Textile yarn production monitoring, regulating and controlling method

Abstract

The invention relates to the technical field of yarn production monitoring, in particular to a spinning yarn production monitoring, regulating and controlling method, which comprises the following steps of: arranging a plurality of vibration acquisition points, acquiring vibration signals in the running process of a roller, and sequentially performing detrending, filtering, wavelet decomposition, demodulation processing and correlation matching analysis to obtain a plurality of vibration signals; and a weak fault characteristic signal covered by noise is effectively extracted. Furthermore, accurate recognition and grading of the roller abrasion degree are achieved by setting an abrasion grade threshold value interval. And regulation and control instructions such as early warning, speed reduction or shutdown are automatically triggered according to the wear level, and intelligent maintenance and management of equipment are achieved. The technical problems that due to the fact that workshop environment noise is large and vibration signals are mixed, roller abrasion is difficult to recognize in the early stage, and abrasion degree evaluation is inaccurate are solved.

Description

Technical Field

[0001] This invention relates to the field of yarn production monitoring, specifically a method for monitoring and controlling textile yarn production. Background Technology

[0002] The unique environment of textile yarn production workshops constitutes a core technological bottleneck for the effective extraction of vibration signals. The superimposed noise generated by multiple pieces of equipment operating simultaneously in the workshop can reach 85-95 dB, creating a harsh monitoring environment. High noise directly leads to an extremely low signal-to-noise ratio for vibration sensors, and weak characteristic signals such as micron-level vibration fluctuations caused by roller wear are easily masked by equipment operating noise and environmental noise.

[0003] Existing traditional vibration signal filtering algorithms have significant adaptation limitations and cannot meet the complex monitoring needs of textile workshops. These algorithms (such as mean filtering and Fourier domain filtering) are mostly suitable for scenarios with simple noise characteristics and clear separation of signal and noise spectra, but lack the ability to handle complex mixed noise in workshops. They can only achieve basic noise suppression and cannot accurately distinguish between normal equipment vibration, environmental interference noise, and fault characteristic signals. Furthermore, they are unable to separate weak fault features directly related to yarn evenness from strongly coupled mixed signals, making it difficult to detect early fault signs in a timely manner. Summary of the Invention

[0004] (1) Technical problems to be solved

[0005] The purpose of this invention is to solve the technical problems in existing textile yarn production, such as the difficulty in early identification of roller wear and inaccurate assessment of wear degree due to high noise levels and mixed vibration signals in the workshop environment.

[0006] (2) Technical solution

[0007] To achieve the above objectives, in one aspect, the present invention provides a method for monitoring and controlling textile yarn production, the method comprising:

[0008] S1: Multiple acquisition points are deployed along the vibration transmission path of the roller, and the first vibration signal is collected synchronously through vibration sensors; the first vibration signal includes the vibration of normal equipment operation, environmental interference noise, and vibration fluctuation fault characteristics;

[0009] S2: Perform detrending processing on the first vibration signal to eliminate baseline drift caused by sensor installation error; remove power grid frequency and harmonic interference to obtain the second vibration signal; perform multi-scale decomposition on the first vibration signal to obtain the target feature range;

[0010] S3: Based on the target feature range, extract the high-frequency components of the second vibration signal and demodulate them to amplify the low-frequency fault features that were originally masked by noise.

[0011] S4: Using low-frequency fault characteristics as a benchmark template, perform correlation matching analysis on fault characteristic signals from multiple acquisition points; retain signal components in each acquisition point whose correlation with the benchmark template is higher than the first preset threshold, and remove environmental interference noise whose correlation is lower than the first preset threshold to obtain the target fault characteristic signal, thereby achieving accurate purification and enhancement of the fault characteristic signal;

[0012] S5: Extract key parameters of the target fault feature signal, and preset threshold ranges for feature parameters corresponding to multiple roller wear levels; compare the extracted key parameters with the preset threshold ranges to determine the current wear severity level of the roller; automatically trigger corresponding control commands according to preset control rules corresponding to different wear levels.

[0013] Furthermore, the plurality of acquisition points include at least one first acquisition point disposed on the roller bearing housing, at least one second acquisition point disposed on the transmission gearbox, and at least one third acquisition point disposed on the adjacent roller connection support structure.

[0014] Further, the detrending processing of the first vibration signal includes:

[0015] Detrending is achieved by removing the polynomial trend term from the signal:

[0016] For the first vibration signal collected Fit an nth-order polynomial: ;

[0017] The signal obtained after detrending: .

