Intelligent typesetting and tailoring path planning method for clothes

Abstract

The invention relates to the technical field of control systems, and discloses an intelligent clothing typesetting and tailoring path planning method, which comprises the following steps of: acquiring electrical parameters of a tailoring cutter driving motor in real time to generate a characteristic frequency spectrum representing a cutting torque, and further identifying an abnormal cutting state by comparing the frequency spectrum with a reference on line; according to the invention, the control system obtains endogenous physical perception and pre-judgment capability, and can actively avoid risks through prospective fine adjustment of self-control behaviors before potential cutting failure occurs, so that the method has the advantages of high efficiency, high reliability and the like, and is suitable for large-scale popularization and application. And the operation mode of the device is converted from passive response to active adjustment.

Description

Technical Field

[0001] This invention relates to a method for intelligent layout and cutting path planning of clothing, belonging to the field of control system technology. Background Technology

[0002] Currently, especially in high-speed CNC cutting applications for flexible materials such as garment fabric pieces, a common technical approach is to use a servo control system to drive the cutting head to accurately reproduce the planned motion path based on a preset digital geometric model. However, this control strategy essentially simplifies the cutting process into a purely geometric problem, ignoring the complex dynamics of the transient physical interaction between the cutting tool and the fabric being processed. The control system is essentially in an open-loop blind state regarding the real-time fluctuations in cutting load caused by changes in internal material tension or local wrinkling during the cutting process.

[0003] This design approach results in a lack of online perception and adaptive adjustment capabilities for the physical state of the processing, leading to the following technical problems. Specifically, existing technologies suffer from the following shortcomings: 1. Lack of risk foresight: The system cannot provide early warning or intervention by identifying physical precursor signals before failures such as fabric tearing or fraying occur; 2. Lag in control adjustment: Even when anomalies are detected by external sensors, the response is mostly passive emergency shutdown, rather than flexible self-adjustment during the evolution of the anomaly. This not only affects production efficiency but also fails to fundamentally prevent losses. Therefore, how to enable the control system to go beyond simple path reproduction, establish a real-time perception capability for the physical interaction of the cutting process, and based on this perception information, achieve preventative adjustments to future control behavior to proactively avoid potential processing failures, is the technical problem this invention aims to solve. Summary of the Invention

[0004] This invention provides a method for intelligent layout and cutting path planning of clothing. Its main purpose is to solve the problem that existing control systems lack the ability to perceive the physical state of the cutting process in real time, thus failing to proactively avoid potential processing failures.

[0005] To achieve the above objectives, the present invention provides a method for intelligent layout and cutting path planning of clothing, comprising the following steps:

[0006] Step a: During the cutting task performed by the cutting shears, the electrical parameters of the drive motor driving the cutting shears are collected in real time at a sampling frequency.

[0007] Step b: Based on the collected electrical parameters, a characteristic spectrum representing the cutting torque is generated in real time by performing a short-time Fourier transform on the continuous electrical parameter signals.

[0008] Step c: Compare the real-time generated feature spectrum with a reference feature spectrum to identify abnormal cutting conditions;

[0009] Step d: When an abnormal cutting state is identified, the motion control parameters of the subsequent cutting paths that have not yet been executed and are adjacent to the current cutting position are adjusted in a feedforward manner based on the type of the abnormal cutting state, so as to avoid cutting failure caused by the abnormal cutting state.

[0010] Preferably, the electrical parameters are the current feedback signal or torque command signal of the drive motor; the reference characteristic spectrum is obtained by test cutting the fabric sample; the abnormal cutting state includes the fabric tension excessive state characterized by the high frequency energy of the characteristic spectrum being higher than a high frequency energy threshold, and the fabric wrinkling state characterized by the appearance of non-harmonic broadband noise.

[0011] Preferably, the feedforward adjustment specifically involves: when excessive fabric tension is detected, reducing the cutting speed of subsequent cutting paths; and when fabric wrinkling is detected, performing micro-operations such as lifting the blade, shifting along the path direction, and then cutting again.

[0012] Preferably, before step c, a step of establishing and dynamically correcting the reference feature spectrum is included. This step specifically involves: during the cutting process, synchronously acquiring the encoder feedback speed signal of the servo motor that drives the cutting bed to perform X-axis and Y-axis translation; when the fluctuation of the speed signal is lower than a stable reference value within a first time period, the feature spectrum corresponding to this time period is identified as a high-confidence health fingerprint, which is used to establish or to perform weighted updates on the existing reference feature spectrum.

[0013] Preferably, after step d, a decision arbitration step is further included, which specifically involves: after generating the feedforward adjustment command, starting a timer and calculating the instantaneous energy consumption rate of the system based on electrical parameters within a second time period; if the instantaneous energy consumption rate is continuously higher than a unit length average energy consumption benchmark by a set ratio, then the feedforward adjustment command is executed; otherwise, the command is revoked.

