Wire drawing machine state monitoring and early warning method and system based on artificial intelligence

Abstract

The invention belongs to the technical field of wire drawing machines, and particularly relates to a wire drawing machine state monitoring and early warning method and system based on artificial intelligence, and the method comprises the steps: collecting surface vibration signals and material parameters of a metal wire in real time, extracting the amplitude, frequency and phase characteristic parameters of the vibration signals, carrying out the cross comparison with the material parameters, and building a real-time incidence relation; accurate monitoring of the equipment state is realized. According to the method, the vibration abnormity judgment threshold value can be dynamically adjusted to adapt to natural fluctuation in the production process, and when it is detected that the difference value exceeds the preset fault-tolerant interval, the historical fluctuation data is used for correcting the difference value weight coefficient, so that it is ensured that early warning is triggered only when a real abnormal condition occurs. According to the method, the accuracy of anomaly detection is remarkably improved, the false alarm rate is reduced, the production continuity and the product quality stability are guaranteed, and meanwhile the reliability and practicability of an early warning system are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of wire drawing machines, and particularly relates to a method and system for monitoring and warning the state of a wire drawing machine based on artificial intelligence. Background Art

[0002] During the production process of metal wires, the operating state of the wire drawing machine is crucial for product quality. Existing wire drawing machine state monitoring systems usually rely on setting fixed vibration thresholds to monitor whether the equipment is abnormal. Such systems mainly collect vibration signals in real time through sensors installed on the wire drawing machine and judge whether there is abnormal vibration according to the preset fixed thresholds. However, this method has a significant technical problem: when there are slight changes in the metal wire material (such as slight changes in trace element composition or changes in microstructure), it will cause changes in the vibration mode during the wire drawing process, and this change may lead to misjudgment, resulting in unnecessary warning triggers or missed reporting of actual problems.

[0003] Due to the instability of the metal wire material and natural fluctuations in the production process, fixed vibration thresholds are difficult to adapt to all situations. Therefore, the prior art often cannot effectively distinguish between vibration changes caused by normal process fluctuations and vibration abnormalities truly caused by equipment failures or impending failures, thus affecting production efficiency and product quality. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for monitoring and warning the state of a wire drawing machine based on artificial intelligence, which not only improves the accuracy of equipment state monitoring but also ensures product quality without affecting production efficiency, so as to solve the problems proposed in the above background art.

[0005] To achieve the above purpose, the present invention adopts the following technical solutions: A method for monitoring and warning the state of a wire drawing machine based on artificial intelligence, comprising the following steps: Real-time collect the vibration signal on the surface of the metal wire during the operation of the wire drawing machine and the material parameters of the corresponding process section; extract the amplitude, frequency, and phase characteristic parameters of the vibration signal, cross-compare the material parameters with the characteristic parameters, and establish a real-time correlation relationship; dynamically adjust the vibration abnormality determination threshold according to the real-time correlation relationship, calculate the difference between the current characteristic parameters and the dynamic threshold, if the difference exceeds the preset error tolerance interval, then extract the historical fluctuation data of the material parameters; correct the weight coefficient of the difference in combination with the historical fluctuation data, and trigger a warning when the corrected difference continuously exceeds the product of the weight coefficient and the dynamic threshold.

[0006] Preferably, the method for real-time collecting the vibration signal on the surface of the wire and the material parameters of the corresponding process section during the operation of the wire drawing machine includes: contacting the surface of the wire with a piezoelectric sensor array to collect the vibration signal and record the timestamp sequence; performing wavelet packet decomposition on the vibration signal, extracting the decomposition coefficients of each order as characteristic parameters, and ensuring that the characteristic parameters meet the set conditions; reading the real-time temperature and tension values of the process section, synchronizing the decomposition coefficients with the temperature and tension through the timestamp, and calculating the weighted comprehensive value of the synchronized data, where the weight coefficient is determined by the mean value of the previous several sampling periods.

[0007] Preferably, the method for extracting the amplitude, frequency, and phase characteristic parameters of the vibration signal includes: performing a fast Fourier transform on the vibration signal to convert the time-domain signal into a frequency-domain signal; calculating the frequency and corresponding amplitude corresponding to the peak value in the frequency-domain signal; determining the phase information using the frequency-domain signal and the peak frequency, and combining the amplitude, frequency, and phase to form a characteristic vector.

[0008] Preferably, the method for cross-comparing the material parameters with the characteristic parameters and establishing a real-time correlation relationship includes: obtaining the weighted comprehensive value and the characteristic vector, and for each sampling period, calculating the difference degree between the material parameters and the characteristic vector; adjusting the correlation factor based on the difference degree, and using the correlation factor to update the real-time correlation relationship between the material parameters and the characteristic parameters.

