An equipment fault analysis and prediction system and method based on frequency domain characteristics

Abstract

The present invention relates to a device fault analysis and prediction system and method based on frequency domain characteristics, belonging to the field of data processing. It collects vibration signals during the operation of a target device, performs frequency domain processing to obtain sideband frequency parameters, conducts sideband frequency recurrence analysis on historical device frequency domain data to obtain sideband frequency recurrence coefficients; collects workpiece processing images of workpieces during the processing of the target device, conducts workpiece anomaly analysis to obtain workpiece anomaly parameters; configures a fault identification strategy according to the sideband frequency recurrence coefficients and workpiece anomaly parameters, performs fault identification on the sideband frequency parameters to obtain a first failure rate; calculates a second failure rate based on the sideband frequency parameters, and combines the first failure rate to calculate the device failure rate, which is used as the device fault analysis and prediction result, solving the technical problems of low fault detection efficiency and difficulty in accurately predicting the occurrence of faults.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and particularly to a device fault analysis and prediction system and method based on frequency domain characteristics. Background Art

[0002] In modern industrial production, the stable operation of equipment is crucial for production efficiency and product quality. However, as equipment operates for a long time, its internal components often fail due to wear, aging, or external environmental factors, thus affecting the production schedule and product quality. Traditional equipment fault detection methods usually rely on manual inspections and experience-based judgments. This method is not only inefficient but also difficult to accurately predict the occurrence of faults. To overcome the deficiencies of traditional equipment fault detection methods, a fault prediction technology based on data analysis has been proposed in the prior art. Among them, frequency domain analysis, as an effective signal processing means, is widely used in the field of equipment fault diagnosis. Frequency domain analysis can extract frequency domain features related to the operating state of the equipment by converting the vibration signal of the equipment from the time domain to the frequency domain. These frequency domain features can reflect the vibration state of the internal components of the equipment, thus providing a strong basis for fault prediction. However, there are still limitations in fault prediction based solely on frequency domain analysis. In reality, equipment faults are often related to multiple factors. In addition to vibration signals, abnormalities in workpieces during the processing process are also one of the important fault omens. For example, abnormalities such as deformation and cracks in workpieces during the processing process are often closely related to the operating state of the equipment. Therefore, how to further improve the accuracy of fault analysis and ensure the maintenance and servicing conditions of the equipment deserves further attention. Summary of the Invention

[0003] Aiming at the technical problems of low efficiency in fault detection and difficulty in accurately predicting the occurrence of faults in the prior art, the present invention provides a device fault analysis and prediction system and method based on frequency domain characteristics to solve the problems.

[0004] The technical solutions of the present invention for solving the above technical problems are as follows:

[0005] In a first aspect, the present invention provides a device fault analysis and prediction system based on frequency domain characteristics. The system includes: a signal acquisition module for acquiring vibration signals during the operation of a target device, performing frequency domain processing to obtain sideband parameters, and performing sideband recurrence analysis on historical device frequency domain data to obtain a sideband recurrence coefficient; an image acquisition module for acquiring workpiece processing images of workpieces during the processing of the target device, performing workpiece abnormality analysis to obtain workpiece abnormality parameters; a fault identification module for configuring a fault identification strategy based on the sideband recurrence coefficient and the workpiece abnormality parameters, performing fault identification on the sideband parameters to obtain a first failure rate; and a result prediction module for calculating a second failure rate based on the sideband parameters, and combining the first failure rate to calculate a device failure rate as the result of device fault analysis and prediction.

[0006] In a second aspect, the present invention provides a device fault analysis and prediction method based on frequency domain characteristics. The method includes: acquiring vibration signals during the operation of a target device, performing frequency domain processing to obtain sideband parameters, and performing sideband recurrence analysis on historical device frequency domain data to obtain a sideband recurrence coefficient; acquiring workpiece processing images of workpieces during the processing of the target device, performing workpiece abnormality analysis to obtain workpiece abnormality parameters; configuring a fault identification strategy based on the sideband recurrence coefficient and the workpiece abnormality parameters, performing fault identification on the sideband parameters to obtain a first failure rate; calculating a second failure rate based on the sideband parameters, and combining the first failure rate to calculate a device failure rate as the result of device fault analysis and prediction.

[0007] The beneficial effects of the present invention are as follows: By combining the sideband parameters obtained through frequency domain processing with the workpiece abnormality parameters obtained through workpiece abnormality analysis, it is possible to more efficiently and accurately predict device faults, improve the accuracy of fault identification, and obtain the device failure rate by comprehensively calculating the first failure rate and the second failure rate, providing a reliable basis for preventive maintenance and fault warning of the device. Description of the Drawings

[0008] Figure 1 It is a schematic structural diagram of a device fault analysis and prediction system based on frequency domain characteristics provided by the present invention.

[0009] Figure 2 It is a schematic flow diagram of a device fault analysis and prediction method based on frequency domain characteristics provided by the present invention.

[0010] Description of the reference numerals: signal acquisition module 11, image acquisition module 12, fault identification module 13, result prediction module 14. Detailed Embodiments

[0011] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present invention.

[0012] In the description of the present invention, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of the described features. In the description of the present invention, "a plurality" means two or more, unless otherwise specifically defined.

