, it drives the screw to rotate. The screw nut is fixedly connected to the tool holder, converting rotational motion into linear axial displacement of the tool. The advantages of this transmission structure are high transmission efficiency and small backlash, ensuring precise position adjustment. For example, for every 0.003 revolutions of the motor, the tool produces a precise displacement of 0.015mm.

[0100] In step S53, the actual cutting depth is calculated based on the linear axial displacement and the real-time feedback signal from the grating ruler.

[0101] It should be noted that the calculation method is as follows: using the mechanical zero point of the cutting tool as a reference, the position coordinates fed back by the grating ruler are converted into the cutting depth of the cutting tool relative to the cardboard surface. For example, if the position coordinates fed back by the grating ruler are 8.002mm and the mechanical zero point offset is 0.005mm, then the actual cutting depth is 8.002mm - 0.005mm = 7.997mm. This calculation process needs to deduct the effects of mechanical zero point offset and temperature drift. The temperature drift is compensated for by collecting the ambient temperature through a temperature sensor installed on the machine tool guide rail, and correcting the position coordinates according to the relationship between temperature and the thermal expansion coefficient of the material to ensure the accuracy of the actual cutting depth calculation.

[0102] In step S54, the actual cutting depth is compared with the target grooving depth to obtain the real-time depth correction result.

[0103] It should be noted that the target slotting depth is set according to the processing procedure document. For example, the target slotting depth for a certain batch of cardboard is 8.000mm. The comparison method is to subtract the actual cutting depth from the target slotting depth to obtain the depth deviation value. If the actual depth is 7.992mm, the real-time depth correction result is 8.000mm-7.992mm=+0.008mm. This positive correction result indicates that the actual depth is too shallow and the cutting depth needs to be further increased in subsequent cycles. If the actual depth is 8.006mm, the correction result is -0.006mm, indicating that the actual depth is too deep and the cutting depth needs to be reduced. The real-time depth correction result is also fed back to the position loop controller as the source of a new round of error signals, forming a closed-loop adjustment. Even if the cutting edge of the tool slowly dulls during continuous processing, the system can still maintain a stable cutting depth through accumulated small corrections. After processing 800 pieces of cardboard continuously, the depth deviation can be controlled within ±0.004mm.

[0104] In step S6, the depth residual is calculated based on the real-time depth correction result, and the depth residual is fed back to the proportional-integral-derivative control algorithm. Iterative compensation is then performed using the accumulated historical error to obtain the optimized depth adjustment parameters, including:

[0105] S61, take the opposite number of the real-time depth correction result to obtain the current depth residual;

[0106] S62, the current depth residual is transmitted to the integral stage of the proportional-integral-derivative control algorithm, and iteratively calculated with the historical error accumulation to generate the cumulative correction value of the integral term;

[0107] S63, calculate the integral compensation control quantity based on the cumulative correction value of the integral term and the preset integral gain coefficient;

[0108] S64, the basic control commands for the tool are superimposed and synthesized using the integral compensation control quantity, and the optimized depth adjustment parameters are output.

[0109] In step S61, the real-time depth correction result is inversely represented to obtain the current depth residual.

[0110] It should be noted that the current depth residual is the opposite of the depth deviation value in step S54. The real-time depth correction result itself is the depth deviation (target - actual). The current depth residual directly uses this deviation value, that is, the difference between the actual depth and the target depth. For example, if the actual depth is 7.992mm and the target is 8.000mm, the current depth residual is -0.008mm (actual - target). This residual can intuitively reflect the degree of deviation between the current cutting depth and the target value, providing a clear direction for subsequent optimization.

[0111] In step S62, the current depth residual is transmitted to the integral stage of the proportional-integral-derivative control algorithm, and iteratively calculated with the historical error accumulation to generate the cumulative correction value of the integral term.

[0112] It should be noted that the historical cumulative error is the sum of the depth residuals of all previous cycles. The iterative calculation method is to add the current depth residual to the historical cumulative error to obtain a new cumulative correction value for the integral term. This method can remember and gradually amplify the slowly changing depth deviation over a long period of time, avoiding the neglect of small errors in a single cycle. For example, if the historical cumulative error of the previous cycle is -0.005mm and the current depth residual of the current cycle is -0.008mm, the cumulative correction value for the integral term after iterative calculation is -0.005mm + (-0.008mm) = -0.013mm. This value can reflect the cumulative deviation caused by tool wear and provide a basis for integral compensation.