[0009] Preferably, the method for dynamically adjusting the vibration anomaly determination threshold according to the real-time correlation relationship includes: obtaining the real-time correlation relationship, calculating the adjustment factor based on the real-time correlation relationship; using the adjustment factor to adjust the initially set vibration anomaly determination threshold; updating the vibration anomaly determination threshold to the monitoring system so that the subsequent evaluation of the vibration signal is based on the updated threshold for comparison.

[0010] Preferably, the method for calculating the difference between the current characteristic parameters and the dynamic threshold includes: obtaining the characteristic vector and the updated vibration anomaly determination threshold; for each characteristic parameter, calculating the difference between it and the dynamic threshold; comparing each difference with a preset error tolerance interval, and if it exceeds the range, marking it as a potential anomaly point, and then calculating the cumulative value of the differences corresponding to all characteristic parameters.

[0011] Preferably, the method for extracting the historical fluctuation data of the material parameters includes the following steps: when the difference exceeds the preset error tolerance interval, triggering a historical data extraction instruction; according to the trigger instruction, retrieving the historical records of the corresponding material parameters in the past several periods from the database; calculating the standard deviation and mean value of the historical fluctuation data, and determining the fluctuation range of the material parameters based on the standard deviation and mean value.

[0012] Preferably, correcting the weight coefficient of the difference in combination with the historical fluctuation data includes: obtaining the standard deviation, mean value, and the fluctuation range of the material parameters, and calculating a correction factor based on the historical fluctuation data; using the correction factor to adjust the weight of the difference and updating the weight coefficient to ensure dynamic adaptation to material changes over time.

[0013] Preferably, when the corrected difference continuously exceeds the product of the weight coefficient and the dynamic threshold, an alarm is triggered, including: obtaining the adjusted difference and the weight coefficient, as well as the updated vibration anomaly determination threshold, calculating the product of the weight coefficient and the dynamic threshold as a new comparison benchmark; monitoring the differences calculated for each cycle within a continuous plurality of sampling cycles, if the differences in all cycles satisfy the condition of being greater than the new comparison benchmark, then prepare to trigger an alarm signal; when the condition is satisfied within a continuous plurality of cycles, calculate the cumulative over-difference, if the cumulative over-difference exceeds a preset safety limit, then officially issue an alarm signal.

[0014] On the other hand, the present invention proposes an artificial intelligence-based wire drawing machine condition monitoring system, including: A signal and parameter acquisition module for performing real-time acquisition of the vibration signal on the surface of the wire and the material parameters of the corresponding process section during the operation of the wire drawing machine; A feature extraction and correlation analysis module for performing extraction of the amplitude, frequency, and phase characteristic parameters of the vibration signal, cross-comparing the material parameters with the characteristic parameters, and establishing a real-time correlation relationship; A dynamic threshold adjustment and anomaly detection module for performing dynamic adjustment of the vibration anomaly determination threshold according to the real-time correlation relationship, calculating the difference between the current characteristic parameters and the dynamic threshold, if the difference exceeds a preset tolerance interval, then extracting the historical fluctuation data of the material parameters; A weight correction and alarm trigger module for performing correction of the weight coefficient of the difference in combination with the historical fluctuation data and triggering an alarm when the corrected difference continuously exceeds the product of the weight coefficient and the dynamic threshold.

[0015] The technical effects and advantages of the present invention: An artificial intelligence-based wire drawing machine condition monitoring and warning method and system proposed by the present invention has the following advantages compared with the prior art: The method for monitoring and warning the state of a wire drawing machine based on artificial intelligence proposed by the present invention realizes precise monitoring of the equipment state by collecting the vibration signals and material parameters on the surface of the wire in real time, extracting the amplitude, frequency and phase characteristic parameters of the vibration signals, and cross-comparing them with the material parameters to establish a real-time correlation relationship. This method can dynamically adjust the vibration anomaly determination threshold to adapt to the natural fluctuations in the production process, and when the detected difference exceeds the preset error tolerance range, use the historical fluctuation data to correct the difference weight coefficient to ensure that the warning is triggered only when a real anomaly occurs. This method significantly improves the accuracy of anomaly detection, reduces the false alarm rate, ensures the continuity of production and the stability of product quality, and at the same time improves the reliability and practicality of the warning system. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a flowchart of a method for monitoring the state of a wire drawing machine based on artificial intelligence according to the present invention; Figure 2 is a block diagram of a system for monitoring the state of a wire drawing machine based on artificial intelligence according to the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0017] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0018] The present invention provides a method for monitoring and warning the state of a wire drawing machine based on artificial intelligence as shown in Figure 1 By dynamically adjusting the vibration anomaly determination threshold and combining historical fluctuation data to correct the weight coefficient of the difference, it can more accurately identify real abnormal conditions, reduce the misjudgment rate, and improve the reliability of the warning system. This method not only improves the accuracy of equipment state monitoring, but also can ensure the quality of products without affecting production efficiency, as follows: In this embodiment, a method and system for monitoring and warning the state of a wire drawing machine based on artificial intelligence includes the following steps: Step 1: Collect the vibration signals on the surface of the wire and the material parameters of the corresponding process section during the operation of the wire drawing machine in real time, specifically including the following steps: During the operation of the wire drawing machine, contact the surface of the wire through a piezoelectric sensor array. These sensors can convert mechanical vibrations into electrical signals to collect vibration signals and record the time stamp sequence T_s; Perform wavelet packet decomposition on the vibration signal to decompose the original signal into multiple frequency sub-bands. Then extract the K-th order decomposition coefficient C_k as a characteristic parameter, satisfying SUM|C_k|^2 <= 3SIG^2, where SIG is the standard deviation of the vibration signal, and SUM represents the sum of the squares of all selected coefficients not exceeding the square of three times the standard deviation. 3*SIG^2 is a threshold used to limit the energy level of the selected characteristic parameters and avoid the influence of excessive noise.

