Intelligent PET fabric cutting control system

Abstract

This invention discloses an intelligent PET fabric cutting control system, which relates to the field of material cutting technology. Through dynamic signal acquisition, frequency decoupling, edge image recognition, and adaptive control, the system achieves full-process intelligent control of the PET fabric cutting process, precisely suppressing chatter marks, optimizing edge quality, and significantly improving cutting accuracy and product consistency. The system is suitable for high-end applications in optics, electrical engineering, and other fields, and has broad application prospects.

Description

Technical Field

[0001] The present invention relates to the technical field of material cutting, in particular to an intelligent PET base cloth cutting control system. Background Art

[0002] A PET fabric cutting device is a mechanical device specifically designed for cutting polyethylene terephthalate (PET) fabrics into fixed lengths, widths, or specific shapes. This device is typically used in PET fabric production lines or post-processing processes. It features automatic feeding, tension control, positioning calibration, and cutting execution, enabling efficient, precise, and non-destructive cutting of continuous rolls of PET fabric. Due to the high toughness and heat resistance of PET fabric, the cutting device requires wear-resistant cutting tools, hot-cut components, or laser cutting modules, supplemented by an intelligent control system to ensure clean cut edges without burrs or melt deformation, thereby meeting the subsequent processing requirements for packaging substrates, electrical insulation, composite film substrates, and other applications.

[0003] The existing technology has the following deficiencies:

[0004] Existing high-precision cutting processes for PET fabrics involve frequency coupling between the cutter's rotational speed and the feed tension. When the cutter's rotational frequency matches the fabric's natural vibration frequency under tension, a "critical resonance" phenomenon can occur, leading to periodic mechanical disturbances during the cutting process. This creates micro-scale periodic vibration marks along the fabric's edges, manifesting as jagged, fine lines along the cutting direction.

[0005] Due to their minute size, these vibration marks are difficult to detect with normal visual inspection and are easily overlooked. When PET fabrics are used in functional scenarios with extremely high edge quality requirements (such as optical films, electrical insulation films, flexible electronic substrates, etc.), these edge defects may cause serious performance degradation problems. On the one hand, micro vibration marks will cause non-uniform scattering of incident light, affecting the optical transmittance, refractive consistency and imaging quality of the material; on the other hand, jagged edges are prone to forming electric field enhancement zones in high-voltage or high-frequency electric field environments, inducing local corona discharge, insulation breakdown and other phenomena, which in turn lead to device short circuits, performance failures and even systemic damage.

[0006] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0007] The purpose of the present invention is to provide an intelligent PET base cloth cutting control system, which realizes intelligent control of the entire PET base cloth cutting process through dynamic signal acquisition, frequency decoupling, edge image recognition and adaptive control, accurately suppresses vibration marks, optimizes edge quality, and significantly improves cutting accuracy and product consistency. It is suitable for high-end applications such as optics and electrical, and has broad application prospects to solve the problems in the above-mentioned background technology.

[0008] To achieve the above objectives, the present invention provides the following technical solutions: an intelligent PET base fabric cutting control system, comprising a vibration data acquisition module, a frequency coupling analysis module, a dynamic frequency decoupling module, an edge disturbance monitoring module, a tangential correction control module, and an edge quality evaluation and optimization module:

[0009] The vibration data acquisition module obtains the tool rotation frequency, material tension changes and feed speed, and constructs a vibration response map for resonance determination;

[0010] The frequency coupling analysis module identifies the characteristic frequency points of harmonic amplification based on the vibration response spectrum and calculates the resonance intensity index for intervention judgment;

[0011] The dynamic frequency decoupling module synchronously adjusts the tool speed and tension response delay according to the resonance intensity index, dynamically adjusts the frequency ratio between the two, and realizes frequency decoupling;

[0012] The edge disturbance monitoring module, after completing frequency decoupling, collects the structural features of the cutting edge, extracts the sawtooth fluctuation amplitude and morphology change rate, and generates an edge state map;

[0013] The tangential correction control module performs tangential correction based on the abnormal trend of the edge state map, fine-tunes the tool entry angle and feed speed, and compensates for residual transient disturbances;

[0014] The edge quality assessment and optimization module compares the edge state maps before and after tangential correction, identifies the residual disturbance distribution, updates the tool motion trajectory and tension adjustment curve, and builds a closed-loop control mechanism to achieve chatter mark suppression and continuous optimization of edge quality.

[0015] Preferably, constructing the vibration response spectrum includes the following steps:

[0016] Obtain tool rotation frequency, material tension change and feed speed, unify the time base, and perform sampling rate reconstruction and amplitude normalization;

[0017] Adopt high-pass filtering and noise suppression algorithm to eliminate low-frequency interference and non-periodic noise, and improve signal quality;

[0018] Perform spectrum analysis and wavelet decomposition to extract the energy distribution of the signal in different frequency intervals and mark the characteristic points of frequency changes;

[0019] A time period division mechanism and a dynamic sampling window control method are introduced to extract frequency features according to the time sliding window to form a dynamic vibration response spectrum.

[0020] Preferably, identifying the harmonic amplification characteristic frequency point and calculating the resonance intensity index includes the following steps:

[0021] Perform fast Fourier transform on the tool rotation frequency, material tension change frequency and feed speed change frequency respectively to obtain their respective frequency amplitude distribution vectors;

[0022] The frequency overlap and amplitude coupling strength of the three types of spectrum vectors are analyzed within the resonant sensitive frequency range to identify characteristic frequency points with convergent frequency components and enhanced amplitude.

[0023] Calculate the total energy proportion, frequency correlation coefficient, bandwidth amplitude gradient and time stability of the characteristic frequency points to form a resonance intensity score value;

[0024] A multi-level judgment threshold is set according to the resonance intensity score, critical resonance frequency points are marked, and target parameters for subsequent dynamic decoupling control are output.

[0025] Preferably, the specific steps of dynamically adjusting the ratio between the tool speed and the tension response delay frequency to achieve frequency decoupling are as follows:

[0026] Identify high-risk frequency coupling sections, extract corresponding time segments and frequency ranges, and determine whether they are in an integer multiple coupling relationship;

[0027] The weighted objective function control method is used to adjust the tool speed trajectory so that its frequency shifts to the resonance frequency point while maintaining the cutting beat and thermal stress response stable.

[0028] The tension response frequency is adjusted through tension feedback delay control, and a joint offset path with the tool frequency is constructed to break the coupling relationship;

[0029] The spectrum energy and frequency overlap are continuously monitored within the sliding time window to determine whether the frequency decoupling effect meets the resonance suppression requirements.

[0030] Preferably, generating the edge state map comprises the following steps:

[0031] An image acquisition device is set at the cutting exit to collect cutting edge image data and perform grayscale processing;

[0032] Edge detection algorithm is used to extract boundary contours, and binary processing is performed to enhance boundary clarity;

[0033] The multi-scale geometric curve fitting method is used to calculate the sawtooth fluctuation amplitude, edge deviation and morphology mutation rate respectively.

[0034] The extracted eigenvalues ​​are plotted as an edge state map along the time axis, and an edge feature comparison and analysis mechanism is introduced to determine whether the tangential correction control logic should be triggered.

