An abnormal vibration mode identification and state monitoring method for a covering yarn machine spindle

Abstract

The present application relates to the technical field of textile equipment state monitoring. The purpose is to provide an abnormal vibration mode identification and state monitoring method for the spindle of a covering yarn machine. Through full-section multi-section sensing layout, mechanical vibration reduction and anti-interference design, four-stage dynamic precision compensation algorithm and working condition coupling abnormal identification model, the present application realizes accurate collection of spindle vibration signals, interference elimination, fault identification and positioning, and solves the core problems of incomplete coverage, weak anti-interference, low precision, fuzzy positioning and untimely early warning in the prior art.

Description

Technical Field

[0001] This invention relates to the field of textile equipment condition monitoring technology, specifically to a method for identifying abnormal vibration patterns and monitoring the condition of a spindle in a covering yarn machine. Background Technology

[0002] Covering yarn machines are core equipment in the textile industry for producing elastic and functional covered yarns. The spindle, as a key rotating actuator, undertakes the core functions of yarn twisting and winding. Its operating state directly determines key quality indicators such as twist uniformity and breaking strength of the covered yarn, while also affecting equipment operating efficiency and production safety. In the actual production process of covering yarn machines, the spindle needs to operate at high speeds of 10,000-20,000 rpm for extended periods, enduring multiple effects including cyclic alternating loads, yarn tension impacts, and lubrication medium loss. This makes it prone to various typical faults: such as quality imbalance caused by assembly deviations or uneven yarn winding, bearing wear due to long-term high-speed friction, mid-section fatigue cracks caused by vibration fatigue, vibration reduction failure due to aging of damping components, and connection faults such as loose root mounting seats and loose spindle bodies. If these faults are not detected and addressed in a timely manner, they can initially lead to increased yarn breakage rates and decreased product quality. In severe cases, they can cause serious production accidents such as spindle breakage and increased frame vibration, resulting in significant economic losses.

[0003] However, existing spindle vibration detection technology for covering yarn machines has many insurmountable technical pain points, which cannot meet the high-precision monitoring requirements in actual production:

[0004] 1. Limited detection coverage and high rate of missed detection of latent faults: Existing technologies mostly adopt single-point detection schemes at the ends, placing sensors only at both ends of the spindle. This cannot cover critical areas prone to latent faults, such as the mid-section damping component area and the root mounting area. For example, fatigue cracks and damping component failures in the mid-section of the spindle may only manifest as localized abnormal vibrations in the early stages, with no obvious changes in the end-point detection signals. This results in a missed detection rate of over 60% for such faults, which are often only discovered after the fault has worsened.

[0005] 2. Weak ability to suppress interference in complex environments: The working environment of the covering yarn machine is harsh and there are multiple sources of interference: high-frequency electromagnetic interference generated by the operation of the drive motor and frequency converter, electrostatic interference generated by the friction between the high-speed yarn and the spindle and guide wheel, and pollution interference formed by yarn fibers and lubricating oil mist. These interferences will cause vibration signal distortion and a significant drop in signal-to-noise ratio. Traditional detection methods only use simple filtering, which is difficult to effectively eliminate composite interference.

[0006] 3. Measurement distortion caused by frame vibration coupling: The detection sensor is usually fixed to the frame. When the covering yarn machine is running at high speed, the vibration of the motor and the vibration of the spindle will cause the frame to vibrate. This vibration will be directly transmitted to the sensor and superimposed with the actual vibration signal of the spindle, resulting in a significant deviation in the measurement result. The traditional method uses fixed coefficient differential compensation, which does not take into account the dynamic changes of frame vibration, and the compensation effect is limited.

[0007] 4. Poor adaptability to working conditions and insufficient compensation accuracy: Covering yarn machines have various working conditions such as start-up, stop, constant speed, yarn change, and full roll. The spindle load, speed fluctuation, and frame vibration characteristics are different under different working conditions. Traditional compensation methods use fixed thresholds and compensation coefficients, which are prone to problems such as over-compensation in low-speed conditions, under-compensation in high-speed conditions, and miscompensation during yarn change, resulting in measurement errors exceeding 20%.

[0008] 5. Ambiguous fault location and low maintenance efficiency: A single detection point cannot distinguish the axial position (upper section / middle section / lower section) and circumferential position of the fault. Fault identification can only determine whether there is an abnormality, but cannot accurately locate the fault location. This results in the need for a complete disassembly and inspection during maintenance, which is time-consuming, labor-intensive, and inefficient.

[0009] 6. Insufficient early warning timeliness: Traditional detection methods are weak in capturing early weak vibration signals and can only issue warnings when the fault develops to the middle and late stages and the vibration amplitude increases significantly. The warning time is usually only 58 hours, which makes it difficult for maintenance personnel to complete fault diagnosis and repair in a short time, and can easily lead to production interruption.

[0010] To address the aforementioned technical challenges, existing technologies lack an integrated detection and compensation method capable of achieving full-section spindle vibration detection, dynamic adaptation to multiple operating conditions, effective suppression of multiple interferences, and precise fault location. Therefore, developing a full-section spindle vibration detection and accuracy compensation method adapted to the complex operating conditions of covering yarn machines has significant practical application value and industry implications. Summary of the Invention

[0011] The purpose of this invention is to provide a method for identifying and monitoring the abnormal vibration patterns of spindles in a covering yarn machine. By employing a full-section multi-section sensing layout, mechanical vibration reduction and anti-interference design, a four-level dynamic precision compensation algorithm, and a working condition coupling anomaly identification model, the method achieves accurate acquisition of spindle vibration signals, interference removal, fault identification, and location, thus solving the core problems of existing technologies such as incomplete coverage, weak anti-interference, low accuracy, ambiguous positioning, and untimely early warning.