[0019] Simultaneously monitor and record the real-time temperature Temp and tension Str in the process section during the wire drawing process, and synchronize the C_k with Temp and Str at the millisecond level through the time stamp to ensure the consistency and accuracy of data analysis.

[0020] Based on the synchronized data, calculate the weighted composite value W = SUM(C_k^2 / Temp^0.5 + 0.8Str) of the synchronized data. The weight coefficient is obtained from the average value of the data in the past N sampling periods to adapt to changes under different working conditions. C_k^2 / Temp^0.5 takes into account the relationship between vibration energy and temperature. The higher the temperature, the smaller the possible influence of vibration. 0.8*Str introduces the tension factor, reflecting the influence of tension on the vibration state. SUM represents the sum of all relevant terms to obtain the final comprehensive evaluation value.

[0021] Step 2: Extract the amplitude, frequency, and phase characteristic parameters of the vibration signal, specifically including the following steps: First, perform a fast Fourier transform (FFT) on the vibration signal on the surface of the wire of the wire drawing machine collected. This process converts the original time-domain signal into a frequency-domain signal, enabling more clearly identifying the frequency components in the signal. Obtain the spectral distribution S_f, where f represents frequency; Calculate the frequency F_max corresponding to the peak value and the corresponding amplitude A_max in the spectral distribution S_f, satisfying A_max = max(S_f); max(S_f) represents the maximum value found from the spectral distribution S_f, that is, the amplitude of the strongest frequency component in the vibration signal.

[0022] Use the spectral distribution S_f and the F_max to determine the phase information Ph, which is calculated by the formula Ph = arctan(Im(S_f) / Re(S_f)), where Im(S_f) is the imaginary part value at the corresponding frequency, and Re(S_f) is the real part value. This formula is used to calculate the phase angle of the complex spectrum at a specific frequency, reflecting the relative time delay or advance of the signal at this frequency.

[0023] Combine the A_max, F_max, and Ph to form a feature vector V = [A_max, F_max, Ph] for feature analysis in subsequent steps. This feature vector synthesizes the key characteristics of the vibration signal, facilitating subsequent feature analysis and pattern recognition.

[0024] Step 3: Cross-compare the material parameters with the feature parameters to establish a real-time correlation relationship, which specifically includes the following steps: The weighted comprehensive value W obtained from Step 1 (reflecting the comprehensive state of the vibration signal under specific working conditions) and the feature vector V = [A_max, F_max, Ph] extracted in Step 2 (containing amplitude, frequency, and phase information). These data provide the key vibration characteristics during the operation of the wire drawing machine.

[0025] For each sampling period, calculate the difference degree Dif between the material parameter M and the feature vector V through the formula: Dif = sqrt((M - W)^2 + (A_max - M_a)^2 + (F_max - M_f)^2 + (Ph - M_p)^2), where M represents the value of the material parameter in the current sampling period, and M_a, M_f, M_p are the corresponding amplitude, frequency, and phase reference values in the material parameters; (M - W)^2 represents the squared difference between the material parameter and the weighted comprehensive value. (A_max - M_a)^2, (F_max - M_f)^2, and (Ph - M_p)^2 represent the squared differences between the amplitude, frequency, and phase in the feature vector and the corresponding material parameter reference values respectively. sqrt(...) is the square root operation after summing the squares, used to obtain the actual difference degree value.

[0026] Adjust the correlation factor Af based on the difference degree Dif, satisfying Af = 1 / (1 + exp(-ALPHA * Dif)), where ALPHA is a constant used to control the speed at which Af changes with Dif; exp(-ALPHA * Dif) uses the exponential function to adjust the change rate of Af, ensuring that it can quickly respond to significant differences while smoothing out small fluctuations. Af = 1 / (1 + exp(-ALPHA * Dif)) ensures that the value of Af is between 0 and 1, reflecting the influence of the difference degree Dif on the correlation strength.