[0014] Preferably, a global disturbance identification step is included before step d. This step specifically involves: when the change in the characteristic spectrum exceeds a high-amplitude threshold, simultaneously acquiring the electrical parameters of the servo motors driving the cutting bed to perform X-axis and Y-axis translation; if the change in the electrical parameters of the X-axis and Y-axis servo motors at the same point in time does not exceed a disturbance reference value, then the anomaly is confirmed as a local cutting problem and the adjustment in step d is performed; otherwise, the anomaly is determined as a machine-wide disturbance and the current motion control parameters are maintained.

[0015] Preferably, before cutting a fabric, a step of detecting the main texture direction of the fabric is included. This step specifically involves: controlling the cutting shear drive motor to perform a standardized micro-perturbation injection action at at least two detection points with different directions on the fabric; acquiring the transient response signal caused by the micro-perturbation injection action; and determining the direction in which the high-frequency component in the response signal decays the slowest as the main texture direction of the fabric.

[0016] Preferably, in step c, the comparison benchmark used to identify abnormal cutting states is dynamic. Specifically, it is generated as follows: before comparison, the judgment threshold of the benchmark feature spectrum is dynamically compensated based on the angle between the current cutting path and the detected main texture direction. The compensated dynamic threshold... Determined by the following relationship in, The original judgment threshold is the baseline feature spectrum. The angle between the current cutting path vector and the detected main texture direction vector. It is a compensation coefficient characterizing the degree of anisotropy of a fabric.

[0017] Preferably, the system also includes a step of performing long-term trend analysis on the characteristic spectrum. This step specifically involves recording the drift of the main peak frequency in the characteristic spectrum with the cumulative cutting time during multiple trimming tasks. When the drift exceeds a passivation reference value, the system generates and issues a prompt message to replace the trimmer after completing the current task.

[0018] Preferably, the identification of abnormal cutting states in step c and the feedforward adjustment in step d are based on a machine learning model. The machine learning model is trained with the feature spectrum sequence and corresponding cutting success or failure labels in the historical cutting process to establish a mapping relationship between the feature spectrum and the abnormal state type and the adjustment strategy.

[0019] Compared with the prior art, the beneficial effects of the present invention are:

[0020] 1. This invention transforms the control system from a simple path command executor into a regulatory system with endogenous physical perception and predictive capabilities. By acquiring the electrical parameters of the motor driving the cutting shears in real time and converting them into a characteristic spectrum that directly characterizes the cutting torque, it gains direct insight into the real physical process. More importantly, the system does not passively stop after an anomaly occurs, but compares the real-time spectrum with a benchmark to identify potential abnormal cutting states. Based on this, it proactively adjusts the motion control parameters of the adjacent cutting paths that have not yet been executed. Thus, it can proactively avoid risks by adjusting its future control behavior in advance before problems such as cutting failure occur. Its operation mode itself constitutes a shift from reactive response based on failure to proactive self-regulation integrated into the process.

[0021] 2. This invention establishes a context-adaptive benchmark generation and compensation mechanism to avoid misjudging inherent physical property differences in fabric texture as process defects due to reliance on a single static benchmark. It first actively applies micro-perturbations and analyzes the response signals to determine the main texture direction of the fabric before cutting. Then, during cutting, it dynamically compensates the judgment threshold based on the angle between the current cutting path and the texture direction. Simultaneously, during a stable cutting phase, the system uses stable X-axis and Y-axis servo motor speed signals as high-reliability anchor points to capture the torque spectrum at that moment for weighted updates to the existing benchmark. Thus, the control system no longer relies on a rigid ideal model but continuously corrects the definition of the normal state during task execution through a combination of active detection and passive learning, ensuring that every anomaly judgment and control adjustment is based on a realistic and appropriate understanding of the current local environment.

[0022] 3. This invention constructs a hierarchical decision arbitration mechanism to ensure the reliability of control commands. It aims to solve the dilemma of high-sensitivity monitoring systems in complex industrial environments, which cannot effectively distinguish between real process problems and external interferences such as mechanical vibrations originating from the entire equipment, resulting in unnecessary interruptions. When the characteristic spectrum of the cutting motor changes drastically, the system does not react immediately. Instead, it first synchronously compares the electrical parameters of the servo motor driving the translation of the cutting bed. Through cross-validation, it identifies whether the anomaly is a local cutting problem or a machine-wide disturbance. Even if it is confirmed to be a local problem, the system will further calculate the instantaneous energy consumption rate of the system based on the electrical parameters within a short period of time. Only when the rate is continuously and significantly higher than the average benchmark is it finally confirmed that there is a continuous and substantial physical obstacle, and executes the corresponding feedforward adjustment command. This decision logic that integrates multi-physical dimension information for mutual verification enables the entire control system to implement intervention measures after identifying local cutting problems in complex actual working conditions. Attached Figure Description

[0023] Figure 1 This is a flowchart of a method for intelligent layout and cutting path planning of clothing according to the present invention.