[0018] Furthermore, the multi-scale decomposition of the first vibration signal includes:

[0019] Wavelet transform is used to transform the second vibration signal Decomposed into: ;

[0020] in, Here, is the wavelet mother function, a is the scaling factor, and b is the translation factor;

[0021] The target feature interval is determined based on the scale corresponding to the wear feature frequency.

[0022] Furthermore, the demodulation processing of the second vibration signal includes:

[0023] For the extracted high-frequency component signals Perform Hilbert transform to obtain ;

[0024] Constructing analytic signals The amplitude of the analytical signal This refers to the low-frequency fault characteristic signal that has been demodulated and amplified.

[0025] Furthermore, the correlation matching analysis includes:

[0026] Quantification is performed using cross-correlation coefficients. The calculation formula is:

[0027] ;

[0028] Where x is the reference template signal sequence, and y is the acquisition point signal sequence to be matched. and Let N be the mean of each, and N be the signal length.

[0029] Furthermore, the preset control rules include: when it is determined to be slight wear, triggering an early warning prompt; when it is determined to be moderate wear, triggering the equipment to reduce speed and prompting a planned maintenance; when it is determined to be severe wear, triggering the equipment to stop and issuing an emergency maintenance alarm.

[0030] (3) Beneficial effects

[0031] Compared with the prior art, the beneficial effects of the present invention are:

[0032] Through multi-scale decomposition and demodulation processing, low-frequency wear characteristics masked by strong noise can be effectively amplified, significantly improving the signal-to-noise ratio and enabling the detection of early and subtle faults.

[0033] By arranging multiple acquisition points along the vibration transmission path and purifying the signal based on correlation analysis, environmental and equipment interference can be effectively eliminated, thereby enhancing the reliability and consistency of fault signals. Attached Figure Description

[0034] Figure 1 This is a flowchart of the textile yarn production monitoring and control method of Embodiment 1 of the present invention. Detailed Implementation

[0035] 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.

[0036] Before providing examples, it is necessary to describe the application scenarios of this invention. This invention is mainly applied to the real-time status monitoring and control of rollers in the textile yarn production process. As the core component for yarn drafting and conveying, the wear state of the roller directly affects the yarn uniformity and production efficiency. Textile workshops are characterized by high noise levels and numerous vibration sources, making traditional vibration analysis methods ineffective. This invention achieves accurate identification and automatic response to roller wear through multi-level processing and intelligent analysis of vibration signals, and is applicable to various spinning equipment such as ring spinning, rotor spinning, and air-jet spinning.

[0037] Example 1: As Figure 1 As shown in the figure, this embodiment provides a method for monitoring and controlling textile yarn production, the method comprising:

[0038] Step S1: Synchronous acquisition of vibration signals from multiple acquisition points:

[0039] Vibration sensors are installed on the roller bearing housing, transmission gearbox, and adjacent roller connecting support structure, with a sampling frequency set to 10kHz. Vibration signals at each point are collected synchronously to form the first vibration signal set. The acquisition time is no less than 30 seconds to cover multiple operating cycles of the equipment.

[0040] Step S2: Signal preprocessing and multi-scale decomposition:

[0041] The first vibration signal at each acquisition point was detrended, and a third-order polynomial fitting was used to remove baseline drift. Subsequently, power frequency harmonic notch filters of 50Hz, 100Hz, and 150Hz were designed to eliminate power grid interference and obtain the second vibration signal.

[0042] The second vibration signal was decomposed into five layers of wavelet packets using the Db4 wavelet basis. Based on the characteristic frequency range of roller wear (usually between 200-800Hz), the sub-band signal at the corresponding scale was selected as the target characteristic interval.

[0043] Step S3: High-frequency demodulation and amplification fault characteristics:

[0044] High-frequency components (typically above 1000Hz) are extracted from the target characteristic range, subjected to Hilbert transform, and an analytic signal is constructed. The amplitude envelope is then extracted to obtain the demodulated low-frequency fault characteristic signal. This process can amplify the low-frequency impulse characteristics that were originally buried by noise by 10-20dB.

[0045] Step S4: Multi-channel correlation matching and feature extraction:

[0046] Using the demodulated fault characteristic signal from a specific acquisition point (usually the bearing housing) as a reference template, calculate the cross-correlation coefficients between the signals at other points and the reference template. Set a threshold. The signal segments with correlation coefficients higher than the threshold are retained, while the interference segments with correlation coefficients lower than the threshold are removed, and the signals are fused to generate the target fault characteristic signal.

[0047] Step S5: Wear Degree Determination and Automatic Adjustment:

[0048] Key parameters such as peak factor, impulse index, and margin factor of the target fault characteristic signal are extracted. Based on historical data and experimental calibration, parameter threshold ranges corresponding to mild, moderate, and severe wear are set.