[0027] Update the real-time association relationship Rl between the material parameters and the characteristic parameters using the association factor Af, and the update is carried out through the formula Rl = R_old + Af(Dif - R_old), where R_old represents the association relationship in the previous cycle. Af(Dif - R_old)* represents the new contribution part adjusted according to the current degree of difference and the association relationship in the previous cycle. Rl = R_old + Af(Dif - R_old)* ensures that the new association relationship takes into account both historical data and the latest change information.

[0028] Step Four: Dynamically adjust the vibration anomaly determination threshold according to the real-time association relationship, which specifically includes the following steps: Obtain the real-time association relationship Rl from Step Three, which reflects the comprehensive state between the material parameters and the characteristic parameters in the current sampling cycle. This step ensures that the subsequent steps can be analyzed and adjusted based on the latest production data.

[0029] Calculate the adjustment factor Adj based on the real-time association relationship Rl, using the formula Adj = 1 + BETA*(Rl - R_min) / (R_max - R_min), where BETA is a preset proportional constant used to control the change range of the adjustment factor; R_min and R_max are respectively the minimum and maximum association relationship values in the historical data; (Rl - R_min) / (R_max - R_min) represents the proportional position of the current real-time association relationship relative to the historical data range. 1 + BETA(...)* ensures that the adjustment factor fluctuates around 1, making the threshold adjustment neither too drastic nor too conservative.

[0030] Adjust the initially set vibration anomaly determination threshold Th_0 using the adjustment factor Adj, and obtain the new vibration anomaly determination threshold Th through the formula Th = Th_0*Adj; here Adj is the adjustment factor calculated in the previous step, which reflects the degree of adjustment required for the threshold under the current working conditions. Th_0*Adj multiplies the initial threshold by the adjustment factor to obtain the new vibration anomaly determination threshold Th, ensuring that it can meet the specific requirements of the current production environment.

[0031] Update the vibration anomaly determination threshold Th into the monitoring system, so that the subsequent evaluation of the vibration signal is based on the updated threshold Th for comparison. Regularly updating the vibration anomaly determination threshold keeps the monitoring system in the best state, which helps to detect and respond to any potential problems or abnormal situations in a timely manner.

[0032] Step Five: Calculate the difference between the current characteristic parameter and the dynamic threshold, which specifically includes the following steps: The feature vector V = [A_max, F_max, Ph] extracted from Step 2 includes amplitude, frequency, and phase information. At the same time, the updated vibration anomaly determination threshold Th is obtained from Step 4. These data provide a basis for subsequent difference calculations.

[0033] For each feature parameter, calculate the difference Dv between it and the dynamic threshold Th using the formula Dv = |V - Th|, where V represents an element in the feature vector, i.e., A_max, F_max, or Ph; |V - Th| means taking the absolute value of the difference between each element in the feature vector and the dynamic threshold Th to ensure that the difference is always positive.

[0034] Compare each of the differences Dv with a preset tolerance interval E, which is used to define the normal fluctuation range. Determine whether it exceeds the range through the formula: If Dv > E, then mark it as a potential anomaly point; Calculate the cumulative value Sum_Dv of the differences Dv corresponding to all feature parameters using the formula Sum_Dv = SUM(Dv), which is used for subsequent comprehensive evaluation. SUM(Dv) means accumulating the differences of all feature parameters to obtain a comprehensive evaluation value for subsequent analysis.

[0035] Step 6: If the difference exceeds the preset tolerance interval, extract the historical fluctuation data of the material parameters, which specifically includes the following steps: If the difference Dv between each feature parameter calculated in Step 5 and the dynamic threshold Th exceeds the preset tolerance interval E, the system automatically triggers an instruction to extract the historical fluctuation data of the relevant material parameters from the database.

[0036] According to the trigger instruction, retrieve the historical records of the corresponding material parameter Mat in the past n cycles from the database, denoted as Mat_hist = [Mat_1, Mat_2,..., Mat_n]; by analyzing the historical data, it is possible to better understand the change trend of the current material parameter and determine whether the current change belongs to the normal fluctuation range or there is an abnormal situation.

[0037] Calculate the standard deviation SIG_mat and the mean MU_mat of the historical fluctuation data Mat_hist using the formulas SIG_mat = sqrt(SUM(Mat_i - MU_mat)^2 / n) and MU_mat = SUM(Mat_i) / n, where i ranges from 1 to n; MU_mat = SUM(Mat_i) / n represents the average value of all historical data points, reflecting the overall level of the historical data. SIG_mat = sqrt(SUM(Mat_i - MU_mat)^2 / n) represents the degree of dispersion of all historical data points relative to the mean, reflecting the volatility of the data.