[0013] In the description of the present invention, the term "for example" is used to mean "serving as an example, illustration, or explanation". Any embodiment described as "for example" in the present invention is not necessarily construed as being more preferred or having more advantages than other embodiments. In order for any person skilled in the art to implement and use the present invention, the following description is given. In the following description, details are set forth for purposes of explanation. It should be understood that those of ordinary skill in the art can recognize that the present invention can be implemented without the use of these specific details. In other instances, well-known structures and processes are not elaborated in detail to avoid unnecessary details from obscuring the description of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope that conforms to the principles and features disclosed in the present invention.

[0014] Embodiment 1:

[0015] As Figure 1 shown, the embodiment of the present invention provides a device fault analysis and prediction system based on frequency domain characteristics. The system includes:

[0016] A signal acquisition module 11, configured to collect vibration signals during the operation of a target device, perform frequency domain processing to obtain sideband parameters, and perform sideband recurrence analysis on historical device frequency domain data to obtain a sideband recurrence coefficient.

[0017] Exemplarily, this solution is for fault diagnosis of rotating equipment based on sidebands in the frequency domain. As the frequency components on both sides of the carrier signal, the existence and characteristics of sidebands are crucial for analyzing equipment faults. Specifically, in the process of fault diagnosis of rotating equipment, it is first necessary to collect vibration signals during the operation of the target equipment. These vibration signals contain the operation state information of various components inside the equipment and are important bases for diagnosing faults. After the vibration signals are collected, they are subjected to frequency-domain processing. Frequency-domain processing is the process of converting time-domain signals into frequency-domain signals. Through this process, the frequency components of the vibration signals can be obtained, including the fundamental frequency and sidebands, etc. Sidebands refer to the frequency components that appear near the main frequency (i.e., the fundamental frequency, the main vibration frequency) after the frequency-domain transformation of the vibration signals. The changes in parameters such as the spacing and amplitude of these sidebands are often associated with specific fault modes of the equipment. Therefore, the fault conditions of the equipment can be identified by analyzing the sideband parameters. After obtaining the sideband parameters, further sideband recurrence analysis is carried out in the historical equipment frequency-domain data. The purpose of this step is to compare the sideband parameters of the current equipment with those of the known faulty equipment in history to identify whether the current equipment has sideband characteristics similar to those of the known faults in history. Specifically, focus on the key parameter of sideband spacing because when a fault occurs, the sideband spacing often shows similar changes. If a certain spacing appears with a relatively large proportion in the sideband spacings within the historical time, then it is considered that this spacing may be related to a certain fault. Through sideband recurrence analysis, a sideband recurrence coefficient can be obtained. The sideband recurrence coefficient is a quantitative index used to measure the similarity between the sideband parameters of the current equipment and those of the known faulty equipment in history. The larger the sideband recurrence coefficient, the more similar the sideband characteristics of the current equipment are to those of the known faulty equipment in history, and thus the greater the probability of a fault occurring. On the contrary, if the sideband recurrence coefficient is small, it indicates that the sideband characteristics of the current equipment are quite different from those of the known faulty equipment in history, and thus the probability of a fault occurring is also small. At this time, it is more likely that the workpiece is abnormally deformed or the equipment itself has no fault. In summary, the sideband recurrence coefficient provides a method for quantitatively evaluating the probability of equipment faults, which helps to detect potential faults in a timely manner and ensure the normal operation of the equipment.

[0018] The image acquisition module 12 is used to acquire the workpiece processing images of the workpiece during the processing of the target equipment, perform workpiece anomaly analysis, and obtain workpiece anomaly parameters.

[0019] Optionally, in the industrial production process, to ensure product quality and production efficiency, it is crucial to monitor the workpiece during the machining process of the target equipment in real time. This monitoring process is achieved by collecting workpiece machining images and identifying whether there are any abnormalities in the workpiece through image analysis. Using advanced image acquisition devices such as high-definition cameras or industrial cameras, the workpiece during the machining process of the target equipment is photographed in real time. These images capture the real-time state of the workpiece during the machining process and provide basic data for subsequent abnormality analysis. Next, abnormality analysis is performed on the collected workpiece images. This process usually involves steps such as image preprocessing, feature extraction, and abnormality recognition. In the image preprocessing stage, operations such as denoising and enhancing contrast are performed on the image to improve the image quality for subsequent feature extraction. In the feature extraction stage, using image processing algorithms such as edge detection and contour extraction, key features related to workpiece deformation are extracted from the image. These features may include the size, shape, surface texture, etc. of the workpiece. Finally, based on the extracted features, machine learning or deep learning algorithms are used to identify abnormalities in the workpiece. The purpose of this step is to determine whether there are any abnormalities such as deformation in the workpiece. Specifically, the extracted features are compared with the preset normal workpiece feature library, and the abnormality degree of the workpiece is evaluated by calculating the difference or similarity between the features. If the difference exceeds the preset threshold or the similarity is lower than a certain standard, it is considered that the workpiece has an abnormality, and the corresponding workpiece abnormality parameters are output. These parameters include information such as the amplitude, position, and direction of workpiece deformation, which have important reference values for subsequent equipment fault analysis and product quality control. In summary, by collecting workpiece machining images and performing abnormality analysis, abnormalities such as deformation of the workpiece during the machining process can be effectively identified, providing strong support for the monitoring and control of the production process.