[0024] Figure 2 This is a schematic diagram illustrating the change of high-frequency energy over time in this invention.

[0025] Figure 3 This is the state transition diagram of the system control logic of the present invention.

[0026] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the present invention clearer, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. However, it should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention.

[0028] This invention discloses a method for intelligent pattern making and cutting path planning in clothing. Its overall architecture revolves around a closed-loop sensing and feedforward adjustment loop, which consists of four core stages: First, a real-time physical process sensing stage based on the electrical parameters of the drive motor, used to dynamically convert cutting force into analyzable data; second, an online cutting state diagnosis stage based on signal processing and pattern recognition, which identifies potential process risk precursors from the data; third, a decision arbitration and disturbance identification stage based on multi-source information fusion, used to ensure the accuracy of control intervention; and finally, a motion control parameter adjustment stage for future paths, which actively fine-tunes system behavior to avoid predicted failures, thus forming a technical closed loop from passive response to active adjustment. In a specific application scenario, such as high-speed CNC cutting of high-density fabrics, existing technologies often face problems of fraying and tearing due to excessive local tension or micro-wrinkles because they cannot perceive the transient physical interaction between the cutter and the fabric in real time. To address this challenge, the execution flow of this invention begins with the quantification of the physical characteristics of the cutting process, and the cutting shears execute... During the cutting task, the control system acquires the current feedback signal or torque command signal of the drive motor driving the cutting shears in real time at a preset sampling frequency, such as 20kHz. Since these electrical parameters of the motor are directly physically related to the mechanical torque output by the motor, this continuously acquired electrical parameter signal constitutes the original characterization of the actual cutting load fluctuation, providing a data foundation for subsequent analysis. To transform the original electrical parameter signal stream into structured data that can intuitively reflect the cutting state, the system is configured to generate a characteristic spectrum characterizing the cutting torque in real time by performing a short-time Fourier transform on the continuous electrical parameter signals. Specifically, the system divides the signal stream into several overlapping data windows, for example, a window with a length of 1024 sampling points and an overlap rate of 50%, and applies a window function to each data window to suppress spectral leakage. Then, a fast Fourier transform is performed to obtain a series of spectrograms that evolve over time. This dynamic characteristic spectrum reflects the rate of change of cutting force on the frequency axis and the intensity of force fluctuation on the amplitude axis, which constitutes the direct basis for the system to make subsequent judgments and decisions.

[0029] To define the normal state of the cutting process, the system is configured to establish and dynamically correct a baseline feature spectrum. Before cutting, an initial baseline feature spectrum is obtained by trial cutting a fabric sample. During the subsequent cutting process, the system also initiates a dynamic correction procedure. This procedure synchronously acquires the encoder feedback speed signals of the servo motors driving the cutting bed to translate along the X and Y axes. When the fluctuation of the speed signal is consistently lower than a stable reference value within a first time period, the system identifies the corresponding feature spectrum within this period as a high-confidence health fingerprint and uses it to perform weighted updates on the existing baseline feature spectrum. Thus, the baseline feature spectrum is continuously corrected to approximate the true health state under the current operating conditions. With real-time feature spectrum and dynamic baseline, the system can continuously compare the two online to identify abnormal cutting states. The system's built-in diagnostic logic is configured to monitor two typical abnormal cutting conditions. The system identifies two states: First, excessive fabric tension, characterized by high-frequency energy exceeding a high-frequency energy threshold in the characteristic spectrum, a precursor to fiber tearing. Second, fabric wrinkling, characterized by the presence of non-harmonic broadband noise, indicates material buildup in front of the cutter. This allows the system to map spectral energy anomalies to specific physical failure modes. Upon identifying an abnormal cutting state, the system immediately adjusts the motion control parameters of subsequent cutting paths adjacent to the current cutting position. This adjustment strategy is directly related to the identified abnormal state type: when excessive fabric tension is detected, the system reduces the cutting speed of subsequent paths; when fabric wrinkling is detected, the system performs micro-operations such as lifting the cutter, shifting along the path, and then lowering the cutter again. This proactive intervention allows the control system to actively manage risks and avoid potential cutting failures.