[0038] Based on the standard deviation SIG_mat and the mean MU_mat, determine the fluctuation range L_mat of the material parameters, which is defined by the formula L_mat = MU_mat ± k * SIG_mat, where k is a constant used to control the width of the fluctuation range. MU_mat ± k * SIG_mat represents a range that extends k times the standard deviation above and below the mean, defining the boundaries of normal fluctuations. This method can effectively distinguish normal fluctuations from abnormal changes, ensuring that only changes that truly exceed the normal range are marked as abnormal, reducing the false alarm rate.

[0039] Step Seven: Modify the weight coefficient of the difference in combination with the historical fluctuation data, specifically including the following steps: The standard deviation SIG_mat and the mean MU_mat calculated in Step Six, as well as the fluctuation range L_mat of the material parameters determined based on these statistics. These statistical data provide an understanding of the changing trend of the material parameters in the past n cycles, providing a basis for subsequent adjustment of the weight coefficient of the difference.

[0040] Calculate the correction factor Corr_Fac based on the historical fluctuation data, using the formula Corr_Fac = 1 / (1 + GAMMA * (SIG_mat / MU_mat)), where GAMMA is a preset adjustment constant used to control the change amplitude of the correction factor. SIG_mat / MU_mat represents the ratio of the standard deviation to the mean, reflecting the relative volatility of the data. 1 / (1 + GAMMA(...)) ensures that the correction factor fluctuates between 0 and 1, such that data points with larger fluctuations are given smaller weights, while data points with smaller fluctuations are given larger weights.

[0041] Use the correction factor Corr_Fac to adjust the weight of the difference Dv, and calculate the adjusted difference Dv_adj through the formula Dv_adj = Dv * Corr_Fac; here Dv is the difference calculated in Step Five, and Corr_Fac is the correction factor calculated in the previous step. Dv * Corr_Fac multiplies the original difference by the correction factor to obtain the adjusted difference Dv_adj, ensuring that it can reflect the specific requirements of the current production environment.

[0042] Update the weight coefficient W_coef, using the formula: W_coef = W_coef_old + ETA(Dv_adj - W_coef_old), where ETA is the learning rate constant and W_coef_old is the weight coefficient of the previous cycle. W_coef_old + ETA(Dv_adj - W_coef_old) ensures that the new weight coefficient takes into account both historical data and the latest change information, gradually optimizing the weight coefficient to adapt to the actual working conditions.

[0043] Step Eight: When the corrected difference continuously exceeds the product of the weight coefficient and the dynamic threshold, a warning is triggered, which specifically includes the following steps: Obtain the adjusted difference Dv_adj (Step Seven), the weight coefficient W_coef (Step Seven), and the updated vibration anomaly determination threshold Th (Step Four) from the previous steps. These data provide the basis for subsequent comparison and warning.

[0044] Calculate the product Prod of the weight coefficient W_coef and the dynamic threshold Th, using the formula Prod = W_coef * Th as the new comparison benchmark; this product is used as the new comparison benchmark to determine whether the adjusted difference within the current cycle exceeds the normal range.

[0045] Monitor Dv_adj calculated for each cycle within m consecutive sampling cycles. If Dv_adj satisfies the condition Dv_adj > Prod in all these m cycles, then prepare to trigger a warning signal; m is a preset number of cycles used to ensure the persistence of abnormal conditions. If Dv_adj > Prod means that if the adjusted difference exceeds the new comparison benchmark value, it is considered that there is an abnormality in that cycle.

[0046] When the above condition is satisfied within m consecutive cycles, calculate the cumulative excess difference S through the formula S = SUM(Dv_adj - Prod) / m. If S exceeds the preset safety limit L_s, then officially issue a warning signal.

[0047] SUM(Dv_adj - Prod) represents the sum of the differences obtained by subtracting the comparison benchmark value from the adjusted difference within all m cycles.

[0048] S = SUM(Dv_adj - Prod) / m represents calculating the average value of these differences as the cumulative excess difference S.

[0049] If S > L_s means that if the cumulative excess difference S exceeds the preset safety limit L_s, then officially issue a warning signal.

[0050] On the other hand, the present invention proposes an artificial intelligence-based wire drawing machine status monitoring system, as Figure 2 shown, including: A signal and parameter acquisition module, which is used to perform real-time acquisition of the vibration signal on the surface of the metal wire during the operation of the wire drawing machine and the material parameters of the corresponding process section; A feature extraction and correlation analysis module, which is used to perform extraction of the amplitude, frequency, and phase characteristic parameters of the vibration signal, cross-compare the material parameters with the characteristic parameters, and establish a real-time correlation relationship; A dynamic threshold adjustment and anomaly detection module, which is used to perform dynamic adjustment of the vibration anomaly determination threshold according to the real-time correlation relationship, calculate the difference between the current characteristic parameters and the dynamic threshold, and if the difference exceeds the preset error tolerance interval, extract the historical fluctuation data of the material parameters; A weight correction and warning trigger module, which is used to perform correction of the weight coefficient of the difference in combination with the historical fluctuation data, and trigger a warning when the corrected difference continuously exceeds the product of the weight coefficient and the dynamic threshold.