[0020] The fault identification module 13 is used to configure a fault identification strategy according to the sideband recurrence coefficient and the workpiece abnormality parameters, perform fault identification on the sideband parameters, and obtain the first failure rate.

[0021] Furthermore, a fault identification strategy is configured by combining the sideband recurrence coefficient and workpiece anomaly parameters, and this process aims to more precisely evaluate the fault status of the equipment. Specifically, the sideband recurrence coefficient is calculated based on historical data and currently collected sideband parameters, and this coefficient reflects the similarity between the current sideband characteristics of the equipment and those of known faulty equipment in history. The larger the sideband recurrence coefficient, the more similar the current sideband parameters of the equipment are to those of the faulty equipment in history. Therefore, it is more likely that the equipment is considered to have a fault. In this case, to more comprehensively evaluate the fault, the number of identification paths W is increased, that is, more analysis methods and detection means are used to confirm the fault. At the same time, the influence of workpiece anomaly parameters on fault identification is also considered. The workpiece anomaly parameters are obtained by collecting and analyzing workpiece processing images, and they reflect abnormal conditions such as deformation and cracks that may occur during workpiece processing. When the workpiece anomaly parameters are larger, it is more likely that the abnormality of the sideband parameters is caused by workpiece anomalies rather than equipment faults. Therefore, in this case, the number of identification paths W is reduced because the workpiece anomalies already provide a reasonable explanation and there is no need to further explore the possibility of equipment faults. After configuring the fault identification strategy, fault identification is performed on the sideband parameters. This process comprehensively considers the influence of the sideband recurrence coefficient and workpiece anomaly parameters, and obtains the first failure rate through methods such as comparative analysis and pattern recognition. The first failure rate is a quantitative index used to evaluate the probability that the equipment currently has a fault. It comprehensively considers the degree of abnormality of the sideband parameters and the possibility of workpiece anomalies, providing a more accurate basis for fault judgment. In summary, configuring the fault identification strategy according to the sideband recurrence coefficient and workpiece anomaly parameters and performing fault identification on the sideband parameters can more precisely evaluate the fault status of the equipment, providing strong support for subsequent maintenance and repair work. The larger the sideband recurrence coefficient, the more likely it is that the equipment is considered to have a fault, so the number of identification paths W will be increased; while the larger the workpiece anomaly parameters, the more likely it is that the fault is caused by workpiece anomalies, so the number of identification paths W will be reduced. This flexible strategy adjustment improves the efficiency of fault analysis.

[0022] The result prediction module 14 is configured to calculate and obtain a second failure rate according to the sideband parameters, and combine the first failure rate to calculate and obtain the equipment failure rate as the equipment fault analysis prediction result.

[0023] Specifically, considering the influence of sideband frequency parameters and workpiece anomaly parameters comprehensively, two failure rates are obtained through calculation, namely the first failure rate and the second failure rate, and finally the equipment failure rate is obtained by combining the two as the prediction result. Specifically, the second failure rate is calculated according to the sideband frequency parameters. The sideband frequency parameters refer to the occurrence of certain specific frequency components during the operation of the equipment, and these frequency components are often closely related to the operation state of the equipment. By monitoring and analyzing the sideband frequency parameters, the recurrence ratio of the sideband frequency can be obtained, that is, the ratio of the frequency of occurrence of specific frequency components during the monitoring period to the total monitoring time. This recurrence ratio can reflect the stability and reliability of the equipment under certain specific conditions. The second failure rate is calculated by analyzing the recurrence ratio of the sideband frequency parameters. The larger the recurrence ratio, the smaller the second failure rate, that is, the larger the recurrence ratio means the more stable the operation state of the equipment at this frequency, so the second failure rate is smaller. This is because if the equipment can operate stably at a certain frequency, the probability of failure is relatively low, so the failure risk of the equipment at this frequency can be indirectly evaluated through the sideband recurrence ratio. Then, the equipment failure rate is calculated by combining the first failure rate obtained previously according to the workpiece anomaly parameters and the sideband recurrence coefficient. The first failure rate is a basic failure rate obtained comprehensively based on various factors such as the historical failure data, operating environment, and maintenance records of the equipment, and this failure rate reflects the probability of failure of the equipment under normal operating conditions. The first failure rate and the second failure rate are weighted and calculated to obtain the equipment failure rate as the final prediction result. Among them, the weight of the first failure rate is greater than that of the second failure rate because when analyzing, the first failure rate comprehensively considers the influence of the sideband recurrence coefficient and workpiece anomaly parameters and more comprehensively reflects the failure state of the equipment, while although the second failure rate is also calculated based on the sideband frequency parameters, it is affected by the non-linear relationship between the recurrence ratio and the failure probability in a specific situation. The final prediction result provides a quantitative assessment of the current failure state of the equipment, considering both the overall operation state of the equipment and the stability at a specific frequency, so as to more comprehensively evaluate the failure risk of the equipment.

