Cut-off dynamic capture method and system combining reinforcement learning with physical modeling

Abstract

The application belongs to the technical field of textiles, and discloses a broken-end dynamic capturing method and system combining reinforcement learning and physical modeling. The method comprises the following steps: constructing a yarn microstructure evolution model; inputting real-time environment parameters of a textile workshop into the microstructure evolution model to predict real-time microstructure parameters of the yarn; performing nonlinear coupling feature extraction on the real-time environment parameters to calculate an environment coupling feature vector; inputting the real-time microstructure parameters as material properties into a yarn tension dynamics equation to solve a predicted macro stress state of the yarn; splicing the macro stress state of the yarn and the environment coupling feature vector to form a state observation value; outputting a broken-end risk probability value by a reinforcement learning intelligent agent according to the state observation value; and generating and outputting a broken-end early warning signal when the broken-end risk probability value exceeds a preset threshold. The application can realize early and accurate warning of yarn broken-end in a complex dynamic environment.

Description

Technical Field

[0001] This invention relates to the field of textile technology, specifically to a method and system for dynamic capture of severed heads that combines reinforcement learning and physical modeling. Background Technology

[0002] In the textile industry, yarn breakage is a common and serious quality problem, especially in high-speed and high-load production environments. Yarn breakage not only affects production efficiency but also leads to decreased product quality and increased costs.

[0003] Currently, yarn breakage detection technologies in the textile industry mainly fall into two categories: one is a physical modeling method based on mechanical sensors, and the other is monitoring through image recognition and video surveillance technologies. Mechanical sensor methods typically rely on monitoring the physical parameters of the yarn, such as tension and breaking strength. However, these methods struggle to effectively capture changes in the yarn's stress state and its coupling relationship with environmental parameters when faced with fluctuations in complex environmental factors (such as temperature, humidity, and airflow). While image recognition and video surveillance technologies can detect yarn breakage, their limited ability to capture changes in the yarn's microstructure and reliance on post-processing prevent them from providing effective early warnings before breakage occurs. Furthermore, existing sensors and detection systems often cannot monitor subtle changes in the yarn's microstructure in real time, particularly the impact of environmental fluctuations on the yarn's stress state, making it impossible to effectively predict yarn breakage.

[0004] In summary, existing technologies have failed to effectively address the problem of yarn breakage warning in dynamically changing environments. Especially under the combined influence of multiple environmental factors, detecting pre-breakage signs remains a significant challenge. Therefore, there is an urgent need for a technical solution capable of accurately capturing pre-breakage signs of yarn in complex environmental conditions to improve the accuracy of warnings and production efficiency. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for dynamic capture of decapitation by combining reinforcement learning and physical modeling, so as to solve the problems mentioned in the background art.

[0006] This invention provides a method for dynamic capture of decapitation combining reinforcement learning and physical modeling, comprising the following steps:

[0007] Step S1: Construct a yarn microstructure evolution model that can simulate the dynamic changes in the friction coefficient of the yarn surface fibers and the cohesion between fibers caused by fluctuations in environmental parameters; wherein, the environmental parameters include temperature, humidity and airflow velocity;

[0008] Step S2: Input the real-time environmental parameters of the textile workshop into the microstructure evolution model to predict the real-time microstructure parameters of the yarn; perform nonlinear coupling feature extraction on the real-time environmental parameters and calculate the environmental coupling feature vector including the temperature and humidity synergy index and the airflow disturbance intensity spectrum.

[0009] Step S3: Input the real-time microstructure parameters as material properties into the yarn tension dynamics equation to solve for the predicted macroscopic stress state of the yarn; concatenate the macroscopic stress state of the yarn with the environmental coupling feature vector to form the state observation value of the reinforcement learning agent;

[0010] Step S4: The reinforcement learning agent outputs the probability value of the yarn breakage risk based on the state observation value; wherein, the reinforcement learning agent learns through training that when the environmental coupling feature vector changes drastically and the macroscopic stress state of the yarn shows a lag or abnormal response, it determines that the coupling relationship is unstable and increases the probability value of the yarn breakage risk.

[0011] Step S5: When the probability value of decapitation risk exceeds a preset threshold, a decapitation warning signal is generated and output.

[0012] This invention also provides a decapitation dynamic capture system combining reinforcement learning and physical modeling, the system comprising:

[0013] The yarn microstructure evolution model construction module is used to: construct a yarn microstructure evolution model that can simulate the dynamic changes in the friction coefficient of the yarn surface fibers and the cohesion between fibers caused by fluctuations in environmental parameters; wherein, environmental parameters include temperature, humidity and airflow velocity;

[0014] The environmental parameter processing and feature extraction module is used to: input the real-time environmental parameters of the textile workshop into the microstructure evolution model to predict the real-time microstructure parameters of the yarn; perform nonlinear coupling feature extraction on the real-time environmental parameters, and calculate the environmental coupling feature vector including the temperature and humidity synergy index and the airflow disturbance intensity spectrum.

[0015] The yarn state prediction and fusion module is used to: input the real-time microstructure parameters as material properties into the yarn tension dynamics equation, solve for the predicted macroscopic stress state of the yarn; and concatenate the macroscopic stress state of the yarn with the environmental coupling feature vector to form the state observation value of the reinforcement learning agent.

[0016] The reinforcement learning agent early warning module is used to: output a breakage risk probability value based on the state observation value; wherein, the reinforcement learning agent learns through training that: when the environmental coupling feature vector changes drastically and the macroscopic stress state of the yarn shows a lag or abnormal response, it determines that the coupling relationship is unstable and increases the breakage risk probability value.