[0012] To achieve the above-mentioned objectives, the technical solution adopted by this invention is: a method for identifying and monitoring abnormal vibration patterns of a spindle in a covering yarn machine, comprising six core steps: mechanical deployment of the detection system, synchronous signal acquisition, signal preprocessing, full-section vibration feature extraction, accuracy compensation, and abnormal pattern identification. The specific implementation of each step is as follows:

[0013] S1. Mechanical deployment of the detection system: Three detection sections are set along the spindle axis, corresponding to the upper, middle and lower sections of the spindle respectively; two anti-oil mist eddy current detection sensors are evenly arranged in the circumference of each detection section, and the circumferential angle between adjacent sensors is 180°; the same type of main reference sensor is installed synchronously at the rack position corresponding to each detection sensor, and an auxiliary reference sensor is additionally configured in the stable area of ​​the rack for each detection section.

[0014] S2. Synchronous signal acquisition: A multi-channel high-speed acquisition module is used to synchronously acquire vibration signals from the detection sensor, main reference sensor, and auxiliary reference sensor, as well as auxiliary data such as spindle speed and bearing temperature; the sampling frequency is ≥ 5 times the highest operating frequency of the spindle, and the minimum sampling frequency is not lower than 1500Hz;

[0015] S3. Signal preprocessing: Three levels of noise reduction are performed sequentially. The first level uses an RC low-pass filter with a cutoff frequency of 10kHz to filter out high-frequency electromagnetic noise generated by the drive motor and frequency converter. The second level uses a moving average method to remove electrostatic spike pulses generated by yarn friction. The third level uses an extreme value averaging algorithm to remove abnormal data points caused by fiber and lubricating oil mist adhesion.

[0016] S4. Full-section vibration feature extraction: For the preprocessed signals of each detection section, extract the time domain features and frequency domain features respectively, calculate the feature mean of two circumferential detection sensors at the same detection section, and form the vibration feature vector of each section; by comparing the consistency of the feature vectors of different sections, preliminarily determine whether the vibration anomaly is a local fault or a global fault.

[0017] S5. Precision Compensation: The four-level compensation process is executed sequentially, including dual-sensor differential basic compensation, working condition adaptation dynamic weight optimization compensation, three-sensor cross-verification compensation, and full-section consistency calibration compensation, finally obtaining the true vibration signal of the spindle.

[0018] S6. Abnormal Pattern Recognition: Call the preset abnormal vibration feature library of the spindle for pattern matching, and perform cross-validation with real-time speed and bearing temperature auxiliary data to output the fault type, location and level, and trigger the corresponding early warning mechanism.

[0019] The beneficial effects of this invention are mainly reflected in:

[0020] 1. Full-section coverage detection significantly reduces the rate of missed detection of hidden faults: By using a layout of three axial detection sections and dual circumferential sensors, full-section coverage without blind spots is achieved for the upper, middle, and lower sections of the spindle. This solves the problem that traditional single-point end detection cannot cover hidden faults such as failure of damping components in the middle section, fatigue cracks, and loose mounting bases at the root. The fault missed detection rate is reduced from more than 60% in the existing technology to less than 8%, and the fault identification accuracy rate is ≥91%.

[0021] 2. Strong ability to suppress complex interference and significantly improved signal-to-noise ratio: Combining the integrated anti-interference design of "mechanical vibration reduction installation + three-level progressive preprocessing + four-level precision compensation", it effectively eliminates complex interference such as electromagnetic interference, electrostatic interference, fiber contamination, and frame vibration coupling. The signal-to-noise ratio reaches 39dB in complex interference environment, which is 116.7% higher than the traditional method, ensuring the authenticity of vibration signal.

[0022] 3. Dynamic adaptation to working conditions and high measurement accuracy: By constructing a "working condition compensation weight" mapping library, the compensation coefficient can be dynamically adjusted under different working conditions. Combined with phase calibration, cross-validation, and installation gap correction, the vibration measurement error is controlled within ±3%, which is significantly better than the ±20% error level of traditional methods and meets the requirements of high-precision monitoring.

[0023] 4. Precise fault location and improved maintenance efficiency: By comparing multi-section feature vectors, the fault can be accurately located in both the cross section and circumferential direction, with a positioning accuracy of ±15°. Maintenance personnel can directly inspect the faulty part without the need for a complete disassembly and inspection, reducing maintenance time by more than 60% and significantly improving maintenance efficiency.

[0024] 5. Strong early warning timeliness and improved production safety: Through multi-dimensional feature extraction and early fault identification algorithms, it can capture early weak vibration signals of the spindle and provide early warning of faults more than 24 hours in advance. Compared with the traditional method (early warning time of 58 hours), the early warning time is extended by 34.8 times, which provides sufficient maintenance window for maintenance personnel, effectively reduces the incidence of production accidents such as yarn breakage and spindle breakage, and improves production stability by more than 40%.

[0025] 6. Good adaptability and strong practicality: Designed for the compact space, high-speed operation and variable working conditions of covering yarn machines, the sensor adopts an oil mist resistant and small-sized model with a simple installation structure. It can be directly applied to the transformation and upgrading of existing covering yarn machines without replacing core components, and supports parallel monitoring of multiple spindles (expandable to 64 spindles for simultaneous detection), adapting to the batch production needs of textile workshops. Detailed Implementation

[0026] The steps of this invention will be described in detail below:

[0027] 1. Mechanical Deployment of the Detection System

[0028] This step lays the mechanical foundation for accurate detection by optimizing sensor layout, mounting structure, and vibration reduction design, while reducing the impact of frame vibration and environmental pollution on the detection process.

