Wire drawing machine fault detection method based on Internet of Things

Abstract

The invention belongs to the technical field of fault detection, and particularly relates to a wire drawing machine fault detection method based on the Internet of Things. Historical vibration and temperature and real-time data of operation of the wire drawing machine are collected through an Internet of Things sensor and are transmitted to a server through an Internet of Things protocol. After data are subjected to preprocessing denoising such as variational mode decomposition and a FastICA algorithm, a vibration anomaly judgment index (including effective anomaly time period judgment, vibration mode matching degree calculation and index synthesis) and a temperature anomaly judgment index (including correlation index calculation, heat conduction model establishment and index synthesis) are calculated, and then the two indexes are fused to obtain an anomaly index. A prediction model is constructed by using a neural network algorithm, after optimization of a genetic algorithm, a fault probability is output in combination with real-time data, and finally, a probability value and an abnormal index are normalized and averaged to obtain a fault score, so that fault detection and evaluation are realized. The method improves the fault detection accuracy and the equipment operation reliability, and reduces the maintenance cost.

Description

Technical Field

[0001] The present invention belongs to the technical field of fault detection, and in particular relates to a wire drawing machine fault detection method based on the Internet of Things. Background Art

[0002] Wire drawing, a key process for continuously producing slender products from metal or polymer materials under high-temperature, high-tension conditions, is widely used in electronic cables, optical fiber preforms, metal wire, and other fields due to its high production capacity, low cost, and strong adaptability. However, wire drawing machines, operating at high speeds and under heavy loads for long periods of time, are prone to malfunctions such as overheating, abnormal vibration, unstable tension, and equipment wear. If not detected in time, these malfunctions can not only cause product dimensional deviations, surface defects, and even scrap, but can also cause production line downtime, increase maintenance costs, and impact production efficiency.

[0003] Existing wire drawing machine fault detection methods often rely on a single indicator or empirical threshold, making it difficult to distinguish normal fluctuations from potential anomalies. Furthermore, they lack efficient denoising and accurate feature extraction capabilities for the large number of vibration, temperature, and current signals collected by multi-source sensors. This inadequate fusion of multimodal information leads to a high false alarm rate and increased risk of missed alarms. Furthermore, traditional systems often rely on local centralized processing, lacking flexible IoT communication support and online model update mechanisms. This prevents them from fully utilizing historical fault data for intelligent training and adaptive optimization, making predictive maintenance difficult. Summary of the Invention

[0004] In view of the technical problems existing in the above-mentioned background technology, the present invention proposes a wire drawing machine fault detection method based on the Internet of Things.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is: comprising the following steps:

[0006] Data collection: IoT sensors deployed on the wire drawing machine collect vibration data and temperature data during the wire drawing machine's operation; the collected vibration data and temperature data include historical data and real-time data;

[0007] Data transmission: The Internet of Things communication protocol is used to transmit the collected wire drawing machine data to the server;

[0008] Data preprocessing: preprocess the data collected by the server to obtain denoised data;

[0009] Vibration abnormality determination index calculation: obtain the vibration abnormality determination index, including effective abnormal period judgment, vibration pattern matching calculation, and synthetic vibration abnormality determination index;

[0010] Judgment of effective abnormal period: By fitting the vibration amplitude curve and extracting the peaks and valleys, locate the abnormal vibration time period during the operation of the wire drawing machine and exclude the interference of normal process fluctuations;

[0011] Calculation of vibration mode matching degree: Compare the difference between the slope of the vibration curve during the abnormal period and the historical reference slope, and convert the slope deviation into the mode matching degree through an exponential function to quantify the deviation degree of the current vibration mode from the normal mode;

[0012] Synthesis of vibration abnormal judgment index: Integrate the vibration amplitude standardized score, mode matching degree and sample entropy deviation to construct the vibration abnormal judgment index;

[0013] Calculation of temperature abnormal judgment index: Obtain the temperature abnormal judgment index, including correlation index calculation, establish a heat conduction model, and synthesize the temperature abnormal judgment index;

