Yarn fineness self-learning based online detection method and system

Abstract

This invention provides a self-learning-based online yarn fineness detection method and system, relating to the field of textile technology. This method aims to solve the problems of existing technologies that rely on manually preset parameters, have poor adaptability, and low automation. The method includes: collecting raw fineness information through a dynamic yarn fineness detection device; automatically determining whether the yarn has entered a stable operating state; after stable operation, automatically entering a self-learning phase, collecting and analyzing learning samples to calculate the benchmark feature value and legal threshold range representing the current yarn fineness; after self-learning, switching to online detection mode, comparing the real-time fineness information with the threshold range, and determining an anomaly and sending a control signal when the value exceeds the range. This invention, through a self-learning mechanism, can automatically adapt to different yarns without preset parameters, achieving automated detection and stable, reliable real-time monitoring, thus ensuring product quality.

Description

Technical Field

[0001] This invention relates to the field of textile technology, and more specifically, to a method and system for online detection of yarn fineness based on self-learning. Background Technology

[0002] In the yarn production and processing of the textile industry, the uniformity of yarn fineness is one of the key factors determining the quality of the final fabric. Traditional methods for yarn fineness testing are mostly offline sampling inspections, which involve randomly selecting a portion of yarn samples from the production batch and measuring them using specialized instruments. This method has a significant time lag, making it impossible to monitor the production process in real time and comprehensively. Once a problem occurs, it often results in batch-wide quality defects, and it is also inefficient and labor-intensive.

[0003] To overcome the shortcomings of offline inspection, the industry has developed various online inspection technologies. Patent document CN111691028A discloses a roving frame with a monitoring system for producing roving from yarn slivers. The roving frame has multiple roving production positions, each including a drafting device, a shaft for receiving bobbins, and a spindle for winding the roving onto the bobbins. The roving frame includes a measuring device comprising at least one measuring unit arranged between the conveyor rollers of the drafting device and the spindle at each roving production position. Each measuring unit contains at least one sensor, and the measuring device is configured to continuously measure at least one quality and / or production parameter of the roving at each roving production position via the measuring units. This patent document describes setting up a measuring device consisting of optical sensors and a guiding device on the production line to continuously measure the high-speed moving yarn and compare the real-time measurement values ​​with a preset target value and an allowable deviation range. When the measured value exceeds the preset range, the system triggers an alarm or controls the production equipment to stop operating. However, these technical solutions generally share a common drawback: they heavily rely on operators manually setting a standard fineness value and corresponding tolerance threshold based on the specifications of the yarn to be produced. When it is necessary to change to a different type or batch of yarn, manual setting must be repeated. This mode, which relies on preset parameters, is cumbersome to operate, has poor adaptability, and is prone to inaccurate testing due to human error in the current trend of flexible production with small batches and multiple varieties.

[0004] In addition, how to ensure that high-speed moving yarns can stably pass through narrow detection areas, and how to enable the system to automatically determine the start and stop status of production to achieve unmanned monitoring, are also problems that existing technologies have not been able to effectively solve. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the purpose of this invention is to provide a self-learning-based online yarn fineness detection method and system. This addresses the problems of existing online yarn fineness detection schemes relying on manually preset parameters, having poor adaptability, and low automation. The aim is to achieve an online fineness detection scheme that can automatically adapt to different types and batches of yarn, requires no manual intervention, and has a stable and reliable detection process.

[0006] The present invention provides a self-learning-based online yarn fineness detection method, comprising the following steps:

[0007] Step S1: Non-contact optical signal acquisition of the yarn moving in the production line is performed through the photoelectric detection structure, the optical signal is converted into an electrical signal to obtain the original fineness information, and the original fineness information is sampled and quantized at high speed to form a real-time digital signal data stream;

[0008] Step S2: Analyze the characteristic parameters of the digital signal data stream in real time to automatically determine whether the yarn has entered a continuous and stable operating state;

[0009] Step S3: When it is determined that the yarn has entered a stable operating state, the self-learning process is automatically triggered. The digital signal data stream within a preset time period is collected as a learning sample set. The learning sample set is analyzed by statistical algorithms to obtain the benchmark feature value and fluctuation feature parameter of the normal fineness of the current batch of yarn, and a legal threshold range is automatically generated based on the two.