[0049] Real-time comparison of current parameters with thresholds: If in the mild range, the system issues a yellow warning; if in the moderate range, the system controls the equipment to reduce speed by 20% and generates a maintenance work order; if in the severe range, it immediately triggers a shutdown and sends an alarm message to the maintenance personnel's mobile phone.

[0050] Specifically, the specific implementation of this invention is based on the actual production scenario of a ring spinning workshop in a large textile enterprise. This workshop covers an area of ​​8,000 square meters and is equipped with 320 JWF1568 ring spinning machines. Each machine has four core roller components: a front roller, a middle roller, a rear roller, and a feed roller, which are responsible for the critical process of yarn drafting and forming. During daily production, multiple machines operate simultaneously at high speed, resulting in an average environmental noise level of up to 92 dB caused by airflow and mechanical friction. The noise frequency covers a wide range from 50 Hz to 10 kHz. Simultaneously, the workshop temperature is controlled at 22 ± 3 °C and the relative humidity is maintained at 55 ± 5%, meeting the basic environmental requirements for textile production. However, this complex environment with high noise and multiple interferences presents a significant challenge to the acquisition and identification of weak vibration signals generated by roller wear. In addition, the workshop power grid supply voltage is 380V±10%, and the interference of the 50Hz power frequency and its 2nd to 5th harmonics (100Hz, 150Hz, 200Hz, 250Hz) will be directly superimposed on the collected vibration signal. The operating speed of the roller is dynamically adjusted between 600-1200r / min according to different spinning processes. The characteristic frequency of roller wear will also change at different speeds. These factors together constitute the core environmental background for the implementation of this invention.

[0051] To ensure the effective implementation of the monitoring and control methods, the experiment selected a PCB 352C65 piezoelectric accelerometer as the core device for vibration signal acquisition. This sensor has a measurement range of ±50g, a sensitivity of 100mV / g, and a frequency response covering 2Hz-10kHz. It can accurately capture micron-level vibration fluctuations caused by roller wear. Its anti-interference capability and dynamic response speed are perfectly suited to the complex environment of the workshop. During installation, a dual fixing method combining magnetic suction and threaded fastening is used to minimize signal distortion caused by sensor installation errors. The data acquisition stage uses an NI cDAQ-9178 chassis paired with an NI9234 data acquisition card, with a single-channel sampling rate set to 10kHz. This satisfies the requirements of the Nyquist sampling theorem while balancing data storage and processing efficiency. The acquisition system supports 16-channel synchronous acquisition with a synchronization error of less than 1μs between channels, enabling precise synchronous capture of vibration signals from multiple acquisition points. Real-time data transmission and storage are achieved through dual USB 3.0 and Ethernet interfaces, ensuring no loss of the original signal. The signal processing and control module adopts an FPGA+ARM heterogeneous computing architecture. The Xilinx Zynq-7020 FPGA is responsible for high-speed signal preprocessing, such as detrending and filtering operations, while the ARM Cortex-A9 chip focuses on complex algorithm operations such as wavelet decomposition, demodulation, and correlation analysis. The 8GB DDR3 memory and 64GB eMMC flash memory can meet the storage needs of a large amount of raw signals and processing results. It communicates with the spinning machine PLC through RS485 and CAN bus to complete the issuance of control commands. It also supports 4G and WiFi communication for remote monitoring and data uploading.

[0052] To ensure experimental accuracy, a PCB 480C02 sensor calibrator was also provided to ensure that the sensor sensitivity error was controlled within ±1%. An AWA5636 noise tester was used to monitor changes in ambient noise in real time. A customized roller wear simulation device could accurately control the amount of roller wear and generate experimental samples with different wear levels. A 10-inch industrial touch screen served as a data visualization terminal, displaying monitoring data, wear levels, and control status in real time, providing intuitive reference for on-site operators.

[0053] During the vibration signal acquisition phase (corresponding to step S1), the deployment of acquisition points strictly adheres to the core principles of full coverage of the vibration transmission path, no omission of fault characteristics, and strong anti-interference capability. The specific locations and distribution methods are determined based on the mechanical structure and vibration transmission characteristics of the rollers. The vibration transmission path of the rollers begins on the roller body, is transmitted to the equipment frame through the bearing housing, and further diffuses through the transmission gearbox and the connecting support structure of adjacent rollers. Therefore, the multiple acquisition points specifically include a first acquisition point set on the roller bearing housing, a second acquisition point on the transmission gearbox, and a third acquisition point on the connecting support structure of adjacent rollers. Each group of rollers on each piece of equipment is equipped with three acquisition points: one first acquisition point, one second acquisition point, and one third acquisition point, ensuring that signals can be captured from the vibration source to the key nodes of the transmission path. During installation, the first sampling point is located close to the center of the end face of the roller bearing housing. This position is closest to the roller shaft and can directly acquire the vibration signal generated by the roller rotation, with less susceptibility to external interference. The second sampling point is fixed on the top housing of the transmission gearbox. This position can capture the vibration transmitted during gear transmission, indirectly reflecting the operating status of the roller. The third sampling point is installed on the side of the support structure connecting adjacent rollers. This position can collect the vibration generated by the interaction between rollers, supplementing key vibration information.