[0030] To enhance decision robustness, the system also constructs a hierarchical decision-making mechanism. When the change in the characteristic spectrum exceeds a high-amplitude threshold, the system first initiates a global disturbance identification step. This step simultaneously acquires the electrical parameters of the servo motors driving the cutting bed to perform X-axis and Y-axis translation. If the changes in the electrical parameters of the X-axis and Y-axis servo motors at the same time point do not exceed a disturbance reference value, the anomaly is confirmed as a local cutting problem and adjustment is authorized. Otherwise, the anomaly is judged as a system-wide disturbance and the current motion control parameters are maintained. Furthermore, the system can be configured with a decision arbitration step as the final confirmation stage. After generating a feedforward adjustment command, the system calculates the instantaneous energy consumption rate of the system based on the electrical parameters within a second time period. Only when the... The feedforward adjustment command is only executed when the rate continuously exceeds the average energy consumption benchmark per unit length by a set proportion; otherwise, the command is revoked. Furthermore, for fabrics with significant anisotropic characteristics, to avoid misjudging high loads in normal texture directions as abnormalities, the system is configured to perform a step of detecting the main texture direction of the fabric before cutting. This step controls the cutting shear drive motor to perform standardized micro-perturbation injection actions at at least two detection points with dissimilar directions on the fabric, and collects the transient response signal triggered by this action. The direction in which the high-frequency components in the response signal attenuate the slowest is then determined as the main texture direction of the fabric. During the cutting process, the judgment threshold for identifying abnormalities is based on the angle between the current cutting path and the detected main texture direction. Dynamic compensation is performed, and the dynamic threshold after compensation is determined. Through relational formulas Confirmed, among which The original judgment threshold is the baseline feature spectrum. The compensation coefficient characterizes the degree of anisotropy of the fabric. In addition, the present invention also includes a step of performing long-term trend analysis on the characteristic spectrum. In multiple cutting tasks, the system records the drift of the main peak frequency in the characteristic spectrum with the cumulative cutting time. When the drift exceeds a passivation reference value, the system will generate and issue a prompt message to replace the cutting shears after completing the task, thereby realizing predictive maintenance based on the physical state of the equipment.

[0031] To define the optimization objective of the intelligent typesetting algorithm, this invention further includes a standardized procedure for generating the fitness function, which first sets the unit area utilization rate as the optimal value. Total length of unit toolpath and the revenue from unit common edge cutting Three core evaluation primitives were set up, and then at least one hundred layout and cutting experiments covering all typical cut piece combinations were conducted on standard fabric samples. The data for each set of experiments were collected and recorded. , and The objective numerical values ​​were obtained, and finally, the least squares method was used to perform multiple linear regression analysis on the experimental data to obtain a fitness function with a unique solution that can characterize the comprehensive benefits. ,in, , and The weighting coefficients for area utilization, total blade path length, and common-edge cutting benefits are uniquely determined by the regression analysis results. To ensure the high accuracy and long-term stability of this method in actual industrial environments, this invention further initiates a benchmark calibration and online calibration module before layout and path planning. This module first uses a high-resolution vision sensor array deployed along the cutting bed track to perform a full-width scan of the laid-out fabric to obtain its real-time contour data, including local stretching and wrinkles. Based on this, a deformation compensation matrix is ​​generated to correct the coordinates of all cut pieces. Subsequently, after completing a preset number of cutting tasks, the cutting system automatically cuts a preset standard graphic containing straight lines, arcs, and acute angles on a dedicated standard substrate. The deviation between the actual cutting trajectory and the theoretical trajectory is measured by the vision sensor, thereby updating a dynamic error model for real-time compensation of cutting head movement errors in subsequent processing.

[0032] Example 1: In an automated process for producing automotive interior parts, the cutting target is a laminated material composed of highly elastic artificial leather and a thin foam substrate. This material exhibits significant anisotropy and is prone to wrinkling or tearing due to stress concentration during high-speed cutting. When the CNC cutting system receives a geometric path instruction for cutting dashboard coverings, the path includes long straight segments tangential to the main texture direction at different angles and multiple corners with small radii of curvature. Existing control methods face an inherent dilemma here: the high cutting speed set to ensure efficiency will cause material damage at corners or when cutting at an angle because stress cannot be quickly unloaded; while the low speed setting used to ensure quality will sacrifice production cycle time and cause material shrinkage due to frictional heat generated during prolonged contact with the foam substrate.

[0033] When dealing with this working condition, the operation flow of the present invention is as follows: Before the task begins, the system first performs a step of detecting the main texture direction of the fabric, determines the mechanical properties of the composite material, and then, based on the compensation coefficient... A relational formula for dynamically adjusting the judgment threshold was established. After the cutting begins, when the blade moves along the angle with the main texture direction... When traveling at high speed on shorter, longer straight segments, the real-time feature spectrum matches the dynamically updated baseline feature spectrum, maintaining high-speed cutting. When path planning indicates the tool is about to enter a region where the radius of curvature decreases sharply and the path vector forms an angle with the main texture direction... As the system approaches a 90-degree turning angle, its dynamic threshold compensation mechanism has already lowered the high-frequency energy threshold representing excessive fabric tension based on the path information. Consequently, the real-time generated characteristic spectrum begins to capture the high-frequency components that increase due to the turning angle. This causes the spectral energy to reach the situational threshold lowered by the aforementioned mechanism before the cutting force reaches the level of physical damage, thereby triggering a feedforward adjustment to smoothly reduce the cutting speed of the upcoming turning path.