[0014] Calculation of correlation index: By calculating the differences in local and overall correlations between temperature and vibration data, quantify the deviation degree of the temperature change trend from the historical law, and identify abnormal temperature fluctuations not related to vibration;

[0015] Establishment of heat conduction model: Construct a heat conduction prediction model based on ambient temperature, load rate and vibration effective value, and distinguish temperature anomalies through the deviation rate between measured temperature and theoretical temperature;

[0016] Synthesis of temperature abnormal judgment index: Integrate the temperature standardized score, correlation index and heat conduction deviation rate to form the temperature abnormal judgment index;

[0017] Abnormal index: Calculate the abnormal index from the vibration abnormal judgment index and the temperature abnormal judgment index;

[0018] Intelligent prediction: Use a neural network model to predict probability values, including model construction, model training, and model prediction;

[0019] Model construction: Use a neural network algorithm to construct a prediction model suitable for wire drawing machine faults, input temperature and vibration data, and the output is the fault probability;

[0020] Model training: Use historical data to train the constructed model and optimize it through a genetic algorithm;

[0021] Model prediction: Input real-time data into the trained model to perform prediction and obtain the predicted fault probability;

[0022] Fault detection: Calculate the total fault score based on the predicted probability value and the abnormal index, and detect and evaluate the fault condition of the wire drawing machine.

[0023] Preferably, the specific implementation steps for preprocessing the data collected by the server to obtain denoised data are as follows:

[0024] First, for the original data , initialize the number of modes K and perform variational mode decomposition to obtain K intrinsic mode components , define the component energy difference , and make a determination and output intrinsic mode components through ;

[0025] Construct an observation matrix , use the FastICA algorithm to estimate the separation matrix W, satisfying , and output independent components ;

[0026] For each independent component , determine the subsequence length and time delay according to the grid search optimization result, construct a phase space matrix, calculate the permutation entropy, and when the permutation entropy exceeds the set noise component threshold, remove it and output the denoising candidate components ;

[0027] Perform wavelet decomposition on each denoising candidate component to obtain wavelet coefficients at different scales, calculate the noise variance based on the statistics of the fault-free condition data at each scale , where is the scale index. According to the set threshold , where N is the length of the data signal, perform hard threshold processing on the wavelet coefficients, and then reconstruct through wavelet inverse transform to output the denoised mode components ;

[0028] Calculate the ratio of the total energy of the denoised signal to the energy of the original signal and the Pearson correlation coefficient between the denoised signal and the historical normal condition signal. If both the ratio and the correlation coefficient meet the requirements, the verification is passed and the final denoised data is output , otherwise, readjust the subsequence length and time delay parameters

[0029] Preferably, the specific implementation of making a determination and output through intrinsic mode components is as follows:

[0030] First, calculate the component energy difference , and the calculation method is: , where N is the length of the original data;

[0031] Judge that when first exceeds the mutation threshold , determine that , stop decomposition and output intrinsic mode components.

[0032] Preferably, the specific calculation of the vibration anomaly determination index is implemented as follows:

[0033] Perform non-linear least squares fitting on the vibration data, extract the peaks and valleys of the vibration amplitude fitting curve, define the time period from the peak to the adjacent valley as the vibration anomaly candidate time period. At the same time, calculate the energy ratio in the 100 - 500 Hz frequency band through short-time Fourier transform. When the sudden increase in energy ratio exceeds 2 times the historical average and the amplitude drop exceeds 3 times the standard deviation, it is determined as an effective anomaly time period;

[0034] Calculate the absolute difference between the slope of the vibration curve in the anomaly time period and the slope of the normal data, and obtain the vibration mode matching degree through the exponential function ;

[0035] Standardize the vibration amplitude to obtain the standard score , and synthesize the vibration anomaly determination index by combining the mode matching degree and the deviation of the vibration signal sample entropy. The calculation method is: , where represents the sample entropy deviation, is the weight coefficient.

[0036] Preferably, the calculation of the vibration anomaly determination index further includes: constructing a spatial vibration field through multiple vibration sensors, and calculating the spatial gradient of the vibration amplitude , when is greater than the set threshold, amplify the vibration anomaly determination index according to .