[0035] Preferably, performing the tangential correction control comprises the following steps:

[0036] Extract the data segments marked as abnormal areas in the edge state map, obtain the sawtooth fine line distribution sequence and extract the periodic parameters;

[0037] Based on the periodic parameters, the tool entry angle fine-tuning strategy is executed to shift the entry angle within the preset angle range so that it is staggered to avoid the overlapping area of ​​the disturbance wave peaks;

[0038] The feed speed is adjusted collaboratively based on periodic parameters, and the feed speed disturbance compensation value is dynamically calculated using a closed-loop proportional-integral control method;

[0039] Collect the adjusted edge image, generate an updated atlas and perform feature comparison to determine whether a stable edge output state has been achieved.

[0040] Preferably, by comparing the edge state maps before and after the tangential correction, performing the edge quality assessment optimization operation includes the following steps:

[0041] Compare the edge state maps before and after tangential correction to identify the residual disturbance distribution characteristics in the edge area, including amplitude changes, offset trends and periodic fluctuation behaviors;

[0042] According to the spatial position and intensity distribution of the residual disturbance, the motion trajectory of the tool in the next cycle is dynamically adjusted, including the feeding path, speed curve and angle control parameters;

[0043] According to the edge disturbance distribution trend, the feeding tension adjustment curve is synchronously modified, including adjusting the tension loading delay and feedback response coefficient;

[0044] Edge map comparison and control parameter update are continuously performed within multiple cycles to build a closed-loop control mechanism based on frequency dynamic decoupling and edge quality stability, thereby achieving continuous optimization of vibration mark suppression and edge quality.

[0045] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0046] The present invention not only monitors multi-source dynamic signals during the cutting process in real time and implements intelligent frequency decoupling processing based on spectrum analysis and resonance intensity indicators, but also accurately captures the changing trends of microscale vibration marks through high-resolution edge image recognition technology, thereby implementing tangential path fine-tuning and feed rate optimization. In particular, after the system introduces edge quality maps and self-learning evaluation mechanisms, the control system has adaptive adjustment capabilities, can dynamically optimize tool control strategies and feeding parameters, and achieve stable control and continuous improvement of edge quality. Overall, this solution significantly improves the cutting accuracy and consistency of PET base fabrics in high-end application scenarios (such as optics, electrical, flexible electronics, etc.), reduces performance degradation or material scrap caused by edge defects, improves the intelligence level of the cutting process and product yield, and has extremely high engineering application value and industrial promotion prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction to the drawings required for use in the embodiments will be given below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0048] Figure 1 This is a module schematic diagram of the intelligent PET base cloth cutting control system of the present invention. DETAILED DESCRIPTION

[0049] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art.

[0050] The present invention provides Figure 1 The intelligent PET base cloth cutting control system shown in the figure includes a vibration data acquisition module, a frequency coupling analysis module, a dynamic frequency decoupling module, an edge disturbance monitoring module, a tangential correction control module, and an edge quality evaluation and optimization module:

[0051] The vibration data acquisition module obtains the tool rotation frequency, material tension change trend and feed speed curve, and constructs the vibration response spectrum during the PET base fabric cutting process to provide the dynamic data basis required for subsequent resonance judgment;

[0052] In order to effectively predict and intervene in the frequency coupling phenomenon that may occur during the cutting process of PET base fabric, the specific steps are as follows:

[0053] The cutting equipment incorporates multiple sets of high-precision dynamic parameter acquisition devices, each used to monitor the tool rotation frequency, material tension trends, and feed speed curves in real time. Tool rotation frequency is acquired using a magnetoelectric speed sensor mounted on the cutting shaft. This sensor continuously outputs tool shaft rotation period signals with a millisecond sampling period, which is then converted into rotation frequency data per unit time. Material tension trends are captured using a tension detection mechanism located between the feed section and the pre-cutting guide section. This mechanism, comprised of a tension-sensing guide roller and a strain gauge sensor, provides real-time feedback on even the smallest tension fluctuations. Feed speed is acquired by analyzing the encoder signal from the main feed motor and performing data correction based on the servo response signal within the feedback control loop to ensure a realistic and accurate speed curve.

[0054] The data from these three sources are recorded in time series format and uploaded synchronously to the data processing platform. Interpolation correction and amplitude normalization are performed on the three types of data under a unified time base to ensure uniform comparability of the various signals in frequency domain analysis. To this end, high-pass filtering and noise suppression algorithms are used to preprocess the raw data to effectively eliminate high-frequency noise components caused by mechanical jitter, electrical interference, or environmental vibration. Through methods such as spectrum analysis and wavelet decomposition, the rotation frequency sequence, tension fluctuation curve, and feed speed variation trend are deconstructed, and their energy distribution in different frequency ranges is extracted. Significant variation segments are then characterized and labeled to provide basic features for the subsequent resonance judgment model.

[0055] High-pass filtering and noise suppression are commonly used techniques in signal processing. Their primary purpose is to remove low-frequency interference and background noise from the raw acquired signal, retaining high-frequency components useful for analysis, and thus extracting clearer and more reliable feature data. In the vibration data collected during the PET fabric cutting process, signals such as tool speed, tension changes, and feed speed are often mixed with low-frequency noise and random fluctuations caused by equipment operation, mechanical structure resonance, electrical system disturbances, and external environmental factors such as ground vibration or temperature fluctuations. Left untreated, these ineffective components can mask the true high-frequency dynamic characteristics, reducing the accuracy of subsequent frequency coupling identification and resonance analysis. By applying a high-pass filtering algorithm, a suitable cutoff frequency can be set to effectively filter out slowly varying interference components below this frequency, retaining only rapidly varying components above this frequency. These components are more likely to be directly related to cutter vibration, tension fluctuations, or feed anomalies. Furthermore, when combined with an adaptive noise suppression algorithm, highly random but weakly periodic interference components, such as electromagnetic noise and motor gear whistling, can be further identified and suppressed, significantly improving the signal-to-noise ratio and data stability. In summary, the introduction of this algorithm makes the data based on the vibration response spectrum more "clean" and realistic, providing a solid foundation for subsequent spectrum analysis and characteristic frequency extraction, and ensuring the accuracy and robustness of resonance judgment.

[0056] In order to effectively deconstruct the signal and extract the frequency domain features of the rotation frequency sequence, tension fluctuation curve, and feeding speed variation trend, a method combining spectrum analysis and wavelet decomposition is used to comprehensively obtain the energy distribution of the signal in different frequency ranges and realize the feature marking of the significant change section. The specific processing steps are as follows:

[0057] Each type of collected original time series signal is reconstructed and normalized at a unified sampling rate to ensure that the three types of signals remain synchronized and comparable on the time axis;

[0058] Use Fast Fourier Transform (FFT) to perform spectrum analysis on each type of signal, converting it from the time domain to the frequency domain to obtain the amplitude distribution curve within the entire frequency range, which is used to determine whether there is a prominent main frequency, harmonic superposition, or energy accumulation phenomenon;

[0059] To further capture the non-stationary frequency changes of the signal within a local period, the discrete wavelet transform (DWT) is used to perform multi-scale decomposition of the signal. The original signal is decomposed into sub-signals of different frequency bands (such as approximate coefficients and detail coefficients), and the energy intensity at each scale is statistically analyzed to form a multi-band energy distribution map.

[0060] By cross-comparing the peak regions of the spectrum with time periods of concentrated wavelet energy, energy mutation points, frequency transitions, or high-frequency surges are marked as characteristic markers and recorded as timestamps and frequency intervals. These signatures not only reveal abnormal frequency behavior of the signal within certain time periods but also provide precise data support for subsequent frequency coupling identification, resonance intensity assessment, and dynamic control, forming a key foundation for the present invention in vibration response identification.