[0010] Step S4: Continuously collect the real-time digital signal data stream of the yarn, compare the value of each sampling point with the legal threshold range, and if the value of the sampling point exceeds the legal threshold range, it is determined that the yarn fineness is abnormal and the corresponding abnormal handling operation is triggered.

[0011] Preferably, in step S2, the characteristic parameter is the signal amplitude, and the specific judgment logic is as follows:

[0012] A first threshold is preset, which is higher than the background noise level but lower than the normal yarn signal level. When the detected signal value is continuously greater than the first threshold within a preset time, it is determined that the yarn has entered a continuous and stable operating state.

[0013] Preferably, in step S2, the characteristic parameter is signal energy, and the specific judgment logic is as follows:

[0014] After the system is powered on, it first collects background noise signals for a preset duration and calculates the average short-time energy E of the background noise signals. noise Then, the short-time energy E of the real-time digital signal data stream is continuously calculated using a preset sliding time window. realtime , when E realtime continuously exceeding E noiseWhen the preset multiple is reached and the preset duration is maintained, the yarn is determined to enter a continuous and stable operating state.

[0015] Preferably, in step S3, the statistical algorithm includes:

[0016] The arithmetic mean μ of all data points in the learning sample set is calculated as the baseline feature value, and the standard deviation σ is calculated as the fluctuation feature parameter.

[0017] The legal threshold range is set to [μ]. k·σ, μ+k·σ], where k is a preset coefficient.

[0018] Preferably, in the self-learning process of step S3, after collecting the learning sample set, the learning sample set is first subjected to Kalman filtering to filter out signal noise, and then the filtered sample data is analyzed by statistical algorithms to obtain the baseline feature value and fluctuation feature parameter.

[0019] Preferably, the system for implementing the detection method includes a yarn fineness dynamic detection device and a sensor control chip;

[0020] The yarn fineness dynamic detection device is used to realize non-contact optical signal acquisition of yarn, including a guide component, a photoelectric transmitting device, and a photoelectric receiving device; the guide component is used to constrain the movement trajectory of the yarn so that the yarn stably passes through the preset detection area; the photoelectric transmitting device and the photoelectric receiving device are arranged opposite to each other on both sides of the guide component, the photoelectric transmitting device emits a stable and collimated light beam that passes through the detection area, and the photoelectric receiving device receives the light signal after passing through the detection area and converts it into an electrical signal, which is then output to the sensor control chip;

[0021] The sensor control chip, as the core processing unit, integrates a processor, memory, and analog-to-digital converter. It is used to drive the photoelectric transmitting device, sample and quantize the electrical signal output by the photoelectric receiving device, execute the yarn running status judgment, self-learning, online detection and anomaly judgment process, and output anomaly handling control signal.

[0022] Preferably, the guide component is a U-shaped groove;

[0023] The arc-shaped inner wall of the U-shaped groove is used to constrain the radial position of the yarn and prevent the yarn from swinging significantly.

[0024] The yarn to be tested is pulled by the production equipment and passes through the preset testing area from the bottom of the U-shaped groove.

[0025] Preferably, the guiding assembly is a spiral guide tube;

[0026] The inner wall of the spiral guide tube is provided with continuous spiral grooves;

[0027] When the yarn passes through the spiral guide tube, it generates a stable circumferential rotation under the action of the spiral groove, so that the cross-section of the yarn at different angles can be detected.

[0028] Preferably, the photoelectric transmitting device is an infrared light-emitting diode, and the photoelectric receiving device is a photodiode or a phototransistor;

[0029] The sensor control chip, through the light emission control module, precisely controls the light emission intensity and on / off state of the photoelectric transmitting device to ensure the stability of the light source.

[0030] Preferably, it further includes an abnormal execution mechanism, which is electrically connected to the sensor control chip and is used to receive abnormal processing control signals output by the sensor control chip and perform corresponding operations;

[0031] The abnormal execution mechanism includes a motor controller for the yarn conveying equipment and an audible and visual alarm. After receiving a control signal, the motor controller controls the yarn conveying equipment to stop running, and the audible and visual alarm issues an alarm prompt after receiving a control signal.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] 1. This invention, through a self-learning mechanism, can automatically learn the normal fineness characteristics of the currently produced yarn and establish a detection benchmark. No operator needs to pre-set any parameters, which improves the adaptability to different types and batches of yarn and reduces the threshold for use and the complexity of operation.