[0024] In a preferred embodiment, the vibration signal during the operation of the target equipment is collected, frequency-domain processing is performed to obtain sideband frequency parameters, and sideband recurrence analysis is performed on the historical equipment frequency-domain data to obtain the sideband recurrence coefficient. The execution steps further include: collecting the vibration signal during the operation of the target equipment; performing frequency-domain transformation on the vibration signal to obtain a frequency-domain signal; in the frequency-domain signal, performing sideband frequency extraction to obtain sideband frequency parameters, where the sideband frequency parameters include sideband frequency spacing; according to the sideband frequency spacing, performing sideband recurrence analysis on the historical equipment frequency-domain data to obtain the sideband recurrence coefficient.

[0025] Specifically, vibration signals are an important reflection of the operating state of equipment. By collecting these signals, a large amount of information during the operation of the equipment can be obtained. After the vibration signals are collected, they need to be processed for further analysis. Here, a frequency-domain processing method is adopted, that is, the vibration signals are transformed from the time domain to the frequency domain. Frequency-domain analysis can reveal the components and intensities of the signals at different frequencies, which is crucial for understanding the operating state and fault modes of the equipment. In the frequency-domain signals, specific frequency components, namely sidebands, are concerned. Sidebands are specific frequency components generated during the operation of the equipment, and they are often related to certain operating states or fault modes of the equipment. In order to extract these sidebands, sideband extraction needs to be carried out within the frequency-domain signals. In this process, the specific parameters of the sidebands are identified and recorded, and one of the most important parameters is the sideband spacing. The sideband spacing refers to the frequency difference between adjacent sidebands, which reflects the vibration characteristics of some structures or components inside the equipment. After obtaining the sideband parameters, it is necessary to further analyze the recurrence of these parameters in the historical data of the equipment. Sideband recurrence analysis refers to comparing the sideband parameters of the current equipment with the historical frequency-domain data of the equipment to evaluate the stability and reliability of the sidebands in the current operating state of the equipment. In this process, a sideband recurrence coefficient is calculated, which reflects the frequency and stability of the sidebands appearing in the historical data. The higher the sideband recurrence coefficient, the more stable the sidebands are in the operating state of the equipment, and the lower the probability of failure.

[0026] In a preferred embodiment, according to the sideband spacing, sideband recurrence analysis is performed within the historical frequency-domain data of the equipment to obtain the sideband recurrence coefficient. The execution steps further include: within the historical frequency-domain data of the equipment, extracting the set of historical sideband spacings when the equipment fails; obtaining the number of occurrences of the sideband spacing within the set of historical sideband spacings to obtain the recurrence ratio; obtaining the maximum ratio of the occurrences of different historical sideband spacings within the set of historical sideband spacings as the maximum recurrence ratio; calculating the ratio of the recurrence ratio to the maximum recurrence ratio as the sideband recurrence coefficient.

[0027] Further, in order to perform sideband frequency recurrence analysis, it is necessary to extract information from historical device frequency domain data. Specifically, the historical sideband frequency spacing set of the device when a fault occurs is extracted. This set contains the sideband frequency spacing data recorded by the device in the past under fault conditions and is the basis for recurrence analysis. Subsequently, the recurrence situation of the sideband frequency spacing of the current device in this set is calculated. The number of times the sideband frequency spacing of the current device appears in the historical sideband frequency spacing set is counted, and the ratio of this number to the total number of times in the set, that is, the recurrence ratio, is calculated. The recurrence ratio reflects the similarity between the sideband frequency spacing of the current device and the historical fault data. However, since different historical sideband frequency spacings may have different recurrence ratios, in order to evaluate whether the recurrence situation of the current sideband frequency spacing is significant, it is also necessary to obtain the maximum ratio of the occurrences of different historical sideband frequency spacings in the historical sideband frequency spacing set as the maximum recurrence ratio. The maximum recurrence ratio represents the highest level of sideband frequency spacing recurrence in historical data and is an important reference for evaluating the recurrence situation of the current sideband frequency spacing. Calculate the ratio of the recurrence ratio to the maximum recurrence ratio as the sideband frequency recurrence coefficient. The sideband frequency recurrence coefficient is a dimensionless value, which reflects the relative recurrence degree of the sideband frequency spacing of the current device in historical fault data. The larger the sideband frequency recurrence coefficient, the more similar the sideband frequency spacing of the current device is to the historical fault data, and thus the higher the probability of a fault occurring. On the contrary, the smaller the sideband frequency recurrence coefficient, the greater the difference between the sideband frequency spacing of the current device and the historical fault data, and the lower the probability of a fault occurring.

[0028] In a preferred embodiment, workpiece processing images of the workpiece during the processing of the target device are collected for workpiece anomaly analysis to obtain workpiece anomaly parameters. The execution steps further include: collecting a set of sample workpiece processing images according to the processing data of the same type of workpiece, and collecting the deformation amplitude of the workpiece in different sample workpiece processing images, which is marked as a set of sample workpiece anomaly parameters; constructing a workpiece image recognizer based on a convolutional neural network; using the set of sample workpiece processing images and the set of sample workpiece anomaly parameters to perform supervised training on the workpiece image recognizer until the accuracy of the workpiece image recognizer meets the requirements; inputting the workpiece processing image into the workpiece image recognizer to recognize and output workpiece anomaly parameters.