[0061] After completing data standardization and frequency domain decomposition, the three types of signals are mapped to a unified three-dimensional coordinate space, with time as the horizontal axis, frequency as the vertical axis, and signal amplitude as the vertical axis, to construct a vibration response spectrum with dynamic timing characteristics. This vibration response spectrum is visualized in the form of a heat map. Within any given time period, the intensity area where the frequency of different signals overlaps will appear as an increase in color intensity, thereby revealing whether there is a coupling trend between the rotation frequency and the material tension change at a certain frequency point. Through this spectrum, it is possible to clearly observe whether the frequency intersection of the tool speed fluctuation and the tension change curve within a certain frequency band is accompanied by a corresponding resonance of the feed speed response, thereby intuitively identifying the possible starting point and characteristic range of the resonance.

[0062] Building on the construction of the vibration response map described above, to ensure accurate data support for subsequent frequency coupling analysis and intervention control, a time segmentation mechanism and dynamic sampling window control method are further introduced, performing sliding window processing on the entire vibration process in time segments. Within each sliding time window, the rotational frequency fluctuation spectrum, tension micro-variation spectrum, and feed speed discrete spectrum of the current time period are independently extracted, and the inter-spectral coupling intensity factor is calculated to determine whether the current frequency coupling high-sensitivity zone has been entered. Through continuous comparison and trend fitting of multiple time periods, the time point and duration of critical resonance can be predicted in advance, providing accurate and systematic dynamic basic data support for subsequent frequency interference suppression. This process not only realizes multidimensional dynamic modeling of the raw data but also provides a highly integrated frequency space visualization solution. This achieves a systematic technical leap from perception to prediction, and from static determination to dynamic identification of frequency coupling phenomena, significantly improving the responsiveness and control preemptiveness of critical vibration mark risks during the high-precision cutting process of PET fabrics.

[0063] The "period division mechanism and dynamic sampling window control method" is a technical means for local analysis of non-stationary time series signals. Its purpose is to divide continuous, complex vibration data into multiple small segments along the time dimension, and independently perform feature extraction and frequency analysis in each time segment, thereby improving the recognition sensitivity and real-time response capabilities of transient resonance behavior and frequency mutation phenomena. In this invention, the specific implementation steps of this method are as follows:

[0064] Set a basic time division rule to preliminarily divide the entire vibration data into preset time periods (such as 100ms, 200ms) to form multiple basic analysis periods;

[0065] A sliding window processing mechanism is introduced in each analysis period. That is, by defining a sampling window of fixed or adaptive length (such as containing 1000 data points), the window is gradually slid along the time axis with a certain overlap rate (such as 50%) to extract local continuous data segments for analysis.

[0066] To address the inconsistency of signal frequency changes under different working conditions, a dynamic sampling window control method is adopted to automatically adjust the length or sliding step of the next window according to the frequency density and energy jump of the signal in the previous window, thereby enhancing the window's ability to resolve areas with drastic changes and avoiding the averaging or omission of key signals.

[0067] Spectral analysis, wavelet deconstruction, and characteristic energy extraction are performed within each sampling window to generate a vibration feature set for the current period. This feature is then compared with the data from the previous period to determine whether the current signal is in a high-sensitivity frequency coupling region or a precursor to critical resonance. This method overcomes the limitations of performing global static analysis of the entire signal, enabling vibration response analysis to be locally dynamic, evolve in real time, and continuously tracked. This provides precise and timely support for the timely adjustment of subsequent control strategies and the prediction of marginal disturbances.

[0068] This step aims to establish a comprehensive, real-time, and visual data foundation for critical resonance phenomena that may occur during the PET fabric cutting process, supporting the subsequent accurate assessment and active control of frequency coupling behavior. Specifically, under high-speed, high-tension, and high-precision cutting conditions, a complex dynamic coupling relationship exists between the tool's rotational frequency, material tension fluctuations, and the feeder system's speed. Once this coupling relationship reaches a point where frequencies coincide, it can trigger a resonance effect, resulting in periodic micro-vibration marks on the cut edge, seriously impacting product edge quality. Therefore, by setting up high-precision multi-source sensors to separately capture the tool's real-time rotational frequency, material tension fluctuations during the initial cutting phase, and the feeder's speed curve, these raw time-series signals are synchronously processed, normalized, and denoised. A comprehensive vibration response spectrum can be constructed after these raw time-series signals are synchronously processed, normalized, and filtered. This spectrum not only reflects the dynamic variations of each physical quantity in the time dimension but also reveals the combined effects of these three factors in the frequency dimension and potential resonance risk zones. Through this process, early perception of vibration behavior, active warning of frequency conflicts, and the logical basis for formulating control strategies can be achieved, thereby laying the core data support foundation for subsequent key steps such as resonance identification, decoupling intervention and quality control. It is the starting point and foundation of the entire dynamic control chain in this step.

[0069] The frequency coupling analysis module performs frequency coupling analysis based on the vibration response spectrum, identifies characteristic frequency points with harmonic amplification effects in the vibration frequency domain, and calculates the resonance intensity index based on the characteristic frequency points to provide a basis for judging resonance intervention;

[0070] In order to effectively identify the frequency coupling resonance phenomenon that may occur during the cutting process of PET base fabric, a frequency coupling analysis is proposed based on the constructed vibration response spectrum to identify the harmonic amplification effect caused by frequency overlap and further calculate the resonance intensity index to provide an accurate judgment basis for subsequent intervention control strategy. The specific steps are as follows:

[0071] The three frequency sequences contained in the vibration response spectrum—tool rotation frequency, material tension variation frequency, and feed speed variation frequency—are extracted and reconstructed in the frequency domain on a unified time axis. Fast Fourier transform (FFT) is used to perform discrete spectrum analysis on the frequency sequences within each time segment, obtaining the amplitude distribution of the three signal types within their respective frequency intervals and forming multiple independent spectrum vectors. To avoid energy shifts caused by signal amplitude differences, each sequence is amplitude normalized before spectrum analysis. In addition, invalid frequency bands below the background noise threshold are eliminated to enhance the accuracy of spectral feature recognition.

[0072] Fast Fourier Transform (FFT) is a mathematical algorithm that efficiently converts continuous or discrete time domain signals into frequency domain signals. Its core principle is to transform the representation of a complex sequence in the time domain into a superposition of multiple sine waves of different frequencies, thereby revealing the energy distribution of the signal in each frequency component. In this step, the tool rotation frequency, material tension change, and feed speed curve all appear as unstable signals that vary with time, and their potential resonance characteristics are often hidden in high-frequency disturbances or harmonics. Through the fast Fourier transform, these three types of signals can be converted from the time domain to the frequency domain, and then the spectral characteristics can be identified. The specific operation steps are as follows:

[0073] Sampling and reconstructing each type of collected time series signal to ensure that the sampling rate meets the Nyquist criterion (at least twice the highest frequency of the signal), and performing zero padding processing on the length to make its length meet the FFT calculation requirements;

[0074] Input the normalized and preprocessed (such as filtering and denoising) signal into the FFT algorithm, perform discrete Fourier transform operation, and output a complex spectrum sequence;

[0075] Calculate the modulus (i.e. the amplitude of the complex number) of each frequency component in the spectrum sequence to obtain the energy intensity at that frequency point;

[0076] The frequency-amplitude relationship of each signal type is plotted as an amplitude distribution graph, and the amplitude values ​​are extracted as vectors in frequency order, forming rotation frequency spectrum vectors, tension spectrum vectors, and feed speed spectrum vectors, respectively. These spectrum vectors not only reveal the energy distribution patterns of each signal in different frequency ranges but also provide comparable, calculable, and quantifiable frequency domain basic data for subsequent frequency coupling analysis, serving as a key intermediate result for frequency resonance identification and intervention judgment.

