Glass bottle stamping abnormity detection method and related equipment

Abstract

The invention belongs to the technical field of glass bottle production quality detection, and discloses a glass bottle punching abnormity detection method and related equipment, and the method comprises the steps: applying micro-frequency vibration excitation with a specific frequency to a punching head of punching equipment, and collecting a vibration signal of the punching head in a punching process; performing frequency domain analysis on the collected vibration signal to extract a frequency domain characteristic value at the excitation frequency; based on an exponential moving average algorithm, dynamically adjusting an anomaly detection threshold according to the continuously acquired frequency domain characteristic values; the currently obtained frequency domain characteristic value is compared with a dynamically adjusted anomaly detection threshold value, whether the stamping process is abnormal or not is judged according to a comparison result, and an anomaly detection result is output; the abnormal detection result is fed back to a control system of the stamping equipment in real time, so that the control system adjusts stamping parameters according to the abnormal detection result; therefore, sensitive and reliable real-time anomaly detection of the glass bottle stamping process can be realized in a high-speed dynamic change production environment.

Description

Technical Field

[0001] The present application relates to the technical field of glass bottle production quality inspection, and in particular to a glass bottle stamping anomaly detection method and related equipment. Background Art

[0002] High-speed glass bottle production plants generally use multi-station rotary stamping equipment to achieve efficient production of lightweight thin-walled glass bottles. In order to effectively control production costs and ensure product quality, the factory urgently needs to accurately monitor the stamping process to avoid defects such as uneven bottle wall thickness and cracks. To this end, each stamping station usually needs to be equipped with a real-time anomaly detection system to instantly detect potential problems during high-speed production and adjust stamping parameters in a timely manner, thereby maintaining production stability and product quality consistency.

[0003] However, the detection systems in the prior art often face the problem of balancing data processing speed and defect detection sensitivity when dealing with high-speed rotating environments. More importantly, in actual production environments, the operating status of equipment is not constant, and environmental noise and equipment operating conditions will drift slowly over time. Traditional fixed threshold detection methods are difficult to adapt to such dynamic changes. When environmental noise increases or equipment operating conditions change, if a fixed threshold is still used, it is easy to cause a significant increase in the false alarm rate. In order to reduce false alarms, raising the threshold will cause a decrease in detection sensitivity, making it impossible to effectively detect minor defects.

[0004] Therefore, how to achieve sensitive and reliable real-time anomaly detection in the glass bottle stamping process in a high-speed, dynamically changing production environment has become a technical problem that needs to be solved urgently.

[0005] In view of the above problems, the existing technology needs to be improved urgently. Summary of the invention

[0006] The purpose of the present application is to provide a glass bottle stamping anomaly detection method and related equipment, which can realize sensitive and reliable real-time anomaly detection of the glass bottle stamping process in a high-speed dynamically changing production environment.

[0007] In a first aspect, the present application provides a method for detecting anomalies in glass bottle punching, which is used for detecting anomalies in a glass bottle punching process performed by a high-speed rotary punching device, and the method comprises the following steps: A1. While applying micro-frequency vibration excitation of a specific frequency to the punch head of the punching equipment, collect the vibration signal of the punch head during the punching process; A2. Perform frequency domain analysis on the collected vibration signal to extract the frequency domain eigenvalue at the excitation frequency; A3. Based on the exponential moving average algorithm, the anomaly detection threshold is dynamically adjusted according to the continuously acquired frequency domain feature values; A4. Compare the currently acquired frequency domain eigenvalue with the dynamically adjusted anomaly detection threshold, determine whether the stamping process is abnormal based on the comparison result, and output the anomaly detection result; A5. Feedback the abnormal detection results to the control system of the stamping equipment in real time so that the control system can adjust the stamping parameters according to the abnormal detection results.

[0008] This method solves the problem that the fixed threshold detection method in the prior art is difficult to adapt to the high-speed dynamically changing production environment by dynamically adjusting the anomaly detection threshold. It can realize sensitive and reliable real-time anomaly detection of the glass bottle stamping process in the high-speed dynamically changing production environment.

[0009] Furthermore, the present application also proposes that step A3 includes: Obtain the rotation speed of high-speed rotary punching equipment; comparing the rotation speed with a preset speed threshold; When the rotation speed is greater than or equal to a preset speed threshold, setting the frequency of dynamically adjusting the abnormality detection threshold to a first frequency; When the rotation speed is less than a preset speed threshold, setting the frequency of dynamically adjusting the abnormality detection threshold to a second frequency; The first frequency is greater than the second frequency.

[0010] In this way, the frequency of dynamically adjusting the abnormality detection threshold can be adaptively adjusted according to the actual rotation speed of the stamping equipment. When the equipment is running at high speed, the threshold can quickly respond to process changes and maintain detection sensitivity. When the equipment is running at low speed, the threshold update frequency is reduced, reducing unnecessary computing resource consumption and maintaining detection stability.

[0011] Furthermore, the present application also proposes that step A3 includes: After the system is started, it enters the initial threshold setting stage; In the initial threshold setting stage, the frequency domain feature values ​​within the preset time window are continuously collected; Calculate the average value of the frequency domain characteristic values ​​collected within a preset time window; The calculated average value is set as the initial anomaly detection threshold; After the initial anomaly detection threshold is set, the exponential moving average algorithm is started based on the set initial anomaly detection threshold, and the anomaly detection threshold is dynamically adjusted according to the frequency domain eigenvalues ​​subsequently obtained in real time.

[0012] In this way, the initial anomaly detection threshold can be determined based on the actual operating data of the equipment, avoiding the deviation caused by subjective settings and improving the accuracy of anomaly detection in the initial stage of system startup.

[0013] Furthermore, the present application also proposes that based on the set initial anomaly detection threshold, the exponential moving average algorithm is started, and the steps of dynamically adjusting the anomaly detection threshold according to the frequency domain characteristic values ​​subsequently acquired in real time include: Calculate the difference between the frequency domain eigenvalue currently obtained and the frequency domain eigenvalue obtained at the previous moment; Determine whether the difference exceeds a preset mutation amplitude threshold; If the difference exceeds the preset mutation amplitude threshold, it is determined that the currently acquired frequency domain feature value contains a noise spike; When it is determined that the currently acquired frequency domain feature value contains a noise spike, when the anomaly detection threshold is updated based on the exponential moving average algorithm, the weight of the currently acquired frequency domain feature value in the exponential moving average calculation is reduced; Based on the frequency domain feature values ​​after weight reduction, the exponential moving average algorithm is used to update the anomaly detection threshold.

[0014] Furthermore, the present application also proposes that the step of updating the anomaly detection threshold using an exponential moving average algorithm based on the frequency domain eigenvalue after weight reduction includes: If the weight reduction ratio of the currently acquired frequency domain eigenvalue in the exponential moving average calculation exceeds a preset ratio threshold, the smoothing factor of the exponential moving average algorithm is switched to a second smoothing factor, and the second smoothing factor is smaller than the first smoothing factor used when the normal update threshold is used; Using a second smoothing factor, updating the anomaly detection threshold based on an exponential moving average algorithm; After the abnormality detection threshold update based on the second smoothing factor is completed, the smoothing factor of the exponential moving average algorithm is switched back to the first smoothing factor.

[0015] Furthermore, the present application also proposes that if the weight reduction ratio of the currently acquired frequency domain eigenvalue in the exponential moving average calculation exceeds a preset ratio threshold, the smoothing factor of the exponential moving average algorithm is switched to a second smoothing factor, and the second smoothing factor is less than the first smoothing factor used when the normal update threshold is used. The steps include: When the weight reduction ratio exceeds the preset ratio threshold, the noise timer is started to record the noise duration; When the weight reduction ratio drops below a preset ratio threshold, the noise timer is stopped to obtain the noise duration; According to the noise duration, a preset duration and smoothing factor mapping table is searched to determine a second smoothing factor corresponding to the noise duration.