[0017] The warning signal generation and output module is used to generate and output a decapitation warning signal when the probability value of the decapitation risk exceeds a preset threshold.

[0018] This invention constructs an evolutionary model from environmental parameters to yarn microstructure, inputs real-time micro parameters as dynamic attributes into the macroscopic tension equation, and constructs state observations by combining environmental coupling characteristics. It utilizes a reinforcement learning agent to identify instability precursors of environmental disturbances and yarn response mismatch, thereby achieving early and accurate warning of yarn breakage in complex dynamic environments. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating a method for dynamic capture of decapitation combining reinforcement learning and physical modeling, as disclosed in an embodiment of the present invention.

[0020] Figure 2 This is a data flow diagram of a decapitation dynamic capture method combining reinforcement learning and physical modeling disclosed in an embodiment of the present invention;

[0021] Figure 3 This is a schematic diagram of a decapitation dynamic capture system that combines reinforcement learning and physical modeling, as disclosed in an embodiment of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Please see Figure 1 , Figure 2 This invention provides a method for dynamic capture of decapitation combining reinforcement learning and physical modeling, comprising the following steps:

[0024] Step S1: Construct a yarn microstructure evolution model that can simulate the dynamic changes in the friction coefficient of the yarn surface fibers and the cohesion between fibers caused by fluctuations in environmental parameters; wherein, the environmental parameters include temperature, humidity and airflow velocity;

[0025] In this step, in the high-speed textile production environment, yarn breakage often does not occur suddenly, but is the result of the gradual deterioration of its internal structure under continuous or drastic fluctuations in environmental parameters such as temperature, humidity, and airflow. Traditional macroscopic sensors (such as tension sensors) can only capture the state changes at the moment of breakage or near breakage, but cannot detect this gradual microscopic deterioration process.

[0026] Therefore, this invention constructs a yarn microstructure evolution model, which is a mathematical model based on the principles of materials mechanics, tribology, and textile technology. This model is used to quantitatively simulate how dynamic fluctuations in environmental parameters (temperature, humidity, airflow velocity) affect the key intrinsic properties of yarn. Specifically:

[0027] Temperature affects the modulus, elasticity, and thermal expansion of fibrous materials. Increased temperature may cause fibers to soften and reduce their rigidity; while temperature fluctuations can cause changes in the frictional state between fibers and between fibers and yarn guide devices.

[0028] Humidity affects the hygroscopicity of fibers, thereby altering their diameter, strength, and frictional properties. Increased humidity can cause fibers to swell and increase the cohesion between fibers, but excessive humidity can also make yarns heavy and prone to sticking.

[0029] Airflow within the workshop (such as air conditioning air and turbulence generated by equipment operation) will produce continuous and random aerodynamic disturbances on the running yarn. These disturbances are directly converted into lateral forces on the yarn, affecting its running trajectory and dynamic tension.

[0030] This yarn microstructure evolution model takes the aforementioned environmental parameters as input and dynamically calculates and outputs microstructure parameters, including the yarn surface fiber friction coefficient and interfiber cohesion, through built-in physical laws (e.g., the fiber diameter variation equation considering moisture absorption and expansion, the Coulomb friction coefficient correction formula based on temperature and humidity conditions, and the stripping effect model of airflow disturbance on surface fibers). The yarn surface fiber friction coefficient reflects the sliding friction state between the yarn and key contact components such as the yarn guide hook, ring, and traveler, directly affecting the yarn's running resistance and wear rate. The interfiber cohesion reflects the tightness of the entanglement and bonding of fibers within the yarn, determining the yarn's intrinsic strength and its ability to resist disintegration under dynamic loads.

[0031] Step S2: Input the real-time environmental parameters of the textile workshop into the microstructure evolution model to predict the real-time microstructure parameters of the yarn; perform nonlinear coupling feature extraction on the real-time environmental parameters and calculate the environmental coupling feature vector including the temperature and humidity synergy index and the airflow disturbance intensity spectrum.

[0032] In this step, real-time data such as temperature T(t), humidity H(t), and airflow velocity V(t) collected by the textile workshop's sensor network are directly input into the aforementioned constructed yarn microstructure evolution model. Based on the current environmental parameter input, this model calculates and outputs the corresponding real-time microstructure parameters of the yarn, namely, the real-time coefficient of friction of the yarn surface fibers. and interfiber cohesion .

[0033] The effects of environmental factors on yarn are not independent linear additive effects, but rather involve complex nonlinear coupling effects. For example, the synergistic effect of high temperature and high humidity may far exceed the simple sum of their individual effects; airflow disturbances at specific frequencies can resonate with the natural frequencies of the yarn. Therefore, this step performs in-depth processing on the original environmental parameters to extract environmental coupling feature vectors, as follows:

[0034] Temperature and humidity synergy index: This index quantifies the intensity and pattern of synergistic fluctuations by calculating the joint fluctuation entropy and phase space trajectory correlation of temperature and humidity time series. A high temperature and humidity synergy index indicates that temperature and humidity are undergoing drastic and synchronous changes, creating strong composite stress on the yarn structure.

[0035] Airflow disturbance intensity spectrum: By performing time-frequency analysis (such as wavelet transform) on the airflow velocity signal, the energy distribution along the frequency dimension is obtained. The focus is on analyzing the energy intensity within key frequency bands (e.g., 0.5-5Hz) close to the natural frequency of yarn transverse vibration. It can be understood that this airflow disturbance intensity spectrum characterizes the dangerous components in the airflow disturbance that may trigger yarn resonance.