[0113] In step S63, the integral compensation control quantity is calculated based on the cumulative correction value of the integral term and the preset integral gain coefficient.

[0114] It should be noted that the integral gain coefficient is set to 0.65. This value was determined through step response experiments under multiple operating conditions. This ensures that the integral term quickly eliminates steady-state errors while avoiding system overshoot caused by integral saturation, thus balancing the compensation strength and system stability. Specifically, this coefficient is set to balance the strength of integral compensation with system stability. An excessively large coefficient will lead to overcompensation and oscillations, while an excessively small coefficient will lead to insufficient compensation and failure to converge quickly. The calculation method is to multiply the cumulative correction value of the integral term by the integral gain coefficient to obtain the integral compensation control quantity. For example, if the cumulative correction value of the integral term is -0.013mm and the integral gain coefficient is 0.65, then the integral compensation control quantity is -0.013mm × 0.65 = -0.00845mm. This control quantity is the compensation amount that needs to be corrected to the basic control command. The negative sign indicates that the cutting depth needs to be increased to offset the residual.

[0115] In step S64, the basic control commands for the tool are superimposed and synthesized using the integral compensation control quantity to output the optimized depth adjustment parameters.

[0116] It should be noted that the basic control command is the preliminary depth adjustment parameter obtained in step S3. The superposition and synthesis method is to directly add the basic control command and the integral compensation control quantity algebraically. For example, if the basic control command is 2.519mm and the integral compensation control quantity is 0.00845mm, then the optimized depth adjustment parameter is 2.52745mm. This parameter can further offset the cumulative depth deviation caused by tool wear. Through continuous residual feedback iteration, the depth deviation can be quickly converged to the preset accuracy range, significantly improving the long-term stability of machining accuracy.

[0117] In step S7, the dynamic compensation gain is determined based on the wear feature vector, the depth adjustment parameters are modulated, a Z-axis servo drive command is generated and executed, and the depth residual is iteratively corrected until it converges to a preset accuracy range, including:

[0118] S71, calculate the spectral tilt based on the wear feature vector;

[0119] S72, determine the dynamic compensation gain coefficient based on the degree of spectrum tilt, and perform weighted modulation on the depth adjustment parameter to generate Z-axis servo drive command;

[0120] S73, input the Z-axis servo drive command into the servo motor control loop to drive the tool displacement, and calculate the updated depth residual.

[0121] S74. If the updated depth residual exceeds the preset residual threshold, repeat the iterative correction until the depth residual converges to the preset accuracy range.

[0122] In step S71, the degree of spectral tilt is calculated based on the wear feature vector.

[0123] It should be noted that the spectral tilt is calculated by taking the frequency domain signal corresponding to the wear feature vector and calculating the ratio of low-frequency (1st-2nd harmonic) energy to high-frequency (3rd-6th harmonic) energy. This ratio reflects the distribution trend of the spectrum. When the spectrum shows high low-frequency energy and rapid attenuation of high-frequency energy, the tilt value is large, indicating that the current cutting state of the tool is unstable, possibly affected by local hardness changes in the cardboard or slight dulling of the cutting edge. For example, if the low-frequency energy is 0.6 and the high-frequency energy is 0.4, the spectral tilt is 0.6 ÷ 0.4 = 1.5; if the low-frequency energy is 0.3 and the high-frequency energy is 0.7, the tilt value is 0.3 ÷ 0.7 ≈ 0.43.

[0124] In step S72, the dynamic compensation gain coefficient is determined based on the degree of spectral tilt, and the depth adjustment parameter is weighted and modulated to generate a Z-axis servo drive command.