[0016] Furthermore, the present application also proposes that step A5 includes: According to the abnormality detection results, determine the abnormal state level of the stamping process; Select a preset control strategy based on the abnormal state level; Execute preset control strategies and adjust stamping parameters.

[0017] In a second aspect, the present application provides a glass bottle stamping anomaly detection device for detecting anomalies in a glass bottle stamping process performed by a high-speed rotary stamping device, the device comprising: A vibration signal acquisition module, used to apply a micro-frequency vibration excitation of a specific frequency to the punch head of the punching equipment and collect the vibration signal of the punch head during the punching process; A frequency domain analysis module is used to perform frequency domain analysis on the collected vibration signal to extract the frequency domain characteristic value at the excitation frequency; A threshold adjustment module is used to dynamically adjust the anomaly detection threshold based on the exponential moving average algorithm according to the continuously acquired frequency domain feature values; The anomaly detection module is used to compare the currently acquired frequency domain characteristic value with the dynamically adjusted anomaly detection threshold, determine whether the stamping process is abnormal based on the comparison result, and output the anomaly detection result; The system adjustment module is used to feed back the abnormality detection result to the control system of the stamping equipment in real time, so that the control system adjusts the stamping parameters according to the abnormality detection result.

[0018] In a third aspect, the present application provides an electronic device, including a processor and a memory, wherein the memory stores a computer program executable by the processor, and when the processor executes the computer program, it runs the steps in the glass bottle stamping abnormality detection method as described above.

[0019] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps in the glass bottle stamping anomaly detection method as described above are executed.

[0020] Beneficial effect: The present application provides a glass bottle stamping anomaly detection method and related equipment, which solves the problem that the fixed threshold detection method in the prior art is difficult to adapt to the high-speed dynamically changing production environment by dynamically adjusting the anomaly detection threshold. It can realize sensitive and reliable real-time anomaly detection of the glass bottle stamping process in the high-speed dynamically changing production environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A flow chart of a method for detecting abnormal punching of a glass bottle provided in an embodiment of the present application.

[0022] Figure 2 A schematic diagram of the structure of a glass bottle stamping anomaly detection device provided in an embodiment of the present application.

[0023] Figure 3A schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0024] Explanation of reference numerals: 1. Vibration signal acquisition module; 2. Frequency domain analysis module; 3. Threshold value adjustment module; 4. Abnormality detection module; 5. System adjustment module; 301. Processor; 302. Memory; 303. Communication bus. DETAILED DESCRIPTION

[0025] The technical solutions in the present application will be clearly and completely described below in conjunction with the drawings in the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the present application for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0026] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0027] refer to Figure 1 The present application proposes a glass bottle stamping anomaly detection method for detecting anomalies in a glass bottle stamping process performed by a high-speed rotary stamping device, the method comprising the following steps: A1. While applying micro-frequency vibration excitation of a specific frequency to the punch head of the punching equipment, collect the vibration signal of the punch head during the punching process; A2. Perform frequency domain analysis on the collected vibration signal to extract the frequency domain eigenvalue at the excitation frequency; A3. Based on the exponential moving average algorithm, the anomaly detection threshold is dynamically adjusted according to the continuously acquired frequency domain feature values; A4. Compare the currently acquired frequency domain eigenvalue with the dynamically adjusted anomaly detection threshold, determine whether the stamping process is abnormal based on the comparison result, and output the anomaly detection result; A5. Feedback the abnormal detection results to the control system of the stamping equipment in real time so that the control system can adjust the stamping parameters according to the abnormal detection results.

[0028] Specifically, the technical solution aims to solve the problem of abnormal detection in the glass bottle stamping process under high-speed dynamic stamping environment. By applying micro-frequency vibration excitation of a specific frequency to the stamping head and collecting vibration signals during the stamping process, the system can obtain the data basis reflecting the state of the stamping process. Subsequently, the collected vibration signal is subjected to frequency domain analysis to extract the frequency domain eigenvalues ​​at the excitation frequency, which can highlight specific frequency components, reduce the interference of other frequency noises, and improve the detection accuracy. Furthermore, based on the exponential moving average algorithm, the abnormal detection threshold is dynamically adjusted according to the continuously acquired frequency domain eigenvalues, so that the threshold can adaptively track the slow drift of the equipment state and the change of environmental noise, overcome the problem of high false alarm rate or low sensitivity of the traditional fixed threshold method in a dynamic environment, and ensure the reliability of detection. By comparing the current frequency domain eigenvalue with the dynamically adjusted abnormal detection threshold, it is possible to effectively determine whether the current stamping process is abnormal and give a clear detection result. Finally, the abnormal detection results are fed back to the control system of the stamping equipment in real time. The control system can adjust the stamping parameters in real time according to the detection results to form a closed-loop control, correct the abnormalities in the stamping process in time, and maintain the stability of the production process and the consistency of product quality. Therefore, the technical solution realizes sensitive and reliable real-time anomaly detection of the glass bottle stamping process under high-speed dynamic stamping environment.

[0029] Among them, in step A1, applying micro-frequency vibration excitation of a specific frequency to the punching head of the stamping equipment can be achieved by a piezoelectric ceramic exciter. The piezoelectric ceramic exciter is fixedly installed at a specific position of the punching head, such as the top or side of the punching head, and a sinusoidal wave signal of a specific frequency is generated by a signal generator to drive the piezoelectric ceramic exciter to generate micro-frequency vibration. The vibration signal of the punching head during the stamping process can be collected by an acceleration sensor. The acceleration sensor is preferably a high-sensitivity MEMS acceleration sensor, which is installed at a position of the punching head close to the stamping position to monitor the vibration conditions during the stamping process in real time.

[0030] The frequency of the micro-frequency vibration excitation (ie, the specific frequency mentioned above, hereinafter referred to as the excitation frequency) can be set according to actual needs, for example, 1 kHz-2 kHz, but not limited thereto.

[0031] Among them, in step A2, the collected vibration signal is subjected to frequency domain analysis to extract the frequency domain characteristic value at the excitation frequency. The specific implementation process may be: First, the time-domain vibration signal collected (by the acceleration sensor) is preprocessed. The preprocessing includes filtering and amplification, the purpose of which is to remove noise interference and increase the signal amplitude. Then, the fast Fourier transform algorithm is used to perform spectrum analysis on the preprocessed time domain signal to obtain the frequency domain signal; Afterwards, in the frequency domain signal, the spectrum amplitude at the excitation frequency is extracted as the frequency domain eigenvalue.

[0032] In some preferred embodiments, step A3 comprises: Obtain the rotation speed of high-speed rotary punching equipment; comparing the rotation speed with a preset speed threshold; When the rotation speed is greater than or equal to a preset speed threshold, setting the frequency of dynamically adjusting the abnormality detection threshold to a first frequency; When the rotation speed is less than a preset speed threshold, setting the frequency of dynamically adjusting the abnormality detection threshold to a second frequency; The first frequency is greater than the second frequency.

[0033] In the actual operation of high-speed rotary stamping equipment, the rotation speed is not constant. When the equipment is running at high speed, the working conditions change faster, and the abnormal detection system needs to be able to adapt to the changes in working conditions more quickly. The dynamic threshold should also be adjusted more frequently to ensure the sensitivity and real-time performance of the detection; when the equipment is running at low speed, the working conditions are relatively stable, and the adjustment frequency of the dynamic threshold can be appropriately reduced to improve the stability of the threshold and avoid misjudgment caused by frequent fluctuations.