[0037] Preferably, the specific calculation of the temperature anomaly determination index is implemented as follows:

[0038] Calculate the overall correlation of the local correlation of temperature and vibration in the current adjacent time period with the full historical data, and obtain the correlation index of the two through the exponential function ;

[0039] Establish a heat conduction prediction model , where represents the ambient temperature, L is the equipment load rate, is the effective value of vibration; calculate the deviation rate of the measured data from the predicted temperature , when is greater than the set threshold, generate a correction term , and correct the deviation rate to obtain the corrected deviation rate ;

[0040] Standardize the temperature data to obtain the standard score , combining the correlation index and the corrected deviation rate to obtain a synthesized temperature anomaly determination index, the calculation method is .

[0041] Preferably, the implementation of calculating the anomaly index from the vibration anomaly determination index and the temperature anomaly determination index is as follows:

[0042] Calculate and mutual information entropy of , and perform normalization, and construct a weight function based on the mutual information entropy: , , where are respectively and historical variances of is the weight of the temperature anomaly determination index, is the weight of the vibration anomaly determination index, and calculate the anomaly index by weighted summation.

[0043] Preferably, the fault detection: the implementation of calculating the total fault score from the predicted probability value and the anomaly index is to normalize the probability value and the anomaly index, and take their average as the total fault score, and evaluate according to the preset fault threshold.

[0044] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0045] 1. Using variational mode decomposition, FastICA algorithm, permutation entropy analysis, wavelet hard threshold denoising, etc. to preprocess the original data, removing noise and retaining fault features.

[0046] 2. Constructing a vibration anomaly determination index and a temperature anomaly determination index, and dynamically fusing the two based on mutual information entropy and historical variance to obtain an anomaly index, realizing precise quantification of faults.

[0047] 3. Normalizing the predicted probability and the anomaly index and then taking the average to obtain the total fault score, realizing dual fault assessment of physical feature determination and intelligent prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0049] Figure 1 It is a structural flowchart of a wire drawing machine fault detection method based on the Internet of Things. Detailed Implementation Modes

[0050] In order to more clearly understand the above-mentioned objects, features, and advantages of the present invention, the present invention will be further described below with reference to the accompanying drawings and embodiments. It should be noted that, without conflict, the embodiments of the present application and the features in the embodiments can be combined with each other.

[0051] In the following description, many specific details are set forth in order to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Therefore, the present invention is not limited by the specific embodiments disclosed in the following specification.

[0052] Embodiment. During the continuous wire drawing production process of a wire drawing machine for metal or polymer materials, faults such as excessive temperature, abnormal vibration, and unstable tension often occur due to high-speed, high-temperature, and high-load operation. Traditional fault detection often relies on manual experience or simple threshold judgment, lacking the ability to perform real-time comprehensive analysis of multi-source data such as vibration and temperature, and unable to make full use of historical fault data for intelligent modeling and adaptive optimization, resulting in a high false alarm rate, a high risk of missed alarms, frequent production line shutdowns, high maintenance costs, and low production efficiency. In the face of the above problems, the present invention proposes a fault detection method for a wire drawing machine based on the Internet of Things. The specific process is as Figure 1 shown. In order to achieve early perception and accurate warning of potential faults of the wire drawing machine and improve the reliability and maintenance efficiency of equipment operation, the present invention adopts a comprehensive fault detection solution that combines multi-sensor real-time collection and transmission based on the Internet of Things, advanced signal denoising and feature extraction, physical anomaly index determination, and intelligent prediction model fusion, so as to realize efficient, reliable, and scalable fault detection and warning of the wire drawing machine under actual production conditions.

[0053] First, data collection is carried out: vibration data and temperature data during the operation of the wire drawing machine are collected through Internet of Things sensors deployed on the wire drawing machine; the collected vibration data and temperature data include historical data and real-time data.

[0054] Then, the Internet of Things communication protocol is used to transmit the collected wire drawing machine data to the server, thereby ensuring the real-time and reliability of the data; the data collection frequency is set according to the specific working conditions of the wire drawing machine, and preliminary verification and timestamp marking are carried out at the edge end.