[0029] Detection sections and sensor layout: Three detection sections are evenly arranged along the spindle axis, corresponding to the upper section (close to the spindle plate, 50mm from the end face of the spindle plate), the middle section (corresponding to the installation position of the damping component, 30mm from the center of the damping component), and the lower section (close to the bearing seat, 50mm from the end face of the bearing seat), to achieve full coverage of the spindle without dead angles; two anti-oil mist eddy current detection sensors are symmetrically arranged circumferentially in each detection section, with the circumferential angle between adjacent sensors being 180°, to weaken the influence of local roundness errors of the spindle on the measurement results;

[0030] Reference sensor layout: At each rack position corresponding to the detection sensor, a main reference sensor of the same model is synchronously installed to collect rack vibration signals. The detection sensor and the main reference sensor adopt a "coaxial and coplanar" design (the probe axis coincides, and the detection surface of the detection sensor is parallel to the measured surface on the detection sensor). The installation spacing is controlled to ≤80mm to ensure that the rack vibration signals collected by the two are highly consistent. An additional auxiliary reference sensor is configured for each detection section and installed in a stable area of ​​the rack away from the vibration source (such as at the rack load-bearing beam) to cross-verify the reliability of the rack vibration signal.

[0031] Vibration reduction and fixing structure design: All sensors are fixed by a combination structure of "rigid base plate + polyurethane vibration damping pad". The rigid base plate is made of 45# steel to ensure installation accuracy. The polyurethane vibration damping pad has a thickness of 5-8mm and a Shore hardness of 60-70, which can effectively absorb frame vibration and reduce the transmission of vibration to the sensor probe. The sensor mounting base is designed with a short cantilever structure with a cantilever length ≤50mm to improve the rigidity of the mounting base itself and avoid resonance. The contact surface between the mounting base and the sensor is precision milled to a flatness of ≤0.02mm, eliminating vibration transmission amplification caused by contact surface gaps. The sensor is rigidly fixed by locking nuts and anti-loosening washers, and an elastic clamp is added at the tail for axial limit to prevent slight displacement of the sensor under high-speed operation.

[0032] Sensor parameters and installation accuracy: The detection sensor, main reference sensor, and auxiliary reference sensor are all oil-mist-resistant eddy current sensors with a range of 2mm, sensitivity ≥8mV / μm, and linearity ≤±0.5%, suitable for oily and dusty environments on covering yarn machines; the perpendicularity of the sensor probe to the measured surface of the spindle is ≤0.5°, and the detection gap is set at the midpoint of the sensor's linear range (1mm).

[0033] 2. Synchronous signal acquisition

[0034] A 16-channel high-speed acquisition module (sampling rate up to 10kHz, meeting the minimum 1500Hz sampling requirement) is adopted. Through a synchronous triggering mechanism, synchronous sampling of 6 detection sensors, 6 main reference sensors, and 3 auxiliary reference sensors (a total of 15 vibration signals) is achieved, with a time synchronization error ≤1μs, avoiding compensation deviations caused by signal phase differences. At the same time, the spindle speed signal is acquired through an optical fiber speed sensor (measurement range 0-30000rpm, accuracy ±0.1%), and the bearing temperature signal is acquired through a miniature temperature sensor attached to the bearing housing (measurement range 20~120℃, accuracy ±0.3℃). All data are transmitted to the data processing terminal via Ethernet. The sampling frequency is dynamically adjusted according to the highest main frequency of the spindle to ensure that the sampling frequency is ≥5 times the highest main frequency (e.g., when the rated spindle speed is 18000rpm, the main frequency is 300Hz, and the sampling frequency is set to 1500Hz).

[0035] 3. Signal preprocessing

[0036] To address the complex interference characteristics of the covering yarn machine, a three-stage progressive noise reduction process is implemented:

[0037] First stage: RC low-pass filter, with a cutoff frequency set to 10kHz. The hardware circuit filters out high-frequency electromagnetic noise above 10kHz generated by the drive motor and frequency converter, while retaining the effective frequency components (0-10kHz) of the spindle vibration.

[0038] Second stage: Moving average filtering, with the window size dynamically adjusted to 510 sampling points (adaptively switching according to rotation speed: 5 sampling points at high speed to ensure real-time signal; 10 sampling points at low speed to improve noise reduction effect), eliminating instantaneous electrostatic spike pulses generated by yarn friction (the duration of the spike is usually <3 sampling points).

[0039] The third level is the extreme value removal averaging algorithm, which sets the extreme value threshold to 3 times the mean of the signal segment. When the value of a certain sampling point exceeds the threshold, it is determined to be an abnormal data point caused by fiber or lubricating oil mist adhesion. The mean of the two sampling points before and after is used to replace it to avoid abnormal points affecting the accuracy of subsequent feature extraction.

[0040] 4. Extraction of vibration features across the entire segment

[0041] For the preprocessed signals from the three detection sections, multidimensional features are extracted to form a feature vector that comprehensively reflects the vibration state of the spindle:

[0042] Time-domain characteristics include peak-to-peak value, peak value, mean, root mean square (RMS), kurtosis, and skewness. Kurtosis is used to identify impact failures such as early fatigue cracks, while skewness is used to determine the symmetry of the vibration signal (e.g., vibration signals caused by mass imbalance are symmetrically distributed).

[0043] Frequency domain characteristics: The time domain signal is converted into a frequency domain signal through Fast Fourier Transform (FFT) to extract characteristic frequencies (such as spindle main frequency, bearing characteristic frequency, and frame resonant frequency), harmonic amplitude ratio (the ratio of harmonic amplitude to fundamental frequency amplitude), and frequency band energy proportion (the proportion of energy in key frequency bands such as 0-50Hz, 50-100Hz, and 100-200Hz to the total energy).

[0044] Feature fusion: Calculate the average feature values ​​of two circumferential sensors at the same detection section to mitigate measurement fluctuations caused by local roundness errors in the spindle (such as machining deviations and surface wear); By comparing the consistency of feature vectors of three axial sections, the fault type is preliminarily determined: if the difference between the feature vectors of the three sections is ≤5%, it is determined to be an overall fault (such as mass imbalance); if the difference between the feature vector of a single section and the other two sections is ≥20%, it is determined to be a local fault (such as fatigue cracks in the middle section or bearing wear in the lower section).