[0029] Preferably, collect the processing data of similar workpieces, and select a representative set of sample workpiece processing images from them. These image sets cover the performance of the workpieces at different processing stages and in different states. Subsequently, for each sample workpiece processing image, observe and record the amplitude of deformation of the workpiece, and label these data as the set of abnormal parameters of the sample workpiece. This step provides supervision information for the subsequent training process. Subsequently, based on the powerful image processing ability of the convolutional neural network (CNN), construct a workpiece image recognizer that can automatically extract the feature information in the image and identify the abnormal state of the workpiece through these features. The structure design of the CNN makes it highly accurate and robust in processing image data. Next, use the set of sample workpiece processing images and the set of sample workpiece abnormal parameters to supervise and train the workpiece image recognizer. During the training process, the recognizer will continuously try to extract features from the image and match these features with the labeled abnormal parameters. Through multiple iterations and optimizations, the accuracy rate of the recognizer gradually increases until the requirements are met. This step is the key to ensuring that the recognizer can accurately identify the abnormal state of the workpiece. Finally, input the workpiece processing image during the processing of the target device into the trained workpiece image recognizer. The recognizer will automatically process the image data and output the abnormal parameters of the workpiece. These parameters can quickly determine whether there is an abnormality in the processing state of the workpiece, so as to take corresponding measures for correction. In summary, this process not only improves the accuracy of recognition and the monitoring efficiency, but also reduces the failure rate during the production process.

[0030] In a preferred embodiment, according to the sideband recurrence coefficient and the workpiece abnormal parameters, configure a fault recognition strategy to perform fault recognition on the sideband parameters to obtain a first failure rate. The execution steps further include: constructing a sideband fault recognizer, where the sideband fault recognizer includes U sideband fault recognition paths, and U is a positive integer; obtaining the maximum workpiece abnormal parameter during workpiece processing; calculating the ratio of the workpiece abnormal parameter to the maximum workpiece abnormal parameter as the workpiece abnormal coefficient, and calculating to obtain the workpiece equipment fault coefficient; calculating to obtain a fault recognition coefficient according to the sideband recurrence coefficient and the workpiece equipment fault coefficient; multiplying the fault recognition coefficient by U and taking the integer to obtain W, and randomly select and configure W sideband fault recognition paths as the fault recognition strategy; input the sideband parameters into the W sideband fault recognition paths to identify and obtain W sideband failure rates, and calculate the mean value to obtain the first failure rate.

[0031] Further, prepare for the subsequent fault identification strategy configuration based on the previous sideband recurrence coefficient and workpiece anomaly parameters. The workpiece anomaly parameters are obtained by collecting workpiece processing images during the processing of the target device and performing anomaly analysis. At the same time, the maximum workpiece anomaly parameter during workpiece processing is also obtained. Then, calculate the workpiece anomaly coefficient, which is the ratio of the current workpiece anomaly parameter to the maximum workpiece anomaly parameter. This ratio reflects the severity of the current workpiece anomaly state relative to the maximum anomaly state. Then calculate the workpiece equipment fault coefficient according to the workpiece anomaly coefficient, which is obtained by subtracting the workpiece anomaly coefficient from 1. The larger the workpiece equipment fault coefficient, the higher the probability that the equipment has a fault. Construct a sideband fault identifier and design U sideband fault identification paths, where U is a positive integer. These paths represent different fault identification modes and can cover possible fault situations more comprehensively. To determine which paths to use for fault identification, calculate the fault identification coefficient based on the sideband recurrence coefficient and the workpiece equipment fault coefficient. Here, the sideband recurrence coefficient and the workpiece equipment fault coefficient can be combined by weighted calculation or by calculating the mean to obtain a comprehensive fault identification coefficient. After obtaining the fault identification coefficient, multiply this coefficient by U and take the integer to get W. W represents the number of randomly selected and configured sideband fault identification paths. This method can ensure that both the influence of the sideband recurrence coefficient and the workpiece anomaly parameters is considered during the fault identification process, and it also has a certain degree of randomness and flexibility. Then, input the sideband parameters into W sideband fault identification paths, and each path will output a sideband failure rate. Finally, calculate the mean of these sideband failure rates to obtain the first failure rate. This first failure rate reflects the comprehensive probability of the current device having a fault. For example, assume there is a target device, and the sideband recurrence coefficient and workpiece anomaly parameters are collected during the processing. At this time, a sideband fault identifier with 5 sideband fault identification paths (U = 5) is constructed. Through calculation, the workpiece anomaly coefficient is obtained as 0.6 (that is, the current workpiece anomaly parameter is 60% of the maximum anomaly parameter), and then the workpiece equipment fault coefficient is obtained as 0.4 (1 - 0.6 = 0.4). Then, the sideband recurrence coefficient (assumed to be 0.8) and the workpiece equipment fault coefficient (0.4) are combined by weighted calculation to obtain a fault identification coefficient of 0.6. According to the fault identification coefficient, W = 3 is calculated (0.6 × 5 = 3, take the integer). Therefore, 3 sideband fault identification paths are randomly selected for fault identification. Finally, calculate the mean of the sideband failure rates output by these 3 paths, and the first failure rate is obtained as 0.25 (assuming the failure rates output by the three paths are 0.2, 0.3, and 0.25 respectively, and the mean is (0.2 + 0.3 + 0.25) / 3 = 0.25).