[0055] Next, in order to achieve the purpose of signal quality guarantee and removal of environmental noise interference, the collected data is preprocessed to obtain denoised data. Specifically, first, the original data , the number of modes K is initialized and variational mode decomposition is performed to obtain K intrinsic mode components , the component energy difference is defined , and the determination output is carried out through eigenmode component, and the determination output through is as follows The specific implementation of the eigenmode component is: First, calculate the component energy difference , and the calculation method is: , where N is the length of the original data; judge when first exceeds the mutation threshold , it is determined that , stop decomposition, and output eigenmode components; construct the observation matrix , use the FastICA algorithm to estimate the separation matrix W, satisfying , and output independent components ; for each independent component , determine the subsequence length and time delay according to the grid search optimization result, construct the phase space matrix, calculate the permutation entropy, and when the permutation entropy exceeds the set noise component threshold, remove it and output the denoising candidate component ; perform wavelet decomposition on each denoising candidate component to obtain wavelet coefficients at different scales, calculate the noise variance based on the statistics of the fault-free condition data at each scale , where is the scale index, according to the set threshold , where N is the length of the data signal, perform hard threshold processing on the wavelet coefficients, and reconstruct through wavelet inverse transform to output the denoised mode component ; calculate the ratio of the total energy of the denoised signal to the energy of the original signal and the Pearson correlation coefficient between the denoised signal and the historical normal condition signal. If both the ratio and the correlation coefficient meet the requirements, the verification is passed and the final denoised data is output , otherwise, readjust the subsequence length and time delay parameters. This data processing can improve the fineness of signal separation; realize noise elimination based on permutation entropy and wavelet hard threshold, and improve the ability to retain weak fault features; the dynamic parameter adjustment and verification steps ensure reliable denoising and adapt to changes in the equipment state.

[0056] After denoising, in order to accurately locate the vibration abnormal period and calculate the abnormal judgment index, first calculate the vibration abnormal judgment index: obtain the vibration abnormal judgment index, including the judgment of the effective abnormal period, the calculation of the vibration mode matching degree, and the synthesis of the vibration abnormal judgment index. Judgment of the effective abnormal period: By fitting the vibration amplitude curve and extracting the wave peaks and valleys, locate the vibration abnormal time period during the operation of the wire drawing machine and exclude the interference of normal process fluctuations. Calculation of the vibration mode matching degree: Compare the difference between the slope of the vibration curve in the abnormal period and the historical reference slope, and convert the slope deviation into the mode matching degree through an exponential function to quantify the deviation degree of the current vibration mode from the normal mode. Synthesis of the vibration abnormal judgment index: Integrate the vibration amplitude standard score, the mode matching degree, and the sample entropy deviation to construct the vibration abnormal judgment index. Specifically, perform non-linear least squares fitting on the vibration data, extract the wave peaks and valleys of the vibration amplitude fitting curve, define the period from the wave peak to the adjacent wave valley as the vibration abnormal candidate period. At the same time, calculate the energy ratio of the 100 - 500 Hz frequency band through short-time Fourier transform. When the energy ratio suddenly increases by more than 2 times the historical mean and the amplitude drop exceeds 3 times the standard deviation, it is determined as the effective abnormal period; calculate the absolute difference between the slope of the vibration curve in the abnormal period and the slope of the normal data, and obtain the vibration mode matching degree through an exponential function ; Standardize the vibration amplitude to obtain the standard score , combine the mode matching degree and the sample entropy deviation of the vibration signal to synthesize the vibration abnormal judgment index, and the calculation method is: , where represents the sample entropy deviation, is the weight coefficient. The calculation of the vibration abnormal judgment index also includes: constructing a spatial vibration field through multiple vibration sensors and calculating the spatial gradient of the vibration amplitude , when is greater than the set threshold, amplify the vibration abnormal judgment index according to .