[0077] The three types of spectral vectors are aligned along the frequency axis, and a frequency overlap matching analysis is performed within a specified resonance-sensitive frequency range (typically near the natural frequency of the mechanical system). By calculating the frequency overlap and amplitude coupling strength between different frequency components, regions of convergent frequency components and energy superposition are identified. In particular, frequency bands where harmonic overlap (e.g., integer multiples of 1:1, 1:2, 2:3, etc.) between the tool rotation frequency and the tension fluctuation frequency, with significantly enhanced amplitude, are identified as potential resonant characteristic frequencies.

[0078] For each identified characteristic frequency point, a resonance intensity index is further calculated to quantify the vibration amplification trend caused by that frequency point. The resonance intensity index is calculated in multiple dimensions: first, the proportion of the total energy amplitude of the frequency point to the total energy of the entire spectrum; second, the correlation coefficient between the three types of frequency sequences at that frequency point; third, the rate of change of the amplitude gradient within the bandwidth of the frequency components near the frequency point; and fourth, the duration of the stability of the frequency point during the time evolution process. By weightedly integrating the above indicators, a resonance intensity score value with engineering operability is formed to objectively assess whether the frequency point is a high-risk vibration amplification source.

[0079] Based on the resonance intensity score, a multi-level threshold is set to distinguish between the non-resonance zone, the mild resonance zone, and the critical resonance zone. In the case of a critical resonance zone, the corresponding frequency value, amplitude characteristics, time period information, and the signal source involved in the coupling are recorded and marked as the target parameter source for the subsequent dynamic decoupling process.

[0080] This step aims to conduct in-depth frequency-domain analysis of the multi-source dynamic signals collected during the PET web cutting process (including tool rotation frequency, material tension variations, and feed speed curves). This allows for the identification of potential resonances caused by frequency coupling and a quantitative assessment of the resonance intensity, providing a scientific and precise basis for subsequent intervention and control. Specifically, under high-tension, high-speed operating conditions, PET webs are susceptible to frequency coupling between the mechanical excitation source, primarily the cutter, and the material's inherent vibrations. When the frequencies of these two sources converge within a certain frequency band or form harmonic multiples, energy superposition and amplification occur, manifesting as "critical resonance," which in turn causes periodic disturbances and vibration marks on the cut edge. Frequency coupling analysis based on the vibration response spectrum allows for the spectral decomposition of multiple signal types using methods such as fast Fourier transforms and wavelet decomposition. The amplitude distribution of these signals across different frequency ranges can then be extracted, allowing for the identification of frequency points with high overlap, concentrated energy, and abnormal amplitudes in the time-frequency space. These are then identified as potential characteristic frequencies. Resonance intensity indicators are then calculated for these frequency points, such as frequency domain energy proportion, spectral overlap amplitude, frequency bandwidth concentration, and timing stability, to comprehensively determine whether they pose a substantial frequency coupling risk. This process not only improves the accuracy and timeliness of resonance identification but also avoids the judgment bias caused by relying solely on time waveforms or manual experience. This lays a solid foundation for intelligent vibration mark prediction and early warning, as well as dynamic decoupling control. It is a key analysis step in the entire intelligent PET base fabric cutting control system vibration management solution.

[0081] The dynamic frequency decoupling module performs frequency interference elimination operations based on the resonance intensity index. By synchronously adjusting the tool speed curve and the material tension response delay, it dynamically adjusts the frequency ratio between the tool rotation frequency and the natural vibration frequency formed by the material tension to achieve frequency decoupling processing;

[0082] To proactively intervene in the critical resonance phenomenon that may occur during the cutting of PET fabric, a frequency interference elimination method is proposed after identifying the resonance characteristic frequency point and calculating the resonance intensity index. By synchronously adjusting the tool speed curve and the material tension response delay, the ratio between the tool rotation frequency and the natural vibration frequency formed by the material tension is dynamically changed, thereby effectively achieving frequency decoupling and avoiding the energy superposition and structural vibration mark formation caused by harmonic coupling. The following steps are included:

[0083] Based on the high-risk frequency coupling segments identified in the previous step, the corresponding time segments and frequency ranges are extracted from the time-frequency spectrum as the target interval for decoupling. Within this interval, the tool's rotational frequency trajectory and the material tension fluctuation curve fed back by the tension sensor are compared and analyzed. The synchronization coefficient and frequency difference ratio between the two frequencies are calculated to determine whether they are in an integer multiple coupling relationship (such as 1:1, 1:2, etc.). If the coupling strength index exceeds the preset threshold, the time period is determined to be a critical control window for frequency intervention.

[0084] After determining the control window, fine-tuning the tool speed curve is performed. By modifying the speed setting parameters of the tool drive unit, the target speed trajectory is slightly perturbed, causing its frequency to offset the target coupling frequency. In this step, a weighted objective function control method is preferably adopted, comprehensively considering the current cutting rhythm, material thermal stress response, and tool acceleration capability to generate a set of optimized speed change curves. This ensures that the speed regulation behavior can achieve frequency misalignment without compromising cutting quality and rhythm stability.

[0085] The weighted objective function control method is a dynamic control strategy based on multi-objective optimization. Its core concept is to establish a mathematical function model between multiple influencing factors and set weight coefficients for each factor, thereby deriving the optimal control variable through weighted comprehensive calculation. In this step, the function of this control method is to achieve dynamic optimization and adjustment of the tool speed change trajectory, so that it does not interfere with the overall cutting rhythm, material response behavior, and stability while achieving frequency misalignment and avoiding resonance. The reason for adopting this method is that during the PET base fabric cutting process, the change in tool speed not only affects the degree of frequency coupling, but also affects the cutting beat, thermal stress release state, and load response of the tool mechanism. Therefore, it is necessary to balance control among multiple objectives. By establishing a weighted objective function, three key constraint variables are introduced: the current cutting rhythm stability target to ensure that the speed regulation behavior does not cause cutting length deviation or synchronization error; the second is the safety margin of the material thermal stress response to avoid speed changes causing thermal distribution disturbances in the cutting area, causing melting or deformation of the material edge; and the third is the instantaneous acceleration capacity limit of the tool to prevent control instability caused by acceleration and deceleration exceeding the mechanical drive limit. Based on real-time data, the weighting ratios of various parameters are dynamically adjusted. An optimization algorithm that minimizes the objective function is executed within each sliding time period to generate a speed variation curve that meets the current operating conditions. This curve effectively breaks the coupling relationship with the integer multiple frequency of the material tension fluctuation, achieving frequency dislocation while ensuring the continuity of the cutting process, consistent product quality, and mechanical safety of the equipment. It is a key control link in achieving the coordinated unification of frequency decoupling and cutting stability.

[0086] To further enhance frequency decoupling, the material tension response delay is coordinated and adjusted. Response delay correction control logic is introduced into the pre-cut tension control section. By adding a microsecond response delay to the tension feedback channel or adjusting the filter constant of the tension feedback loop, a controllable offset in the tension fluctuation frequency is achieved. Synchronized with the tool speed adjustment behavior, a joint offset frequency path is constructed to decouple the coupling path between the original natural frequency and the excitation frequency, thereby breaking the harmonic enhancement mechanism at the frequency level.