[0034] Among them, the preset speed threshold, the first frequency and the second frequency can be set according to actual needs.

[0035] Specifically, the acquisition of the rotation speed can be achieved via an encoder installed on the rotating shaft of the stamping equipment. The encoder monitors the rotation speed of the rotating shaft in real time and transmits the rotation speed data to the control system. The preset speed threshold is stored in the control system as a pre-set parameter. For example, the preset speed threshold can be set to 100 revolutions per minute. The control system compares and judges the rotation speed obtained in real time with the preset speed threshold. If the current rotation speed is higher than or equal to 100 revolutions per minute, the system determines that the stamping equipment is in a high-speed operation state. At this time, the frequency of dynamically adjusting the abnormality detection threshold is set to the first frequency. For example, the first frequency can be set to update the threshold once every second. Conversely, if the current rotation speed is lower than 100 revolutions per minute, the system determines that the stamping equipment is in a low-speed operation state, and the frequency of dynamically adjusting the abnormality detection threshold is set to the second frequency. For example, the second frequency can be set to update the threshold every 5 seconds.

[0036] In this way, the frequency of dynamically adjusting the abnormality detection threshold can be adaptively adjusted according to the actual rotation speed of the stamping equipment. When the equipment is running at high speed, the threshold can quickly respond to process changes and maintain detection sensitivity. When the equipment is running at low speed, the threshold update frequency is reduced, reducing unnecessary computing resource consumption and maintaining detection stability.

[0037] Specifically, this solution aims to solve the problem of poor detection sensitivity and real-time performance that may be caused by the dynamic adjustment of the abnormal detection threshold at a fixed adjustment frequency at different speeds of high-speed rotary stamping equipment. By linking the adjustment frequency of the dynamically adjusted abnormal detection threshold with the rotation speed of the equipment, the adaptability of the detection method to changes in the equipment working conditions is achieved. Under high-speed operation, a higher first frequency is used to dynamically adjust the abnormal detection threshold, so that the abnormal detection system can quickly track changes in the stamping process, promptly discover and respond to potential abnormal situations, and ensure the detection sensitivity and real-time performance under high-speed production. Conversely, under low-speed operation, a lower second frequency is used to dynamically adjust the abnormal detection threshold, which reduces the frequency of threshold update, avoids the threshold being too sensitive to normal operating fluctuations, improves the stability of detection, and reduces unnecessary computing resource consumption. As a result, regardless of whether the stamping equipment is in a high-speed or low-speed operation state, the abnormal detection system can adaptively adjust the update frequency of the dynamic threshold according to the actual operating speed, so that good detection performance can be maintained at various speeds, and the overall performance and adaptability of the abnormal detection system are improved.

[0038] For example, in some specific embodiments, the preset speed threshold is set to 80 revolutions per minute. The first frequency is set to update the threshold twice per second, that is, the update cycle is 0.5 seconds. The second frequency is set to update the threshold once every 2 seconds, that is, the update cycle is 2 seconds. When the real-time monitoring value of the rotation speed of the stamping equipment is higher than or equal to 80 revolutions per minute, the dynamic threshold is updated every 0.5 seconds. When the real-time monitoring value of the rotation speed of the stamping equipment is lower than 80 revolutions per minute, the dynamic threshold is updated every 2 seconds. Through the differentiated setting of speed thresholds and frequencies, the linkage adaptation of the dynamic threshold adjustment frequency and the equipment rotation speed is achieved.

[0039] In some preferred embodiments, step A3 further comprises: After the system is started, it enters the initial threshold setting stage; In the initial threshold setting stage, the frequency domain feature values ​​within the preset time window are continuously collected; Calculate the average value of the frequency domain characteristic values ​​collected within a preset time window; The calculated average value is set as the initial anomaly detection threshold; After the initial anomaly detection threshold is set, the exponential moving average algorithm is started based on the set initial anomaly detection threshold, and the anomaly detection threshold is dynamically adjusted according to the frequency domain eigenvalues ​​subsequently obtained in real time.

[0040] In the actual production process of high-speed rotary stamping equipment, when the anomaly detection system is started, there is a lack of prior knowledge of the current working conditions. If the initial threshold is set too high, the system may be insensitive to minor anomalies that occur in the early stage, resulting in missed detections; if the initial threshold is set too low, it is easy to misjudge fluctuations under normal working conditions as anomalies, resulting in an increase in the false alarm rate. This problem can be solved by the above method.

[0041] Among them, after the system is started, the anomaly detection method enters the initial threshold setting stage to prepare the initial threshold for subsequent anomaly detection work. In the initial threshold setting stage, the stamping equipment operates under normal working conditions, and the system continuously collects vibration signals within the preset time window, processes the collected vibration signals, and extracts the frequency domain eigenvalues ​​at the excitation frequency. The length of the preset time window can be set according to actual needs, for example, it can be set to 10 minutes or longer to ensure that enough data samples are collected to accurately reflect the characteristics of the equipment under normal operating conditions. After completing the frequency domain eigenvalue collection within the preset time window, the average value of all frequency domain eigenvalues ​​collected during this period is calculated. The calculated average value is set as the initial anomaly detection threshold for subsequent anomaly detection judgment. After the initial anomaly detection threshold is set, the system officially enters the normal anomaly detection stage. The exponential moving average algorithm is started, with the set initial anomaly detection threshold as the starting value, and the anomaly detection threshold is dynamically updated according to the subsequent real-time acquired frequency domain eigenvalues. As a result, the anomaly detection threshold can be adaptively adjusted as the equipment operating state changes slowly, ensuring the effectiveness and accuracy of anomaly detection.

[0042] For example, in a glass bottle stamping production line using high-speed rotary stamping equipment, in order to ensure the quality of glass bottle products, it is necessary to perform abnormal detection on the stamping process. In view of the problem that the fixed threshold detection method in the prior art has a high false alarm rate or low sensitivity in a dynamically changing production environment, a dynamic threshold adjustment method of the present application is proposed. When the system starts, it first enters the initial threshold setting stage. For example, the preset time window is set to 15 minutes. During these 15 minutes, the system continuously collects the vibration signal of the punch head during the normal stamping process and extracts the frequency domain eigenvalues ​​at the excitation frequency. Assume that within these 15 minutes, a total of N frequency domain eigenvalues ​​are collected, namely f1, f2,..., fN. Calculate the average value of these frequency domain eigenvalues ​​F_avg=(f1+f2+...+fN) / N. Set the calculated average value F_avg as the initial abnormal detection threshold T0. After the initial threshold setting is completed, the system begins to dynamically adjust the abnormal detection threshold using the exponential moving average algorithm. Each subsequent time a new frequency domain eigenvalue is obtained, the anomaly detection threshold will be updated based on the exponential moving average algorithm and compared with the dynamically adjusted anomaly detection threshold to determine whether the stamping process is abnormal. In this way, the initial anomaly detection threshold can be determined based on the actual operation data of the equipment, avoiding the deviation caused by subjective settings and improving the accuracy of anomaly detection in the initial stage of system startup (when using the dynamic exponential moving average algorithm to dynamically adjust the anomaly detection threshold, it is necessary to set the initial threshold. If the initial threshold is set improperly, it may lead to a high false alarm rate or insufficient sensitivity in the initial stage of system startup, affecting the accuracy and timeliness of anomaly detection). At the same time, the introduction of the exponential moving average algorithm enables the anomaly detection threshold to adapt to the dynamic changes in the equipment's operating status, ensuring the long-term effectiveness of anomaly detection.