[0057] Then calculate the temperature abnormal judgment index to obtain the temperature abnormal judgment index, including the calculation of the correlation index, establishing a heat conduction model, and synthesizing the temperature abnormal judgment index; Calculation of the correlation index: By calculating the differences in the local correlation and the overall correlation between the temperature and the vibration data, quantify the deviation degree of the temperature change trend from the historical law and identify the abnormal temperature fluctuations not related to vibration; Establishing a heat conduction model: Construct a heat conduction prediction model based on the ambient temperature, load rate, and vibration effective value, and distinguish the temperature abnormality through the deviation rate between the measured temperature and the theoretical temperature; Synthesis of the temperature abnormal judgment index: Integrate the temperature standard score, the correlation index, and the heat conduction deviation rate to form the temperature abnormal judgment index. Specifically, calculate the local correlation between the temperature and the vibration in the current adjacent period and the overall correlation of the full historical data, and obtain the correlation index of the two through an exponential function ; A heat conduction model is constructed using machine learning methods , specifically, the random forest regression algorithm is selected, and the input parameters include: . The model is trained with historical data to establish the mapping relationship between the input parameters and the measured temperature, and the predicted temperature is output, providing the input for the deviation rate . Then, the deviation rate between the measured data and the predicted temperature is calculated , are the measured temperature and the predicted temperature respectively. When is greater than the set threshold, a correction term is generated. The implementation of the correction term is , where is the reference temperature. The deviation rate is corrected to obtain the corrected deviation rate ; The temperature data is standardized to obtain the standard score . Combining the correlation index and the corrected deviation rate, a synthetic temperature anomaly determination index is obtained, and the calculation method is .

[0058] Among them, the calculation of the correlation index is as follows: First, a sliding window is set, and the Pearson correlation coefficient between the current temperature and the vibration effective value within the window is calculated to obtain the local correlation; then, the overall correlation coefficient of the historical data is calculated. The absolute difference between the local and overall correlations is obtained, and it is normalized by dividing it by the standard deviation of the historical correlation difference, and then substituted into the exponential function to obtain the correlation index, where is the absolute difference between the local and overall correlations, is the standard deviation of the historical correlation difference.

[0059] Next, in order to comprehensively reflect the dual anomaly risks of vibration and temperature, an anomaly index fusion scheme is adopted. The anomaly index: The anomaly index is calculated for the vibration anomaly determination index and the temperature anomaly determination index. Specifically, first, the historical statistics of the vibration anomaly determination index and the temperature anomaly determination index are performed, and their historical variances and distribution characteristics are calculated; then, the mutual information entropy between the two is calculated, and a normalized weight function is constructed based on the mutual information entropy and the variance information, so that in different operating conditions and fault types, the fusion weights of the two anomaly determination indexes can adaptively change according to their statistical information; the vibration anomaly determination index and the temperature anomaly determination index are fused according to this weight function to obtain the anomaly index. The implementation of calculating the anomaly index for the vibration anomaly determination index and the temperature anomaly determination index is: Calculate and of the mutual information entropy , and perform normalization. Based on the mutual information entropy, a weight function is constructed: , , where respectively and historical variances, and calculate the anomaly index R by weighted summation, .

[0060] Then, in order to achieve the purpose of intelligent prediction and adaptive optimization based on historical annotation data, a neural network prediction scheme is adopted. Intelligent prediction: Use a neural network model to predict probability values, including model construction, model training, and model prediction; Model construction: Use a neural network algorithm to construct a prediction model applicable to the faults of the wire drawing machine, input temperature and vibration data, and the output is the fault probability; Model training: Use historical data to train the constructed model and optimize it through a genetic algorithm. Specifically, use the genetic algorithm to optimize network hyperparameters such as the number of layers, the number of nodes, the learning rate, and the regularization coefficient; Model prediction: Input real-time data into the trained model to obtain the predicted fault probability through prediction.

[0061] Finally, fault detection: Calculate the total fault score by calculating the predicted probability value and the anomaly index, and detect and evaluate the fault situation of the wire drawing machine. In order to achieve the purpose of collaborative verification and fusion warning of the prediction result and physical anomaly determination, a prediction probability and anomaly index fusion scheme is adopted: The implementation of calculating the total fault score by calculating the predicted probability value and the anomaly index is to normalize the probability value and the anomaly index and take their average as the total fault score, and evaluate according to the preset fault threshold.