[0034] During the corner cutting process, due to the unevenness of the underlying foam, tiny material accumulations appeared in front of the blade. This was identified by the system as non-harmonic broadband noise and determined to be a state of fabric wrinkling. At this time, the system did not stop, but immediately performed micro-operations of lifting the blade, moving along the path direction, and then cutting again, thus overcoming the physical obstacle. Under this scheme, the system no longer relies on a global static speed setting, but through continuous perception, prediction, and adjustment cycles, the cutting speed is transformed into a dynamic variable that responds to local working conditions in real time. After the task is completed, the contour edges of the dashboard covering are smooth, without burrs, frayed edges, or deformation caused by stress. Its final geometric dimensions are consistent with the digital model, and the entire cutting process is completed automatically without human intervention.

[0035] Example 2: To quantitatively verify the effectiveness and response speed of the method of the present invention in identifying and avoiding different types of potential cutting failures, this experiment was established. The purpose of the experiment was to objectively evaluate the system's ability to distinguish abnormal state types and perform corresponding feedforward adjustments when faced with two controllable, simulated real-world failure precursor events. The experimental platform was modified from a standard three-axis CNC cutting machine, on which the control system of the present invention was integrated. The cutting shear drive motor was a servo motor that could provide real-time feedback of torque command signals. The fabric to be cut was denim from the same batch, and an electromagnetic tensioner with programmable tension was installed downstream of its travel path. At the same time, a miniature pneumatic actuator was installed below the cutting path to create instantaneous pleats on the fabric as needed. The control group of the experiment used the same hardware platform, but its feedforward adjustment function was disabled, allowing it to operate as a traditional, open-loop execution geometric path control system.

[0036] In the experiment, a key parameter—the high-frequency energy threshold used to characterize excessive fabric tension—required a balancing act. Excessive sensitivity might trigger unnecessary speed-reduction interventions, while insufficient sensitivity might miss opportunities to prevent failure. Therefore, the setting procedure was as follows: First, a series of destructive pre-tests were conducted on the test fabric sample. Cutting was performed at an initial speed, and the tension of the electromagnetic tensioner was gradually increased until the first fiber broke. The characteristic spectral energy integral value in the 3kHz to 5kHz frequency band was recorded just before breakage. This process was repeated five times, and the average value was taken. Finally, 70% of this average value was set as the high-frequency energy threshold for this experiment. The value was determined to be 0.85 normalized energy units under this condition to reduce the probability of false triggering while ensuring reaction margin. The test was divided into two groups. The first group was designed to simulate the state of excessive fabric tension. The cutting bed made a straight cut at an initial speed of 2 m / s. When the cut reached the midpoint of the path, the electromagnetic tensioner was activated to apply a step tension to the fabric. The second group was designed to simulate the state of fabric wrinkling. Under the same initial conditions, when the cut reached the midpoint of the path, the micro pneumatic actuator was activated to create an instantaneous wrinkle with a height of 2 mm 5 mm in front of the cutter. Both groups of tests were performed with the feedforward adjustment function disabled and enabled respectively. The key process data are recorded in the table below.

[0037] Table 1: Comparison of Experimental Data

[0038]

[0039] Referring to Table 1, in test sequence A-2, the system identified the high-frequency energy exceeding the limit within 45 milliseconds after tension was applied and performed a speed reduction operation to avoid failure. In test sequence B-2, the system correctly interpreted non-harmonic noise as fabric wrinkling and performed a micro-operation of lifting the blade to avoid obstacles within 62 milliseconds, similarly preventing the production of defective products. The data shows that the characteristic spectrum analysis mechanism adopted in this invention can produce a discriminative response to changes in cutting resistance caused by different physical factors, and transform this identification result into timely, type-matched, and risk-avoiding feedforward control behavior. The results of this experiment confirm that the intelligent garment layout and cutting path planning method can effectively distinguish potential failure risks caused by excessive fabric tension or fabric wrinkling based on real-time analysis of the electrical parameters of the drive motor, and perform targeted feedforward adjustments with sufficiently low latency. The results show that this method has the ability to proactively avoid cutting failures without interrupting the production process.