[0043] In some specific embodiments, the length of the preset time window can be adjusted according to the operating characteristics and noise level of the actual production line. For example, for a production line with a high noise level or large fluctuations in the equipment state, the preset time window can be appropriately extended to obtain a more stable initial threshold. Conversely, for a production line with a stable operating state, the preset time window can be appropriately shortened to speed up the setting of the initial threshold. As a preferred embodiment, the length of the preset time window is set to cover at least one production shift, such as 8 hours, to fully consider the periodic changes that may exist in the production process. In the initial threshold setting stage, a low-speed or no-load operation mode can be used to reduce abnormal interference in the initial stage and obtain more representative normal state data. In addition, when calculating the average value, a weighted average or an outlier elimination method can be used to further improve the accuracy and robustness of the initial threshold. For example, a median filter or a standard deviation elimination method can be used to remove the frequency domain feature values ​​that are obviously deviated from the normal range collected within the preset time window, and then the average value is calculated.

[0044] Further, based on the set initial anomaly detection threshold, starting the exponential moving average algorithm, and dynamically adjusting the anomaly detection threshold according to the frequency domain characteristic value subsequently acquired in real time may include: Calculate the difference between the frequency domain eigenvalue currently obtained and the frequency domain eigenvalue obtained at the previous moment; Determine whether the difference exceeds a preset mutation amplitude threshold; If the difference exceeds the preset mutation amplitude threshold, it is determined that the currently acquired frequency domain feature value contains a noise spike; When it is determined that the currently acquired frequency domain feature value contains a noise spike, when the anomaly detection threshold is updated based on the exponential moving average algorithm, the weight of the currently acquired frequency domain feature value in the exponential moving average calculation is reduced; Based on the frequency domain feature values ​​after weight reduction, the exponential moving average algorithm is used to update the anomaly detection threshold.

[0045] In an actual production environment, stamping equipment may be subject to various unpredictable noise interferences during operation, such as equipment vibration, electromagnetic interference, etc. These sudden noises may cause short spikes in the collected vibration signal, which in turn appear as mutations in the frequency domain eigenvalues. If the dynamic threshold adjustment algorithm directly processes the frequency domain eigenvalues ​​containing noise spikes, the threshold may be quickly raised or lowered by noise interference, resulting in false alarms and reducing the reliability of the detection system. This problem can be solved in the above way.

[0046] Specifically, the frequency domain eigenvalue at the current moment is subtracted from the frequency domain eigenvalue at the previous moment to obtain the difference. Subsequently, the difference is compared with the preset mutation amplitude threshold, which is a pre-set value used to define the degree of mutation of the frequency domain eigenvalue and can be set according to actual needs. When the difference exceeds the preset mutation amplitude threshold, the system determines that the current frequency domain eigenvalue is interfered by noise spikes. In order to reduce the impact of noise spikes on the dynamic adjustment of the anomaly detection threshold, the weight of the current frequency domain eigenvalue interfered by the noise spike is reduced when the exponential moving average algorithm updates the threshold. The weight reduction can be achieved by introducing a weight factor, and the weight factor value is less than or equal to 1 and greater than or equal to 0. The frequency domain eigenvalue after the weight reduction is used in the subsequent exponential moving average algorithm calculation to update the anomaly detection threshold. As a result, the dynamic adjustment process of the anomaly detection threshold can reduce the adverse effects of noise spikes and ensure the accuracy of the threshold adjustment.

[0047] Specifically, in the glass bottle stamping anomaly detection system, in order to solve the threshold misadjustment problem that may be caused by noise spikes, a dynamic threshold adjustment scheme with noise spike suppression is adopted. The system first collects the vibration signal in the stamping process in real time, performs frequency domain analysis, and extracts the frequency domain eigenvalue at the excitation frequency. In the dynamic threshold adjustment link, the system calculates the difference between the frequency domain eigenvalue at the current moment and the frequency domain eigenvalue at the previous moment. For example, the preset mutation amplitude threshold is set to an empirical value, such as 0.5. If the calculated difference is greater than 0.5, the system determines that the current frequency domain eigenvalue contains a noise spike. When it is determined that there is a noise spike, the weight of the current frequency domain eigenvalue is reduced when the exponential moving average algorithm is used to update the anomaly detection threshold. For example, under normal circumstances, the weight of the current frequency domain eigenvalue is 1, and when a noise spike is detected, the weight is adjusted to 0.2. Then, the frequency domain eigenvalue after weight adjustment is substituted into the exponential moving average algorithm formula to calculate the new anomaly detection threshold. Through this weight reduction strategy, the impact of noise spikes on threshold updates is effectively reduced, preventing the anomaly detection threshold from being unreasonably raised due to noise spike interference, and ensuring that the anomaly detection system maintains a high sensitivity.

[0048] In some specific implementations, the preset mutation amplitude threshold can be adjusted according to the noise environment of the actual production line and the system performance requirements. The value of the weight reduction factor can be adaptively adjusted according to the intensity of the noise spike. The greater the noise spike intensity, the smaller the value of the weight reduction factor. As a preferred implementation, a piecewise function can be used to determine the weight reduction factor. For different degrees of noise spikes, different weight reduction ratios are used to achieve more refined noise suppression.

[0049] For example, in some embodiments, when it is determined that the currently acquired frequency domain feature value contains a noise spike, when updating the abnormality detection threshold based on the exponential moving average algorithm, the step of reducing the weight of the currently acquired frequency domain feature value in the exponential moving average calculation includes: Determine the relationship between the absolute value Δf of the difference between the currently acquired frequency domain eigenvalue and the frequency domain eigenvalue acquired at the previous moment and the preset low mutation amplitude threshold Δf_l and high mutation amplitude threshold Δf_h, where Δf_h>Δf_l>0; According to the judgment result, the dynamic weight reduction factor β is determined, where: When Δf≤Δf_l, β=1; When Δf_l<Δf≤Δf_h, β=β_min+(1-β_min)*(Δf_h-Δf) / (Δf_h-Δf_l); When Δf>Δf_h, β=β_min; Among them, β_min is the preset minimum weight reduction factor, 0≤β_min<1; The dynamic weight reduction factor β is multiplied by the weight of the currently acquired frequency domain eigenvalue in the exponential moving average calculation to achieve the reduction of the weight of the currently acquired frequency domain eigenvalue in the exponential moving average calculation.

[0050] Among them, judging the absolute value Δf of the difference between the frequency domain eigenvalue currently obtained and the frequency domain eigenvalue obtained at the previous moment and the size relationship between the preset low mutation amplitude threshold Δf_l and the high mutation amplitude threshold Δf_h is to evaluate the intensity level of the noise spike. Specifically, the low mutation amplitude threshold Δf_l and the high mutation amplitude threshold Δf_h are pre-set to divide the different intensity ranges of the noise spike. When the absolute value of the difference Δf is less than or equal to Δf_l, it indicates that the noise spike is in a low amplitude or no noise spike state. When Δf is between Δf_l and Δf_h, it indicates that the noise spike is in a medium amplitude state. When Δf is greater than Δf_h, it indicates that the noise spike is in a high amplitude state.

[0051] According to the different ranges of Δf, the dynamic weight reduction factor β is determined. The dynamic weight reduction factor β is determined in the form of a piecewise function to achieve refined weight adjustment for noise spikes of different intensity levels. When the noise spike amplitude is small, the β value is set to 1 to ensure that the frequency domain eigenvalue weight is not reduced and the threshold is updated normally. When the noise spike amplitude increases, the β value is dynamically adjusted between 1 and β_min according to a linear function to achieve a gradual reduction in weight. When the noise spike amplitude reaches a high level, the β value is set to the preset minimum weight reduction factor β_min to achieve the maximum reduction in weight.