[0125] It should be noted that the value rule of the dynamic compensation gain coefficient is determined through multi-condition orthogonal optimization experiments and system closed-loop response characteristic analysis. The core purpose is to balance the response speed of depth adjustment and system stability, and avoid overshoot oscillation or compensation response lag caused by fixed gain. The calculation method of spectrum tilt is to divide the total energy of the 1st to 2nd harmonics in the low frequency band by the total energy of the 3rd to 6th harmonics in the high frequency band. The smaller the value, the higher the proportion of high frequency energy, the more severe the tool wear, and the worse the cutting state stability. This invention uses a three-factor, three-level orthogonal experiment covering three core influencing factors: spectrum tilt, cardboard material hardness, and tool wear. The steady-state error, overshoot, and convergence period of depth control are used as evaluation indicators. The optimal value rule is determined through multiple sets of repeated experiments.

[0126] For example, through orthogonal experimental data regression analysis, the fitting relationship between the gain coefficient and the spectral tilt is obtained as follows: the gain coefficient equals 0.2 times the spectral tilt plus 0.3. When the spectral tilt ≤ 0.5, corresponding to the working condition of moderate to heavy tool wear and strong nonlinear cutting impact, the gain coefficient is taken as 0.4, which can control the system overshoot within 2%, balancing compensation accuracy and system stability. When 0.5 < spectral tilt ≤ 0.8, corresponding to the normal mass production working condition of light tool wear, the gain coefficient is taken as 0.6, which has the fastest convergence speed and the best overall control effect. When the spectral tilt > 0.8, corresponding to the tool wear... Under good working conditions and a stable cutting process, a gain coefficient of 0.8 can be used to achieve rapid convergence of small deviations. For special processing conditions such as ultra-thick cardboard and high-speed grooving, those skilled in the art can use steady-state error ≤ ±0.005mm and overshoot <5% as constraints to fine-tune the gain coefficient within the range of ±0.1. The specific implementation of weighted modulation is to compare the optimized depth adjustment parameters with the final drive command of the previous cycle to obtain the expected depth adjustment amount for this round. After multiplying the adjustment amount by the dynamic compensation gain coefficient to complete the sensitivity adjustment, it is synthesized with the drive command of the previous cycle to generate the final Z-axis servo drive command.

[0127] Specifically, to verify the reliability and universality of the above value selection rules, a three-factor, three-level orthogonal experiment was conducted. The factors and levels are set as shown in the table below:

[0128]

[0129] The experiment was arranged using an L9(3^3) orthogonal array, with each experiment repeated three times and the average value taken as the response value. The evaluation indicators were steady-state error (Es), overshoot (σ), and convergence period (Tc). The experimental results and range analysis are shown in the table below:

[0130]

[0131] Taking the overshoot σ as an example, the range of factor A is 2.6, the range of factor B is 0.5, and the range of factor C is 0.4, indicating that the degree of spectral tilt has the most significant impact on the overshoot and is the main basis for adjusting the dynamic compensation gain. Further analysis of variance of the experimental results shows that the value of factor A is much larger than the critical value, confirming its significance. The gain coefficient corresponding to the optimal combination of levels of each factor (A2, B1, C2) is used as the benchmark, and the gain coefficient is optimized in combination with different levels of factor A. A linear relationship is established through regression analysis, in which the gain coefficient is equal to 0.2 times the degree of spectral tilt plus 0.3, with a determination coefficient R²=0.96. This relationship has good fitting accuracy in the range of spectral tilt [0.3, 1.2]. According to the actual control requirements, the rounding range of the gain coefficient segment is finally determined. The average steady-state error of this segmentation rule on the validation set (another 10 sets of working conditions) is ≤0.004mm, the overshoot is <3%, and the convergence period is <2.5, proving its universality and reliability.

[0132] In step S73, the Z-axis servo drive command is input into the servo motor control loop to drive the tool displacement, and the updated depth residual is calculated.

[0133] It should be noted that the execution process of the servo motor control loop is consistent with step S5. After the driving tool generates a corresponding micro-displacement, the updated actual cutting depth value is collected by a grating ruler. The calculation method is to subtract the target grooving depth from the updated actual cutting depth value to obtain the updated depth residual. For example, if the target grooving depth is 8.000mm and the updated actual depth is 7.996mm, then the updated depth residual is -0.004mm, which is significantly converged from the previous -0.008mm.

[0134] In step S74, if the updated depth residual exceeds the preset residual threshold, the iterative correction is repeated until the depth residual converges to the preset accuracy range.