[0054] All sensors underwent sensitivity calibration using a calibrator before installation. During installation, a torque wrench was used to control the tightening force, ensuring consistent sensor orientation at each acquisition point and avoiding signal acquisition discrepancies due to installation angle deviations. During signal acquisition, the 16-channel synchronous trigger function of the data acquisition card enabled simultaneous acquisition of vibration signals from all acquisition points. The sampling rate was set to 10kHz, and the acquisition duration was determined based on the spinning process cycle, with each batch lasting 30 minutes, continuously acquiring the original vibration signal, i.e., the first vibration signal.

[0055] The first vibration signal contains three core components: vibrations from normal equipment operation, mainly from regular vibrations generated by the uniform rotation of rollers and normal meshing of gears, with frequencies concentrated in the range of 100Hz-1kHz; environmental interference noise, mainly including irregular signals such as vibrations generated by other equipment in the workshop, airflow noise, and ground vibration, with frequencies distributed in the range of 50Hz-10kHz; and vibration fluctuation fault characteristics, namely weak vibrations generated by roller wear, which are manifested as low-frequency pulse signals with frequencies concentrated in the range of 0.1-50Hz and amplitudes only at the micrometer level. These signals are easily masked by the first two types of signals, and are the target signals that this invention needs to focus on extracting.

[0056] The signal preprocessing stage (corresponding to step S2) first performs detrending processing, the core purpose of which is to eliminate baseline drift caused by sensor installation errors. During sensor installation, even with rigorous calibration and fixation, slight installation tilt or uneven tightening may still occur, causing a slow baseline drift in the acquired first vibration signal. This manifests as a linear or non-linear trend in the overall signal. If this drift is not eliminated, it will severely affect the accuracy of subsequent fault feature extraction. This embodiment achieves detrending processing by removing polynomial trend terms from the signal. For the acquired first vibration signal x(t), the order n of the polynomial needs to be determined first. Through extensive experimental verification, when n=3, polynomial fitting can most accurately approximate the baseline drift trend while avoiding signal distortion caused by overfitting. Therefore, a 3rd-order polynomial is chosen for fitting, and the fitting formula is: ,in For constant terms, The coefficient of the linear term, The coefficient of the quadratic term, The coefficient of the cubic term.

[0057] The polynomial coefficients are calculated using the least squares method. The specific process is as follows: Select a first vibration signal x(t) with a duration of 10 seconds, with N = 100,000 sampling points (sampling rate 10kHz), and construct the objective function. By calculating Q respectively , , , Taking the partial derivatives of x(t) and setting them to zero, we obtain a system of linear equations. Solving this system of equations yields the values ​​of the coefficients. For example, in a segment of acquired x(t) signal, hour , hour , hour , hour The result was obtained by calculating using the least squares method. , , , The corresponding 3rd order polynomial Signals after trend reversal The processed data showed that the baseline, which originally showed a slow upward trend, was completely eliminated, and the signal fluctuated around zero. The drift amplitude was reduced from 0.06mV to less than 0.002mV, effectively improving the stability of the signal.

[0058] After detrending, the next step is to remove the power grid frequency and its harmonic interference to obtain the second vibration signal. The workshop power grid frequency is 50Hz, and its 2nd to 5th harmonics (100Hz, 150Hz, 200Hz, 250Hz) are transmitted to the spinning equipment through the power supply lines and are superimposed on the vibration signal. This type of interference signal has the characteristics of fixed frequency and stable amplitude. If it is not removed, it will mask the low-frequency fault characteristic signal. This embodiment uses an adaptive notch filter algorithm to remove interference. This algorithm can automatically track the frequency changes of the power frequency and harmonics and dynamically adjust the notch parameters. Compared with traditional fixed-frequency notch filtering, it has stronger anti-interference adaptability. The filter parameters are set as follows: the notch center frequencies correspond to 50Hz, 100Hz, 150Hz, 200Hz, and 250Hz, respectively; the notch bandwidth is set to 2Hz to ensure that the effective signals of adjacent frequencies are not affected while removing the interference frequencies; the adaptive adjustment step size is set to 0.01 to ensure the speed and stability of frequency tracking. During the filtering process, the detrended signal is first analyzed using Fast Fourier Transform (FFT) to identify the peak frequencies and amplitudes of the power frequency and harmonics. Then, adaptive notch filtering is initiated, performing notch filtering on each interference frequency. The filtered signal is then converted back to the time domain signal, i.e., the second vibration signal, using inverse FFT. Experimental data shows that the signal before filtering exhibits significant peaks at frequencies such as 50Hz and 100Hz, with amplitudes reaching 0.1mV. After filtering, these peaks are effectively suppressed, and the amplitudes are reduced to below 0.005mV, while the amplitudes of the vibrations during normal equipment operation and fault characteristic signals remain essentially unchanged, successfully eliminating power grid interference.