[0052] The dynamic weight reduction factor β is used to multiply the weight of the currently acquired frequency domain eigenvalue in the exponential moving average calculation, which means that in the exponential moving average calculation formula (specifically, the exponential moving average calculation formula is EMA_y=(1-a)*k*P+a*EMA_y', EMA_y is the adjusted anomaly detection threshold, a is the smoothing factor, P is the currently acquired frequency domain eigenvalue, k is the weight of P, and EMA_y' is the anomaly detection threshold before adjustment), the weight of the currently acquired frequency domain eigenvalue needs to be multiplied by the dynamic weight reduction factor β. Through the multiplication operation, the weight of the current frequency domain eigenvalue in the exponential moving average calculation is dynamically adjusted according to the noise peak amplitude (expressed by the formula k=k*β).

[0053] Specifically, in the glass bottle stamping anomaly detection method, in order to solve the problem of noise spikes interfering with the dynamic adjustment of the anomaly detection threshold, a weight reduction method is used. First, the system collects the vibration signal of the punch head in the stamping process in real time, and performs frequency domain analysis on the vibration signal to obtain the frequency domain eigenvalue at the excitation frequency. Then, the absolute value Δf of the difference between the current frequency domain eigenvalue and the frequency domain eigenvalue at the previous moment is calculated. The low mutation amplitude threshold Δf_l and the high mutation amplitude threshold Δf_h are preset to judge the intensity of the noise spike. When Δf≤Δf_l, it indicates that the signal is stable, and the weight reduction factor β is set to 1, and the current frequency domain eigenvalue weight is not reduced. When Δf_l<Δf≤Δf_h, it indicates that a medium-intensity noise spike occurs, and the β value is calculated by a linear function to achieve a moderate weight reduction. When Δf>Δf_h, it indicates that a high-intensity noise spike occurs, and the β value is set to β_min to achieve the maximum weight reduction. Finally, the dynamic weight reduction factor β is applied to the exponential moving average algorithm. By reducing the frequency domain eigenvalue weight corresponding to the noise spike, the influence of the noise spike on the update of the anomaly detection threshold is reduced, and the accuracy and robustness of the dynamic adjustment of the threshold are improved. In this way, the system can suppress noise interference while maintaining a sensitive response to actual abnormal situations, ensuring the reliability of anomaly detection in the glass bottle stamping process.

[0054] By setting the specific values ​​of the parameters Δf_l, Δf_h and β_min, the weight reduction strategy can be flexibly adjusted according to the actual noise characteristics to achieve the optimal noise suppression effect and anomaly detection performance.

[0055] When a noise spike is detected, the influence of noise interference on the threshold is suppressed by reducing the weight of the current frequency domain eigenvalue in the exponential moving average calculation. However, simply reducing the weight may slow down the response speed of the threshold to changes in the actual working conditions, especially when noise occurs frequently. Continuous weight reduction may cause threshold update lag and reduce the system's detection sensitivity to real anomalies. To this end, in some embodiments, based on the frequency domain eigenvalue after the weight reduction, the step of using the exponential moving average algorithm to update the anomaly detection threshold includes: If the weight reduction ratio of the currently acquired frequency domain eigenvalue in the exponential moving average calculation exceeds a preset ratio threshold, the smoothing factor of the exponential moving average algorithm is switched to a second smoothing factor, and the second smoothing factor is smaller than the first smoothing factor used when the normal update threshold is used; Using a second smoothing factor, updating the anomaly detection threshold based on an exponential moving average algorithm; After the abnormality detection threshold update based on the second smoothing factor is completed, the smoothing factor of the exponential moving average algorithm is switched back to the first smoothing factor.

[0056] Among them, for the weight reduction ratio, a preset ratio threshold is set (which can be set according to actual needs). When the weight reduction ratio exceeds the preset ratio threshold, the system determines that the noise impact is more serious. At this time, the smoothing factor of the exponential moving average algorithm is switched from the first smoothing factor to a smaller second smoothing factor. The smaller smoothing factor makes the abnormal detection threshold more sensitive to the change of the frequency domain characteristic value of the new input, thereby accelerating the threshold update speed and quickly adapting to the signal change in the noise environment. After completing the abnormal detection threshold update based on the second smoothing factor, the smoothing factor is switched back to the normal first smoothing factor to ensure the smooth update of the threshold under normal conditions. Through this dynamic switching of the smoothing factor, the adjustment of the abnormal detection threshold can be accelerated when the noise impact is serious, and the system's adaptability to the dynamic noise environment and the real-time performance of abnormal detection can be improved. Switching the smoothing factor and using a smaller second smoothing factor can speed up the speed of the threshold falling or rising, quickly respond to noise changes, and is conducive to quickly adjusting the threshold to an appropriate level to ensure the sensitivity and accuracy of detection. After the noise impact is weakened, switch back to the first smoothing factor to ensure the stability of the threshold update.

[0057] Specifically, the system monitors in real time the weight reduction ratio of the currently acquired frequency domain eigenvalues ​​in the exponential moving average calculation. The preset ratio threshold is used as a boundary value to judge the severity of the noise impact. For example, the preset ratio threshold can be set to 50%. When the weight reduction ratio calculated by the system exceeds 50%, it indicates that the current noise spike has a greater impact on the frequency domain eigenvalue, or the noise duration is long, and the adjustment speed of the abnormal detection threshold needs to be accelerated. At this time, the system automatically switches the smoothing factor of the exponential moving average algorithm from the first smoothing factor to the second smoothing factor. The first smoothing factor and the second smoothing factor are both pre-set parameters, and the second smoothing factor is less than the first smoothing factor. For example, the first smoothing factor can be set to 0.2, and the second smoothing factor can be set to 0.05. After using the second smoothing factor, the exponential moving average algorithm will consider the currently acquired frequency domain eigenvalues ​​more when updating the abnormal detection threshold, thereby speeding up the threshold's response speed to noise changes. After completing the abnormal detection threshold update based on the second smoothing factor, the system switches the smoothing factor back to the first smoothing factor to ensure that under normal stamping conditions, the abnormal detection threshold can smoothly adapt to the slow drift of the equipment working conditions.

[0058] In some specific embodiments, when the weight reduction ratio exceeds a preset ratio threshold, the noise timer is started and the noise duration is recorded. When the weight reduction ratio drops below the preset ratio threshold, the noise timer is stopped and the noise duration is obtained. According to the noise duration, a preset duration and smoothing factor mapping table is searched, and a second smoothing factor corresponding to the noise duration is determined. For example, the preset duration and smoothing factor mapping table can be set to: when the noise duration is less than 1 second, the second smoothing factor is 0.1; when the noise duration is 1 second to 5 seconds, the second smoothing factor is 0.05; when the noise duration is greater than 5 seconds, the second smoothing factor is 0.02. In this way, the appropriate second smoothing factor can be adaptively selected according to the length of the noise duration, so as to more finely adjust the update speed of the abnormality detection threshold and improve the robustness of the system in complex noise environments.

[0059] In some possible implementations, if the weight reduction ratio of the currently acquired frequency domain eigenvalue in the exponential moving average calculation exceeds a preset ratio threshold, the step of switching the smoothing factor of the exponential moving average algorithm to a second smoothing factor, where the second smoothing factor is less than the first smoothing factor used when the normal update threshold is reached, includes: When the weight reduction ratio exceeds the preset ratio threshold, the noise timer is started to record the noise duration; When the weight reduction ratio drops below a preset ratio threshold, the noise timer is stopped to obtain the noise duration; According to the noise duration, a preset duration and smoothing factor mapping table is searched to determine a second smoothing factor corresponding to the noise duration.