[0135] It should be noted that the preset residual threshold is set to ±0.005mm. This threshold is set based on the industry precision requirements for cardboard slotting. Exceeding this threshold will affect subsequent folding or assembly processes. The converged precision range is set to ±0.002mm to ensure high processing quality standards. The judgment method is as follows: if the absolute value of the updated depth residual is greater than 0.005mm, return to step S71 and repeat the spectrum analysis, dynamic gain determination, weighted modulation, and driving process to form a closed-loop iterative correction; if the absolute value of the residual is ≤0.005mm, continue iterating until the residual converges to within ±0.002mm. For example, if the initial updated residual is -0.004mm, which is already within the threshold range, after two more iterations, the residual shrinks to -0.001mm, meeting the preset precision requirements, and the correction stops. This dynamic gain modulation method based on the degree of spectral tilt can provide targeted compensation for the real-time cutting status of the tool, avoiding problems such as slow adjustment or oscillation that may occur under fixed gain. In long-term production where the thickness of the cardboard material fluctuates slightly or the tool wears gradually, it can maintain a high degree of consistency in the slotting depth.

[0136] Reference Figure 2 The second embodiment of the present invention provides an automated control system for a digital cardboard slotting machine, comprising:

[0137] The data processing module is used to synchronously collect the vibration simulation signal and cutting force simulation signal of the tool during the cardboard slotting process, and calculate the initial state index after preprocessing.

[0138] The wear analysis module is used to perform frequency domain analysis on the initial state index to obtain the octave amplitude sequence, and construct a spectral feature matrix based on the octave amplitude sequence as the wear feature vector;

[0139] The parameter preliminary module is used to calculate the energy ratio based on the wear feature vector. If the energy ratio meets the preset wear feature threshold, a corresponding status identifier is generated. The module retrieves the historical depth deviation set from the preset wear depth mapping library based on the status identifier and calculates the benchmark compensation amount. The module then calculates the preliminary depth adjustment parameters based on the benchmark compensation amount.

[0140] The integral calculation module is used to calculate the motor adjustment command by using the proportional-integral-derivative control algorithm based on the preliminary depth adjustment parameters.

[0141] The dynamic correction module is used to parse the motor adjustment command, generate electromagnetic torque and execute it, generate linear axial displacement, and calculate the actual cutting depth in combination with the real-time feedback signal of the grating ruler. The actual cutting depth is compared with the target grooving depth to obtain the real-time depth correction result.

[0142] The parameter optimization module is used to calculate the depth residual based on the real-time depth correction result, and feed the depth residual back to the proportional-integral-derivative control algorithm. Iterative compensation is performed in combination with the historical error accumulation to obtain the optimized depth adjustment parameters.

[0143] The iterative correction module is used to determine the dynamic compensation gain based on the wear feature vector, modulate the depth adjustment parameters, generate and execute Z-axis servo drive commands, and iteratively correct the depth residual until it converges to a preset accuracy range.