[0040] Example 3: This example combines Figures 1 to 3 This paper describes the implementation of a method for intelligent pattern layout and cutting path planning in clothing, such as... Figure 1As shown, the process begins with real-time data acquisition from the drive motor. Then, through a feature spectrum generation stage, the acquired electrical parameter signals are processed using a short-time Fourier transform to generate a dynamic spectrum characterizing the cutting torque. The subsequent online status diagnosis stage compares this real-time spectrum with a dynamic, contextualized benchmark to determine if any potential anomalies are detected. If no anomalies are detected, or if adjustments have been made, the system enters the optimized cutting execution stage to proactively avoid potential cutting failures. If a potential anomaly is detected, a hierarchical decision arbitration mechanism first confirms and adjusts the adjustment instructions. This mechanism eliminates machine-wide disturbances through global disturbance identification and confirms the existence of persistent physical obstacles through decision arbitration. After confirmation, the system initiates feedforward adjustments for subsequent cutting paths that have not yet been executed. The diagnostic benchmark used in this process is provided by a dynamic benchmark generation and compensation mechanism. This mechanism determines the main texture direction through fabric texture detection and performs dynamic threshold compensation based on the angle between the cutting path and this direction. Simultaneously, it captures high-reliability health fingerprints during the smooth cutting stage to dynamically correct the benchmark, thus forming a complete closed-loop sensing and feedforward adjustment circuit.

[0041] like Figure 2 As shown, the horizontal axis represents time (s), and the vertical axis represents the normalized spectral energy. The anomaly detection threshold, indicated by the dashed line, is preset to a normalized energy unit of 0.85. The data points connected by solid lines depict the trajectory of the high-frequency energy as monitored in real time over time. As can be seen from the figure, the high-frequency energy begins to rise continuously after t=1.5s and crosses the anomaly detection threshold for the first time at approximately t=2.4s. It does not fall back below the threshold until approximately t=3.2s. This process precisely illustrates how the system can identify potential precursors to pruning failure in real time and quantitatively by comparing a key physical characteristic (high-frequency energy) with a dynamic or preset threshold.

[0042] like Figure 3 As shown, after the task is started, the system defaults to normal cutting state. When an event occurs where the characteristic spectrum change exceeds the threshold, the process enters the decision arbitration state. In the decision arbitration stage, if the condition is determined to be a machine-wide disturbance or an unconfirmed persistent physical obstacle, the system will return to normal cutting state. If the condition is identified as excessive tension and energy consumption is confirmed, the system will switch to deceleration adjustment state to deal with excessive fabric tension. If the condition is identified as fabric wrinkling and energy consumption is confirmed, the system will switch to micro-operation adjustment state to deal with fabric wrinkling. In the deceleration adjustment or micro-operation adjustment state, the system will return to normal cutting state after the adjustment is completed. In addition, starting from the normal cutting state, if an event occurs where the main peak frequency drift exceeds the limit in the long-term trend analysis, the system will trigger a maintenance prompt to suggest replacing the cutting shears, thereby realizing predictive maintenance based on equipment status.

[0043] Example 4: In a production workshop containing multiple CNC cutting machines, the foundation and equipment bases experience instantaneous mechanical vibrations due to the operation of nearby heavy equipment. When this vibration is transmitted to the frame of the cutting bed, a technical challenge arises: the vibration is picked up by the servo motors driving the cutting bed's X-axis and Y-axis translation, as well as the motors driving the cutting shears, inducing a synchronous and high-amplitude signal disturbance in their respective electrical parameters. If a control system that only monitors the characteristic spectrum of the cutting shears is used, this machine-wide disturbance may be misjudged as a severe local cutting anomaly, leading to incorrect adjustments. To identify such global disturbances, the technical solution of this invention, after identifying a change in the characteristic spectrum of the cutting shear drive motor exceeding a high-amplitude threshold, does not immediately perform adjustments but first enters a global disturbance identification step. This step quantifies the correlation of the signals of each motor in the frequency domain. The system synchronously acquires electrical parameter signals from the X-axis and Y-axis servo motors and the cutting shear drive motor, and performs spectral coherence calculations on these three sets of signals within a 20-millisecond time window to obtain a normalized coherence coefficient value between 0 and 1. Based on this, a disturbance reference value used to distinguish between local problems and global disturbances is defined as a coherence threshold. The specific calibration process is as follows: during the no-load debugging phase of the equipment, a calibration vibration source with a known bandwidth is applied to the equipment base, and the coherence coefficient values ​​between the motor signals are recorded during this period. After repeating this process multiple times, 90% of the minimum observed value is taken as the disturbance reference value. This value was determined to be 0.75 in this calibration. If the coherence calculation results of the three sets of signals are higher than this threshold during actual operation, the system determines the anomaly as a whole-machine disturbance and maintains the current motion control parameters; otherwise, it is confirmed as a local cutting problem.