[0087] After combining the two aforementioned adjustments, the newly generated vibration response spectrum is continuously monitored within multiple sliding time windows to verify whether the frequency overlap has decreased and whether the resonance intensity index is less than the control threshold. The changes in spectral energy before and after decoupling are also recorded. If the frequency decoupling is determined to have reached the target state, the frequency stability maintenance state is entered. If residual coupling still exists, slight frequency perturbations and tension delay optimization are continued until the vibration energy at the critical frequency point drops to a safe range.

[0088] This step proactively intervenes in identified frequency coupling risk areas. By precisely adjusting the frequency ratio between the tool rotation frequency and the material tension response, harmonic coupling between the two can be disrupted within specific frequency ranges. This effectively suppresses critical resonance and prevents periodic edge vibration marks and mechanical disturbances during the PET fabric cutting process. Under high-speed and high-tension conditions, the tool rotation frequency and material tension fluctuation frequency often exhibit dynamic variations. If they form an integer-multiple frequency relationship within a certain period, this can easily trigger frequency overlap and energy superposition, leading to instability. To mitigate this, this step proposes a synchronous adjustment strategy. Based on a resonance intensity index, it identifies time intervals with high coupling intensity and fine-tunes the time delay of the tool rotation curve and the material tension response within these intervals. By adjusting the target speed curve of the tool drive unit, the rotation frequency can be appropriately offset within the critical frequency range, offsetting the coupling region with the material's natural vibration frequency. Furthermore, to maintain coordination, a response delay adjustment mechanism is introduced into the tension feedback path to dynamically adjust the response time of the tension adjustment command, resulting in a corresponding shift in the main frequency component of the material tension. The combination of these two methods achieves frequency dislocation in the frequency domain, disrupting the coupling path and achieving frequency decoupling. This process not only boasts a high degree of real-time and adaptive capabilities, but also maintains stable operation while ensuring the cutting cycle and process requirements. It effectively suppresses mechanical damage and product quality issues caused by frequency resonance, and is a key technical link in achieving the coordinated integration of dynamic steady-state control and precision cutting.

[0089] The edge disturbance monitoring module performs edge disturbance monitoring after completing the frequency decoupling process, obtains the structural change characteristic values ​​of the cutting edge in real time, extracts the sawtooth fluctuation amplitude and morphology change rate, and generates an edge state map to describe the microscopic vibration state of the cutting edge of the PET base fabric;

[0090] To further enhance the real-time monitoring capabilities of PET fabric cutting quality, an edge disturbance monitoring method is proposed after completing the aforementioned frequency decoupling process. By dynamically collecting and modeling the microscopic features of the cutting edge structure changes, a graph data describing the edge state is generated, providing a reliable basis for subsequent quality correction and control optimization. The method specifically includes the following steps:

[0091] A high-resolution image acquisition device is installed at the cutting exit, and its field of view is calibrated to focus on the contour of the cut edge of the PET fabric. The image acquisition device uses an industrial line scan camera or a high-frame-rate area array camera with micron-level resolution, combined with a high-brightness, flicker-free backlight for background enhancement, ensuring that edge jagged textures and contour transitions are clearly captured. The image acquisition cycle is synchronized with the tool speed control logic, ensuring that within each cycle, an image data sequence is obtained that perfectly matches the actual cutting action.

[0092] Structural feature extraction is performed on the collected edge image sequence. First, edge detection algorithms (such as the Sobel operator and the Canny edge operator) are used to extract the edge contours in the image. The detection results are then binarized to enhance edge clarity. Multi-scale geometric curve fitting methods are then used to quantitatively describe the extracted edge contours and extract key geometric feature parameters such as sawtooth fluctuation amplitude, edge deviation, and edge morphology mutation rate. These parameters reflect the current microscopic vibration state of the cut edge, particularly the formation and development trend of vibration marks.

[0093] An edge detection algorithm is an image processing method whose primary function is to identify locations within an image where pixel grayscale values ​​undergo sudden changes, specifically where the boundaries between objects and background or between different structures lie. In this invention, an edge detection algorithm is used to extract contour boundaries from captured images of cut edges of PET fabric to identify any jagged fluctuations, structural changes, or other edge defects, thereby providing a data foundation for subsequent geometric analysis and edge state mapping. The specific implementation steps for this process are as follows:

[0094] Grayscale processing is performed on the original image obtained, and the color image is converted into a single-channel grayscale image to reduce data complexity;

[0095] Apply an edge detection algorithm, such as the Sobel operator or the Canny edge operator, to calculate the grayscale gradient of the image. Taking the Canny algorithm as an example, this method first performs a Gaussian filter on the image to remove noise. It then calculates the horizontal and vertical gradient changes for each pixel. It then uses non-maximum suppression to accurately locate edge points. Finally, it uses a double threshold method to connect the edges to form a continuous boundary.

[0096] The detected edge map is binarized, that is, the points with pixel values ​​higher than a set threshold are set to "1", and the rest are set to "0", thereby enhancing the contrast between the boundary contour and the background, making the edge clearer and easier to extract;

[0097] The processed binary image is fed into a geometric contour fitting algorithm to further extract microscopic features such as sawtooth amplitude and edge deviation. This edge detection and binarization process not only improves the clarity and quantification of image edge structures, but also provides accurate and visual input for real-time monitoring of cutting quality changes. It is a key technical step in achieving intelligent edge disturbance monitoring and anomaly identification.

[0098] The multi-scale geometric curve fitting method is a mathematical process that models and analyzes edge contours at different spatial scales. Its core concept is to perform step-by-step approximation fitting on the same edge contour by setting multiple spatial resolutions or fitting granularities, thereby revealing geometric features that include both overall morphological trends and local detail changes. In this step, the method is used to quantitatively describe the PET base fabric cutting boundary contour obtained after edge detection and binarization processing, extracting key geometric feature parameters including sawtooth fluctuation amplitude, edge deviation, and edge morphology mutation rate to determine whether there are any abnormalities in the micro-perturbation behavior during the cutting process. The specific steps are as follows:

[0099] Extract the contour point set and arrange the two-dimensional coordinate data of the boundary points in the edge image in sequence to form the original edge point sequence;

[0100] Multiple fitting scales (e.g., coarse, medium, and fine) are set, and curve fitting is performed on the point sequence at each scale. Coarse-scale fitting uses spline interpolation or B-spline curve modeling to describe overall edge trends; medium-scale fitting focuses on smaller fluctuations and uses Bezier curves or sliding polynomial approximations; fine-scale fitting uses high-order Fourier curves or piecewise linear approximations to identify local sawtooth features.

[0101] At each fitting scale, the maximum vertical offset of the edge contour is calculated to extract the sawtooth fluctuation amplitude; the average deviation between the contour fitting curve and the ideal straight edge is calculated to obtain the edge deviation; and by comparing the contour change rate in adjacent time frames, the mutation rate of the edge morphology is extracted to identify rapid local disturbance behavior.

[0102] The three types of geometric features are organized into a sequence of feature vectors in chronological order and used to construct an edge state map. Compared to traditional single-scale fitting methods, this method has stronger hierarchical expression capabilities and anomaly detection sensitivity. It can accurately capture the evolution of edge defects under microscopic perturbations and is a core technical link for achieving high-precision edge quality assessment and vibration mark analysis.

[0103] The extracted structural change characteristic values ​​are mapped to the time axis to establish an edge state map. The specific method is: with the acquisition time as the horizontal axis and the characteristic value as the vertical axis, the sawtooth fluctuation amplitude curve, the morphology change rate curve, and the contour deviation trend graph are plotted respectively, thereby constructing a multi-dimensional edge state evolution map. The map not only reflects the changing trend of edge quality in the time dimension, but also identifies the starting point, peak point, and duration interval of vibration mark increase, contour anomaly, or structural instability. For time periods with obvious abnormal trends, they are marked as high-sensitivity disturbance areas and serve as key inputs for subsequent control logic adjustments.