[0036] Step S3: Input the real-time microstructure parameters as material properties into the yarn tension dynamics equation to solve for the predicted macroscopic stress state of the yarn; concatenate the macroscopic stress state of the yarn with the environmental coupling feature vector to form the state observation value of the reinforcement learning agent;

[0037] In this step, the real-time microstructure parameters predicted in step S2 are... , As a key material property, it is input into the preset yarn tension dynamics equation. This yarn tension dynamics equation is a differential equation describing the forces acting on the yarn along the spinning path, typically considering inertial forces, damping forces, elastic restoring forces, and external excitation forces. Its expression is as follows:

[0038]

[0039] in, —Inertial force term. The equivalent mass of the yarn within the monitoring interval (a constant determined by the yarn linear density and the interval length); This is the acceleration of the yarn's transverse (or axial) vibration. This reflects the inertial effect caused by changes in the yarn's motion.

[0040] —Damping force term. The speed of yarn vibration; The damping coefficient is the input of real-time microstructure parameters, i.e., the fiber friction coefficient. and interfiber cohesion .

[0041] It should be noted that the damping of yarn mainly comes from the frictional slippage between the internal fibers (and) , (Strong correlation) and friction with the yarn guiding device. By adjusting the microstructure parameters , Substitute into function This allows for the real-time mapping of changes in the internal state of the yarn directly to macroscopic dynamic damping characteristics. For example, cohesion force. A decrease in the damping coefficient may lead to increased interfiber slippage, manifested as an increase in the damping coefficient. The changes.

[0042] —Elastic restoring force term. This represents the displacement of the yarn relative to its equilibrium position. This is the stiffness coefficient (or elastic coefficient), which is also a dynamic parameter, representing the interfiber cohesion force. The function.

[0043] It should be noted that the transverse or axial stiffness of a yarn largely depends on the tightness of the fiber bonding (i.e., cohesion). When the cohesion weakens, the yarn's ability to resist deformation decreases, which manifests as a decrease in the stiffness coefficient. The reduction is understandable, as this design directly links the microstructural strength to the macroscopic mechanical stiffness.

[0044] —External incentive term. The calculated environmental coupling feature vector includes temperature and humidity synergy index, airflow disturbance intensity spectrum, etc.

[0045] It should be noted that the right-hand side of the equation is a complex function related to the coupling characteristics of the real-time environment. For example, the energy at a specific frequency in the airflow disturbance intensity spectrum can be used as the excitation amplitude input at that frequency. This setting ensures that the source of the model's driving force corresponds to the complex disturbances in a real workshop.

[0046] After solving the yarn tension dynamics equations using numerical methods (such as the Runge-Kutta method), the predicted macroscopic stress state of the yarn is obtained. This state is represented by a set of vectors. It means that among them This is the predicted value of dynamic tension. This is the predicted value of vibration amplitude. Characteristic frequencies, etc. It can be understood that the macroscopic stress state of the yarn is reflected in the current real-time microstructural parameters. , and environmental coupling excitation The mechanical behavior that the yarn system should exhibit under the combined action.

[0047] Next, the macroscopic force state vector Feature vectors coupled with the environment (e.g.) ,in The temperature and humidity synergy index, The specific frequency band energy characteristics of the airflow disturbance intensity spectrum are merged according to dimension to form a higher-dimensional state observation. .

[0048] Step S4: The reinforcement learning agent outputs the probability value of the yarn breakage risk based on the state observation value; wherein, the reinforcement learning agent learns through training that when the environmental coupling feature vector changes drastically and the macroscopic stress state of the yarn shows a lag or abnormal response, it determines that the coupling relationship is unstable and increases the probability value of the yarn breakage risk.

[0049] In this step, before deployment to the actual production system, the reinforcement learning agent needs to complete sufficient learning in a training environment consisting of historical production datasets, a high-fidelity yarn dynamic simulation platform, and risk labels annotated by experts. Its training objective is to learn the optimal policy to maximize long-term cumulative rewards. The reinforcement learning agent can be constructed using either Deep Deterministic Policy Gradient (DDPG) or Proximal Policy Optimization (PPO) algorithms.

[0050] It should be noted that during the training of the reinforcement learning agent, the reward function is designed to guide the agent to capture key precursors to instability. Specifically, a small positive reward is given during stable operation; when a sharp jump is detected in the environmental coupling feature vector (such as the temperature and humidity co-existence index, the key frequency band energy of the airflow disturbance intensity spectrum) (e.g., its rate of change exceeds a threshold), a positive reward is given. Meanwhile, the macroscopic stress state obtained from solving the yarn tension dynamics equation (such as the tension prediction value) Compared to a stable baseline, there is a significant lag (e.g., response time delay exceeding [a certain value]). When an agent exhibits abnormal fluctuations (e.g., a sudden increase in variance), a large negative reward is given. This reward mechanism forces the agent to comprehensively consider the dynamic matching relationship between environmental inputs and system responses when making decisions.

[0051] When a reinforcement learning agent makes online decisions, the state observations output in step S3 are used. As input, its internal network outputs a scalar value, namely the probability value of decapitation risk, through forward propagation. Understandably, the core calculation logic for this decapitation risk probability value stems from what was learned during training:

[0052] The agent continuously analyzes the correlation between environmental features and macroscopic state components in the observed state values. When it identifies drastic changes in environmental features while the macroscopic state component exhibits lag or abnormal patterns, it determines that the current yarn system's environment-structure-mechanical coupling relationship is unstable. This instability is a high-risk precursor to impending yarn breakage. Therefore, the agent dynamically adjusts the output risk probability value accordingly. For example, when the energy of the airflow disturbance intensity spectrum in the 1-3Hz frequency band suddenly increases by 50%, but the predicted yarn tension does not respond synchronously and instead shows low-frequency oscillations, It rose rapidly from 0.1 to over 0.7.