[0062] The above is only a preferred embodiment of the present invention, and it is not a limitation of the present invention in other forms. Any person skilled in the art may use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A method for detecting faults of a wire drawing machine based on the Internet of Things, characterized in that, It includes the following steps: Data acquisition: Vibration data and temperature data during the operation of the wire drawing machine are collected through Internet of Things sensors deployed on the wire drawing machine; the collected vibration data and temperature data include historical data and real-time data; Data transmission: The data of the wire drawing machine collected is transmitted to the server using the Internet of Things communication protocol; Data preprocessing: The data collected by the server is preprocessed to obtain denoised data; Calculation of vibration anomaly determination index: Obtain the vibration anomaly determination index, including judgment of effective anomaly period, calculation of vibration mode matching degree, and synthesis of vibration anomaly determination index; Calculation of temperature anomaly determination index: Obtain the temperature anomaly determination index, including calculation of correlation index, establishment of heat conduction model, and synthesis of temperature anomaly determination index; Anomaly index: Calculate the anomaly index by calculating the vibration anomaly determination index and the temperature anomaly determination index; Intelligent prediction: Use a neural network model to predict probability values, including model construction, model training, and model prediction; Model construction: Use a neural network algorithm to construct a prediction model suitable for wire drawing machine faults, input temperature and vibration data, and the output is the fault probability; Model training: Use historical data to train the constructed model and optimize it through a genetic algorithm; Model prediction: Input real-time data into the trained model to perform prediction and obtain the predicted fault probability; Fault detection: Calculate the total fault score through the probability value and anomaly index obtained by prediction, and detect and evaluate the fault situation of the wire drawing machine.

2. The method for detecting faults of a wire drawing machine based on the Internet of Things according to claim 1, characterized in that, The implementation of the calculation of the vibration anomaly determination index: Obtain the vibration anomaly determination index, including judgment of effective anomaly period, calculation of vibration mode matching degree, and synthesis of vibration anomaly determination index is as follows: Judgment of effective anomaly period: Locate the vibration anomaly time period during the operation of the wire drawing machine by fitting the vibration amplitude curve and extracting wave peaks and valleys, and exclude the interference of normal process fluctuations; Calculation of vibration mode matching degree: Compare the difference between the slope of the vibration curve during the anomaly period and the historical reference slope, and convert the slope deviation into a mode matching degree through an exponential function to quantify the deviation degree of the current vibration mode from the normal mode; Synthesis of vibration anomaly determination index: Integrate the vibration amplitude standardized score, mode matching degree, and sample entropy deviation to construct the vibration anomaly determination index.

3. The method for detecting faults of a wire drawing machine based on the Internet of Things according to claim 1, characterized in that, The implementation of the calculation of the temperature anomaly determination index: Obtain the temperature anomaly determination index, including calculation of correlation index, establishment of heat conduction model, and synthesis of temperature anomaly determination index is as follows: Calculation of correlation index: Quantify the deviation degree of the temperature change trend from the historical law by calculating the differences in local correlation and overall correlation between temperature and vibration data, and identify abnormal temperature fluctuations not related to vibration; Establishment of heat conduction model: Construct a heat conduction prediction model based on ambient temperature, load rate, and vibration effective value, and distinguish temperature anomalies through the deviation rate between measured temperature and theoretical temperature; Synthesis of temperature anomaly determination index: Integrate the temperature standardized score, correlation index, and heat conduction deviation rate to form the temperature anomaly determination index.