[0045] 5. Precision Compensation

[0046] A four-level progressive compensation strategy is adopted to gradually eliminate measurement deviations caused by frame vibration, changes in operating conditions, and installation errors, ultimately restoring the true vibration signal of the spindle:

[0047] (1) Dual-sensor differential foundation compensation

[0048] Amplitude matching coefficient calibration: Under the condition of uniform speed and no-load operation of the spindle (no yarn load, and the speed is stable at ±1% of the rated speed), the actual vibration signal S of the spindle under no-load is obtained through standard vibration table calibration. det0 Simultaneously, the unloaded frame vibration signal S from the main reference sensor is acquired. ref0 Calculate the amplitude matching coefficient K1=S det0 / S ref0 To ensure the amplitude matching accuracy of the frame vibration signal;

[0049] Phase calibration: Three reflective marks are evenly set along the circumference of the spindle end face. A synchronization reference signal is triggered by a laser photoelectric sensor, and the detection sensor signal S is calculated in real time. det With the main reference sensor signal S ref phase difference Δ φ The main reference sensor signal is corrected using a phase shift algorithm. ， ref =S ref ×sin(ωt+Δ φ ), where ω is the signal angular frequency, t is the instantaneous time, and the phase compensation accuracy is ≤0.5°;

[0050] Differential operation: The phase-corrected main reference sensor signal and the detection sensor signal are differentially processed to cancel the common-mode interference of the frame vibration and obtain the preliminary compensation signal S.comp1 =S det -K1×S ， ref .

[0051] (2) Dynamic weight optimization compensation for working conditions

[0052] Construction of the working condition mapping library: Based on the actual operating data of the covering yarn machine, the amplitude matching coefficient adjustment rules under different working conditions are preset to form a "working condition compensation weight" mapping library:

[0053] During start-up and shutdown: If the speed fluctuation is ≥10% (e.g., the speed increases from 0 to 18000 rpm during startup, resulting in large fluctuations), the frame will vibrate violently. Increase K1 by 20%-30% to enhance the frame vibration compensation effect.

[0054] Full roll stage: The roll capacity is ≥80% of the rated capacity (e.g., if the rated capacity is 5kg, the roll weight is ≥4kg). The spindle load is stable and the frame vibration is reduced. K1 is reduced by 10%-15% to avoid overcompensation masking the true vibration of the spindle.

[0055] Uniform speed and stable stage: speed fluctuation ≤1%, roll capacity <80%, keep K1 unchanged;

[0056] Real-time operating condition identification: The speed fluctuation amplitude is determined by the 30-second sliding standard deviation of the speed signal, and the roll capacity is obtained by the roll diameter detection device (such as a laser rangefinder sensor) to comprehensively identify the current operating condition;

[0057] Dynamic adjustment and anomaly removal: K1 is dynamically adjusted based on the identified operating conditions to obtain the optimized compensation signal S. ， comp1 For sudden vibrations (signal slope ≥ 0.1 mm / ms) caused by sudden changes in yarn tension during the yarn changing stage, differential compensation is paused and the average signal value of the first 5 sampling cycles is used instead to avoid miscompensation caused by sudden changes in tension.

[0058] (3) Three-sensor cross-validation compensation

[0059] Deviation calculation: Acquire auxiliary reference sensor signal S ref2 (Installed on the frame support beam, vibration amplitude ≤ 10% of the frame vibration at the sensor location), calculate the main reference sensor correction signal S. ， ref With auxiliary reference sensor signal S ref2 The absolute deviation ΔS = |S ， ref - S ref2 |;

[0060] Compensation optimization:

[0061] If ΔS ≤ 3% × sensor full scale (e.g., if the sensor range is 2mm, 3% of the full scale is 0.06mm), the frame vibration signal is considered stable, and the average of the two values, S, is taken. ref avg =(S ， ref + S ref2 Substituting ) / 2 into the difference formula yields the second-level compensation signal S. comp2 =S det -k1'×S ref avg (k1' is the dynamically adjusted amplitude matching coefficient).

[0062] If ΔS > 3% × sensor full scale, it is determined that there is local resonance in the rack (such as resonance caused by loose fasteners in a certain part). An abnormality mark is activated and a rack vibration warning is triggered (audio and visual alarm, prompting maintenance personnel to check the rack). At the same time, the compensation weight of the historical working conditions (such as the same time period last week and the same roll capacity) is used to replace the current k1' for compensation to ensure the continuity of compensation.

[0063] (4) Consistency calibration compensation across the entire segment

[0064] Peak-to-peak value calculation: Calculate the secondary compensation signal S for the three detection sections. comp2 The peak values ​​are denoted as A1 for the upper segment, A2 for the middle segment, and A3 for the lower segment, respectively.

[0065] Consistency coefficient calculation: The overall vibration consistency coefficient C is calculated by the ratio of the standard deviation to the mean. The formula is as follows:

[0066]

[0067] in =(A1+A2+A3) / 3 is the average peak-to-peak value of the entire segment, and C reflects the uniformity of vibration of the three sections;

[0068] Deviation correction:

[0069] If C≤5%, the overall state of the spindle is considered stable, and there is no significant deviation in the installation gap. This serves as a representative value of the vibration across the entire segment at that moment.

[0070] If C > 5%, locate the abnormal section where the peak value deviates the most (e.g., A2-). (If the absolute value is the largest, the middle section is determined to be an abnormal section). The actual installation gap d of the detection sensor at this section is measured using a laser rangefinder. act Calculate the standard gap d std The difference Δd = d (1 mm) act -d std Through the sensor sensitivity coefficient K s (Obtained from standard vibration table calibration, e.g., K) s=8mV / μm) Correction deviation, finally obtaining the true vibration signal S of the spindle. final =S comp2 +Δd×K s .