[0059] Next, the first vibration signal is decomposed into multiple scales to obtain the target feature range. The core purpose of multiple scale decomposition is to decompose the complex mixed signal into different scale spaces, separate signal components in different frequency ranges, and thus accurately locate the fault feature frequency range corresponding to roller wear. This embodiment uses wavelet transform for multiple scale decomposition. Wavelet transform has good time-frequency localization characteristics, which can provide high frequency resolution in the low-frequency band and high time resolution in the high-frequency band, making it very suitable for processing non-stationary vibration signals.

[0060] First, a wavelet mother function was selected. By comparing the decomposition effects of various wavelet mother functions such as db1, db4, db6, and sym4, it was found that the db4 wavelet has a moderate support length, good symmetry, and the strongest adaptability for extracting fault features from mechanical vibration signals. Therefore, the db4 wavelet was chosen as the mother function. The scaling factor 'a' and the translation factor 'b' were determined based on the signal frequency range and sampling rate. The sampling rate was 10kHz, and the signal frequency range was 0.1Hz-10kHz. This was based on the scale-frequency correspondence of wavelet transform. (Where c is the center frequency of the wavelet mother function, and the center frequency of the db4 wavelet is approximately 0.714 Hz), the frequency ranges corresponding to different scales are calculated: Scale At that time, the frequency range was 2.5kHz-5kHz; scale At that time, the frequency range was 1.25kHz-2.5kHz; scale At that time, the frequency range was 625Hz-1.25kHz; the scale was... At that time, the frequency range was 312.5Hz-625Hz; the scale... At that time, the frequency range was 156.25Hz-312.5Hz; scale At that time, the frequency range was 78.125Hz-156.25Hz; the scale... At that time, the frequency range was 39.0625Hz-78.125Hz; scale When the scale a=256, the frequency range is 19.53125Hz-39.0625Hz; when the scale a=512, the frequency range is 9.765625Hz-19.53125Hz; when the scale a=512, the frequency range is 4.8828125Hz-9.765625Hz. At that time, the frequency range was 0.1Hz-4.8828125Hz.

[0061] Experiments on rollers with different degrees of wear revealed that the fault characteristic signals caused by roller wear are mainly concentrated in the low-frequency range of 0.1Hz-50Hz, corresponding to the wavelet transform scale. to Therefore, the target feature interval is determined as the scale. to The corresponding signal components. During the multi-scale decomposition process, the second vibration signal... Input the wavelet transform module and set the decomposition level to 10 to obtain 10 detail signals (d1-d10) and 1 approximate signal (a10). Among them, d1-d5 correspond to high-frequency components (frequency > 50Hz), and d6-d10 and a10 correspond to low-frequency components (frequency ≤ 50Hz), which are the signal components of the target feature range. Subsequent processing will be based on these signal components.

[0062] The core objective of the signal demodulation stage (corresponding to step S3) is to amplify the low-frequency fault features that were originally masked by noise. Based on the target feature range obtained from multi-scale decomposition, high-frequency components are first extracted from the second vibration signal. Here, high-frequency components refer to relatively high-frequency components within the target feature range, specifically signals in the 20Hz-50Hz range. These frequency components contain the modulation information of the low-frequency fault features. Demodulation can separate and amplify the low-frequency fault features from the high-frequency carrier. The extraction of high-frequency components is achieved using bandpass filtering. The passband frequency of the bandpass filter is set to 20Hz-50Hz, and the stopband attenuation is set to 80dB to ensure effective filtering of other frequency components.