[0034] 2. This invention can automatically determine the running status of the yarn, realize unmanned start and stop of the detection process, reduce manual intervention, and improve production efficiency and the intelligence level of the system.

[0035] 3. By setting specific guiding components, this invention can effectively constrain the trajectory of yarns moving at high speeds, ensuring that they stably pass through the optimal detection area, thereby improving the stability and reliability of signal acquisition and providing a high-quality data source for subsequent accurate judgment.

[0036] 4. This invention can monitor the yarn fineness online in real time for 100% of the time during the production process. Once defects such as thick or thin spots that exceed the normal range are detected, it can immediately respond and trigger control actions such as stopping the machine, thereby effectively intercepting defects and ensuring product quality from the source. Attached Figure Description

[0037] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0038] Figure 1This is a schematic diagram of the structure of a yarn fineness dynamic detection device provided in an embodiment of the present invention;

[0039] Figure 2 A hardware system functional module block diagram provided in an embodiment of the present invention;

[0040] Figure 3 A schematic diagram of the basic process of an online yarn fineness detection method based on self-learning provided in an embodiment of the present invention;

[0041] Figure 4 This is a detailed flowchart of the self-learning method in Embodiment 2 of the present invention. Detailed Implementation

[0042] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0043] Example 1

[0044] This invention provides a basic implementation scheme for a self-learning-based online yarn fineness detection method and system. This scheme aims to achieve fully automatic and adaptive fineness anomaly detection of yarns moving on a production line without requiring manual pre-setting of any yarn specification-related parameters.

[0045] Reference Figure 1 and Figure 2 The system involved in this embodiment includes a yarn fineness dynamic detection device and a sensor control chip as the core processing unit. The yarn fineness dynamic detection device is the physical basis for non-contact measurement, and its structure includes a guide component for guiding the yarn's movement trajectory. In this embodiment, the guide component is specifically a U-shaped groove. A photoelectric transmitting device and a photoelectric receiving device are arranged opposite each other on both sides of the opening of the U-shaped groove. Specifically, in actual operation, the yarn to be detected passes through the bottom of the U-shaped groove at high speed under the traction of production equipment (such as a winding machine). The arc-shaped inner wall of the U-shaped groove effectively constrains the radial position of the yarn, preventing it from swaying significantly due to vibration or airflow, thereby ensuring that the yarn can stably pass through the preset detection area between the photoelectric transmitting device and the photoelectric receiving device.

[0046] The photoelectric transmitting device can be an infrared light-emitting diode, which, driven by a sensor control chip, emits a stable and collimated beam of light that passes laterally through the opening of the U-shaped groove. Correspondingly, the photoelectric receiving device can be a photodiode or phototransistor, located on the other side of the beam path, to receive the light signal after it passes through the detection area. When no yarn passes through, the photoelectric receiving device receives the full light intensity; when a yarn passes through, it blocks part of the beam, causing the light intensity received by the photoelectric receiving device to decrease. The thicker the yarn diameter, the more light is blocked, and the weaker the received light intensity; conversely, the thinner the yarn, the stronger the received light intensity. Therefore, the amplitude of the electrical signal (e.g., voltage or current) output by the photoelectric receiving device is inversely proportional to the real-time fineness of the yarn. This time-varying electrical signal is what this invention refers to as the original fineness information.

[0047] Reference Figure 2 The entire hardware system operates under the unified coordination of a sensor control chip. This sensor control chip can be a microcontroller or a digital signal processor, integrating a processor, memory, and necessary peripherals such as an analog-to-digital converter. The sensor control chip precisely controls the luminous intensity and switching state of the light-emitting module (i.e., the photoelectric transmitting device) via the light-emitting control module to ensure the stability of the light source. Simultaneously, the sensor control chip utilizes its built-in analog-to-digital converter to sample and quantize the analog electrical signal output from the receiving module (i.e., the photoelectric receiving device) at a fixed sampling frequency (e.g., 1000 times per second, or 1 kHz), converting it into a series of discrete digital values ​​to obtain a raw, detailed information data stream suitable for subsequent software processing.