[0053] Step S5: When the probability value of decapitation risk exceeds a preset threshold, a decapitation warning signal is generated and output.

[0054] In this step, the real-time decapitation risk probability value is output to the agent. Continuous monitoring will be conducted, once The value exceeds the preset threshold This immediately triggers a head breakage warning signal. (Preset threshold) It can be flexibly configured according to different yarn types, process speeds and quality requirements. For example, it can be calibrated based on the statistical distribution of risk probability values ​​in the retrospective analysis of historical yarn breakage events (such as setting it to the 95th percentile of the risk probability distribution) and combined with the tolerance for false alarms and missed alarms in actual production.

[0055] It is understandable that the generated decapitation warning signal is a structured data packet, which may contain information such as:

[0056] Warning level: According to Exceeding The severity level is divided into different grades, such as Attention, Warning, and Severe. Spindle / Yarn Path Identifier: Precisely indicates the location where the risk of yarn breakage occurs. Timestamp: The precise time the warning was triggered. Key Data Snapshot: Real-time environmental parameters at the moment the warning was triggered (…). ), predicted microstructure parameters ( ), predicted macroscopic stress state ( and the final risk probability value wait.

[0057] In addition, the generated early warning signals are output through a communication interface, enabling multi-level responses. For example, the signals can drive audible and visual alarms installed near the corresponding spindles or be pushed to the operator's smart handheld terminal, achieving precise on-site alarms and manual intervention guidance; the signals can be transmitted to the main control system of the spinning machine, automatically triggering preset deceleration programs, making minor adjustments to process parameters (such as adjusting the draft ratio), or activating backup cleaning devices to attempt intervention and adjustment before the spindle breakage occurs; the signals are simultaneously uploaded to the workshop manufacturing execution system (MES) or cloud data platform for production quality traceability, equipment health status analysis, and big data analysis for process optimization, forming a closed-loop improvement.

[0058] This invention constructs an evolutionary model from environmental parameters to yarn microstructure, inputs real-time micro parameters as dynamic attributes into the macroscopic tension equation, and constructs state observations by combining environmental coupling characteristics. It utilizes a reinforcement learning agent to identify instability precursors of environmental disturbances and yarn response mismatch, thereby achieving early and accurate warning of yarn breakage in complex dynamic environments.

[0059] As an example, nonlinear coupling feature extraction is performed on real-time environmental parameters to calculate an environmental coupling feature vector, including the temperature and humidity synergy index and the airflow disturbance intensity spectrum, including:

[0060] Step S21: By analyzing the joint probability distribution of the temperature time series and the humidity time series and their dynamic trajectories in the reconstructed phase space, the joint fluctuation entropy and the average trajectory divergence are calculated respectively, and the two are weighted and fused to generate the temperature and humidity synergy index.

[0061] In this step, a joint probability distribution is constructed for the temperature and humidity time series collected within the monitoring time window. Specifically, the temperature and humidity values ​​are divided into several discrete intervals, and the frequency of their combinations falling within each joint interval is counted, forming a joint probability distribution table for temperature and humidity. Based on this joint probability distribution, its Shannon entropy is calculated as the joint fluctuation entropy characterizing the disorder of the temperature and humidity joint state. The lower the entropy value, the more concentrated the temperature and humidity state combinations are, and the stronger the coordination; the higher the entropy value, the more dispersed the state combinations are, and the weaker the coordination.

[0062] For example: Obtain the temperature sequence T=[t1,t2,...,t] for N consecutive sampling times. N ], and the humidity sequence H=[h1,h2,...,h] for N consecutive sampling times. N Normalization is performed on the following:

[0063] , where μ T μ H σ is the mean. T σ HThe standard deviation is denoted as .

[0064] The normalized temperature and humidity values ​​are discretized separately. Let the temperature be divided into M... T Humidity is divided into several ranges, M. H There are several intervals. Calculate the probability distribution of the temperature value falling in interval i and the humidity value falling in interval j simultaneously:

[0065]

[0066] Furthermore, the joint Shannon entropy is calculated as the joint fluctuation entropy JWE:

[0067] .

[0068] Simultaneously, phase space reconstruction techniques from time series analysis were employed to reconstruct the dynamic system phase space of the temperature and humidity series, respectively, yielding their respective dynamic evolution trajectories. The distance between corresponding points on these two trajectories was calculated, and the average value over the entire time window was obtained to determine the average trajectory divergence. The smaller this divergence value, the more similar the dynamic change patterns of temperature and humidity are.

[0069] For example: the phase space of a temperature sequence is reconstructed using the delayed embedding method. Let the embedding dimension be m and the delay time be τ, then the phase space vector of the temperature sequence is obtained: Similarly, the phase space vector of the humidity sequence is reconstructed. .

[0070] For each time point k, calculate the cosine distance between the phase space vectors of the temperature and humidity sequences:

[0071] .

[0072] Calculate the average trajectory divergence:

[0073] ,in This represents the number of points in the reconstructed phase space.

[0074] Finally, the joint fluctuation entropy and the average trajectory divergence obtained above are weighted and summed to generate a comprehensive temperature and humidity synergy index. The weighting coefficient , Determined through regression analysis of historical severed head data, and satisfying the following conditions: Understandably, the higher the temperature and humidity synergy index, the more intense and highly synchronized the combined changes in temperature and humidity are.

[0075] Step S22: The time-frequency energy distribution is obtained by performing time-frequency transformation on the airflow velocity sequence, and the energy characteristics of the corresponding key frequency bands are extracted from the time-frequency energy distribution according to the natural frequency range determined by the yarn tension dynamics equation, so as to form the airflow disturbance intensity spectrum.