[0071] 6. Anomaly Pattern Recognition

[0072] A dual identification mechanism of "feature database matching + auxiliary data cross-validation" is adopted to achieve accurate fault identification and location:

[0073] Abnormal vibration feature library access: A pre-set feature library containing 6 types of typical spindle faults, with each type of fault corresponding to a clear time-domain / frequency-domain feature threshold and associated auxiliary features.

[0074] For example, six typical faults are identified: mass imbalance, bearing wear, spindle loosening, damping component failure, root mounting loosening, and mid-section fatigue cracking. Each fault type corresponds to a preset time-domain characteristic threshold, frequency-domain characteristic frequency range, and associated auxiliary features.

[0075] Mass imbalance: The peak value of the full-section vibration increases with the square of the rotational speed, the characteristic frequency is the main frequency of the spindle, and the bearing temperature is ≤55℃ with no obvious abnormalities;

[0076] Bearing wear: The peak-to-peak value of vibration in the lower section is ≥0.08mm, multiple harmonic peaks appear in the low frequency band, and the bearing temperature is ≥65℃;

[0077] Spindle loosening: The vibration amplitude of the upper section fluctuates by ≥30%, with no fixed characteristic frequency, and is synchronized with the rotational speed fluctuation;

[0078] Damping component failure: peak-to-peak vibration value of the mid-section ≥ 0.07 mm, vibration decay time ≥ 0.5 s, and severe vibration during start-up and shutdown.

[0079] The root mounting bracket is loose: the peak-to-peak value of the vibration of the lower section is ≥0.09mm, the characteristic frequency includes the resonant frequency component of the frame, and the temperature is not obviously abnormal;

[0080] Mid-section fatigue crack: Mid-section cross-section vibration peak-to-peak value ≥ 0.06 mm, high-frequency harmonic amplitude ratio ≥ 0.3, and irregular vibration waveform.

[0081] The vibration feature vector after full-segment consistency calibration is matched with the feature library for similarity (using the cosine similarity algorithm, a similarity of ≥85% is considered a successful match).

[0082] Cross-validation with auxiliary data: The matching results are validated by combining real-time speed and bearing temperature data to avoid misjudgments.

[0083] Example 1: If the feature vector matches "mass imbalance", and the speed fluctuation is ≤1% and the bearing temperature is ≤55℃, it is determined to be a mass imbalance fault;

[0084] Example 2: If the feature vector matches "bearing wear", and the peak-to-peak value of the vibration of the lower section is ≥0.08mm and the bearing temperature is ≥65℃, it is determined to be a bearing wear fault;

[0085] Fault Location and Severity Classification: Based on the successfully matched fault type, combined with the abnormal cross-section location (upper / middle / lower section) and the difference in circumferential sensor signals (e.g., if the difference between two circumferential sensor signals at a certain cross-section is ≥20%, the circumferential fault location is determined), the system outputs "Fault Type + Cross-Section Location + Circumferential Location"; simultaneously, based on the proportion of vibration peak-to-peak values ​​exceeding the threshold and the degree of temperature exceedance, the fault level is classified (minor / moderate / severe), triggering the corresponding early warning mechanism.

[0086] Minor anomalies (10%-20% above the threshold): The control box provides audible and visual alerts, and the data terminal marks the anomaly; these do not affect production.

[0087] Moderate abnormality (exceeding the threshold by 20%-50%): A shutdown inspection reminder will be pushed to the administrator's mobile APP. It is recommended to shut down the machine for maintenance after the current roll is completed.

[0088] Severe abnormality (exceeding the threshold by ≥50%): Directly triggers emergency shutdown, cuts off the power supply to the covering yarn machine, and prevents the fault from escalating.

[0089] The technical solution of the present invention will be further described in detail below with reference to specific experimental data and practical application scenarios, to verify the effectiveness and practicality of the method.

[0090] I. Experimental Conditions

[0091] 1. Experimental subject: A certain brand of covering yarn machine (model: BY980), spindle parameters: total length 600mm, rated speed 18000rpm (main frequency 300Hz), rated package capacity 5kg, bearing model: 6203ZZ;

[0092] 2. Testing equipment and parameters:

[0093] Sensors: Oil mist eddy current sensor (range 2mm, sensitivity 8mV / μm, linearity ±0.5%), fiber optic speed sensor (measurement range 0-30000rpm, accuracy ±0.1%), miniature temperature sensor (measurement range 20~120℃, accuracy ±0.3℃).

[0094] Data acquisition and processing: 16-channel high-speed acquisition module (sampling rate 10kHz), data processing terminal (CPU: Intel i7 12700H, memory 32GB, equipped with Python 3.9 algorithm platform).

[0095] Calibration equipment: standard vibration table (model: JZK50, accuracy ±0.001mm), laser rangefinder (accuracy ±0.01mm).

[0096] 3. Experimental environment: Simulate the actual production environment of the covering yarn machine, with a temperature of 25±2℃ and a humidity of 50±5%RH. Turn on the frequency converter (frequency 0-50Hz) to generate electromagnetic interference, generate electrostatic interference through high-speed friction of the yarn, and manually sprinkle fiber dust to simulate a polluted environment.

[0097] 4. Fault Simulation: Six typical fault types are simulated using a fault simulation device:

[0098] Mass imbalance: Add a 0.5g counterweight to the upper section of the spindle;

[0099] Bearing wear: Replace the worn bearing with a new one if it has been in operation for 5000 hours.

[0100] Loose spindle: Loosen the connecting bolt between the spindle and the spindle disc by 1 turn (torque decreases from 20 N·m to 10 N·m);

[0101] Damping component failure: Replace the middle section damping rubber ring with a low-quality rubber ring (Shore hardness 40) that is not hard enough.

[0102] Loose root mounting: Loosen the bearing housing fixing bolts by 1 turn (torque decreases from 25 N·m to 12 N·m);

[0103] Mid-section fatigue cracks: Micro-cracks 0.5 mm deep are machined in the mid-section of the spindle by wire cutting.