[0048] Based on the above hardware system, the method flow of this embodiment is as follows: Figure 3 As shown, the specific steps include:

[0049] Step S1: Sample yarn fineness information. After the system is powered on, the sensor control chip continuously samples the output signal of the receiving module at high speed, forming a real-time data stream. When no yarn passes through the detection area, the collected signal value fluctuates at a low level, which mainly reflects the background noise of the circuit.

[0050] Step S2: Automatic detection of yarn running status. This step aims to enable the system to automatically determine whether production has started, i.e., whether the yarn has entered and begun to stably pass through the detection area. As a specific implementation, this embodiment adopts a judgment method based on signal amplitude analysis. The system software monitors the collected raw fineness information in real time. In the initial silent state, the signal value is usually very low, for example, in a 12-bit analog-to-digital converter system, its value may be between 50 and 100. When the yarn enters the detection area, due to the light being blocked, the signal value will undergo a significant jump, for example, jumping to the range of 2000 to 3000, and will continue to fluctuate at this high level due to the inherent slight unevenness of the yarn itself.

[0051] To ensure the reliability of the judgment, a first threshold is preset within the system. This threshold is significantly higher than the background noise level but significantly lower than the level of normal yarn signals; for example, it can be set to 500. The system continuously analyzes the fluctuation amplitude of the raw fineness information. When the signal value is detected to be consistently greater than the first threshold for a preset time period (e.g., 1 consecutive second), the system determines that the yarn has entered a continuous and stable operating state. This judgment method effectively distinguishes the yarn's entry into a stable operating state from occasional noise spikes, ensuring that the triggering of subsequent self-learning steps is based on a real and stable production state.

[0052] Step S3: Perform automatic learning of yarn feature values. Once the system determines in step S2 that the yarn has entered a stable operating state, it will automatically trigger the self-learning phase. The purpose of this phase is to automatically learn the normal fineness characteristics of the currently produced batch of yarn without any prior knowledge. The system will start an internal timer to continuously collect raw fineness information for a preset duration (e.g., 10 seconds) and use all data points collected during this period (a total of 10,000 data points at a 1kHz sampling rate) as a learning sample set.

[0053] After the training samples are collected, the system will analyze the sample set based on a preset statistical algorithm. Specifically, the system will calculate the arithmetic mean of the 10,000 data points. and standard deviation Understandably, this arithmetic mean... The standard deviation is a benchmark characteristic value that characterizes the fineness of the normal yarn in the current batch, reflecting the statistically average fineness level of the yarn. This quantifies the inherent and acceptable degree of fineness fluctuation in normal yarn during the production process.

[0054] Subsequently, the system used the calculated and This automatically sets a legal threshold range that allows for fluctuations. In this embodiment, this range is set to... ,in This is a preset coefficient used to control the width of the threshold range, i.e., the detection sensitivity. According to the three sigma criterion in statistics, normal data points have a 99.7% probability of falling within the threshold range. Within the range. Therefore, as a preferred setting, it can be... Set it to 3. This specifically defines the legal threshold range as follows: At this point, the learning phase has ended, and the system has established personalized testing standards for the current batch of yarn.

[0055] Step S4: Perform online yarn fineness detection and judgment. After the self-learning phase, the system seamlessly and automatically switches to online fineness detection mode. In this mode, the system continues to collect raw fineness information in real time at a frequency of 1kHz. For each newly collected data point, the system compares its value with the legal threshold range established in step S3. Compare them.

[0056] If the value of the current sampling point is within this range, the yarn fineness is determined to be normal, the system takes no action, and continues monitoring the next sampling point. If the value of the current sampling point exceeds this range, i.e., is less than... (This indicates the presence of abnormally thick sections, with more light blocking and lower signal values) or greater than (Indicating abnormal details, less light blocking, and higher signal values), the system immediately determines that the yarn fineness is abnormal.