[0060] Wherein, the noise timer is configured to start timing when the weight reduction ratio exceeds a preset ratio threshold, and stop timing when the weight reduction ratio drops below the preset ratio threshold, thereby obtaining the noise duration. The preset duration and smoothing factor mapping table stores the correspondence between the noise duration and the second smoothing factor, wherein the longer the noise duration, the smaller the corresponding second smoothing factor is set to. By looking up this mapping table, a suitable second smoothing factor can be determined according to different noise durations. The determined second smoothing factor is used in the exponential moving average algorithm to update the anomaly detection threshold. Since the second smoothing factor is less than the first smoothing factor used when the threshold is normally updated, the update speed of the anomaly detection threshold is accelerated during the noise duration, so that the anomaly detection threshold can adapt to changes in the noise environment more quickly.

[0061] Specifically, when the system detects that the weight reduction ratio exceeds the preset ratio threshold, it indicates that the frequency domain feature value currently obtained may be affected by the noise spike. At this time, the noise timer is immediately started to record the duration of the noise. During the duration of the noise, the weight reduction ratio may continue to be higher than the preset ratio threshold, or it may rise again after a short reduction. Only when the weight reduction ratio is stably reduced to below the preset ratio threshold, the noise impact is considered to be basically over, the noise timer is stopped, and the complete noise duration from start to stop is recorded. After obtaining the noise duration, the system will search for the corresponding second smoothing factor in the preset duration and smoothing factor mapping table according to this duration. In the mapping table, the duration is divided into multiple intervals, and each interval corresponds to a preset second smoothing factor value. The longer the duration of the interval, the smaller the corresponding second smoothing factor value. By searching the mapping table, the system can adaptively select the appropriate second smoothing factor according to the length of the noise duration, and use this second smoothing factor in the exponential moving average algorithm to update the anomaly detection threshold, so as to achieve dynamic adjustment of the threshold update speed.

[0062] In some specific embodiments, the preset ratio threshold is set to 80% of the weight reduction ratio. The duration and smoothing factor mapping table is preset as follows: when the noise duration is less than 0.1 seconds, the second smoothing factor is set to 0.8; when the noise duration is between 0.1 seconds and 0.5 seconds, the second smoothing factor is set to 0.5; when the noise duration is greater than 0.5 seconds, the second smoothing factor is set to 0.2. When it is detected that the weight reduction ratio exceeds 80%, the noise timer starts. If the weight reduction ratio drops below 80% after the noise lasts for 0.3 seconds, the noise timer stops and the noise duration is obtained as 0.3 seconds. According to the noise duration of 0.3 seconds, the mapping table is searched to determine that the corresponding second smoothing factor is 0.5. Then, the second smoothing factor of 0.5 is used to update the anomaly detection threshold based on the exponential moving average algorithm. In this way, the anomaly detection threshold can adaptively adjust the update speed according to the noise duration, thereby improving the accuracy of anomaly detection in a noisy environment.

[0063] In this embodiment, step A4 includes: A401. Calculate the difference between the currently acquired frequency domain feature value and the dynamically adjusted anomaly detection threshold, and record it as the feature difference; A402. Compare the characteristic difference with the preset first deviation threshold and the second deviation threshold; A403. If the characteristic difference is less than or equal to the first deviation threshold, the stamping process is determined to be normal, and the normal state detection result is output; A404. If the characteristic difference is greater than the first deviation threshold and less than or equal to the second deviation threshold, the stamping process is determined to be a potential abnormal state, and the potential abnormal state detection result is output; A405. If the characteristic difference is greater than the second deviation threshold, the stamping process is determined to be in a serious abnormal state, and the serious abnormal state detection result is output.

[0064] Among them, in order to more finely evaluate the abnormal state of the stamping process, the abnormal state is divided into multiple levels. Specifically, after obtaining the current frequency domain eigenvalue and obtaining the dynamically adjusted abnormal detection threshold, the system first calculates the difference between the two to obtain the characteristic difference. This characteristic difference represents the degree to which the current stamping state deviates from the normal state. Subsequently, the calculated characteristic difference is compared with two preset thresholds, namely the first deviation threshold and the second deviation threshold. The first deviation threshold and the second deviation threshold are pre-set (can be set according to actual needs) to define different levels of abnormal states. For example, the first deviation threshold can be set to a relatively small value, indicating that there is a slight abnormality in the stamping process. The second deviation threshold is set to a relatively large value, indicating that there is a serious abnormality in the stamping process. The comparison process divides the state of the stamping process into three levels according to the size of the characteristic difference. When the difference is less than or equal to the first deviation threshold, the system determines that the stamping process is in a normal state. This means that the current stamping parameters and equipment are in good operating condition and the product quality is stable. In this case, the system outputs a normal state detection result, indicating that the production process can continue to maintain the current state. When the characteristic difference is between the first deviation threshold and the second deviation threshold, the system determines that the stamping process is in a potential abnormal state. This means that the stamping process may begin to have slight abnormalities, but has not yet reached a serious level. The system outputs the potential abnormal state detection result, which can trigger an early warning mechanism to remind the operator to pay attention, or the control system can make small adjustments to prevent the abnormal state from further deteriorating. When the characteristic difference is greater than the second deviation threshold, the system determines that the stamping process is in a serious abnormal state. This indicates that the stamping process has undergone significant abnormalities, which may seriously affect product quality and even equipment safety. The system outputs the serious abnormal state detection result, and the control system can take immediate measures, such as stopping the machine for inspection or significantly adjusting the stamping parameters to avoid greater losses. Through this hierarchical abnormality judgment method, the abnormality detection system can provide richer and more operationally instructive detection results. The control system can adopt corresponding control strategies according to different levels of abnormal states to achieve more refined process control and optimization.

[0065] Specifically, on the glass bottle stamping production line, high-speed rotary stamping equipment is running continuously. In order to monitor the stamping quality in real time, the anomaly detection system is deployed at each stamping station. The system first collects the vibration signal of the punch head during the stamping process, and performs frequency domain analysis to extract the frequency domain eigenvalues ​​at specific excitation frequencies. These frequency domain eigenvalues ​​reflect the subtle changes in the stamping process. In order to adapt to the dynamic changes in the production environment, the system uses the exponential moving average algorithm to dynamically adjust the anomaly detection threshold. When making anomaly judgments, the system calculates the difference between the currently acquired frequency domain eigenvalues ​​and the dynamically adjusted anomaly detection threshold to obtain the characteristic difference. The first deviation threshold and the second deviation threshold are pre-set. For example, the first deviation threshold can be set to 3 units, and the second deviation threshold can be set to 8 units. When the calculated characteristic difference is less than or equal to 3 units, the system determines that the stamping process is in a normal state and outputs a "normal" test result. After receiving the "normal" result, the control system maintains the current stamping parameters unchanged. When the characteristic difference is greater than 3 units and less than or equal to 8 units, the system determines that the stamping process is in a potential abnormal state and outputs the detection result of "potential abnormality". After receiving the "potential abnormality" result, the control system can issue a warning signal and slightly adjust parameters such as stamping pressure or stamping time to observe whether the potential abnormality can be eliminated. When the characteristic difference is greater than 8 units, the system determines that the stamping process is in a serious abnormal state and outputs the detection result of "serious abnormality". After receiving the "serious abnormality" result, the control system immediately shuts down and alarms the operator to check and maintain. Through the division and processing of this three-level abnormal state, the monitoring of the stamping process is more refined, and the control system can take corresponding measures according to the severity of the abnormality, which not only ensures the stability of production but also improves product quality.