[0104] II. Experimental Procedure

[0105] 1. Detection System Deployment: Sensors are installed according to the method of this invention. The three detection sections are respectively 50mm from the spindle plate (upper section), 30mm from the center of the damping component (middle section), and 50mm from the bearing seat (lower section). Each section has two circumferential sensors (180° included angle). The main reference sensor (60mm spacing) is installed on the corresponding frame, and the auxiliary reference sensor is installed on the frame load-bearing beam. The sensors are fixed by "45# steel rigid base plate + 6mm polyurethane vibration damping pad (Shore hardness 65)". The cantilever length of the mounting base is 40mm, the flatness of the contact surface is 0.015mm, the perpendicularity of the probe to the spindle is 0.3°, and the detection gap is 1mm.

[0106] 2. Sensor and Algorithm Calibration: The sensor sensitivity coefficient K_s = 8mV / μm was calibrated using a standard vibration table; S was collected under uniform speed and no-load conditions of the spindle. det0 =0.02mm, S ref0 =0.0235mm, calculate the amplitude matching coefficient K1=0.02 / 0.0235≈0.85;

[0107] 3. Data acquisition: Vibration, speed and temperature data were collected under normal conditions and six types of fault conditions, including start-stop, constant speed, yarn change and full roll. Each condition lasted for 30 minutes, with a sampling frequency of 1500Hz. Data from the traditional end single-point detection method was recorded simultaneously for comparison.

[0108] 4. Data processing and analysis: The preprocessing, feature extraction, accuracy compensation and anomaly identification algorithms of this invention are used to process the data, and statistical indicators such as measurement error, fault identification rate, positioning accuracy and early warning time are calculated.

[0109] 5. Anti-interference performance verification: Vibration signals under normal conditions were collected under five environments: no interference, electromagnetic interference, electrostatic interference, frame vibration, and combined interference. The signal-to-noise ratio under different environments was calculated to verify the anti-interference effect of the method of the present invention.

[0110] 6. Long-term stability test: The detection system will run continuously for 1000 hours, and vibration measurement data will be recorded 3 times a day. The fluctuation range of measurement error will be statistically analyzed to verify the long-term stability of the system.

[0111] III. Experimental Data and Results Analysis

[0112] (I) Measurement accuracy verification results

[0113] Table 1. Comparison of peak-to-peak measurement errors under different operating conditions (unit: mm)

[0114]

[0115] As shown in Table 1, the measurement error of the traditional end detection method exceeds 20% under various working conditions, while the measurement error of the method of the present invention is controlled within ±3%, and the measurement accuracy is significantly better than that of the traditional method. Moreover, the error fluctuation is small when switching working conditions, and the adaptability is strong.

[0116] (II) Fault Identification and Location Accuracy Verification Results

[0117] Table 2 Comparison of recognition rate and positioning accuracy for 6 typical fault types

[0118]

[0119] As shown in Table 2, traditional end-point detection methods have an extremely low recognition rate (≤42%) for mid-section and root-related faults (such as damping component failure, mid-section fatigue cracks, and root mounting looseness), while the method of this invention has a recognition rate of ≥91% for all six types of faults, especially significantly improving the ability to identify latent faults. At the same time, the method of this invention can achieve precise positioning of "section + circumferential direction", with a positioning error of ≤±15° and a feature matching similarity of ≥85%, providing clear direction for maintenance.

[0120] (III) Results of Anti-interference Performance Verification

[0121] Table 3. Comparison of signal-to-noise ratio under different interference environments (unit: dB)

[0122]

[0123] A higher signal-to-noise ratio (SNR) indicates less interference affecting the signal and more reliable data. Table 3 shows that in a single interference environment, the SNR of the method described in this invention is more than 50% higher than that of the traditional method; in a complex interference environment, the SNR of the method described in this invention still reaches 39 dB, an improvement of 116.7% compared to the traditional method (18 dB). Its anti-interference performance is significantly better than that of the traditional method, and it can adapt to the complex working environment of the covering yarn machine.

[0124] (iv) Results of verification of the timeliness of early warning

[0125] Table 4 Comparison of Early Warning Times for Typical Faults

[0126]

[0127] As shown in Table 4, the method of the present invention can detect abnormal signals in the early stage of the fault (when the vibration amplitude is only 25%-35% of the shutdown threshold) and issue an early warning more than 24 hours in advance. The warning time is extended by 34.8 times compared with the traditional method, which provides maintenance personnel with sufficient maintenance time and can effectively avoid production accidents caused by the deterioration of the fault.

[0128] (v) Results of long-term stability tests

[0129] After 1000 hours of continuous operation, the vibration measurement error fluctuation range of the method of the present invention was within ±0.5%, the sensor mounting structure was secure, signal acquisition was stable, and no data loss or abnormal interruption occurred. In contrast, after 500 hours of operation, the measurement error of the traditional end-detection method gradually increased to over ±30%, and gap shifts occurred on the sensor surface due to fiber contamination, leading to data distortion. This demonstrates that the method of the present invention has good long-term stability and can meet the monitoring requirements of continuous production of covered yarn machines.

[0130] IV. Experimental Conclusions

[0131] Experimental data shows that the full-section vibration detection and accuracy compensation method for the spindle of the covering yarn machine described in this invention effectively solves the core pain points of existing technologies, such as incomplete detection coverage, weak anti-interference, low accuracy, ambiguous positioning, and untimely early warning, through a full-section multi-section sensing layout, mechanical vibration reduction and anti-interference design, a four-level dynamic accuracy compensation algorithm, and a working condition coupling anomaly identification model. The method controls the vibration measurement error within ±3%, achieves a fault identification accuracy of ≥91%, and a positioning accuracy of "section + circumferential ±15°". In complex interference environments, the signal-to-noise ratio is improved by over 100%, and faults can be predicted more than 24 hours in advance. Furthermore, it is stable and reliable in long-term operation, adapting to the actual needs of high-speed operation, complex working conditions, and harsh environments of covering yarn machines. It can be directly applied to the upgrading and transformation of existing covering yarn machines, demonstrating significant practical application value and promising prospects for promotion.