[0057] Once an anomaly is detected, the system immediately generates and sends a control signal. In a specific application, this control signal can be a high-level pulse or a specific serial communication command, sent to the yarn conveying equipment connected to the system (such as the motor controller of a winding machine). Upon receiving this control signal, the yarn conveying equipment immediately stops operation, thereby cutting off the defective yarn and preventing it from continuing to wind or enter the next process. Simultaneously, the system can also trigger an audible and visual alarm to alert on-site operators to take appropriate action.

[0058] In summary, the technical solution of this embodiment achieves complete self-adaptation and automation. When changing to different specifications of yarn, operators do not need to make any settings; they only need to start production normally, and the system will automatically complete the learning of the new yarn and the establishment of testing standards, thereby greatly improving production flexibility and testing reliability.

[0059] Example 2

[0060] This embodiment, based on Embodiment 1, further proposes an optimized scheme for the self-learning algorithm, aiming to improve the accuracy and anti-interference capability of the established benchmark feature values ​​and legal threshold ranges. The hardware structure and overall method flow of this embodiment are basically the same as those of Embodiment 1; the core difference lies in the improvement of the internal implementation method of step S3, i.e., automatically learning yarn feature values.

[0061] Reference Figure 4 This demonstrates the internal flow of the optimized self-learning steps in this embodiment.

[0062] Once the system determines that the yarn has entered a stable operating state and triggers self-learning, it first executes step S301 to receive the original signal. Similar to Example 1, the system collects the original fineness information within a preset time period (e.g., 10 seconds) to form a learning sample set.

[0063] Subsequently, before calculating the statistical parameters, a crucial preprocessing step S302 is added: performing Kalman filtering. A Kalman filter is a highly efficient recursive filter capable of estimating the optimal state of a dynamic system from a series of noisy observations. In the technical scenario of this invention, the actual fineness of the yarn can be considered as the system state, while the raw fineness information acquired by the sensor represents observations with measurement noise. The acquired learning samples (a time series of data) are input into a pre-configured Kalman filter. The filter generates the optimal estimate of the state at the current moment based on the state estimate from the previous moment and the observation at the current moment. This process is repeated recursively until the entire learning sample sequence has been processed.

[0064] After Kalman filtering, a smoother signal sequence that more closely resembles the true variation in fineness is obtained. High-frequency random noise, glitches, and abrupt changes caused by minute vibrations in the original signal are effectively filtered out.

[0065] Next, in step S303, the mean of the filtered signal is calculated. and standard deviation It should be noted that at this point, the system is not based on the original learning sample data, but rather on the smoothed signal sequence obtained after filtering in step S302 to calculate the arithmetic mean and standard deviation. Since most of the noise has been filtered out, the calculated mean... It can more accurately represent the average fineness benchmark of yarn, while the standard deviation This more accurately reflects the inherent volatility of the yarn itself, rather than the volatility introduced by noise.

[0066] Finally, in step S304, the system sets the threshold range. Similar to Example 1, the system sets the threshold range based on the newly calculated... and To set the legal threshold range, i.e. ,in Similarly, it is a preset coefficient (e.g.) ).

[0067] After completing the aforementioned optimized self-learning steps, the system switches to online detail detection mode. During online comparison, the system still compares the real-time acquired, unfiltered raw detail information with this more accurate legal threshold range established based on filtered data.

[0068] The beneficial effect of this embodiment is that by introducing Kalman filtering during the self-learning phase, the quality of benchmark establishment is significantly improved. This is because the benchmark eigenvalues... and standard deviation The calculation eliminates most of the noise interference, and the generated legal threshold range can more closely and accurately encompass the fluctuation range of normal yarn. The direct benefit of this is that, during online detection, it can effectively reduce the probability of misjudgment caused by normal signal jitter or random noise spikes, that is, reduce the false alarm rate of the system, thereby improving the stability and reliability of the entire detection system.

[0069] Example 3

[0070] This embodiment proposes a more robust optimization scheme for the judgment logic of automatically detecting the yarn running status (step S2) in Embodiment 1. In some industrial environments, electromagnetic interference may be severe, resulting in high background noise levels or occasional large-amplitude interference pulses, which may cause misjudgments by the judgment method based on simple amplitude thresholds (as described in Embodiment 1). This embodiment adopts a method based on signal energy analysis to improve the accuracy and anti-interference capability of the running status judgment.