[0066] In some specific embodiments, the first deviation threshold is set to 3, and the second deviation threshold is set to 8. In the actual stamping process, the system collects frequency domain eigenvalues ​​in real time. Assume that at a certain moment, the collected frequency domain eigenvalue is 15. At this time, the dynamically adjusted abnormality detection threshold is 10. The system first calculates the characteristic difference and obtains 15-10=5. Then, the system compares the characteristic difference 5 with the first deviation threshold 3 and the second deviation threshold 8. Since 5 is greater than 3 and less than or equal to 8, the system determines that the stamping process is in a potential abnormal state and outputs a detection result of "potential abnormality". After receiving the "potential abnormality" result, the control system can start the preset control strategy, for example, reduce the stamping pressure by 5%, and continue to monitor the stamping process. If the subsequent test results return to normal, the new stamping parameters are maintained; if the abnormal state persists or worsens, further control measures are taken. This hierarchical threshold comparison method enables the abnormality detection system to more accurately reflect the true state of the stamping process and provide a more refined control basis for the control system.

[0067] In some embodiments, step A5 comprises: A501. Determine the abnormal state level of the stamping process based on the abnormal detection results; A502. Select a preset control strategy based on the abnormal state level; A503. Execute preset control strategy and adjust stamping parameters.

[0068] Among them, in order to solve the problem of how the control system adjusts the stamping parameters according to different degrees of abnormal conditions to achieve effective control, after receiving the abnormal detection results, the control system will first divide the abnormal conditions of the stamping process into different levels according to the results, such as normal state, potential abnormal state, severe abnormal state, etc. Then, for each abnormal state level, the corresponding control strategy is pre-set in the system. For example, for the normal state, the control system can maintain the current stamping parameters unchanged; for the potential abnormal state, the control system can fine-tune parameters such as stamping pressure or stamping speed to observe whether the potential abnormality can be eliminated; for the severe abnormal state, the control system can immediately suspend the stamping process and send out an alarm signal to avoid product defects or equipment damage. Finally, the control system executes the preset control strategy that matches the current abnormal state level, and realizes closed-loop control of the stamping process by adjusting the stamping parameters, thereby ensuring product quality and production stability.

[0069] The classification of abnormal state levels can be based on the comparison results of deviation thresholds. For example, when the difference between the frequency domain eigenvalue and the dynamically adjusted abnormal detection threshold is less than or equal to the first preset deviation threshold, it is determined to be a normal state; when the difference is greater than the first preset deviation threshold and less than or equal to the second preset deviation threshold, it is determined to be a potential abnormal state; when the difference is greater than the second preset deviation threshold, it is determined to be a serious abnormal state. The preset control strategy can be stored in the memory of the control system and implemented in the form of a lookup table or a conditional judgment statement. The control system retrieves and executes the corresponding control strategy according to the determined abnormal state level, thereby realizing automatic adjustment of the stamping parameters.

[0070] Specifically, on the glass bottle stamping production line, after the anomaly detection system detects an abnormality in the stamping process, it first needs to determine the severity of the anomaly. For example, if the anomaly detection result shows that the deviation between the frequency domain eigenvalue and the anomaly detection threshold exceeds the second preset deviation threshold, the control system determines that the stamping process is in a serious abnormal state and sets the abnormal state level to "serious". Subsequently, the control system searches the preset control strategy table, and the control strategy corresponding to the "serious abnormality" level in the table is "immediately stop stamping and alarm". As a result, the control system immediately issues a shutdown command to stop the punching head of the current station, and starts the alarm device to prompt the operator to perform inspection and maintenance. If the anomaly detection result shows that the deviation between the frequency domain eigenvalue and the anomaly detection threshold is between the first preset deviation threshold and the second preset deviation threshold, the control system determines that the stamping process is in a potential abnormal state and sets the abnormal state level to "potential". The control strategy corresponding to the "potential abnormality" level in the control strategy table can be "fine-tuning the stamping pressure". As a result, the control system sends an instruction to the pressure control unit of the stamping equipment to slightly reduce the stamping pressure by a certain percentage, such as 5%, and then continues to monitor the stamping process to observe whether the abnormal state is eliminated. Through this hierarchical control strategy, the control system can take different countermeasures according to the severity of the abnormality, avoid misjudgment and missed judgment, and achieve precise control of the stamping process.

[0071] In some specific embodiments, the control system can be implemented by a PLC programmable logic controller, which receives the abnormal detection results fed back by the abnormal detection system and executes the corresponding control strategy according to the pre-programmed control logic. For example, a mapping table of abnormal state levels and control strategies can be established in the PLC in advance, and the table defines the stamping parameter adjustment schemes corresponding to different abnormal state levels, including the adjustment amount or adjustment mode of parameters such as stamping pressure, stamping speed, and stamping time. When the PLC receives the abnormal state level signal output by the abnormal detection system, for example, through a digital signal or bus communication method, the PLC searches for the corresponding control strategy in the mapping table according to the signal, and converts the control strategy into a specific control instruction, which is sent to the actuator of the stamping equipment, such as a pressure valve, a motor driver, etc., so as to realize the automatic adjustment of the stamping parameters. As a preferred embodiment, the control system can also have a human-computer interaction interface, and the operator can view the current abnormal state level, the execution of the control strategy, and the real-time data of the stamping parameters through the human-computer interaction interface, and can manually intervene in the control process, such as modifying the preset control strategy or manually adjusting the stamping parameters to cope with complex production situations.

[0072] refer to Figure 2 The present application also provides a glass bottle punching anomaly detection device for detecting anomalies in a glass bottle punching process performed by a high-speed rotary punching device, the device comprising: The vibration signal acquisition module 1 is used to apply a micro-frequency vibration excitation of a specific frequency to the punch head of the punching equipment and collect the vibration signal of the punch head during the punching process (for the specific process, refer to step A1 above); Frequency domain analysis module 2, used to perform frequency domain analysis on the collected vibration signal to extract the frequency domain eigenvalue at the excitation frequency (for the specific process, refer to step A2 above); Threshold adjustment module 3, used to dynamically adjust the anomaly detection threshold based on the exponential moving average algorithm according to the continuously acquired frequency domain feature values ​​(for the specific process, refer to step A3 above); Anomaly detection module 4, used to compare the currently acquired frequency domain eigenvalue with the dynamically adjusted anomaly detection threshold, determine whether the stamping process is abnormal according to the comparison result, and output the anomaly detection result (for the specific process, refer to step A4 above); The system adjustment module 5 is used to feed back the abnormal detection result to the control system of the stamping equipment in real time, so that the control system adjusts the stamping parameters according to the abnormal detection result (for the specific process, refer to step A5 above).

[0073] Please refer to Figure 3 , Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application, the present application provides an electronic device, including: a processor 301 and a memory 302, the processor 301 and the memory 302 are interconnected and communicate with each other through a communication bus 303 and / or other forms of connection mechanisms (not shown), the memory 302 stores a computer program executable by the processor 301, and when the electronic device is running, the processor 301 executes the computer program to execute the glass bottle stamping abnormality detection method in any optional implementation of the above embodiment to achieve the following functions: while applying a micro-frequency vibration excitation of a specific frequency to the stamping head of the stamping equipment, collecting the vibration signal of the stamping head during the stamping process; performing frequency domain analysis on the collected vibration signal to extract the frequency domain eigenvalue at the excitation frequency; based on the exponential moving average algorithm, dynamically adjusting the abnormality detection threshold according to the continuously acquired frequency domain eigenvalues; comparing the currently acquired frequency domain eigenvalue with the dynamically adjusted abnormality detection threshold, judging whether the stamping process is abnormal according to the comparison result, and outputting the abnormality detection result; feeding back the abnormality detection result to the control system of the stamping equipment in real time, so that the control system adjusts the stamping parameters according to the abnormality detection result.