[0144] It should be noted that the automated control system for a digital cardboard slotting machine provided in this embodiment of the invention is used to execute all the process steps of the automated control method for a digital cardboard slotting machine in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0145] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the system embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0146] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. An automated control method for a digital cardboard slotting machine, characterized in that, include: Simultaneously collect the vibration simulation signal and cutting force simulation signal of the tool during the cardboard slotting process, and calculate the initial state index after preprocessing; Frequency domain analysis is performed on the initial state index to obtain the harmonic amplitude sequence. A spectral feature matrix is ​​constructed based on the harmonic amplitude sequence as the wear feature vector. The energy percentage is calculated based on the wear feature vector. If the energy percentage meets the preset wear feature threshold, a corresponding status identifier is generated. The historical depth deviation set is obtained by searching the preset wear depth mapping library based on the status identifier and the benchmark compensation amount is calculated. The preliminary depth adjustment parameters are calculated based on the benchmark compensation amount. Based on the preliminary depth adjustment parameters, a proportional-integral-derivative control algorithm is used to calculate and obtain the motor adjustment command; The motor adjustment command is analyzed, electromagnetic torque is generated and executed, linear axial displacement is generated, and the actual cutting depth is calculated by combining the real-time feedback signal of the grating ruler. The actual cutting depth is compared with the target grooving depth to obtain the real-time depth correction result. The depth residual is calculated based on the real-time depth correction result, and the depth residual is fed back to the proportional-integral-derivative control algorithm. Iterative compensation is performed in combination with the historical error accumulation to obtain the optimized depth adjustment parameters. The dynamic compensation gain is determined based on the wear feature vector, and the depth adjustment parameters are modulated to generate and execute Z-axis servo drive commands. The depth residual is iteratively corrected until it converges to a preset accuracy range. The step of determining the dynamic compensation gain based on the wear feature vector, modulating the depth adjustment parameters, generating and executing Z-axis servo drive commands, and iteratively correcting the depth residual until it converges to a preset accuracy range includes: Calculate the spectral tilt based on the wear feature vector; The dynamic compensation gain coefficient is determined based on the degree of spectral tilt, and the depth adjustment parameter is weighted and modulated to generate Z-axis servo drive commands. The Z-axis servo drive command is input into the servo motor control loop to drive the tool displacement, and the updated depth residual is calculated. If the updated depth residual exceeds the preset residual threshold, the iterative correction is repeated until the depth residual converges to the preset accuracy range. The energy percentage is calculated as follows: based on the power spectral density integral, the energy integral value of the high-frequency band signal is divided by the total energy integral value of the entire signal to obtain the high-frequency energy percentage value. The calculation method for the spectrum tilt is as follows: the sum of the energy of the 1st to 2nd harmonics in the low-frequency band divided by the sum of the energy of the 3rd to 6th harmonics in the high-frequency band. The smaller the value, the higher the proportion of high-frequency energy, the more severe the tool wear, and the worse the stability of the cutting state. The optimal value rule was determined by a three-factor, three-level orthogonal experiment on the three core influencing factors of spectrum tilt, cardboard material hardness, and tool wear, with steady-state error, overshoot, and convergence period of depth control as evaluation indicators. The optimal value rule was determined by multiple sets of repeated experiments.

2. The method according to claim 1, characterized in that, The vibration simulation signal and cutting force simulation signal of the tool during the synchronous acquisition of cardboard slotting process are preprocessed and the initial state index is calculated, including: Simultaneously acquire vibration simulation signals and real-time cutting force simulation signals, and generate an original multi-dimensional synchronization signal sequence by aligning them with a unified clock pulse; The original multidimensional synchronization signal sequence was denoised to separate pure vibration waveform data and pure cutting force variation curve. The root mean square amplitude in the time domain is calculated based on the pure vibration waveform data, the peak value of the cutting force is identified from the pure cutting force variation curve, and a dynamic feature dataset is generated. Based on the dynamic feature dataset, the amplitude change rate and force fluctuation deviation values ​​are calculated to generate initial state indicators.

3. The method according to claim 1, characterized in that, The initial state index is subjected to frequency domain analysis to obtain an octave amplitude sequence. A spectral feature matrix is ​​constructed based on the octave amplitude sequence as a wear feature vector, including: Extract the vibration signal segment and cutting force signal segment associated with the initial state index, and perform frequency domain transformation to generate the vibration frequency domain signal and the cutting force frequency domain spectrum; Energy values ​​are extracted from the vibration frequency domain signal and the cutting force frequency domain spectrum to generate a harmonic amplitude sequence; Calculate the frequency doubling amplitude and harmonic distortion coefficients of the spindle rotation frequency based on the frequency doubling amplitude sequence; The frequency doubling growth amplitude and harmonic distortion coefficient are combined to construct a spectral feature matrix, and the wear feature vector is determined.

4. The method according to claim 1, characterized in that, The process involves calculating the energy percentage based on the wear feature vector. If the energy percentage meets a preset wear feature threshold, a corresponding status identifier is generated. A preset wear depth mapping library is retrieved based on the status identifier to obtain a historical depth deviation set, and a baseline compensation amount is calculated. Preliminary depth adjustment parameters are then calculated based on the baseline compensation amount, including: The wear feature vector is analyzed to separate the high-frequency signal components and calculate the ratio of the high-frequency signal components to the total signal energy. If the ratio exceeds the preset wear characteristic threshold, the tool is determined to be in a moderate wear state, and a moderate wear state identifier is generated. The preset wear depth mapping library is retrieved based on the moderate wear state identifier to obtain the historical depth deviation set, and a weighted average calculation is performed to obtain the baseline compensation amount of the depth deviation. The reference compensation amount is superimposed on the standard cutting depth command to obtain the preliminary depth adjustment parameters.