[0044] After confirming the anomaly as a localized cutting problem, the system then executes a decision arbitration step to eliminate false anomalies caused by transient sensor noise or self-healing minor physical interactions. The physical basis for this is that any persistent physical obstacle will inevitably be accompanied by a continuous increase in the system's energy consumption rate. Therefore, the system starts a timer while generating a feedforward adjustment command. In the subsequent second time period, it calculates the instantaneous energy consumption rate based on the electrical parameters of the cutting shear drive motor and compares it with a benchmark of average energy consumption per unit length. The benchmark is statistically derived from a test cut of the current fabric at a stable cutting speed, and the second time period is 100 milliseconds. The time interval is set to cover 10 energy calculation cycles. Only when the continuously calculated instantaneous energy consumption rate is consistently higher than the energy threshold determined by the benchmark and a ratio set to 130% within the second time interval will the system finally confirm the existence of a substantial physical obstacle and execute the generated feedforward adjustment command; otherwise, the command will be revoked. Through the hierarchical decision-making logic that integrates multi-source signal coherence analysis and subsequent energy consumption trend arbitration, the system can isolate real and threatening local cutting problems from industrial environmental noise and benign instantaneous disturbances, thereby focusing the control intervention on the confirmed risks.

[0045] Example 5: When applying the technical solution of the present invention to a previously unprocessed high-strength aramid fabric, a standardized offline calibration and parameter setting procedure needs to be executed to ensure the accuracy of identification and the stability of control in subsequent mass production. This procedure begins with probing the main texture direction of the fabric on the new fabric sample. After determining the main texture direction, a compensation coefficient characterizing its anisotropy is calibrated. The system will perform standard straight-line cuts at a constant cutting speed along multiple discrete angles from 0 degrees, 15 degrees, 30 degrees to 90 degrees relative to the main texture direction, and record the average cutting torque corresponding to the stable cutting segment at each angle. This set of empirical data points containing angles and torques will then be compared with the relevant formula using the least squares method. By performing a fitting process, the compensation coefficient that best describes the characteristics of the fabric can be calculated. The value is used as the initial reference feature spectrum of the fabric, and the stable feature spectrum recorded when cutting at zero degree angle is used as the initial reference feature spectrum of the fabric.

[0046] After calibrating the fabric properties, the newly installed cutting shears also need to be calibrated for wear characteristics to determine their passivation reference value. This procedure uses the same batch of aramid fabric for long-distance continuous cutting. Every 500 meters of cumulative cutting length, the system automatically performs a diagnostic cut on a standard sample and records the characteristic spectrum at that moment. At the same time, a digital microscope is used to check the edge quality of the diagnostic cut. When fiber burrs that meet the microscopic wear standard are observed, the system records the cumulative drift of the main peak frequency in the characteristic spectrum at that moment relative to the initial state, and sets this drift as the passivation reference value for this type of cutter and this material combination. In addition, the structured data generated and recorded during all the above offline calibration and wear testing processes, which includes complete working condition parameters, high-fidelity characteristic spectrum sequences, and corresponding physical events or quality results, are uniformly archived to form a physically meaningful and traceable empirical dataset necessary for training machine learning models.

[0047] Example 6: When deploying a machine learning model in an all-weather unattended cutting system for processing aerospace-grade composite materials, a standardized engineering procedure including model specification and system self-testing needs to be implemented to ensure the reliability of its decision-making and the safety of dealing with potential sensor failures. This procedure first defines the input and output structure of the machine learning model. Instead of directly using the original feature spectrum sequence, the input of each frame of feature spectrum is divided into preset frequency bands, namely 0 to 1kHz, 1kHz to 3kHz, and 3kHz to 5kHz. The total energy, energy standard deviation, and main peak frequency in each frequency band are calculated. The above feature values ​​generated by the spectrum of the last five consecutive moments constitute a fixed-dimensional input vector. The output of the model is designed as a two-branch structure. One branch is a classification output, which is used to identify whether the current working condition belongs to a defined abnormal type such as excessive fabric tension or fabric wrinkling. The other branch is a regression output, which is used to directly output a standardized motion control parameter adjustment value.

[0048] This procedure then defines an online self-checking and fault-tolerance mechanism to ensure the integrity of the system input signals. Before each trimming task, the control system instructs the trimmer drive motor to run at a standard speed for one second under no-load conditions and compares the resulting electrical signal spectrum with a pre-stored reference spectrum of the motor under healthy no-load conditions. If the cosine similarity of the two spectra is higher than a preset start-up threshold, the system is allowed to start the task; otherwise, it is determined that there is an initial fault in the sensor or the motor itself and an alarm is issued. During the task execution, the system also continuously monitors the total energy of the electrical parameter signals. If the total energy is lower than a low-level hardware noise threshold representing complete signal loss for a certain period of time, the system determines that the sensor signal is interrupted and immediately stops the trimming task. At the same time, the system state is switched to a safe mode to avoid making control decisions based on failed or falsified input data. In this way, the autonomous operation of the system is based on continuous verification of its own perception capabilities.