[0132] V. Other Application Examples

[0133] Textile factory A and textile factory B used the method of this invention to modify 10 and 6 BY980 type covering yarn machines respectively. After installing the corresponding detection system, the overall application data after 6 months of operation is as follows:

[0134] 1. The yarn breakage rate decreased from 3.2 times / thousand spindles·hour before the renovation to 1.1 times / thousand spindles·hour, a reduction of 65.6%;

[0135] 2. Downtime due to spindle failure decreased from 4.5 hours / month / unit before the upgrade to 0.8 hours / month / unit, a reduction of 82.2%;

[0136] 3. The frequency of spindle replacement has been reduced from 12 spindles / month to 3 spindles / month, resulting in a 75% reduction in maintenance costs;

[0137] 4. The pass rate of twist uniformity of covered yarn products increased from 92% before the improvement to 98.5%, and the product quality was significantly improved.

[0138] The above application examples further verify the practicality and economic benefits of the method of the present invention, which can bring significant improvements in production efficiency and cost savings to textile enterprises.

Claims

1. A method for identifying abnormal vibration mode and monitoring the condition of a package yarn machine spindle, characterized by, The method comprises the following steps: S1, detecting system mechanical deployment: three detection sections are arranged along the spindle shaft, corresponding to the upper, middle and lower sections of the spindle respectively; two oil mist-proof eddy current detection sensors are uniformly arranged in the circumferential direction of each detection section, and the circumferential included angle between adjacent sensors is 180°; a main reference sensor of the same type is synchronously installed at the rack position corresponding to each detection sensor, and an auxiliary reference sensor is additionally arranged at each detection section and installed on the stable area of the rack; S2, signal synchronous acquisition: a multi-channel high-speed acquisition module is used to synchronously acquire the vibration signals of the detection sensors, the main reference sensors and the auxiliary reference sensors, and the spindle speed and bearing temperature auxiliary data; the sampling frequency is greater than or equal to 5 times the highest working frequency of the spindle, and the minimum sampling frequency is not less than 1500 Hz; S3, signal preprocessing: three levels of denoising are sequentially performed, the first level is to filter out the high-frequency electromagnetic clutter generated by the driving motor and the frequency converter through RC low-pass filtering with a cutoff frequency of 10 kHz; the second level is to remove the static peak pulse generated by the yarn friction by using the sliding average method; the third level is to remove the abnormal data points caused by the attachment of fibers and lubricating oil mist by using the de-extreme value average algorithm: the extreme value threshold is set to 3 times the average value of the signal segment, when the value of a sampling point exceeds the threshold, it is determined that the abnormal data point is caused by the attachment of fibers and lubricating oil mist, and the average value of the front and rear sampling points is used to replace it, so as to avoid the influence of abnormal points on the subsequent feature extraction accuracy; S4, full-section vibration feature extraction: the preprocessed signals of each detection section are extracted to obtain the time domain features and the frequency domain features, the feature mean values of the two circumferential detection sensors of the same detection section are calculated, and the vibration feature vectors of each section are formed; the consistency of the feature vectors of different sections is compared to preliminarily judge whether the vibration abnormality is a local fault or a whole fault; S5, precision compensation: four levels of compensation processes are sequentially performed, including double-sensor differential basic compensation, working condition adaptive dynamic weight optimization compensation, three-sensor cross-validation compensation and full-section consistency calibration compensation, and finally the real vibration signal of the spindle is obtained; S6, abnormal mode recognition: the preset abnormal vibration feature library of the spindle is called to perform pattern matching, and cross-validation is performed in combination with real-time speed and bearing temperature auxiliary data, and the fault type, occurrence position and fault level are output, and the corresponding early warning mechanism is triggered.

2. The method of identifying abnormal vibration pattern and condition monitoring of a covered yarn machine spool as claimed in claim 1 wherein, In the step S1, the detection sensor and the main reference sensor adopt a "coaxial coplanar + short distance installation" structure, the distance is less than or equal to 80 mm, and all the sensors are fixed through a "rigid base plate + polyurethane damping pad"; the perpendicularity between the sensor probe and the measured surface of the spindle is less than or equal to 0.5°, the detection gap is set as the midpoint of the linear range of the sensor, the mounting seat adopts a short cantilever structure, the cantilever length is less than or equal to 50 mm, and the flatness of the contact surface is less than or equal to 0.02 mm.

3. The method of identifying abnormal vibration pattern and condition monitoring of a covered yarn machine spool as claimed in claim 2, wherein, The polyurethane damping pad in the step S1 has a thickness of 5-8 mm and a Shore hardness of 60-70; the sensor is rigidly fixed by a locking nut + a lock washer, and an elastic clamp is additionally installed at the tail of the sensor for axial limiting to prevent slight displacement under high-speed operation; the detection sensor, the main reference sensor and the auxiliary reference sensor are the same type of oil mist-proof eddy current sensors, with a range of 2 mm, a sensitivity of ≥8 mV / μm and a linearity of ≤±0.5%.

4. The method of identifying abnormal vibration pattern and condition monitoring of a covered yarn machine spool as claimed in claim 1 wherein, The double-sensor differential basic compensation in the step S5 specifically includes: S5.11 amplitude matching coefficient calibration: in the state of uniform speed and no-load of the spindle rod, the real no-load vibration signal S of the detection sensor is collected det0 , the no-load frame vibration signal S of the main reference sensor is collected ref0 , and the amplitude matching coefficient K1=S det0 / S ref0 is calculated S5.12 Phase calibration: real-time calculation of detection sensor signal S triggered by spindle end face preset mark synchronous reference signal det and the phase difference Δ φ of the main reference sensor signal S ref , the phase offset correction of the main reference sensor signal, the formula is: ， ref =S ref ×sin(ωt+Δ φ ), where ω is the signal angular frequency, t is the instantaneous time, and the phase compensation accuracy is ≤0.5°; S5.13 Difference operation: difference operation between the corrected main reference sensor signal and the detection sensor signal, to obtain a preliminary compensation signal S comp1 =S det -K1×S ， ref where S ， ref is the phase-corrected main reference sensor signal.