[0063] High-frequency component signals were extracted. Then, a Hilbert transform is applied. The essence of the Hilbert transform is to perform a 90° phase shift on the signal, and its calculation formula is as follows: , where P represents the Cauchy principal value. To more clearly illustrate the transformation process, a high-frequency component signal y(t) is selected for example calculation. The number of sampling points N=1000 for this signal, and the sampling interval is... For a time interval t from 0 to 0.0999 s, the amplitude of y(t) ranges from 0.001 mV to 0.005 mV. The Hilbert transform is calculated using numerical integration for each... Calculate the integral The Cauchy principal value is obtained to get the corresponding ,For example hour, Calculations yielded ; hour, Calculations yielded Constructing analytic signals Where j is the imaginary unit, representing the amplitude of the analytic signal. This amplitude is the low-frequency fault characteristic signal amplified after demodulation. Through demodulation, the low-frequency fault characteristic signal, which originally had an amplitude of only 0.001mV-0.005mV, is amplified to 0.003mV-0.007mV, an increase of more than 60%. At the same time, the signal-to-noise ratio of the signal increases from the original 10dB to more than 25dB. The low-frequency fault characteristics that were originally masked by noise become clearly distinguishable. For example, the pulse signal generated by slight wear on the roller surface was almost completely submerged in noise before demodulation, but after demodulation, the peak value, period and other characteristic parameters of the pulse can be clearly observed.

[0064] The correlation matching analysis stage (corresponding to step S4) aims to achieve accurate purification and enhancement of fault feature signals. First, a benchmark template needs to be established. The benchmark template is constructed based on a large amount of experimental data. Vibration signals from rollers with different wear levels from multiple devices are selected. After detrending, filtering, multi-scale decomposition, and demodulation, typical low-frequency fault feature signals are extracted as benchmark templates. Specifically, 100 rollers are selected, including 20 with slight wear, 30 with moderate wear, 30 with severe wear, and 20 in normal condition. The collected signals from each roller undergo the aforementioned preprocessing and demodulation to extract low-frequency fault feature signals. Through cluster analysis, the most representative signals under each wear state are selected as benchmark templates. Finally, three benchmark templates are determined, corresponding to the typical fault features of slight, moderate, and severe wear, respectively. The first preset threshold was determined through statistical analysis. Correlation calculations were performed on 1000 sets of normal and fault signals, and the distribution range of cross-correlation coefficients was statistically analyzed. It was found that the cross-correlation coefficients between normal signals and the reference template were all less than 0.6, while the cross-correlation coefficients between fault signals and the corresponding reference template were all greater than 0.8. Therefore, the first preset threshold was determined to be 0.8, that is, signal components with cross-correlation coefficients greater than 0.8 were retained, and environmental interference noise with a cross-correlation coefficient less than 0.8 was eliminated.

[0065] Correlation matching analysis is quantified by the cross-correlation coefficient ρ. x The formula for calculating ᵧ is: Where x is the reference template signal sequence and y is the acquisition point signal sequence to be matched. ̄ and ȳ are the mean values ​​of x and y, respectively, and N is the signal length.

[0066] Taking the signal sequence y at a certain acquisition point as an example, , , Calculate the molecular part First calculate each and The product of and is then summed, resulting in a numerator of 0.025; the denominator is . Calculations yielded , The product is 0.000125, and the square root is 0.01118, therefore the cross-correlation coefficient is... This value is greater than the preset threshold of 0.8, therefore this signal component is retained. For the signal sequence from another acquisition point... The cross-correlation coefficient was calculated. If the value is less than 0.8, the signal component is determined to be environmental noise and is removed.

[0067] By analyzing signals from multiple acquisition points one by one, signal components with a correlation higher than 0.8 with the benchmark template were retained, while low-correlation interference noise was eliminated, ultimately yielding the target fault characteristic signal. After correlation matching analysis, the signal-to-noise ratio of the target fault characteristic signal was further improved to over 35dB, significantly enhancing the identification of fault features. For example, multiple interference pulses were successfully eliminated, retaining only characteristic pulses related to roller wear, greatly improving signal purity and providing accurate data support for subsequent wear level determination.

[0068] The wear level determination and control stage (corresponding to step S5) is the final implementation stage of this invention. First, it is necessary to extract key parameters of the target fault characteristic signal. These key parameters include peak value, RMS value, kurtosis, skewness, impulse factor, and waveform factor, each of which is closely related to the degree of roller wear. Peak value is the maximum amplitude of the signal, directly reflecting the vibration intensity generated by wear; the more severe the wear, the larger the peak value. RMS value reflects the average energy of the signal and can reflect the continuous impact of wear. Kurtosis is a parameter describing the steepness of the signal amplitude distribution. When the roller is operating normally, the signal amplitude distribution is relatively uniform, with a kurtosis close to 3. As wear intensifies, the pulse component in the signal increases, and the kurtosis gradually increases. Skewness describes the asymmetry of the signal amplitude distribution; the pulse signal generated by wear will cause the skewness to shift in the positive direction. Impulse factor is the ratio of peak value to RMS value, which can amplify the influence of pulse characteristics. Waveform factor is the ratio of RMS value to average value, reflecting the change in the waveform shape of the signal.