[0032] In a preferred embodiment, an edge frequency fault identifier is constructed, and the execution steps further include: according to the fault diagnosis data of the device within the historical time, collecting a set of sample edge frequency parameters, and collecting the probability of the device having a fault when different sample edge frequency parameters appear, which is labeled as a set of sample edge frequency fault rates; randomly extracting U training samples with replacement from within the set of sample edge frequency parameters and the set of sample edge frequency fault rates; using a feedforward neural network to construct U edge frequency fault recognition paths, and respectively using the U training samples to perform supervised training and testing on the U edge frequency fault recognition paths until the accuracy meets the requirements; combining the U edge frequency fault recognition paths to obtain an edge frequency fault identifier.

[0033] Optionally, a set of sample edge frequency parameters is collected from the fault diagnosis data of the device within the historical time. These parameters are obtained through sensors or other monitoring means during the operation of the device, and they can reflect the operating state and possible fault conditions of the device. At the same time, it is also necessary to collect the fault probabilities corresponding to these sample edge frequency parameters, that is, the set of sample edge frequency fault rates. These fault probabilities are calculated through the actual fault records in the historical data, and they provide label information for subsequent model training. Subsequently, U training samples are randomly extracted from them. Sampling with replacement is used here, which means that the same sample may be selected multiple times, which helps to improve the generalization ability of the model. U represents the number of edge frequency fault recognition paths constructed, and it can be adjusted according to actual needs. Next, a feedforward neural network is used to construct U edge frequency fault recognition paths. A feedforward neural network is a commonly used machine learning model that can handle complex non-linear relationships and has strong generalization ability. For each edge frequency fault recognition path, a training sample is used for supervised training and testing. During the training process, the model will continuously adjust its parameters to minimize the difference between the predicted fault rate and the actual fault rate. The testing process is used to verify the performance of the model to ensure that it can also maintain a high accuracy on new data. When all the edge frequency fault recognition paths have been fully trained and tested, they are combined to form a complete edge frequency fault identifier. This identifier can process multiple edge frequency parameters simultaneously and output a comprehensive fault rate prediction result. Since multiple paths are combined, even if the prediction result of a certain path deviates, other paths can compensate, thereby improving the overall prediction accuracy. By constructing an edge frequency fault identifier, the historical fault diagnosis data of the device can be fully utilized to train a model that can accurately predict the fault probability of the device. This model can not only improve the accuracy and reliability of fault recognition, but also, due to the use of multiple edge frequency fault recognition paths for combination, the robustness and generalization ability of the model are also significantly improved.

[0034] In a preferred embodiment, according to the sideband frequency parameter, a second failure rate is calculated, and the execution steps further include: obtaining the average proportion of different historical sideband frequency spacings appearing in the historical sideband frequency spacing set as the average recurrence proportion; when the recurrence proportion is less than the average recurrence proportion, calculating the ratio of the difference between the recurrence proportion and the average recurrence proportion to the average recurrence proportion as the second failure probability; when the recurrence proportion is greater than or equal to the average recurrence proportion, outputting the second failure probability as 0.

[0035] Specifically, the sideband frequency spacing set is obtained from the historical data of the device. These sideband frequency spacings are important parameters reflecting the vibration characteristics of the device obtained through monitoring means under different operating states of the device. Then, calculate the average proportion of different historical sideband frequency spacings appearing in this historical sideband frequency spacing set. This proportion is the average recurrence proportion, which reflects the universality and stability of each sideband frequency spacing appearing under normal conditions. Now, assume there is a recurrence proportion of a new sideband frequency spacing (i.e., the proportion of this sideband frequency spacing appearing in the new data). Compare this recurrence proportion with the previously calculated average recurrence proportion. If the recurrence proportion is less than the average recurrence proportion, it means that the number of times this sideband frequency spacing appears in the new data is less than normal. The smaller the recurrence proportion, the fewer the number of times the sideband frequency spacing appears, the higher the randomness, and the more likely it is that the change in the sideband frequency spacing is caused by workpiece deformation or other abnormal conditions. Therefore, calculate the ratio of (average recurrence proportion - recurrence proportion) / average recurrence proportion, and use this ratio as the second failure probability. The larger this ratio, the greater the difference between the recurrence proportion of this sideband frequency spacing and the average recurrence proportion, and the higher the probability of failure. If the recurrence proportion is greater than or equal to the average recurrence proportion, it is considered that the number of times this sideband frequency spacing appears in the new data is normal and does not show abnormal failure characteristics. Therefore, directly output the second failure probability as 0.

[0036] In a preferred embodiment, in combination with the first failure rate, a device failure rate is calculated as the device failure analysis and prediction result, and the execution steps further include: performing weighted calculation on the first failure rate and the second failure rate to obtain the device failure rate; using the device failure rate as the device failure analysis and prediction result.