5. The method of identifying abnormal vibration pattern and condition monitoring of a wrapped yarn machine spool as claimed in claim 4 wherein, The working condition adaptive dynamic weight optimization compensation in the step S5 specifically includes: S5.21, a "working condition compensation weight" mapping library is constructed, and preset amplitude matching coefficient adjustment rules under different working conditions are: when the speed fluctuation is ≥10%, it is the start-stop stage, K1 is adjusted up by 20%-30%; when the package capacity is ≥80% of the rated capacity, it is the full package stage, K1 is adjusted down by 10%-15%; and in the uniform speed stable stage, K1 remains unchanged; S5.22, real-time working condition recognition: the current working condition is judged through the speed signal fluctuation amplitude and the package capacity detection data; S5.23 Dynamic adjustment and abnormality rejection: dynamically adjust K1 according to the identified working conditions to obtain an optimized compensation signal S ， comp1 ; for the yarn changing stage, the sudden vibration caused by the sudden change of yarn tension is identified through the signal slope, the differential compensation is suspended, and the signal mean value of the previous 5 sampling periods is replaced to avoid false compensation.

6. The method of identifying abnormal vibration pattern and condition monitoring of a wrapped yarn machine spool as claimed in claim 5 wherein, The three-sensor cross-verification compensation in the step S5 specifically includes: S5.31 acquiring an auxiliary reference sensor signal S ref2 , calculating a main reference sensor signal S ， ref deviation ΔS = |S ref2 | of the main reference sensor signal S ， ref - S ref2 |; S5.31 deviation judgment and compensation optimization: if ΔS≤3%×full scale of sensor, it is judged that the rack vibration signal is stable, and S ref avg =(S ， ref + S ref2 ) / 2, the difference formula is obtained by substituting the secondary compensation signal S comp2 =S det -k1'×S ref avg , k1' is the amplitude matching coefficient adjusted dynamically; if ΔS>3%×full scale of sensor, it is judged that the rack has local resonance, the abnormal marker is started and the rack vibration warning is triggered, and the compensation weight of the historical same period working condition is used to replace the current k1' for compensation.

7. The method of abnormal vibration mode identification and condition monitoring of a wrapped yarn machine spool as claimed in claim 6, wherein, The full-segment consistency calibration compensation in the step S5 specifically includes: S5.41 Calculate the peak-to-peak value of the three detected section secondary compensation signals S comp2 to obtain the peak-to-peak values of each section, the upper section is A1, the middle section is A2, and the lower section is A3. S5.41, consistency coefficient calculation: the full-segment vibration consistency coefficient C is calculated, ； wherein = (A1 + A2 + A3) / 3 is the full section peak to peak average value; S5.41 Deviation Correction: If C≤5%, the overall state of the spindle is considered stable, and the following applies: As the representative value of the entire vibration at that moment; if C > 5%, locate the abnormal section where the peak value deviates the most, and measure the actual installation gap d of the sensor at that section. act Calculate the standard gap d std The difference Δd=d act -d std Through formula S final = S comp2 +Δd×K s Correction bias, where K s The sensor sensitivity coefficient is obtained from the calibration of a standard vibration table.

8. The abnormal vibration mode identification and condition monitoring method of a wrapped yarn machine spool as claimed in claim 1, wherein, The preset spindle abnormal vibration feature library in the step S6 includes six typical faults of mass imbalance, bearing wear, spindle loosening, damping element failure, root mounting seat loosening and middle segment fatigue crack, each type of fault corresponds to a preset time domain feature threshold, a frequency domain feature frequency interval and an associated auxiliary feature, wherein: Mass imbalance: the full cross-section vibration peak-to-peak value increases with the square of the speed, the characteristic frequency is the spindle main frequency, and the bearing temperature is ≤55℃ without obvious abnormality; Bearing wear: the lower segment cross-section vibration peak-to-peak value is ≥0.08 mm, multiple harmonic peak values appear in the low frequency band, and the bearing temperature is ≥65℃; Spindle loosening: the upper segment cross-section vibration amplitude fluctuation amplitude is ≥30%, without fixed characteristic frequency, and synchronized with the speed fluctuation; Damping element failure: the middle segment cross-section vibration peak-to-peak value is ≥0.07 mm, the vibration decay time is ≥0.5 s, and the vibration is severe in the start-stop stage; Root mounting seat loosening: the lower segment cross-section vibration peak-to-peak value is ≥0.09 mm, the characteristic frequency contains the machine frame resonance frequency component, and the temperature has no obvious abnormality; Middle segment fatigue crack: the middle segment cross-section vibration peak-to-peak value is ≥0.06 mm, the high frequency band harmonic amplitude ratio is ≥0.3, and the vibration waveform is irregular.

9. The method of identifying abnormal vibration pattern and condition monitoring of a covered yarn machine spool as claimed in claim 1 wherein, The abnormal mode recognition in the step S6 further includes fault level division: according to the proportion of the vibration peak-to-peak value exceeding the threshold value and the temperature exceeding the threshold value, the fault is divided into: 10%-20% exceeding the threshold value is slight abnormality, 20%-50% exceeding the threshold value is moderate abnormality, and ≥50% exceeding the threshold value is serious abnormality; the slight abnormality triggers an audible light prompt warning, the moderate abnormality triggers a stop inspection reminder, and the serious abnormality directly triggers an emergency stop.

Images