[0051] In addition, when the above modules are executing, they are also used to implement other steps of the above-mentioned method for monitoring the state of a wire drawing machine based on artificial intelligence, as follows: Step 1: Real-time acquisition of vibration signals and process parameters during the operation of the wire drawing machine Operation instructions: Install a piezoelectric sensor array on the surface of the metal wire of the wire drawing machine, and the sampling frequency is 1000Hz per second.

[0052] Synchronously record the time stamps Ts = [t1, t2,..., tn], and each time point corresponds to a vibration signal sample.

[0053] Perform wavelet packet decomposition on the original signal s(t), extract the K-th order coefficient C_k, satisfying SUM(|C_k|^2) ≤ 3σ² (σ is the standard deviation of the original signal). At the same time, record the current temperature Temp = 65°C and the tension Str = 120N. Synchronize C_k, Temp, and Str according to the time stamp.

[0054] Calculation example: Assume σ = 0.5, then 3σ² = 0.75; if a group of C_k = [0.3, -0.4, 0.2] is extracted, then SUM(|C_k|²) = (0.3)^2 + (-0.4)^2 + (0.2)^2 = 0.29 < 0.75 → meets the requirements.

[0055] Then calculate the weighted comprehensive value W: W = SUM(C_k^2 / sqrt(Temp) + 0.8 * Str) = (0.3^2) / sqrt(65) + (0.4^2) / sqrt(65) + (0.2^2) / sqrt(65) + 0.8 * 120 =(0.09 + 0.16 + 0.04) / 8.06 + 96 = 0.036 + 96 = 96.036。

[0056] Step 2: Extract the amplitude, frequency, and phase characteristics of the vibration signal Operating instructions: Perform an FFT transformation on C_k to obtain the frequency spectrum S_f. Find the maximum amplitude A_max and its corresponding frequency F_max. Calculate the phase Ph = arctan(Im(S_f) / Re(S_f)) using the complex frequency spectrum. Form the feature vector V = [A_max, F_max, Ph].

[0057] Calculation example: Assume that after FFT, the main frequency F_max = 280 Hz and its amplitude A_max = 5.2 V; The corresponding complex number S_f = 4.8 + j3.6 → Ph = arctan(3.6 / 4.8) = 36.87°; The feature vector V = [5.2, 280, 36.87].

[0058] Step 3: Cross - comparison of material parameters and feature parameters Operating instructions: The material parameter M is the current set value. For example, M = 95 (representing the mass coefficient of a certain copper alloy).

[0059] M_a, M_f, and M_p are the standard reference values of this material. For example, M_a = 5.0, M_f = 270, M_p = 35°.

[0060] Calculate the difference degree: Dif = sqrt((M - W)^2 + (A_max - M_a)^2 + (F_max - M_f)^2 + (Ph - M_p)^2).

[0061] Calculation example: W = 96.036, A_max = 5.2, F_max = 280, Ph = 36.87; M = 95, M_a = 5.0, M_f = 270, M_p = 35; Dif = sqrt((95 - 96.036)^2 + (5.2 - 5.0)^2 + (280 - 270)^2 + (36.87 - 35)^2) = sqrt((-1.036)^2 + (0.2)^2 + (10)^2 + (1.87)^2) = sqrt(1.073 + 0.04 + 100 + 3.5) =sqrt(104.613)=10.23。

[0062] Then calculate the correlation factor Af = 1 / (1 + exp(-α*Dif)), taking α = 0.1: Af = 1 / (1 + e^(-0.1*10.23)) = 1 / (1 + e^-1.023) = 1 / (1 + 0.36) = 0.735。

[0063] Update the real-time correlation relationship Rl = R_old + Af*(Dif - R_old), assuming the previous R_old = 12: Rl = 12 + 0.735*(10.23 - 12) = 12 - 1.305 = 10.695。

[0064] Step Four: Dynamically adjust the vibration anomaly determination threshold Th Operation instructions: Use historical data R_min = 8, R_max = 15, BETA = 2; Calculate the adjustment factor Adj = 1 + BETA*(Rl - R_min) / (R_max - R_min); Adj = 1 + 2*(10.695 - 8) / (15 - 8) = 1 + 2*(2.695) / 7 = 1 + 0.77 = 1.77; Initial threshold Th0 = 0.6 → New threshold Th = Th0*Adj = 0.6*1.77 = 1.062。