[0049] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0050] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for smart layout and cutting path planning of a garment, characterized in that, The method comprises the following steps: Step a, collecting electrical parameters of a driving motor driving the cutting knife in real time at a sampling frequency during the cutting task of the cutting knife; Step b, generating a characteristic spectrum representing the cutting torque in real time based on the collected electrical parameters by performing a short-time Fourier transform on the continuous electrical parameter signals; Step c, comparing the generated characteristic spectrum in real time with a reference characteristic spectrum to identify an abnormal cutting state; Step d, when the abnormal cutting state is identified, feeding forwardly adjusting the motion control parameters of a subsequent cutting path next to the current cutting position which has not been executed based on the type of the abnormal cutting state to avoid cutting failure caused by the abnormal cutting state.

2. The method of claim 1, wherein, The electrical parameters are current feedback signals or torque instruction signals of the driving motor; The reference characteristic spectrum is obtained by trial cutting on a fabric sample; the abnormal cutting state includes a fabric tension too large state represented by high energy in a high frequency region of the characteristic spectrum and a fabric wrinkling state represented by wideband noise of non-harmonics.

3. The method of claim 2, wherein, The feeding forwardly adjusting specifically includes: when the fabric tension too large state is identified, reducing the cutting speed of the subsequent cutting path; and when the fabric wrinkling state is identified, performing a micro-operation of lifting the cutting knife, displacing along the path direction and lowering the cutting knife again.

4. The method of claim 1, wherein, Before step c, there is also a step of establishing and dynamically correcting the reference characteristic spectrum, which specifically includes: synchronously collecting encoder feedback speed signals of servo motors driving the X-axis and Y-axis translation of the cutting bed during the cutting process; when the fluctuation of the speed signals is lower than a stable reference value within a first time length, the corresponding characteristic spectrum in this period is identified as a high-confidence health fingerprint.

5. The method of claim 1, wherein, After step d, there is also a decision arbitration step, which specifically includes: after the feeding forwardly adjusting instruction is generated, starting a timer and calculating the instantaneous energy consumption rate of the electrical parameter calculation system within a second time length; if the instantaneous energy consumption rate continuously exceeds a unit length average energy consumption reference by a set proportion, the feeding forwardly adjusting instruction is executed; otherwise, the instruction is cancelled.

6. The method of claim 1, wherein, Before step d, there is also a global disturbance identification step, which specifically includes: when the change of the characteristic spectrum exceeds a high amplitude threshold, synchronously acquiring electrical parameters of servo motors driving the X-axis and Y-axis translation of the cutting bed; if the changes of the electrical parameters of the X-axis and Y-axis servo motors at the same time point do not exceed a disturbance reference value, it is confirmed that the abnormality is a local cutting problem and the adjustment of step d is performed; otherwise, the abnormality is determined as a whole machine disturbance and the current motion control parameters are maintained.

7. The method of claim 1, wherein, Before cutting a fabric, there is also a step of detecting the main texture direction of the fabric, which specifically includes: controlling the cutting knife driving motor to perform a standardized micro-disturbance injection action at at least two detection points of the fabric in different directions; collecting transient response signals caused by the micro-disturbance injection action and determining the direction with the slowest attenuation of high frequency components in the response signals as the main texture direction of the fabric.

8. The method of claim 7, wherein, In step c, the comparison reference for identifying the abnormal cutting state is dynamic, and the specific generation manner is: before comparison, the judgment threshold of the reference characteristic spectrum is dynamically compensated according to the included angle between the current cutting path and the detected main texture direction, and the dynamic threshold after compensation is The dynamic threshold is determined by the following relationship: Wherein, is the original judgment threshold of the reference characteristic spectrum, is the included angle between the current cutting path vector and the detected main texture direction vector, is a compensation coefficient representing the anisotropy degree of the fabric.

9. The method of claim 1, wherein, Also included is a step of long-term trend analysis of the characteristic spectrum, which is specifically: in multiple cutting tasks, record the drift of the main peak frequency in the characteristic spectrum with the cumulative cutting time; when the drift exceeds a passivation reference value, the system generates and issues a prompt information to replace the cutting knife after completing the current task.

10. The method of claim 1, wherein, The identification of the abnormal cutting state in step c and the feedforward adjustment in step d are based on a machine learning model; the machine learning model is trained with the characteristic spectrum sequence and the corresponding cutting success or failure label in the historical cutting process to establish a mapping relationship between the characteristic spectrum and the abnormal state type and the adjustment strategy.

Images