[0104] In order to ensure the stability and adaptability of the edge state map, this step further introduces an edge feature comparison and analysis mechanism to perform a differential analysis on the edge state map of the current cycle and the historical normal sample map, and calculate the change amplitude threshold and offset rate. If the change trend of the current edge vibration state exceeds the set threshold range, the tangential control correction logic will be automatically triggered to provide an early warning of possible disturbance amplification or local resonance backtracking. In addition, through the superposition analysis of multi-cycle maps, it is also possible to determine whether the edge disturbance has periodic characteristics, thereby inferring whether the vibration source is stable and whether the resonance intervention is effective. This edge disturbance monitoring method not only breaks through the traditional passive quality assessment path based on end-product inspection, but also realizes forward-looking perception and closed-loop feedback control of cutting edge quality changes through real-time visual recognition and multi-dimensional feature modeling.

[0105] The edge feature comparison and analysis mechanism is an intelligent analysis method used to identify edge quality trends and sudden anomalies. Its basic principle is to compare the currently collected edge state characteristics with historical normal sample patterns point by point and item by item, extracting the degree of change in edge state in terms of morphology, fluctuation amplitude, stability, and other aspects, and based on this, determine whether quality degradation or potential vibration backtracking issues have occurred. In the present invention, the role of this mechanism is to provide closed-loop feedback for continuous monitoring after frequency decoupling, ensuring the continuity and stability of edge quality over time. The specific implementation method is as follows: First, the fitted and quantified edge feature parameters, including sawtooth fluctuation amplitude, edge deviation, and morphology mutation rate, are extracted from the current cycle, and an edge state map for the current cycle is constructed along the time axis. Second, several sample maps that have been manually calibrated or identified as qualified are retrieved from a historical database as references for normal edge states. Third, a difference analysis is performed between the current map and the reference map. Using a sliding window comparison method, the absolute difference and percentage deviation rate are calculated for each feature dimension, and a change amplitude threshold is set based on the historical statistical fluctuation range. Finally, sections where the difference exceeds the threshold are marked as abnormal regions, and the growth trend of the deviation rate is combined to determine whether they represent critical quality risks. This mechanism not only dynamically identifies edge morphology anomalies but also detects quality risks caused by residual vibration at an early stage, promptly guiding the fine-tuning of subsequent control parameters. This achieves a transition from "frequency control" to "result verification" closed-loop monitoring, which is a key technical support for ensuring the stability of the cut edge quality of PET base fabrics.

[0106] This step, after completing the frequency decoupling process, further confirms the effectiveness of vibration intervention measures by monitoring the structural changes of the cut PET fabric edge in real time and in detail. It also dynamically identifies residual disturbances or potential anomalies, thereby ensuring the microscopic quality stability of the cut edge. Although the previous step decouples the coupling between tool rotation and material tension at the frequency level, nonlinear factors such as slight variations in material thickness, tool wear, and thermal stress accumulation can still induce localized microscale vibrations or critical backtracking vibration marks during the actual cutting process. Therefore, this step utilizes a high-precision image acquisition device to capture real-time images of the PET fabric edge at the cut exit and uses an edge detection algorithm to extract clear edge contour data. Furthermore, a multi-scale geometric curve fitting method is used to quantitatively analyze the edge contour, extracting microstructural features such as sawtooth fluctuation amplitude, edge deviation, and topographic change rate. These characteristic parameters are then organized into a continuous data stream along the time axis to construct a time-resolved edge state map, which clearly reflects the edge's stability, smoothness, and disturbance trends over different time periods. This spectrum can not only be used to identify whether abnormal vibrations still exist in the current cycle, but also serve as historical data input for comparison with normal samples to determine whether further tangential correction or frequency readjustment is necessary. In summary, this step constitutes the core of the "result-oriented control logic", embodying the closed-loop control concept from dynamic frequency intervention to real-time quality verification, significantly improving the adaptability to high-precision cutting tasks and the control capability of edge quality consistency.

[0107] The tangential correction control module performs tangential correction control based on the changing trend of abnormal areas in the edge state map. It also performs micro-calibration of the tool entry angle and feed speed according to the distribution period of the sawtooth fine lines to compensate for the residual transient disturbance behavior after frequency decoupling.

[0108] To further improve edge quality control during PET fabric cutting, a tangential correction control method was proposed after completing frequency decoupling and acquiring an edge state map. This method dynamically identifies the periodic characteristics of sawtooth fine lines caused by residual disturbances based on the changing trends of abnormal regions in the edge state map. It also performs micro-calibration of the tool's feed angle and feed speed, achieving precise compensation for transient disturbances while maintaining frequency decoupling. The method includes the following steps:

[0109] Based on the edge state map constructed in the previous step, the data segments marked as abnormal areas in the map are extracted, and the corresponding sawtooth fine line distribution sequence is obtained in the form of a time window. Through signal reconstruction and morphological analysis methods, the periodic parameters of the sawtooth fine lines are extracted, including indicators such as peak spacing, trough depth variation, and repeatability amplitude. This periodic structural feature can be regarded as a spatiotemporal representation of the residual disturbance, reflecting the unstable echo during the cutting process. This periodic information is then converted into a mathematical feature vector and bound to the tool control command system as the input basis for subsequent compensation calculations.

[0110] Integrating the aforementioned sawtooth periodicity, a fine-tuning strategy for the cutter's entry angle is implemented. By analyzing the sensitivity curve between the cutter's entry angle and the disturbances of the cutting edge, a slight offset of the angle, typically between 0.1° and 0.5°, is applied within the permissible mechanical adjustment range. This offset strategy matches the sawtooth period to the multiples of the entry frequency, ensuring that the cutter avoids the overlapping region of disturbance peaks at the moment of contact, thereby dispersing instantaneous load concentration and suppressing the formation of secondary resonance.

[0111] Based on the adjustment of the fabric feed angle, the feed speed is coordinated and corrected. Since the formation of sawtooth fine lines is often related to instantaneous fluctuations in the feed speed, after capturing edge features with significant periodic disturbances, the feed speed curve is slightly adjusted, particularly during the cut-in and cut-out stages. A rate buffer or ramp compensation mechanism is implemented to establish a temporal relationship between the feed rate and the tool feed frequency that offsets disturbances. To ensure accuracy, this correction strategy employs a closed-loop PI control method based on disturbance response feedback. The optimal rate disturbance value is calculated within each vibration mark cycle, achieving disturbance-reversing intervention.

[0112] The closed-loop PI control method based on disturbance response feedback is a precision control method that uses edge disturbance results as feedback input and implements real-time adjustments through proportional (P) and integral (I) control strategies. Its core purpose is to dynamically correct the tool feed rate or other key parameters based on the actual degree of edge disturbance detected during the cutting process, thereby achieving reverse intervention and gradual suppression of the disturbance source. In this step, this method is mainly used to identify and correct the periodicity of sawtooth fine lines caused by residual vibration. The specific implementation steps are as follows:

[0113] In each cutting cycle, the disturbance characteristic values ​​such as sawtooth fluctuation amplitude, vibration period and edge deviation are obtained based on the edge image analysis results, and compared with the set ideal edge stability threshold to calculate the disturbance error value as the control deviation input;

[0114] Based on this deviation, the proportional (P) term is used to calculate the direct response adjustment caused by the current error, quickly generating a preliminary adjustment command. At the same time, the integral (I) term is activated to correct the trend of error accumulation over time, avoiding adjustment lag or residual accumulation.