[0071] The hardware structure of this embodiment is the same as that of Embodiment 1. The main difference in the method flow is that a preparatory step for calculating the background noise energy is added, and the specific implementation algorithm of step S2 is replaced.

[0072] Specifically, after the system is powered on, before starting the formal testing, a preliminary step is performed: acquiring and calculating the background noise energy. At this point, it is assumed that production has not yet started and no yarn is passing through the detection device. The system acquires a signal for a period of time (e.g., preset to 2 seconds), representing the pure circuit and environmental background noise in the current environment. Then, the system divides this background noise signal into frames with a short sliding time window (e.g., 100 milliseconds, corresponding to 100 sampling points) and calculates the short-time energy of each frame. It should be noted that one frame of signal... The short-time energy can be defined as the sum of the squares of the amplitudes of all sampling points within the frame, i.e. Finally, the system calculates the average short-time energy of all noisy frames, denoted as . .this The value then serves as the benchmark for subsequent judgments.

[0073] After this preparatory step is completed, the system enters standby mode and begins executing the optimized step S200, namely, the energy-based operation status judgment. The system continues to calculate the short-time energy of the raw fineness information acquired in real time within the same sliding time window (100 milliseconds), denoted as... .

[0074] The system will use this real-time short-term energy With pre-calculated background noise energy Continuous comparison is performed. When the yarn enters the detection area, the signal amplitude increases overall, resulting in a significant and continuous increase in signal energy. The specific judgment logic is as follows: when the system detects real-time short-term energy... Continuously exceeding the background noise energy The preset multiplier (e.g., 5 times, i.e.) Only when this state lasts for a preset time (e.g., 0.5 seconds) can it be finally determined that the yarn has entered a stable operating state.

[0075] This method, based on continuous energy growth, offers stronger anti-interference capabilities compared to a simple amplitude threshold method. While a single noise spike might momentarily cause the signal amplitude to exceed the threshold, its energy contribution within a short time window is limited and its duration is extremely short, thus it will not be misinterpreted as yarn intrusion. Only when the signal energy experiences a sustained, order-of-magnitude jump will the system confirm a change in operating state.

[0076] Once it is determined that the yarn has entered a stable operating state, the subsequent self-learning step S3 and online detection step S4 can be executed using the method described in Example 1 or Example 2.

[0077] Furthermore, the energy analysis method used in this embodiment can also be used to determine yarn stopping or breakage events. During online detection, the system can continue monitoring. If found From high levels back to near The level (e.g., less than for a continuous period of time) The system can then determine that the yarn has stopped running and can automatically return to the initial standby state to prepare for the next running status detection, thereby realizing fully automated closed-loop control of start and stop.

[0078] Example 4

[0079] This embodiment demonstrates a scheme that achieves deep collaborative optimization of hardware structure and software algorithm, aiming to fundamentally improve the accuracy of detection and the reliability of operational status judgment.

[0080] First, at the hardware structure level, this embodiment features a specially designed guide component in the yarn fineness dynamic detection device. Instead of using the U-shaped groove from Embodiment 1, this embodiment employs a specially machined spiral guide tube. One or more continuous, shallow spiral grooves are machined on the inner wall of this spiral guide tube. When the yarn passes through the guide tube at high speed under axial tension, the yarn surface makes slight contact with the sidewall of the spiral groove. Due to the helix angle of the spiral groove, this contact applies a stable tangential force component to the yarn, causing the yarn to generate a stable circumferential rotation related to its axial velocity while undergoing high-speed axial movement. This design adds a detectable and regular rotational motion to the yarn.

[0081] Correspondingly, at the software algorithm level, the method of this embodiment makes full use of the rotational characteristics introduced by the hardware design and adopts fundamental algorithmic innovations for the two core steps of running state judgment (step S2) and self-learning (step S3).