4. The method for detecting faults of a wire drawing machine based on the Internet of Things according to claim 1, wherein, The specific implementation steps of preprocessing the data collected by the server to obtain denoised data are as follows: First, for the original data , initialize the number of modes K and perform variational mode decomposition to obtain K intrinsic mode components , define the component energy difference , and make a determination and output through intrinsic mode components; Construct the observation matrix , and use the FastICA algorithm to estimate the separation matrix W, satisfying , and output independent components ; For each independent component , determine the subsequence length and time delay according to the optimized result of grid search, construct a phase space matrix, calculate the permutation entropy, and when the permutation entropy exceeds the set noise component threshold, remove it and output the denoising candidate component ; For each denoising candidate component perform wavelet decomposition to obtain wavelet coefficients at different scales, and calculate the noise variance based on the statistics of the fault-free condition data at each scale , where is the scale index. According to the set threshold , where N is the length of the data signal, perform hard threshold processing on the wavelet coefficients, and then reconstruct through inverse wavelet transform to output the denoised modal component ; Calculate the ratio of the total energy of the denoised signal to the energy of the original signal and the Pearson correlation coefficient between the denoised signal and the historical normal condition signal. If both the ratio and the correlation coefficient meet the requirements, the verification passes and the final denoised data is output. , otherwise, readjust the subsequence length and time delay parameters.

5. A fault detection method for a wire drawing machine based on the Internet of Things according to claim 4, characterized in that, The determination output through is specifically implemented as follows: the following eigenmode components First, calculate the component energy difference , and the calculation method is as follows: , where N is the length of the original data; Judge when first exceeds the mutation threshold , determine , stop decomposition, and output intrinsic mode components.

6. The method for detecting faults of a wire drawing machine based on the Internet of Things according to claim 2, characterized in that, The specific calculation of the vibration anomaly determination index is as follows: Perform non-linear least squares fitting on the vibration data, extract the peaks and valleys of the vibration amplitude fitting curve, define the time period from the peak to the adjacent valley as the vibration anomaly candidate time period. At the same time, calculate the energy ratio of the 100 - 500 Hz frequency band through short-time Fourier transform. When the sudden increase in energy ratio exceeds 2 times the historical mean and the amplitude drop exceeds 3 times the standard deviation, it is determined as an effective anomaly time period; Calculate the absolute difference between the slope of the vibration curve during the abnormal period and the slope of the normal data, and obtain the vibration mode matching degree through the exponential function ; Normalize the vibration amplitude to obtain a standard score , and combine the pattern matching degree and the vibration signal sample entropy deviation to synthesize a vibration anomaly determination index. The calculation method is as follows: , where represents the sample entropy deviation, is the weight coefficient.

7. The method for detecting faults of a wire drawing machine based on the Internet of Things according to claim 6, wherein The calculation of the vibration anomaly determination index further includes: constructing a spatial vibration field through multiple vibration sensors and calculating the spatial gradient of the vibration amplitude , when is greater than a set threshold value, the vibration anomaly determination index is amplified according to ​ 8. A fault detection method for a wire drawing machine based on the Internet of Things according to claim 3, characterized in that, The specific calculation of the temperature anomaly determination index is as follows: Calculate the overall correlation of the local correlation full history data of temperature and vibration in the current adjacent time period, and obtain the correlation index of the two through an exponential function ; Establish a heat conduction prediction model , where represents the ambient temperature, L is the device load rate, is the effective vibration value; calculate the deviation rate between the measured data and the predicted temperature , when is greater than the set threshold, generate a correction term , and correct the deviation rate to obtain the corrected deviation rate ; Standardize the temperature data to obtain the standard score , combine the correlation index and the corrected deviation rate to obtain the synthetic temperature anomaly determination index, and the calculation method is .

9. A method for detecting faults of a wire drawing machine based on the Internet of Things according to claim 1, characterized in that, The calculation of the anomaly index by calculating the vibration anomaly determination index and the temperature anomaly determination index is as follows: Calculation and mutual information entropy , and perform normalization, and construct a weight function based on the mutual information entropy: , , where are respectively and historical variances, is the weight of the temperature anomaly determination index, is the weight of the vibration anomaly determination index, and calculate the anomaly index by the method of weighted summation.

10. A method for detecting faults of a wire drawing machine based on the Internet of Things according to claim 1, characterized in that, The fault detection: The implementation of calculating the total fault score by calculating the predicted probability value and the anomaly index is to normalize the probability value and the anomaly index, and take their average as the total fault score, and evaluate according to the preset fault threshold.

Images