[0115] The control outputs of the P and I terms are weighted and synthesized to obtain the real-time control value, which is then used to fine-tune the tool feed rate or angle disturbance compensation parameters.

[0116] In the next cycle, edge disturbance feedback data is retrieved and the error input is updated, forming a closed-loop control process. Through continuous iteration, this method can achieve rapid suppression and adaptive correction of disturbance responses, ensuring stable output of high-quality cutting edges under different operating conditions. This is the key technical means for improving control accuracy and stability in this invention.

[0117] After adjusting the feed angle and feed speed, the edge image of the calibration area is recaptured and an updated edge state map is generated. This new map is compared with the abnormal area map from the previous cycle. If core indicators such as fluctuation amplitude, deviation rate, and vibration period all show a downward trend, the calibration is considered effective. If significant residuals still exist, parameters are adjusted and compensation is iteratively performed until a stable edge output is achieved.

[0118] The purpose of this step is to, after completing the frequency decoupling process and identifying the abnormal change trend of the PET base fabric cutting edge, implement a set of tangential correction control strategies with fine adjustment capabilities based on the real-time feedback information in the edge state map. By micro-calibrating the tool's cloth entry angle and feed speed, it effectively compensates for the transient disturbance behavior remaining after frequency decoupling. Although the previous frequency adjustment and tension control have significantly reduced the risk of systemic resonance, due to the continuous effect of nonlinear interference factors such as material thickness fluctuations, tool wear, and thermal stress accumulation, small but periodic micro-perturbations may still occur within the local time window, resulting in the repeated distribution and slight offset of edge sawtooth fine lines. To suppress the edge morphology fluctuations caused by such disturbances, this step analyzes the periodic vibration mark characteristics of the abnormal area in the edge state map, extracts the spatial distribution period and fluctuation amplitude information of the fine lines, and uses this as the basis for dynamic control. Specifically, fine-tuning the tool's entry angle can change its contact direction when cutting into the material, causing the tool's cutting trajectory to be misaligned with the disturbance peak, thereby dispersing the impact energy; while micro-calibration of the feed speed adjusts the feeding rhythm to match the time the cutter enters the cloth, creating a "disturbance-offsetting" phase intervention effect, preventing periodic vibrations from forming a continuous structural accumulation at the edge. This tangential correction process uses a closed-loop PI control method, using the edge disturbance response as feedback, to continuously optimize parameter settings to ensure dynamic compensation for perturbation residues while maintaining cutting efficiency. Through this strategy, adaptive correction of fine-scale quality issues can be achieved while maintaining overall operational stability, greatly improving the consistency and smoothness of the PET base cloth edge, and is an indispensable key link in achieving ultra-precision cutting quality control.

[0119] The edge quality assessment and optimization module performs edge quality assessment after completing the tangential correction control operation. By comparing the changes in the edge state map before and after the tangential correction, it identifies the distribution of residual disturbance characteristics, updates the tool motion trajectory and feed tension adjustment curve for the next cycle, and establishes a closed-loop control mechanism based on frequency dynamic decoupling and edge quality stability. This achieves continuous optimization of vibration mark suppression and edge consistency during the high-precision cutting process of PET base fabrics.

[0120] To achieve continuous optimization of edge quality during high-precision cutting of PET fabrics, an edge quality assessment method is proposed after completing the preceding tangential correction control operation. This method compares the edge state maps before and after tangential correction to identify the spatial distribution and evolution trend of residual disturbance features. Based on this, the tool motion trajectory and feed tension adjustment curve for the next cycle are dynamically updated. Ultimately, a closed-loop control mechanism is established, based on frequency dynamic decoupling and aiming for stable edge quality. The method specifically includes the following steps:

[0121] Within a specified time window after the tangential correction control operation is completed, image data of the cut edge of the PET substrate is recaptured and a corresponding edge state map is generated at a preset sampling frequency. This map should include key geometric characteristic parameters such as sawtooth fluctuation amplitude, edge morphology change rate, and contour offset, and should be mapped one-to-one with the pre-correction map on the time axis. By analyzing the difference in feature vectors between the maps, it is possible to accurately determine whether the tangential correction operation has achieved substantial suppression of edge disturbances within the target time period.

[0122] A multi-dimensional comparison of the residual disturbance characteristics in the pre- and post-correction maps is performed, including but not limited to metrics such as peak distribution density, cyclic stability, amplitude gradient, and spatial propagation path. Through difference calculation and trend modeling, it is determined whether the residual disturbance exhibits positive evolutionary trends such as weakening, convergence, and smoothing, or whether it exhibits unstable characteristics such as new local concentrations or cyclical repetitions. This comparison not only assesses the immediate effectiveness of the current control strategy but also provides a forward-looking assessment of the need for subsequent intervention.

[0123] Based on the spatial distribution and evolution of the residual disturbances, adaptive updates are made to the cutting parameters for the next cycle. These include subtle adjustments to the tool trajectory, such as optimizing the acceleration and deceleration curves of the fabric feed path, the cutting starting point position, and the angular offset path. Furthermore, updates to the feed tension control curve are made, such as adjusting the time delay for tension loading, the balance buffer coefficient, or the feedback control sensitivity. These updated parameters serve as the initial settings for the next cutting cycle, enabling a coordinated upgrade of strategy optimization and execution behavior to ensure that disturbances are not accumulated or amplified.

[0124] After running for multiple cycles, the continuous comparison and trend fitting of historical edge state maps are used to determine whether the edge quality is stable within the acceptable threshold. If the edge vibration mark index is lower than the set intervention threshold for multiple consecutive cycles, it is determined to have entered the steady-state operation zone, maintain the current parameter combination, and conduct regular reviews. If periodic fluctuations in edge quality are found or the recovery after intervention is unsatisfactory, the control strategy optimization process will be automatically triggered, and the closed-loop control process of frequency analysis-interference adjustment-map monitoring-strategy update will be re-entered. Through this method, an edge quality self-feedback optimization mechanism driven by intervention control is realized. It can maintain the consistency and stability of edge quality in a dynamic and complex cutting environment, significantly reducing the need for manual intervention. It is a key closed-loop technical guarantee for achieving high-precision and intelligent PET base cloth cutting.

[0125] This step aims to establish a closed-loop feedback mechanism between cutting edge quality and control parameters after the tangential correction control operation is implemented. By comparing the changes in the edge state maps before and after the tangential correction in real time, it is possible to identify any residual disturbance characteristics and make strategic corrections based on their distribution, thereby accurately updating the tool motion trajectory and feed tension adjustment curve for the next cycle. Although macroscopic resonance and transient disturbances have been effectively reduced through frequency decoupling and tangential fine-tuning in the early stages, the actual cutting process may still cause a certain degree of edge quality fluctuation due to local material elastic fluctuations, equipment micro-seismicity, or thermal stress changes. In this case, a single intervention is insufficient to ensure long-term cutting stability, and an edge quality-based feedback evaluation mechanism must be introduced. By comparing the edge state maps before and after the correction, it is possible to quantify the changes in the sawtooth fluctuation amplitude, offset rate, and periodic residual, and determine whether the edge disturbance is converging spatially or shifting to a new frequency band or location. Once residual disturbances or areas of trending instability are identified, the dynamic response curves for the tool's feed angle, path planning, and feed tension for the next cycle are adjusted based on the disturbance distribution characteristics, thereby establishing an intelligent closed-loop control chain based on frequency decoupling and quality perception as feedback. This mechanism not only strengthens the continuous tracking and adaptive response to edge anomalies, but also opens up the closed-loop logic of the entire "control-monitoring-evaluation-recontrol" process. This ensures that PET base fabrics can maintain high-quality cutting results with neat edges, no chatter marks, and no melt edges during long-term, high-intensity continuous cutting tasks. This is a key link in achieving quality stability throughout the entire process.