[0037] Specifically, the first failure rate is calculated by a sideband fault identifier based on sideband parameters and a fault identification strategy; while the second failure rate is calculated based on the difference between the recurrence ratio of sideband spacing and the average recurrence ratio. These two failure rates reflect the fault possibility of the device from different perspectives. To obtain a comprehensive device failure rate, the first failure rate and the second failure rate need to be weighted and calculated. The purpose of weighting is to allocate different weights according to the importance and reliability of the two failure rates. The allocation of weights can be determined according to the actual situation and expert experience, or can be automatically optimized by methods such as machine learning. The formula for weighted calculation can be expressed as: device failure rate = α × first failure rate + β × second failure rate, where α and β are the weights of the first failure rate and the second failure rate respectively, and α + β = 1, for example, 0.7 and 0.3 respectively. After weighted calculation, a comprehensive device failure rate is obtained. This failure rate is a quantitative indicator, which reflects the fault possibility of the device in the current operating state. This failure rate can be used as the result of device fault analysis and prediction for device maintenance, management and decision-making. By combining the first failure rate and the second failure rate to calculate the device failure rate, the fault possibility of the device can be evaluated more comprehensively. The first failure rate reflects the vibration characteristics and fault characteristics of the device from the perspective of sideband parameters, while the second failure rate provides additional fault information from the perspective of the recurrence ratio of sideband spacing. The combined use of these two failure rates can complement and verify each other, thereby improving the accuracy and reliability of device fault analysis. At the same time, the weighted calculation method also allows adjusting the weights of the two failure rates according to the actual situation and requirements, so as to more flexibly respond to different application scenarios and device types. This makes the result of device fault analysis and prediction more in line with the actual needs and can provide more powerful support for device maintenance and management.

[0038] The device fault analysis and prediction system based on frequency domain characteristics provided by the embodiments of the present invention has at least the following technical effects:

[0039] 1. It combines the frequency domain vibration signal analysis during the operation of the target device and the abnormal analysis of the workpiece processing image. By collecting vibration signals and performing frequency domain processing to obtain sideband parameters and sideband recurrence coefficients, and at the same time collecting workpiece processing images and using a convolutional neural network for abnormal analysis to obtain workpiece abnormal parameters. The fusion of this multi-dimensional data enables the system to more comprehensively capture the early signs of device faults and improve the accuracy and reliability of fault analysis.

[0040] 2. The sideband frequency recurrence coefficient and the workpiece anomaly coefficient are introduced as key indicators for fault identification. The sideband frequency recurrence coefficient evaluates the fault possibility by comparing the recurrence of the current sideband frequency spacing in historical data, while the workpiece anomaly coefficient quantifies the degree of workpiece deformation based on the recognition result of the workpiece image. According to these two coefficients, the fault identification strategy is intelligently configured, and an appropriate sideband frequency fault identification path is selected for fault identification, thus realizing the personalization and intelligence of fault identification.

[0041] 3. Not only the first failure rate is obtained through the sideband frequency fault identifier, but also the second failure rate is calculated according to the recurrence ratio of the sideband frequency spacing. These two failure rates respectively reflect the fault state of the equipment from different perspectives. By weighted calculation of these two failure rates, a comprehensive equipment failure rate is obtained as the result of equipment fault analysis and prediction. This method comprehensively considers various fault factors, improves the accuracy and practicality of equipment fault prediction, and provides strong support for the maintenance and management of the equipment.

[0042] Embodiment 2:

[0043] As Figure 2 shown, based on the same inventive concept as the equipment fault analysis and prediction system based on frequency domain characteristics provided in Embodiment 1, the embodiment of the present invention further provides a method for equipment fault analysis and prediction based on frequency domain characteristics, and the method includes:

[0044] Collect the vibration signal during the operation of the target equipment, perform frequency domain processing to obtain sideband frequency parameters, and perform sideband frequency recurrence analysis in the historical equipment frequency domain data to obtain the sideband frequency recurrence coefficient.

[0045] Collect the workpiece processing image of the workpiece during the processing of the target equipment, perform workpiece anomaly analysis to obtain workpiece anomaly parameters.

[0046] According to the sideband frequency recurrence coefficient and the workpiece anomaly parameters, configure a fault identification strategy, perform fault identification on the sideband frequency parameters, and obtain the first failure rate.

[0047] According to the sideband frequency parameters, calculate to obtain the second failure rate, and combine the first failure rate to calculate the equipment failure rate as the result of equipment fault analysis and prediction.

[0048] Further, collecting the vibration signal during the operation of the target equipment, performing frequency domain processing to obtain sideband frequency parameters, and performing sideband frequency recurrence analysis in the historical equipment frequency domain data to obtain the sideband frequency recurrence coefficient includes: collecting the vibration signal during the operation of the target equipment; performing frequency domain transformation on the vibration signal to obtain a frequency domain signal; in the frequency domain signal, perform sideband frequency extraction to obtain sideband frequency parameters, where the sideband frequency parameters include sideband frequency spacing; according to the sideband frequency spacing, perform sideband frequency recurrence analysis in the historical equipment frequency domain data to obtain the sideband frequency recurrence coefficient.

[0049] Further, according to the sideband frequency spacing, perform sideband frequency recurrence analysis on the historical device frequency domain data to obtain a sideband frequency recurrence coefficient, including: in the historical device frequency domain data, extract the set of historical sideband frequency spacings when the device fails; obtain the number of occurrences of the sideband frequency spacing in the set of historical sideband frequency spacings to obtain a recurrence ratio; obtain the maximum ratio of the occurrences of different historical sideband frequency spacings in the set of historical sideband frequency spacings as the maximum recurrence ratio; calculate the ratio of the recurrence ratio to the maximum recurrence ratio as the sideband frequency recurrence coefficient.