[0069] The extraction of key parameters is based on the target fault characteristic signal. The parameter values ​​are obtained through numerical calculation. For example, the peak value of a target fault characteristic signal is 0.008mV, the effective value is 0.003mV, the kurtosis is 4.8, the skewness is 1.2, the impulse factor is 2.67, and the waveform factor is 1.5. The threshold ranges of the characteristic parameters corresponding to multiple roller wear levels were determined through statistical analysis of a large amount of experimental data. Accelerated wear experiments were conducted on rollers from 200 machines. Starting from normal conditions, vibration signals were collected every 10 hours until the rollers completely failed, resulting in 1000 sets of experimental data. The aforementioned key parameters were extracted from each set of data. Through cluster analysis and statistical modeling, the parameter threshold ranges corresponding to the three wear levels—light, moderate, and heavy—were determined. The specific threshold ranges are as follows: For mild wear, peak value 0.003mV-0.005mV, RMS value 0.001mV-0.002mV, kurtosis 3.2-4.5, skewness 0.5-1.0, impulse factor 2.0-2.5, and waveform factor 1.2-1.4; For moderate wear, peak value 0.005mV-0.008mV, RMS value 0.002mV-0.004mV, kurtosis 4.6-6.0, skewness 1.1-1.5, impulse factor 2.6-3.0, and waveform factor 1.5-1.7; For severe wear, peak value >0.008mV, RMS value >0.004mV, kurtosis >6.0, skewness >1.5, impulse factor >3.0, and waveform factor >1.7.

[0070] By comparing the extracted key parameters with preset threshold ranges, the current wear severity level of the roller can be determined. For example, among the parameters extracted above, the peak value of 0.008mV, effective value of 0.003mV, kurtosis of 4.8, skewness of 1.2, impulse factor of 2.67, and waveform factor of 1.5 all fall within the threshold range for moderate wear, therefore the roller is determined to be moderately worn. Based on preset control rules corresponding to different wear levels, the system automatically triggers corresponding control commands: When light wear is detected, the system displays a warning prompt on the industrial touchscreen and simultaneously sends a warning message to the operator's mobile APP via workshop broadcast, prompting the operator to closely monitor the roller's operating status and plan to check it during the next equipment maintenance; when moderate wear is detected, the system sends a speed reduction command to the spinning machine PLC via the CAN bus, reducing the roller speed from the current 1000r / min to 800r / min to slow down the rate of wear aggravation, and simultaneously triggers a planned maintenance prompt, displayed on the touchscreen and sent to the equipment management system, arranging maintenance and replacement within 24 hours; when severe wear is detected, the system immediately sends a stop command to the PLC, and the spinning machine stops running within 3 seconds of receiving the command, simultaneously activating an emergency maintenance alarm, flashing red indicator lights on the corresponding equipment in the workshop, issuing a rapid alarm sound via broadcast, and sending an emergency maintenance notification to maintenance personnel via the APP, ensuring timely replacement of the worn roller and avoiding yarn quality problems and equipment damage caused by roller failure.

[0071] To verify the effectiveness of the method of this invention, a comparative experiment was conducted over a period of 6 months. Sixty pieces of equipment in the workshop were selected and divided into an experimental group and a control group, with 30 pieces of equipment in each group. The experimental group used the monitoring and control method of this invention, while the control group used the traditional mean filtering + Fourier domain filtering monitoring method. The experimental results showed that the early wear identification accuracy of the rollers in the experimental group reached 98%, while that in the control group was only 75%; the wear level determination accuracy of the experimental group reached 95%, while that in the control group was only 80%; the cumulative downtime due to equipment failure in the experimental group was 12 hours, while that in the control group was 48 hours; the average yarn evenness CV value of the yarn produced in the experimental group was 1.8%, while that in the control group was 2.5%; the roller replacement cost in the experimental group was 30% lower than that in the control group, and the yarn scrap rate was reduced by 25%. During the experiment, the experimental group successfully identified 35 cases of mild wear, 28 cases of moderate wear, and 12 cases of severe wear, all of which triggered the corresponding control commands in a timely manner, and no serious production accidents caused by roller wear occurred. In contrast, the control group missed 12 cases of mild wear and 8 cases of moderate wear, and falsely reported 5 cases of normal conditions as malfunctions, resulting in a decline in yarn quality in some equipment due to increased wear. In fact, two pieces of equipment even experienced roller breakage due to severe wear not being addressed in time, causing serious downtime losses.