5. The method according to claim 1, characterized in that, The step of calculating the motor adjustment command based on the preliminary depth adjustment parameters using a proportional-integral-derivative control algorithm includes: The current position feedback value output by the servo motor encoder is collected, and the difference between the current position feedback value and the preliminary depth adjustment parameter is calculated to obtain the position error signal. Based on a preset integral time constant, discrete-time integration is performed on the position error signal, and the integral cumulative compensation, proportional adjustment component and differential prediction correction are obtained by combining proportional and differential operations. The total control output is generated by linearly superimposing the integral cumulative compensation, the proportional adjustment component, and the derivative prediction correction. The total control output is converted into a pulse sequence format to obtain a motor adjustment command that includes cumulative compensation of the integral term.

6. The method according to claim 1, characterized in that, The process involves parsing the motor adjustment command, generating and executing the electromagnetic torque, generating a linear axial displacement, and calculating the actual cutting depth using the real-time feedback signal from the grating ruler. The actual cutting depth is then compared with the target grooving depth to obtain a real-time depth correction result, including: The system collects real-time feedback signals from the grating ruler, receives and analyzes the motor adjustment commands, and generates electromagnetic torque. The electromagnetic torque drives the ball screw to rotate, which in turn causes the tool to produce a linear axial displacement. The actual cutting depth is calculated based on the linear axial displacement and the real-time feedback signal from the grating ruler. The actual cutting depth is compared with the target grooving depth to obtain the real-time depth correction result.

7. The method according to claim 1, characterized in that, The step of calculating the depth residual based on the real-time depth correction result and feeding the depth residual back to the proportional-integral-derivative control algorithm, combined with the historical error accumulation for iterative compensation, to obtain the optimized depth adjustment parameters includes: The inverse of the real-time depth correction result is used to obtain the current depth residual. The current depth residual is transmitted to the integral stage of the proportional-integral-derivative control algorithm, and iteratively calculated with the historical error accumulation to generate the cumulative correction value of the integral term. The integral compensation control quantity is calculated based on the cumulative correction value of the integral term and the preset integral gain coefficient; The basic control commands for the tool are superimposed and synthesized using the integral compensation control quantity to output optimized depth adjustment parameters.

8. An automated control system for a digital cardboard slotting machine, characterized in that, For implementing the method as described in any one of claims 1-7, comprising: The data processing module is used to synchronously collect the vibration simulation signal and cutting force simulation signal of the tool during the cardboard slotting process, and calculate the initial state index after preprocessing. The wear analysis module is used to perform frequency domain analysis on the initial state index to obtain the octave amplitude sequence, and construct a spectral feature matrix based on the octave amplitude sequence as the wear feature vector; The parameter preliminary module is used to calculate the energy ratio based on the wear feature vector. If the energy ratio meets the preset wear feature threshold, a corresponding status identifier is generated. The module retrieves the historical depth deviation set from the preset wear depth mapping library based on the status identifier and calculates the benchmark compensation amount. The module then calculates the preliminary depth adjustment parameters based on the benchmark compensation amount. The integral calculation module is used to calculate the motor adjustment command by using the proportional-integral-derivative control algorithm based on the preliminary depth adjustment parameters. The dynamic correction module is used to parse the motor adjustment command, generate electromagnetic torque and execute it, generate linear axial displacement, and calculate the actual cutting depth in combination with the real-time feedback signal of the grating ruler. The actual cutting depth is compared with the target grooving depth to obtain the real-time depth correction result. The parameter optimization module is used to calculate the depth residual based on the real-time depth correction result, and feed the depth residual back to the proportional-integral-derivative control algorithm. Iterative compensation is performed in combination with the historical error accumulation to obtain the optimized depth adjustment parameters. The iterative correction module is used to determine the dynamic compensation gain based on the wear feature vector, modulate the depth adjustment parameters, generate and execute Z-axis servo drive commands, and iteratively correct the depth residual until it converges to a preset accuracy range.

Images