[0126] The PET web cutting control system constructed using the aforementioned solution achieves a breakthrough in closed-loop control, encompassing a full range of processes, from vibration sensing, frequency identification, dynamic decoupling, to edge quality feedback. This system effectively addresses the critical resonance and edge vibration marks caused by frequency coupling between cutter speed and feed tension. This control system not only monitors multi-source dynamic signals during the cutting process in real time and implements intelligent frequency decoupling based on spectrum analysis and resonance intensity metrics, but also uses high-resolution edge image recognition to accurately capture microscale vibration mark trends, enabling fine-tuning of the tangential path and optimizing feed rates. In particular, the implementation of an edge quality map and self-learning evaluation mechanism enables adaptive adjustment, dynamically optimizing tool control strategies and feed parameters for stable and continuous improvement of edge quality. Overall, this solution significantly improves the cutting accuracy and consistency of PET webs in high-end applications (such as optics, electrical engineering, and flexible electronics), reduces performance degradation and material scrap caused by edge defects, and enhances the intelligence of the cutting process and product yield. The system possesses significant engineering application value and promising industrial expansion prospects.

[0127] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0128] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

[0129] It should be noted that, in this document, if there are relational terms such as first and second, etc., they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprises", "comprising" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device that includes the element.

[0130] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0131] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0132] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0133] 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 these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0134] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0135] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0136] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

Claims

1. Intelligent PET base cloth cutting control system, characterized by: It includes vibration data acquisition module, frequency coupling analysis module, dynamic frequency decoupling module, edge disturbance monitoring module, tangential correction control module and edge quality evaluation and optimization module: The vibration data acquisition module obtains the tool rotation frequency, material tension changes and feed speed, and constructs a vibration response map for resonance determination; The frequency coupling analysis module identifies the characteristic frequency points of harmonic amplification based on the vibration response spectrum and calculates the resonance intensity index for intervention judgment; The dynamic frequency decoupling module synchronously adjusts the tool speed and tension response delay according to the resonance intensity index, dynamically adjusts the frequency ratio between the two, and realizes frequency decoupling; The edge disturbance monitoring module, after completing frequency decoupling, collects the structural features of the cutting edge, extracts the sawtooth fluctuation amplitude and morphology change rate, and generates an edge state map; The tangential correction control module performs tangential correction based on the abnormal trend of the edge state map, fine-tunes the tool entry angle and feed speed, and compensates for residual transient disturbances; The edge quality assessment and optimization module compares the edge state maps before and after tangential correction, identifies the residual disturbance distribution, updates the tool motion trajectory and tension adjustment curve, and builds a closed-loop control mechanism to achieve chatter mark suppression and continuous optimization of edge quality.

2. The intelligent PET base cloth cutting control system according to claim 1 is characterized in that: Constructing a vibration response map involves the following steps: Obtain tool rotation frequency, material tension change and feed speed, unify the time base, and perform sampling rate reconstruction and amplitude normalization; Adopt high-pass filtering and noise suppression algorithm to eliminate low-frequency interference and non-periodic noise, and improve signal quality; Perform spectrum analysis and wavelet decomposition to extract the energy distribution of the signal in different frequency intervals and mark the characteristic points of frequency changes; A time period division mechanism and a dynamic sampling window control method are introduced to extract frequency features according to the time sliding window to form a dynamic vibration response spectrum.

3. The intelligent PET base cloth cutting control system according to claim 1 is characterized in that: Identifying the harmonic amplification characteristic frequency point and calculating the resonance intensity index includes the following steps: Perform fast Fourier transform on the tool rotation frequency, material tension change frequency and feed speed change frequency respectively to obtain their respective frequency amplitude distribution vectors; The frequency overlap and amplitude coupling strength of the three types of spectrum vectors are analyzed within the resonant sensitive frequency range to identify characteristic frequency points with convergent frequency components and enhanced amplitude. Calculate the total energy proportion, frequency correlation coefficient, bandwidth amplitude gradient and time stability of the characteristic frequency points to form a resonance intensity score value; A multi-level judgment threshold is set according to the resonance intensity score, critical resonance frequency points are marked, and target parameters for subsequent dynamic decoupling control are output.

4. The intelligent PET base cloth cutting control system according to claim 1 is characterized in that: Dynamically adjust the ratio between tool speed and tension response delay frequency to achieve frequency decoupling. The specific steps are as follows: Identify high-risk frequency coupling sections, extract corresponding time segments and frequency ranges, and determine whether they are in an integer multiple coupling relationship; The weighted objective function control method is used to adjust the tool speed trajectory so that its frequency shifts to the resonance frequency point while maintaining the cutting beat and thermal stress response stable. The tension response frequency is adjusted through tension feedback delay control, and a joint offset path with the tool frequency is constructed to break the coupling relationship; The spectrum energy and frequency overlap are continuously monitored within the sliding time window to determine whether the frequency decoupling effect meets the resonance suppression requirements.

5. The intelligent PET base cloth cutting control system according to claim 1 is characterized in that: Generating an edge state graph involves the following steps: An image acquisition device is set at the cutting exit to collect cutting edge image data and perform grayscale processing; Edge detection algorithm is used to extract boundary contours, and binary processing is performed to enhance boundary clarity; The multi-scale geometric curve fitting method is used to calculate the sawtooth fluctuation amplitude, edge deviation and morphology mutation rate respectively. The extracted eigenvalues ​​are plotted as an edge state map along the time axis, and an edge feature comparison and analysis mechanism is introduced to determine whether the tangential correction control logic should be triggered.

6. The intelligent PET base cloth cutting control system according to claim 1 is characterized in that: Executing tangential correction control includes the following steps: Extract the data segments marked as abnormal areas in the edge state map, obtain the sawtooth fine line distribution sequence and extract the periodic parameters; Based on the periodic parameters, the tool entry angle fine-tuning strategy is executed to shift the entry angle within the preset angle range so that it is staggered to avoid the overlapping area of ​​the disturbance wave peaks; The feed speed is adjusted collaboratively based on periodic parameters, and the feed speed disturbance compensation value is dynamically calculated using a closed-loop proportional-integral control method; Collect the adjusted edge image, generate an updated atlas and perform feature comparison to determine whether a stable edge output state has been achieved.

7. The intelligent PET base cloth cutting control system according to claim 1 is characterized in that: By comparing the edge state maps before and after tangential correction, edge quality assessment optimization is performed, including the following steps: Compare the edge state maps before and after tangential correction to identify the residual disturbance distribution characteristics in the edge area, including amplitude changes, offset trends and periodic fluctuation behaviors; According to the spatial position and intensity distribution of the residual disturbance, the motion trajectory of the tool in the next cycle is dynamically adjusted, including the feeding path, speed curve and angle control parameters; According to the edge disturbance distribution trend, the feeding tension adjustment curve is synchronously modified, including adjusting the tension loading delay and feedback response coefficient; Edge map comparison and control parameter update are continuously performed within multiple cycles to build a closed-loop control mechanism based on frequency dynamic decoupling and edge quality stability, thereby achieving continuous optimization of vibration mark suppression and edge quality.

Images