[0065] Step Five: Difference calculation and preliminary judgment Operation instructions: Compare V = [5.2, 280, 36.87] with Th = 1.062; Calculate the difference Dv = |Vi - Th| for each parameter respectively; Dv1 = |5.2 - 1.062| = 4.138; Dv2 = |280 - 1.062| = 278.938; Dv3 = |36.87 - 1.062| = 35.808; Fault tolerance interval E = 0.5, obviously Dv1~3 are all > E, marked as potential abnormal points, Sum_Dv = 4.138 + 278.938 + 35.808 = 318.884。

[0066] Step Six: Extract historical fluctuation data of material parameters Operation instructions: If Dv > E holds, extract the material parameters Mat_hist within the past n = 5 cycles from the database: Mat_hist = [94.5, 95.0, 94.8, 95.2, 95.1]; Calculate MU_mat = (94.5 + 95 + 94.8 + 95.2 + 95.1) / 5 = 94.92; SIG_mat = sqrt(SUM(Mat_i - MU_mat)^2 / n) SIG_mat = sqrt[(94.5 - 94.92)^2 + (95 - 94.92)^2 +...] / 5 = sqrt(0.1764 + 0.0064 + 0.0144 + 0.0784 + 0.0324) / 5 = sqrt(0.308) / 5 = 0.062.

[0067] L_mat = MU_mat ± k * SIG_mat, k = 2 → L_mat = 94.92 ± 0.124.

[0068] Step Seven: Modify the difference weight coefficient Operation description: Calculate the correction factor Corr_Fac = 1 / (1 + γ * (SIG_mat / MU_mat)), γ = 0.5; Corr_Fac = 1 / (1 + 0.5 * (0.062 / 94.92)) = 1 / (1 + 0.000327) = 0.9997; Adjust the difference Dv_adj = Dv * Corr_Fac; Dv_adj1 = 4.138 * 0.9997 = 4.137; Dv_adj2 = 278.938 * 0.9997 = 278.854; Dv_adj3 = 35.808 * 0.9997 = 35.797; Update the weight coefficient W_coef: W_coef_new = W_coef_old + η * (Dv_adj - W_coef_old); Assume W_coef_old = 0.5, η = 0.1: W_coef_new = 0.5 + 0.1 * (4.137 - 0.5) = 0.5 + 0.3637 = 0.8637.

[0069] Step Eight: Continuously trigger an alarm for exceeding the limit Operation description: For three consecutive cycles where m = 3, Dv_adj > Prod = W_coefTh = 0.8637×1.062 = 0.917; Assume that Dv_adj for three consecutive periods are respectively: Cycle1: 4.137 > 0.917; Cycle2: 4.201 > 0.917; Cycle3: 4.253 > 0.917; Calculate the cumulative over - difference value: S=(4.137 - 0.917 + 4.201 - 0.917 + 4.253 - 0.917) / 3 =(3.22 + 3.284 + 3.336) / 3 = 9.84 / 3 = 3.28.

[0070] If the safety limit Ls = 2.5, then 3.28 > 2.5 → trigger a warning!

[0071] This embodiment fully demonstrates how to apply artificial intelligence technology to the condition monitoring of traditional industrial equipment. By combining physical sensors, signal - processing algorithms, machine - learning models, and real - time warning mechanisms, it realizes the all - round perception and intelligent decision - making of the running state of the wire - drawing machine, with high practicality and popularization value. At the same time, the system can also adapt to different materials (such as copper, aluminum, steel wire, etc.), without the need for frequent manual threshold adjustment, greatly improving the versatility and intelligent level of the system.

[0072] Finally, it should be noted that the above - mentioned are only the preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, for those skilled in the art, they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent substitution on some of the technical features. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An artificial intelligence-based method for monitoring and warning the status of a wire drawing machine, characterized in that, Including the following steps: Collect in real time the vibration signals on the surface of the wire and the material parameters of the corresponding process section during the operation of the wire drawing machine; Extract the amplitude, frequency and phase characteristic parameters of the vibration signals, cross-compare the material parameters with the characteristic parameters, and establish a real-time correlation relationship; Dynamically adjust the vibration anomaly determination threshold according to the real-time correlation relationship, calculate the difference between the current characteristic parameters and the dynamic threshold. If the difference exceeds the preset tolerance interval, extract the historical fluctuation data of the material parameters; Combine the historical fluctuation data to correct the weight coefficient of the difference. When the corrected difference continuously exceeds the product of the weight coefficient and the dynamic threshold, trigger an alarm.

2. The method for monitoring and warning the state of a wire drawing machine based on artificial intelligence according to claim 1, characterized in that The real-time collection of the vibration signals on the surface of the wire and the material parameters of the corresponding process section during the operation of the wire drawing machine includes: Contact the surface of the wire through a piezoelectric sensor array, collect vibration signals and record the timestamp sequence; Perform wavelet packet decomposition on the vibration signals, extract the decomposition coefficients of the order as characteristic parameters, and ensure that the characteristic parameters meet the set conditions; Read the real-time temperature and tension values of the process section, synchronize the decomposition coefficients with the temperature and tension through the timestamp, and calculate the weighted comprehensive value of the synchronized data, where the weight coefficient is determined by the mean value of the previous several sampling periods.