[0076] As an example, the inherent frequency range is obtained by performing eigenvalue analysis on the yarn tension dynamics equation.

[0077] In this step, time-frequency transformation analysis is performed on the real-time acquired airflow velocity time series, preferably using continuous wavelet transform. Through continuous wavelet transform, the one-dimensional airflow velocity signal is converted into a two-dimensional time-frequency energy distribution map, which can clearly show the disturbance energy intensity at different times and at different frequency components.

[0078] For example: for the airflow velocity sequence V=[v1,v2,...,v...] N Perform continuous wavelet transform:

[0079] ,in Here, is the Morlet wavelet basis function, a is the scaling parameter (corresponding to frequency), and b is the translation parameter (corresponding to time).

[0080] To focus on the disturbance components that significantly affect yarn vibration, it is necessary to determine the natural frequency range of the yarn system. In this embodiment, the natural frequency range is obtained by linearizing the yarn tension dynamics equation and solving for its eigenvalues; the frequencies corresponding to these eigenvalues ​​are the main natural frequencies of the yarn system.

[0081] For example: linearization is performed near the equilibrium point based on the yarn tension dynamics equation: ,in , Let be the damping and stiffness coefficients of the yarn under reference conditions. Solve the characteristic equation: Eigenvalues ​​are obtained: The imaginary part That is, angular frequency. .

[0082] Therefore, the natural frequency of the system is: The frequency range is [f0-Δf, f0+Δf], where Δf can be set to 10%-20% of f0 depending on the yarn type.

[0083] Based on the obtained inherent frequency range, energy characteristics within the corresponding frequency bands (e.g., multiple key frequency bands formed by extending a certain bandwidth above and below each inherent frequency) are extracted from the aforementioned time-frequency energy distribution map. These characteristics can be the average energy, peak energy, or energy variance of each key frequency band during the monitoring period. Organizing these extracted energy characteristics by frequency band constitutes the airflow disturbance intensity spectrum. It can be understood that this airflow disturbance intensity spectrum directly quantifies the intensity of the dangerous frequency components in the airflow disturbance that may excite yarn resonance.

[0084] For example: the wavelet transform result W(a,b) is converted into a time-frequency energy distribution S(f,t)=|W(a,b)| 2 The relationship between frequency f and scale a is f = f c / a,f c The center frequency of the wavelet.

[0085] Within the inherent frequency range [f0-Δf, f0+Δf], the frequency band is divided into K equal sub-bands, each with a center frequency of f. k Calculate the energy characteristics of each sub-band. Taking the k-th sub-band as an example, calculate:

[0086] Average energy: ;

[0087] Peak energy: ;

[0088] Energy variance: .

[0089] The intensity spectrum of airflow disturbance is represented in the following vector form:

[0090] .

[0091] The above calculation results are combined into an environmental coupling feature vector: .

[0092] As an example, when the environmental coupling characteristic vector changes drastically and the macroscopic stress state of the yarn exhibits a lag or abnormal response, the coupling relationship is determined to be unstable, including:

[0093] The rate of change of the environmental coupling feature vector in adjacent time windows is calculated in real time. When the rate of change of the temperature and humidity synergy index exceeds the first threshold, and / or the rate of change of energy in any key frequency band of the airflow disturbance intensity spectrum exceeds the second threshold, it is determined that the environmental coupling feature vector has changed drastically.

[0094] In this step, the rate of change of each component in the environmental coupling feature vector is calculated in real time, using a fixed time window (e.g., 10 seconds). Specifically, the relative rate of change or absolute difference rate between the feature values ​​of the current time window and the previous time window is calculated.

[0095] Regarding the temperature and humidity synergy index Set the first threshold .when rate of change per unit time ( )Exceed At that time, a drastic change in the temperature and humidity synergy is determined. It is understandable that this first threshold... It can be calibrated based on the index fluctuation range during extreme weather events or the start-up and shutdown of air conditioning systems in historical data.

[0096] For the intensity spectrum of airflow disturbance Set a second threshold Examine the rate of change of energy (e.g., average energy) in any key frequency band of the vector. If the rate of change of energy in any frequency band exceeds... This means that the intensity of the airflow disturbance has changed drastically. This second threshold... It can be set according to the typical spectrum energy fluctuations caused by sudden changes in fan speed or the opening and closing of doors and windows.

[0097] When the energy of any key frequency band in the temperature and humidity synergy index or airflow disturbance intensity spectrum is determined to be drastic, it is a sign that a drastic change in the environmental coupling characteristic vector has been triggered.

[0098] The macroscopic stress state of the yarn obtained by solving the yarn tension dynamics equation is compared with the benchmark macroscopic stress state calculated under stable environmental conditions based on the yarn tension dynamics equation. When the dynamic tension response delay in the macroscopic stress state of the yarn exceeds a preset delay threshold and / or the vibration amplitude exceeds a preset safe range, it is determined that the macroscopic stress state of the yarn has a hysteresis or abnormal response.

[0099] In this step, during the initial system calibration or stable production period, a set of benchmark macroscopic stress states is obtained by inputting steady-state environmental parameters and corresponding benchmark microstructure parameters through the yarn tension dynamics equation, denoted as... This represents the healthy mechanical performance of the yarn under conditions of no severe disturbance.

[0100] Calculate the macroscopic stress state of the yarn (e.g., dynamic tension) from the moment the environmental characteristics are determined to be drastic changes to the real-time value. The time interval elapsed until a significant deviation (e.g., a deviation from the baseline exceeding 3%) begins to occur is defined as the response delay time. .like Exceeding the preset delay threshold (For example, 1.5 times the normal thermal inertia response time), then it is determined that the macroscopic stress state response of the yarn is lagging.