[0072] During long-term operation, the method of this invention has demonstrated good stability and adaptability, capable of handling complex situations such as fluctuations in workshop environmental noise, changes in power grid voltage, and adjustments to roller speed. When environmental noise suddenly increases to 95dB, fault characteristic signals can still be accurately extracted by dynamically adjusting filter parameters and correlation thresholds. When power grid voltage fluctuations cause a ±1Hz shift in power frequency interference, adaptive notch filtering can quickly track frequency changes and effectively eliminate interference. When the roller speed is adjusted to 600r / min or 1200r / min, the accurate extraction of fault characteristic signals is ensured by adjusting the scale factor and target feature range of the multi-scale decomposition. Furthermore, the method of this invention has good scalability and can be adapted to different models of spinning machines and rollers. By adjusting the reference template and threshold range, it can be applied to fault monitoring and control of other textile equipment.

[0073] In summary, the textile yarn production monitoring and control method of the present invention effectively solves the technical problems of difficulty in early identification of roller wear and inaccurate assessment of wear degree in high-noise and multi-interference environments in textile workshops through scientific deployment of collection points, multi-level signal preprocessing, fault feature enhancement, correlation purification, precise grading and intelligent control. It significantly improves the stability of yarn production and product quality, reduces equipment maintenance costs and production losses, and has important practical application value and promotion significance.

[0074] It should be noted that the specific methods of the system in the above embodiments have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0075] Finally, it should be noted that although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for monitoring and controlling textile yarn production, characterized in that, The method includes: S1: Multiple acquisition points are deployed along the vibration transmission path of the roller, and the first vibration signal is collected synchronously through vibration sensors; the first vibration signal includes the vibration of normal equipment operation, environmental interference noise, and vibration fluctuation fault characteristics; S2: Perform detrending processing on the first vibration signal to eliminate baseline drift caused by sensor installation error; remove power grid frequency and harmonic interference to obtain the second vibration signal; perform multi-scale decomposition on the first vibration signal to obtain the target feature range; S3: Based on the target feature range, extract the high-frequency components of the second vibration signal and demodulate them to amplify the low-frequency fault features that were originally masked by noise. S4: Using low-frequency fault characteristics as a benchmark template, perform correlation matching analysis on fault characteristic signals from multiple acquisition points; retain signal components in each acquisition point whose correlation with the benchmark template is higher than the first preset threshold, and remove environmental interference noise whose correlation is lower than the first preset threshold to obtain the target fault characteristic signal, thereby achieving accurate purification and enhancement of the fault characteristic signal; S5: Extract key parameters of the target fault feature signal, and preset threshold ranges for feature parameters corresponding to multiple roller wear levels; compare the extracted key parameters with the preset threshold ranges to determine the current wear severity level of the roller; automatically trigger corresponding control commands according to preset control rules corresponding to different wear levels.

2. The method according to claim 1, characterized in that, The plurality of acquisition points include at least one first acquisition point disposed on the roller bearing housing, at least one second acquisition point disposed on the transmission gearbox, and at least one third acquisition point disposed on the adjacent roller connection support structure.

3. The method according to claim 1, characterized in that, The detrending process for the first vibration signal includes: Detrending is achieved by removing the polynomial trend term from the signal: For the first vibration signal collected Fit an nth-order polynomial: ; The signal obtained after detrending: .

4. The method according to claim 1, characterized in that, The multi-scale decomposition of the first vibration signal includes: Wavelet transform is used to transform the second vibration signal Decomposed into: ; in, Here, is the wavelet mother function, a is the scaling factor, and b is the translation factor; The target feature interval is determined based on the scale corresponding to the wear feature frequency.

5. The method according to claim 1, characterized in that, The demodulation process for the second vibration signal includes: For the extracted high-frequency component signals Perform Hilbert transform to obtain ; Constructing analytic signals The amplitude of the analytical signal This refers to the low-frequency fault characteristic signal that has been demodulated and amplified.

6. The method according to claim 1, characterized in that, The correlation matching analysis includes: Quantification is performed using cross-correlation coefficients. The calculation formula is: ； Where x is the reference template signal sequence, and y is the acquisition point signal sequence to be matched. and Let N be the mean of each, and N be the signal length.

7. The method according to claim 1, characterized in that, The preset control rules include: when the wear is determined to be slight, a warning prompt is triggered; when the wear is determined to be moderate, the equipment is slowed down and a planned maintenance prompt is issued; when the wear is determined to be severe, a shutdown command is triggered and an emergency maintenance alarm is issued.

Images