[0082] For step S2, namely the automatic detection of yarn movement, this embodiment employs a frequency domain analysis-based method. The system processes the real-time acquired raw fineness information data stream using a short-time Fourier transform, that is, continuously performing a fast Fourier transform on the signal with a sliding window to obtain the dynamic spectrum of the signal. When there is no yarn movement or the yarn is stationary, the signal is mainly random noise, and its spectrum exhibits a roughly flat characteristic over a wide frequency range. However, when the yarn begins to move steadily and rotate under the guidance of the helical guide tube, due to the possibility that the yarn cross-section itself is not perfectly circular (e.g., slightly elliptical) or has periodic textures on its surface, its obstruction of the light beam during rotation will produce a periodic change. This periodic signal change is reflected in the spectrum, forming a sharp and stable characteristic frequency peak. The frequency of this characteristic frequency peak... It directly corresponds to the rotational speed of the yarn.

[0083] Therefore, the operating status judgment logic in this embodiment is transformed into: determining whether the yarn has entered a stable operating state by detecting whether there is a significant characteristic frequency peak with a high signal-to-noise ratio in the frequency spectrum. Compared with the amplitude method or energy method, this detection method based on deterministic frequency characteristics generated by a specific physical phenomenon (rotation) is almost completely unaffected by broadband noise or random interference, and has high accuracy and reliability.

[0084] For step S3, namely the automatic learning of yarn characteristic values, this embodiment also utilizes the yarn rotation information to optimize the calculation of the reference value. This is achieved by confirming stable yarn operation and identifying characteristic frequencies through frequency domain analysis. Then, the system can calculate the time required for the yarn to rotate once, i.e., the rotation period. .

[0085] After collecting training samples (e.g., 10 seconds of data), when calculating the benchmark characteristic value representing the average fineness of the yarn, the system no longer simply calculates the arithmetic mean of all sampling points, but instead uses a calculation method that better reflects the average cross-sectional characteristics of the yarn. A preferred implementation is to calculate the benchmark characteristic value representing the average fineness of the yarn over one or more complete rotation cycles (…). The system integrates or averages the signal over a given period. For example, the system can average the entire sequence of learning samples. Since the learning time (10 seconds) is much longer than a single rotation cycle (usually in the millisecond range), this average is actually equivalent to averaging the signal over thousands of complete rotation cycles.

[0086] Understandably, the physical significance of this averaging method based on rotation cycles lies in its ability to eliminate measurement bias caused by the irregular (non-circular) shape of the yarn cross-section. During one rotation of the yarn, its diameter at different angles is measured. By averaging the signal over the entire cycle, the result more accurately represents the equivalent average circular cross-sectional diameter of the yarn, thus obtaining a more precise and fair benchmark characteristic value. The valid threshold range calculated based on this more accurate benchmark characteristic value is naturally better suited for subsequent precise detection tasks.

[0087] In this embodiment, by improving the hardware aspect of the spiral guide tube, a new and reliable physical information dimension (i.e., rotation frequency) is introduced into the software algorithm, and a matching frequency domain analysis algorithm is designed, thereby achieving deep coupling and synergistic effect between hardware and software. This design not only provides a highly robust method for judging the operating status, but also significantly improves the measurement accuracy of the fineness benchmark, thereby comprehensively improving the performance of the entire online detection system.