[0101] If the amplitude of vibration Exceeding the benchmark amplitude The safety range or characteristic frequency is set by a safety factor (e.g., 1.8 times). Deviation from reference frequency If the percentage exceeds a certain percentage (e.g., 15%), it is determined that the macroscopic stress state of the yarn is abnormal.

[0102] When the macroscopic stress state of the yarn meets one or more of the conditions of delayed response or abnormal response, it is a sign that the macroscopic stress state has a delayed or abnormal response.

[0103] When both of the above conditions are met, the decision coupling relationship of the reinforcement learning agent is unstable.

[0104] The reinforcement learning agent continuously monitors the two judgment indicators mentioned above. Only when both indicators—a drastic change in the environmental coupling feature vector and a delayed or abnormal response in the macroscopic stress state of the yarn—are triggered simultaneously within the same or adjacent monitoring time window, will the agent ultimately determine that the current environment-structure-mechanical coupling relationship of the yarn system is unstable, i.e., it is a precursor to yarn breakage.

[0105] As an example, reinforcement learning agents are pre-trained in the following manner:

[0106] A training dataset is constructed based on historical production data, where each training sample contains a sequence of continuous state observations for a historical time period and the corresponding label of the decapitation result.

[0107] Design the training reward function: give a basic positive reward at each step where no yarn breakage occurs, give a negative reward when the environmental coupling feature vector changes drastically and the macroscopic stress state of the yarn shows a lag or abnormal response, and give a large penalty at the step where yarn breakage finally occurs.

[0108] Based on the training dataset, the reinforcement learning agent learns with the goal of maximizing cumulative reward until its policy converges.

[0109] In this implementation, historical production data from the textile workshop is collected and preprocessed, including environmental sensor data (temperature, humidity, airflow velocity), yarn condition data (obtained indirectly or generated through simulation), and key yarn breakage event records. The historical production data is divided into several training samples in chronological order. Each training sample contains a complete information sequence within a continuous time window (e.g., 60 seconds before the yarn breakage occurs), as detailed below:

[0110] State observation sequence: Each point in this sequence contains a complete state vector, which is the concatenation of the predicted macroscopic stress state of the yarn at the corresponding time and the coupled feature vector of the environment. The "head-breaking" result label indicates whether a head break occurred at the end of the time window. For head-breaking samples, the end of the window is marked as "head-breaking"; for normal samples, the entire window is marked as "normal". The constructed sample set is divided into training, validation, and test sets for model training, tuning, and performance evaluation.

[0111] The specific design of the reward function directly determines the quality of the policy ultimately learned by the agent. The reward function in this embodiment... At each time step Calculated according to the following rules:

[0112] Basic positive reward: If no break occurs at this time step, a small, constant positive reward is given. (e.g., +0.1) to encourage agents to survive longer under normal circumstances.

[0113] Risk-based negative reward: When the current state is detected to meet the aforementioned definition of an unstable coupling pattern (i.e., drastic changes in environmental characteristics, coupled with lag or anomalies in macroscopic force states), a large negative reward is given. (e.g., -1) to force the agent to actively identify and be alert to this high-risk state, even if the decapitation has not yet occurred.

[0114] Decapitation Penalty: If decapitation occurs at this time step, a very large negative reward is given as a penalty. (e.g., -10), and terminate the current training round so that the agent must prioritize avoiding decapitation.

[0115] Deep reinforcement learning algorithms suitable for continuous states and action spaces, such as Proximal Policy Optimization (PPO), Deep Deterministic Policy Gradient (DDPG), or Soft Actor-Critic (SAC) algorithms, can be used to construct and train the agent. The agent's goal is to learn a policy. This allows for maximizing the cumulative discounted reward obtained from the environment when interacting with a simulation environment built from the training dataset. ,in It is the discount factor (0 < γ < 1).

[0116] The agent learns from repeated trials, gaining experience from both success (positive rewards) and failure (negative rewards). In particular, through extensive exposure to intermediate states marked by negative rewards in risk-based scenarios, the agent is forced to learn that in the early stages of a mismatch between environmental disturbances and mechanical responses, even if absolute values ​​such as tension have not yet reached dangerous levels, it foreshadows an extremely high risk of decapitation. Understandably, this learning process is a quantification and internalization of human expert experience.

[0117] Training continues until the agent's policy stabilizes. Convergence criteria can be: performance evaluated on the validation set (such as accuracy and recall for head-cutting prediction) no longer shows significant improvement, or the average cumulative reward curve obtained by the agent in the training environment tends to plateau.

[0118] Please see Figure 3 This invention also provides a decapitation dynamic capture system 200 that combines reinforcement learning and physical modeling, the system comprising:

[0119] The yarn microstructure evolution model construction module 2001 is used to: construct a yarn microstructure evolution model that can simulate the dynamic changes in the friction coefficient of the yarn surface fibers and the cohesion between fibers caused by fluctuations in environmental parameters; wherein, the environmental parameters include temperature, humidity and airflow velocity;

[0120] The environmental parameter processing and feature extraction module 2002 is used to: input the real-time environmental parameters of the textile workshop into the microstructure evolution model to predict the real-time microstructure parameters of the yarn; perform nonlinear coupling feature extraction on the real-time environmental parameters, and calculate the environmental coupling feature vector including the temperature and humidity synergy index and the airflow disturbance intensity spectrum.

[0121] The yarn state prediction and fusion module 2003 is used to: input the real-time microstructure parameters as material properties into the yarn tension dynamics equation, solve for the predicted macroscopic stress state of the yarn; and concatenate the macroscopic stress state of the yarn with the environmental coupling feature vector to form the state observation value of the reinforcement learning agent.