[0074] The embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the glass bottle stamping abnormality detection method in any optional implementation of the above-mentioned embodiment is executed to achieve the following functions: while applying a micro-frequency vibration excitation of a specific frequency to the stamping head of the stamping equipment, the vibration signal of the stamping head during the stamping process is collected; frequency domain analysis is performed on the collected vibration signal to extract the frequency domain eigenvalue at the excitation frequency; based on the exponential moving average algorithm, the abnormality detection threshold is dynamically adjusted according to the continuously acquired frequency domain eigenvalues; the currently acquired frequency domain eigenvalue is compared with the dynamically adjusted abnormality detection threshold, and whether the stamping process is abnormal is determined according to the comparison result, and the abnormality detection result is output; the abnormality detection result is fed back to the control system of the stamping equipment in real time, so that the control system adjusts the stamping parameters according to the abnormality detection result.

[0075] Among them, the computer-readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read-Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0076] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0077] In addition, the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0078] Furthermore, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist independently, or two or more modules may be integrated to form an independent part.

[0079] In this document, relational terms such as first and second, etc. are used merely to distinguish one entity or operation from another entity or operation, but do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0080] The above description is only an embodiment of the present application and is not intended to limit the protection scope of the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for detecting anomalies in glass bottle punching, used for detecting anomalies in a glass bottle punching process performed by a high-speed rotary punching device, characterized in that: The method comprises the following steps: A1. While applying a micro-frequency vibration excitation of a specific frequency to the punch head of the punching equipment, collecting the vibration signal of the punch head during the punching process; A2. Perform frequency domain analysis on the collected vibration signal to extract the frequency domain eigenvalue at the excitation frequency; A3. Based on the exponential moving average algorithm, the anomaly detection threshold is dynamically adjusted according to the continuously acquired frequency domain feature values; A4. Compare the currently acquired frequency domain eigenvalue with the dynamically adjusted anomaly detection threshold, determine whether the stamping process is abnormal based on the comparison result, and output the anomaly detection result; A5. Feedback the abnormality detection result to the control system of the stamping equipment in real time, so that the control system adjusts the stamping parameters according to the abnormality detection result.

2. A method for detecting abnormal punching of a glass bottle according to claim 1, characterized in that: Step A3 includes: Obtaining the rotation speed of the high-speed rotary punching device; comparing the rotation speed with a preset speed threshold; When the rotation speed is greater than or equal to the preset speed threshold, setting the frequency of the dynamically adjusted abnormality detection threshold to a first frequency; When the rotation speed is less than the preset speed threshold, setting the frequency of the dynamically adjusted abnormality detection threshold to a second frequency; The first frequency is greater than the second frequency.

3. A method for detecting abnormal punching of a glass bottle according to claim 1, characterized in that: Step A3 includes: After the system is started, it enters the initial threshold setting stage; In the initial threshold setting stage, the frequency domain feature values ​​within the preset time window are continuously collected; Calculating the average value of the frequency domain characteristic values ​​collected within the preset time window; The calculated average value is set as the initial anomaly detection threshold; After the initial anomaly detection threshold is set, the exponential moving average algorithm is started based on the set initial anomaly detection threshold, and the anomaly detection threshold is dynamically adjusted according to the frequency domain eigenvalues ​​subsequently obtained in real time.

4. A method for detecting abnormal punching of a glass bottle according to claim 3, characterized in that: The step of starting the exponential moving average algorithm based on the set initial anomaly detection threshold and dynamically adjusting the anomaly detection threshold according to the frequency domain characteristic value subsequently acquired in real time includes: Calculate the difference between the frequency domain eigenvalue currently obtained and the frequency domain eigenvalue obtained at the previous moment; Determine whether the difference exceeds a preset mutation amplitude threshold; If the difference exceeds the preset mutation amplitude threshold, it is determined that the currently acquired frequency domain feature value contains a noise peak; When it is determined that the currently acquired frequency domain feature value contains a noise spike, when updating the abnormality detection threshold based on the exponential moving average algorithm, reducing the weight of the currently acquired frequency domain feature value in the exponential moving average calculation; Based on the frequency domain feature values ​​after weight reduction, the exponential moving average algorithm is used to update the anomaly detection threshold.

5. A method for detecting abnormal punching of a glass bottle according to claim 4, characterized in that: The step of updating the anomaly detection threshold using an exponential moving average algorithm based on the frequency domain eigenvalue after weight reduction includes: If the weight reduction ratio of the currently acquired frequency domain eigenvalue in the exponential moving average calculation exceeds a preset ratio threshold, the smoothing factor of the exponential moving average algorithm is switched to a second smoothing factor, which is smaller than the first smoothing factor used when the normal update threshold is used; Using the second smoothing factor, updating the anomaly detection threshold based on an exponential moving average algorithm; After the abnormality detection threshold is updated based on the second smoothing factor, the smoothing factor of the exponential moving average algorithm is switched back to the first smoothing factor.

6. A method for detecting abnormal punching of a glass bottle according to claim 5, characterized in that: The step of switching the smoothing factor of the exponential moving average algorithm to a second smoothing factor if the weight reduction ratio of the currently acquired frequency domain eigenvalue in the exponential moving average calculation exceeds a preset ratio threshold, wherein the second smoothing factor is less than the first smoothing factor used when the normal update threshold is: When the weight reduction ratio exceeds the preset ratio threshold, the noise timer is started to record the noise duration; When the weight reduction ratio drops below a preset ratio threshold, the noise timer is stopped to obtain the noise duration; According to the noise duration, a preset duration and smoothing factor mapping table is searched to determine a second smoothing factor corresponding to the noise duration.

7. A method for detecting abnormal punching of a glass bottle according to claim 1, characterized in that: Step A5 includes: Determining the abnormal state level of the stamping process according to the abnormality detection result; According to the abnormal state level, select a preset control strategy; Execute the preset control strategy and adjust the stamping parameters.

8. A glass bottle punching anomaly detection device, used for anomaly detection in a glass bottle punching process performed by a high-speed rotary punching device, characterized in that: The device includes: A vibration signal acquisition module, used to apply a micro-frequency vibration excitation of a specific frequency to the punch head of the punching equipment and collect the vibration signal of the punch head during the punching process; A frequency domain analysis module is used to perform frequency domain analysis on the collected vibration signal to extract the frequency domain characteristic value at the excitation frequency; A threshold adjustment module is used to dynamically adjust the anomaly detection threshold based on the exponential moving average algorithm according to the continuously acquired frequency domain feature values; The anomaly detection module is used to compare the currently acquired frequency domain characteristic value with the dynamically adjusted anomaly detection threshold, determine whether the stamping process is abnormal based on the comparison result, and output the anomaly detection result; The system adjustment module is used to feed back the abnormality detection result to the control system of the stamping equipment in real time, so that the control system adjusts the stamping parameters according to the abnormality detection result.

9. An electronic device, characterized in that: The invention comprises a processor and a memory, wherein the memory stores a computer program executable by the processor, and when the processor executes the computer program, the steps in the method for detecting abnormal punching of a glass bottle as claimed in any one of claims 1 to 7 are executed.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, the steps of the glass bottle stamping abnormality detection method as described in any one of claims 1 to 7 are executed.

Images