[0088] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, making the system and its various devices, modules, and units of this invention function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided by this invention can be considered as a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0089] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A self-learning-based online yarn fineness detection method, characterized in that, Includes the following steps: Step S1: Non-contact optical signal acquisition of the yarn moving in the production line is performed through the photoelectric detection structure, the optical signal is converted into an electrical signal to obtain the original fineness information, and the original fineness information is sampled and quantized at high speed to form a real-time digital signal data stream; Step S2: Analyze the characteristic parameters of the digital signal data stream in real time to automatically determine whether the yarn has entered a continuous and stable operating state; Step S3: When it is determined that the yarn has entered a stable operating state, the self-learning process is automatically triggered. The digital signal data stream within a preset time period is collected as a learning sample set. The learning sample set is analyzed by statistical algorithms to obtain the benchmark feature value and fluctuation feature parameter of the normal fineness of the current batch of yarn, and a legal threshold range is automatically generated based on the two. Step S4: Continuously collect the real-time digital signal data stream of the yarn, compare the value of each sampling point with the legal threshold range, and if the value of the sampling point exceeds the legal threshold range, it is determined that the yarn fineness is abnormal and the corresponding abnormal handling operation is triggered. The characteristic parameter is the signal amplitude, and the specific judgment logic is as follows: A first threshold is preset, which is higher than the background noise level and lower than the normal yarn signal level. When the detected signal value is continuously greater than the first threshold within a preset time, it is determined that the yarn has entered a continuous and stable operating state. Alternatively, the characteristic parameter may be signal energy, and the specific judgment logic is as follows: After the system is powered on, it first collects background noise signals for a preset duration and calculates the average short-time energy E of the background noise signals. noise Then, the short-time energy E of the real-time digital signal data stream is continuously calculated using a preset sliding time window. realtime , when E realtime continuously exceeding E noise When the preset multiple is reached and the preset duration is maintained, the yarn is determined to enter a continuous and stable operating state. In step S3, the statistical algorithm includes: The arithmetic mean μ of all data points in the learning sample set is calculated as the baseline feature value, and the standard deviation σ is calculated as the fluctuation feature parameter. The legal threshold range is set to [μ]. k·σ, μ+k·σ], where k is a preset coefficient.

2. The online yarn fineness detection method based on self-learning according to claim 1, characterized in that, In the self-learning process of step S3, after collecting the learning sample set, the learning sample set is first processed by Kalman filtering to remove signal noise, and then the filtered sample data is analyzed by statistical algorithms to obtain the baseline feature value and fluctuation feature parameter.

3. A self-learning-based online yarn fineness detection system, characterized in that, The system for implementing the detection method according to any one of claims 1-2 includes a yarn fineness dynamic detection device and a sensor control chip; The yarn fineness dynamic detection device is used to realize non-contact optical signal acquisition of yarn, including a guide component, a photoelectric transmitting device, and a photoelectric receiving device; the guide component is used to constrain the movement trajectory of the yarn so that the yarn stably passes through the preset detection area; the photoelectric transmitting device and the photoelectric receiving device are arranged opposite to each other on both sides of the guide component, the photoelectric transmitting device emits a stable and collimated light beam that passes through the detection area, and the photoelectric receiving device receives the light signal after passing through the detection area and converts it into an electrical signal, which is then output to the sensor control chip; The sensor control chip, as the core processing unit, integrates a processor, memory, and analog-to-digital converter. It is used to drive the photoelectric transmitting device, sample and quantize the electrical signal output by the photoelectric receiving device, execute the yarn running status judgment, self-learning, online detection and anomaly judgment process, and output anomaly handling control signal.

4. The online yarn fineness detection system based on self-learning according to claim 3, characterized in that, The guide component is a U-shaped groove; The arc-shaped inner wall of the U-shaped groove is used to constrain the radial position of the yarn and prevent the yarn from swinging significantly. The yarn to be tested is pulled by the production equipment and passes through the preset testing area from the bottom of the U-shaped groove.

5. The online yarn fineness detection system based on self-learning according to claim 3, characterized in that, The guiding component is a spiral guide tube; The inner wall of the spiral guide tube is provided with continuous spiral grooves; When the yarn passes through the spiral guide tube, it generates a stable circumferential rotation under the action of the spiral groove, so that the cross-section of the yarn at different angles can be detected.

6. The online yarn fineness detection system based on self-learning according to claim 3, characterized in that, The photoelectric transmitting device is an infrared light-emitting diode, and the photoelectric receiving device is a photodiode or a phototransistor. The sensor control chip, through the light emission control module, precisely controls the light emission intensity and on / off state of the photoelectric transmitting device to ensure the stability of the light source.

7. The online yarn fineness detection system based on self-learning according to claim 3, characterized in that, It also includes an abnormal execution mechanism, which is electrically connected to the sensor control chip and is used to receive abnormal processing control signals output by the sensor control chip and perform corresponding operations; The abnormal execution mechanism includes a motor controller for the yarn conveying equipment and an audible and visual alarm. After receiving a control signal, the motor controller controls the yarn conveying equipment to stop running, and the audible and visual alarm issues an alarm prompt after receiving a control signal.

Images