[0122] The reinforcement learning agent early warning module 2004 is used for: the reinforcement learning agent outputs a breakage risk probability value based on the state observation value; wherein, the reinforcement learning agent learns through training that: when the environmental coupling feature vector changes drastically and the macroscopic stress state of the yarn shows a lag or abnormal response, it determines that the coupling relationship is unstable and increases the breakage risk probability value.

[0123] The warning signal generation and output module 2005 is used to generate and output a decapitation warning signal when the probability value of the decapitation risk exceeds a preset threshold.

[0124] As an example, the environmental parameter processing and feature extraction module 2002 is specifically used for:

[0125] By analyzing the joint probability distribution of temperature time series and humidity time series and their dynamic trajectories in the reconstructed phase space, the joint fluctuation entropy and the average trajectory divergence are calculated respectively, and the two are weighted and fused to generate the temperature and humidity synergy index.

[0126] The time-frequency energy distribution is obtained by performing time-frequency transformation on the airflow velocity sequence, and the energy characteristics of the corresponding key frequency bands are extracted from the time-frequency energy distribution based on the natural frequency range determined by the yarn tension dynamics equation to form the airflow disturbance intensity spectrum.

[0127] As an example, the inherent frequency range is obtained by performing eigenvalue analysis on the yarn tension dynamics equation.

[0128] As an example, the reinforcement learning agent early warning module 2004 is specifically used for:

[0129] The rate of change of the environmental coupling feature vector in adjacent time windows is calculated in real time. When the rate of change of the temperature and humidity synergy index exceeds the first threshold, and / or the rate of change of energy in any key frequency band of the airflow disturbance intensity spectrum exceeds the second threshold, it is determined that the environmental coupling feature vector has changed drastically.

[0130] The macroscopic stress state of the yarn obtained by solving the yarn tension dynamics equation is compared with the benchmark macroscopic stress state calculated under stable environmental conditions based on the yarn tension dynamics equation. When the dynamic tension response delay in the macroscopic stress state of the yarn exceeds a preset delay threshold and / or the vibration amplitude exceeds a preset safe range, it is determined that the macroscopic stress state of the yarn has a hysteresis or abnormal response.

[0131] When both of the above conditions are met, the decision coupling relationship of the reinforcement learning agent is unstable.

[0132] As an example, reinforcement learning agents are pre-trained in the following manner:

[0133] A training dataset is constructed based on historical production data, where each training sample contains a sequence of continuous state observations for a historical time period and the corresponding label of the decapitation result.

[0134] Design the training reward function: give a basic positive reward at each step where no yarn breakage occurs, give a negative reward when the environmental coupling feature vector changes drastically and the macroscopic stress state of the yarn shows a lag or abnormal response, and give a large penalty at the step where yarn breakage finally occurs.

[0135] Based on the training dataset, the reinforcement learning agent learns with the goal of maximizing cumulative reward until its policy converges.

[0136] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for dynamically capturing decapitations by combining reinforcement learning and physical modeling, characterized in that, The methods and steps include the following: Step S1: Construct a yarn microstructure evolution model that can simulate the dynamic changes in the friction coefficient of the yarn surface fibers and the cohesion between fibers caused by fluctuations in environmental parameters; wherein, the environmental parameters include temperature, humidity and airflow velocity; Step S2: Input the real-time environmental parameters of the textile workshop into the microstructure evolution model to predict the real-time microstructure parameters of the yarn; perform nonlinear coupling feature extraction on the real-time environmental parameters and calculate the environmental coupling feature vector including the temperature and humidity synergy index and the airflow disturbance intensity spectrum. Step S3: Input the real-time microstructure parameters as material properties into the yarn tension dynamics equation to solve for the predicted macroscopic stress state of the yarn; concatenate the macroscopic stress state of the yarn with the environmental coupling feature vector to form the state observation value of the reinforcement learning agent; Step S4: The reinforcement learning agent outputs the probability value of the yarn breakage risk based on the state observation value; wherein, the reinforcement learning agent learns through training that when the environmental coupling feature vector changes drastically and the macroscopic stress state of the yarn shows a lag or abnormal response, it determines that the coupling relationship is unstable and increases the probability value of the yarn breakage risk. Step S5: When the probability value of decapitation risk exceeds a preset threshold, a decapitation warning signal is generated and output.

2. The method for dynamic capture of decapitation combining reinforcement learning and physical modeling according to claim 1, characterized in that: Nonlinear coupling feature extraction is performed on real-time environmental parameters to calculate environmental coupling feature vectors, including temperature and humidity synergy index and airflow disturbance intensity spectrum, including: Step S21: By analyzing the joint probability distribution of the temperature time series and the humidity time series and their dynamic trajectories in the reconstructed phase space, the joint fluctuation entropy and the average trajectory divergence are calculated respectively, and the two are weighted and fused to generate the temperature and humidity synergy index. Step S22: The time-frequency energy distribution is obtained by performing time-frequency transformation on the airflow velocity sequence, and the energy characteristics of the corresponding key frequency bands are extracted from the time-frequency energy distribution according to the natural frequency range determined by the yarn tension dynamics equation, so as to form the airflow disturbance intensity spectrum.

3. The method for dynamic capture of decapitation combining reinforcement learning and physical modeling according to claim 2, characterized in that: The inherent frequency range is obtained by performing eigenvalue analysis on the yarn tension dynamics equation.