3. The method for monitoring and warning the state of a wire drawing machine based on artificial intelligence according to claim 2, wherein, Extracting the amplitude, frequency and phase characteristic parameters of the vibration signals includes: Perform a fast Fourier transform on the vibration signals to convert the time-domain signals into frequency-domain signals; Calculate the frequency and corresponding amplitude corresponding to the peak value in the frequency-domain signals; Use the frequency-domain signals and the peak frequency to determine the phase information, and combine the amplitude, frequency and phase to form a characteristic vector.

4. The state monitoring and early warning method of a wire drawing machine based on artificial intelligence according to claim 3, wherein, Cross-compare the material parameters with the characteristic parameters and establish a real-time correlation relationship, including: Obtain the weighted comprehensive value and the characteristic vector. For each sampling period, calculate the difference degree between the material parameters and the characteristic vector; Adjust the correlation factor based on the difference degree, and use the correlation factor to update the real-time correlation relationship between the material parameters and the characteristic parameters.

5. The state monitoring and early warning method for a wire drawing machine based on artificial intelligence according to claim 1, characterized in that, Dynamically adjust the vibration anomaly determination threshold according to the real-time correlation relationship, including: Obtain the real-time correlation relationship and calculate the adjustment factor based on the real-time correlation relationship; Use the adjustment factor to adjust the initially set vibration anomaly determination threshold; Update the vibration anomaly determination threshold to the monitoring system so that the subsequent evaluation of the vibration signals is based on the updated threshold for comparison.

6. The method for monitoring and warning the state of a wire drawing machine based on artificial intelligence according to claim 5, characterized in that, Calculate the difference between the current characteristic parameters and the dynamic threshold, including: Obtain the characteristic vector and the updated vibration anomaly determination threshold; For each characteristic parameter, calculate the difference between it and the dynamic threshold; Compare each difference with the preset tolerance interval. If it exceeds the range, mark it as a potential anomaly point, and then calculate the cumulative value of the differences corresponding to all characteristic parameters.

7. A method for monitoring and warning the state of a wire drawing machine based on artificial intelligence according to claim 6, characterized in that, Extracting the historical fluctuation data of the material parameters includes the following steps: When the difference exceeds the preset tolerance interval, trigger a historical data extraction instruction; According to the trigger instruction, retrieve the historical records of the corresponding material parameters in the past several cycles from the database; Calculate the standard deviation and mean of the historical fluctuation data, and determine the fluctuation range of the material parameters based on the standard deviation and mean.

8. The method for monitoring and warning the state of a wire drawing machine based on artificial intelligence according to claim 7, wherein, Correct the weight coefficient of the difference in combination with the historical fluctuation data, including: Obtain the standard deviation, mean, and material parameter fluctuation range, and calculate the correction factor based on the historical fluctuation data; Use the correction factor to adjust the weight of the difference, update the weight coefficient, and ensure dynamic adaptation to material changes over time.

9. The state monitoring and early warning method for a wire drawing machine based on artificial intelligence according to claim 8, characterized in that, When the corrected difference continuously exceeds the product of the weight coefficient and the dynamic threshold, trigger an alarm, including: Obtain the adjusted difference and weight coefficient, as well as the updated vibration anomaly determination threshold, and calculate the product of the weight coefficient and the dynamic threshold as the new comparison benchmark; Monitor the differences calculated for each cycle within a continuous number of sampling cycles. If the differences in all cycles meet the condition of being greater than the new comparison benchmark, then prepare to trigger an alarm signal; When the condition is met within a continuous number of cycles, calculate the cumulative excess difference. If the cumulative excess difference exceeds the preset safety limit, then officially issue an alarm signal.

10. An artificial intelligence-based wire drawing machine condition monitoring system for implementing the method according to any one of claims 1-9, characterized in that, Including: A signal and parameter acquisition module for performing real-time acquisition of the vibration signal on the surface of the wire during the operation of the wire drawing machine and the material parameters of the corresponding process section; A feature extraction and correlation analysis module for performing extraction of the amplitude, frequency, and phase characteristic parameters of the vibration signal, cross-comparing the material parameters with the characteristic parameters, and establishing a real-time correlation relationship; A dynamic threshold adjustment and anomaly detection module for performing dynamic adjustment of the vibration anomaly determination threshold according to the real-time correlation relationship, calculating the difference between the current characteristic parameter and the dynamic threshold. If the difference exceeds the preset error tolerance range, then extract the historical fluctuation data of the material parameters; A weight correction and alarm trigger module for performing correction of the weight coefficient of the difference in combination with the historical fluctuation data and triggering an alarm when the corrected difference continuously exceeds the product of the weight coefficient and the dynamic threshold.

Images