4. The method for dynamic capture of decapitation combining reinforcement learning and physical modeling according to claim 1, characterized in that: When the environmental coupling characteristic vector changes drastically and the macroscopic stress state of the yarn exhibits a lag or abnormal response, the coupling relationship is determined to be unstable, including: The rate of change of the environmental coupling feature vector in adjacent time windows is calculated in real time. When the rate of change of the temperature and humidity synergy index exceeds the first threshold, and / or the rate of change of energy of any key frequency band in the airflow disturbance intensity spectrum exceeds the second threshold, it is determined that the environmental coupling feature vector has changed drastically. The macroscopic stress state of the yarn obtained by solving the yarn tension dynamics equation is compared with the benchmark macroscopic stress state calculated under stable environmental conditions based on the yarn tension dynamics equation. When the dynamic tension response delay in the macroscopic stress state of the yarn exceeds a preset delay threshold and / or the vibration amplitude exceeds a preset safe range, it is determined that the macroscopic stress state of the yarn has a hysteresis or abnormal response. When both of the above conditions are met, the decision coupling relationship of the reinforcement learning agent is unstable.

5. The method for dynamic capture of decapitation combining reinforcement learning and physical modeling according to claim 1, characterized in that: Reinforcement learning agents are pre-trained in the following ways: A training dataset is constructed based on historical production data, where each training sample contains a sequence of continuous state observations for a historical time period and the corresponding label of the decapitation result. Design the training reward function: give a basic positive reward at each step where no yarn breakage occurs, give a negative reward when the environmental coupling feature vector changes drastically and the macroscopic stress state of the yarn shows a lag or abnormal response, and give a large penalty at the step where yarn breakage finally occurs. Based on the training dataset, the reinforcement learning agent learns with the goal of maximizing cumulative reward until its policy converges.

6. A decapitation dynamic capture system combining reinforcement learning and physical modeling, characterized in that: The system includes: The yarn microstructure evolution model construction module is used to: construct a yarn microstructure evolution model that can simulate the dynamic changes in the friction coefficient of the yarn surface fibers and the cohesion between fibers caused by fluctuations in environmental parameters; wherein, environmental parameters include temperature, humidity and airflow velocity; The environmental parameter processing and feature extraction module is used to: input the real-time environmental parameters of the textile workshop into the microstructure evolution model to predict the real-time microstructure parameters of the yarn; perform nonlinear coupling feature extraction on the real-time environmental parameters, and calculate the environmental coupling feature vector including the temperature and humidity synergy index and the airflow disturbance intensity spectrum. The yarn state prediction and fusion module is used to: input the real-time microstructure parameters as material properties into the yarn tension dynamics equation, solve for the predicted macroscopic stress state of the yarn; and concatenate the macroscopic stress state of the yarn with the environmental coupling feature vector to form the state observation value of the reinforcement learning agent. The reinforcement learning agent early warning module is used to: output a breakage risk probability value based on the state observation value; wherein, the reinforcement learning agent learns through training that: when the environmental coupling feature vector changes drastically and the macroscopic stress state of the yarn shows a lag or abnormal response, it determines that the coupling relationship is unstable and increases the breakage risk probability value. The warning signal generation and output module is used to generate and output a decapitation warning signal when the probability value of the decapitation risk exceeds a preset threshold.

7. A decapitation dynamic capture system combining reinforcement learning and physical modeling according to claim 6, characterized in that: The environmental parameter processing and feature extraction module is specifically used for: By analyzing the joint probability distribution of temperature time series and humidity time series and their dynamic trajectories in the reconstructed phase space, the joint fluctuation entropy and the average trajectory divergence are calculated respectively, and the two are weighted and fused to generate the temperature and humidity synergy index. The time-frequency energy distribution is obtained by performing time-frequency transformation on the airflow velocity sequence, and the energy characteristics of the corresponding key frequency bands are extracted from the time-frequency energy distribution based on the natural frequency range determined by the yarn tension dynamics equation to form the airflow disturbance intensity spectrum.

8. The decapitation dynamic capture system combining reinforcement learning and physical modeling according to claim 7, characterized in that: The inherent frequency range is obtained by performing eigenvalue analysis on the yarn tension dynamics equation.

9. A decapitation dynamic capture system combining reinforcement learning and physical modeling according to claim 6, characterized in that: The reinforcement learning agent early warning module is specifically used for: The rate of change of the environmental coupling feature vector in adjacent time windows is calculated in real time. When the rate of change of the temperature and humidity synergy index exceeds the first threshold, and / or the rate of change of energy of any key frequency band in the airflow disturbance intensity spectrum exceeds the second threshold, it is determined that the environmental coupling feature vector has changed drastically. The macroscopic stress state of the yarn obtained by solving the yarn tension dynamics equation is compared with the benchmark macroscopic stress state calculated under stable environmental conditions based on the yarn tension dynamics equation. When the dynamic tension response delay in the macroscopic stress state of the yarn exceeds a preset delay threshold and / or the vibration amplitude exceeds a preset safe range, it is determined that the macroscopic stress state of the yarn has a hysteresis or abnormal response. When both of the above conditions are met, the decision coupling relationship of the reinforcement learning agent is unstable.

10. A decapitation dynamic capture system combining reinforcement learning and physical modeling according to claim 6, characterized in that: Reinforcement learning agents are pre-trained in the following ways: A training dataset is constructed based on historical production data, where each training sample contains a sequence of continuous state observations for a historical time period and the corresponding label of the decapitation result. Design the training reward function: give a basic positive reward at each step where no yarn breakage occurs, give a negative reward when the environmental coupling feature vector changes drastically and the macroscopic stress state of the yarn shows a lag or abnormal response, and give a large penalty at the step where yarn breakage finally occurs. Based on the training dataset, the reinforcement learning agent learns with the goal of maximizing cumulative reward until its policy converges.

Images