[0050] Further, collect workpiece processing images of the workpiece during the processing of the target device, perform workpiece anomaly analysis to obtain workpiece anomaly parameters, including: according to the processing data of similar workpieces, collect a set of sample workpiece processing images, and collect the deformation amplitude of the workpiece in different sample workpiece processing images, which is marked as a set of sample workpiece anomaly parameters; based on a convolutional neural network, construct a workpiece image recognizer; use the set of sample workpiece processing images and the set of sample workpiece anomaly parameters to perform supervised training on the workpiece image recognizer until the accuracy of the workpiece image recognizer meets the requirements; input the workpiece processing image into the workpiece image recognizer to recognize and output workpiece anomaly parameters.

[0051] Further, configure a fault recognition strategy according to the sideband frequency recurrence coefficient and the workpiece anomaly parameters, perform fault recognition on the sideband frequency parameters to obtain a first failure rate, including: construct a sideband frequency fault recognizer, where the sideband frequency fault recognizer includes U sideband frequency fault recognition paths, and U is a positive integer; obtain the maximum workpiece anomaly parameter during workpiece processing; calculate the ratio of the workpiece anomaly parameter to the maximum workpiece anomaly parameter as the workpiece anomaly coefficient, and calculate and obtain the workpiece equipment fault coefficient; calculate and obtain a fault recognition coefficient according to the sideband frequency recurrence coefficient and the workpiece equipment fault coefficient; multiply the fault recognition coefficient by U and round down to obtain W, randomly select and configure W sideband frequency fault recognition paths as the fault recognition strategy; input the sideband frequency parameters into the W sideband frequency fault recognition paths to recognize and obtain W sideband frequency failure rates, and calculate the mean value to obtain the first failure rate.

[0052] Further, construct a sideband frequency fault recognizer, including: according to the fault diagnosis data of the device in the historical time, collect a set of sample sideband frequency parameters, and collect the probability of the device failing when different sample sideband frequency parameters appear, which is marked as a set of sample sideband frequency failure rates; randomly extract U training samples with replacement from the set of sample sideband frequency parameters and the set of sample sideband frequency failure rates; use a feedforward neural network to construct U sideband frequency fault recognition paths, and respectively use the U training samples to perform supervised training and testing on the U sideband frequency fault recognition paths until the accuracy meets the requirements; combine the U sideband frequency fault recognition paths to obtain a sideband frequency fault recognizer.

[0053] Further, according to the side frequency parameter, a second failure rate is calculated, including: obtaining the average proportion of different historical side frequency spacings appearing in the historical side frequency spacing set as the average recurrence proportion; when the recurrence proportion is less than the average recurrence proportion, calculating the ratio of the difference between the recurrence proportion and the average recurrence proportion to the average recurrence proportion as the second failure probability; when the recurrence proportion is greater than or equal to the average recurrence proportion, outputting the second failure probability as 0.

[0054] Further, in combination with the first failure rate, a device failure rate is calculated as the device failure analysis and prediction result, including: performing weighted calculation on the first failure rate and the second failure rate to obtain the device failure rate; using the device failure rate as the device failure analysis and prediction result.

[0055] It should be noted that the above-mentioned sequence of embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above describes specific embodiments of this specification. The processes depicted in the drawings do not necessarily require the specific order and continuous order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0056] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

[0057] This specification and the drawings are only exemplary descriptions of the present application and are considered to have covered any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. A device fault analysis and prediction system based on frequency domain characteristics, characterized in that The system includes: A signal acquisition module, which is used to acquire vibration signals during the operation of the target device, perform frequency-domain processing to obtain sideband parameters, perform sideband recurrence analysis in the historical device frequency-domain data, and obtain the sideband recurrence coefficient; An image acquisition module, which is used to acquire workpiece processing images of workpieces during the processing of the target device, perform workpiece anomaly analysis, and obtain workpiece anomaly parameters; A fault identification module, which is used to configure a fault identification strategy according to the sideband recurrence coefficient and workpiece anomaly parameters, perform fault identification on the sideband parameters, and obtain the first failure rate; A result prediction module, which is used to calculate a second failure rate according to the sideband parameters, and combine the first failure rate to calculate the device failure rate as the device fault analysis and prediction result; Among them, configuring a fault identification strategy according to the sideband recurrence coefficient and workpiece anomaly parameters, performing fault identification on the sideband parameters, and obtaining the first failure rate includes: Constructing a sideband fault identifier, where the sideband fault identifier includes U sideband fault identification paths, and U is a positive integer; Obtaining the maximum value of the workpiece anomaly parameters during workpiece processing as the maximum workpiece anomaly parameter; Calculating the ratio of the workpiece anomaly parameter to the maximum workpiece anomaly parameter as the workpiece anomaly coefficient, and calculating the workpiece device fault coefficient; Calculating a fault identification coefficient according to the sideband recurrence coefficient and the workpiece device fault coefficient; Multiplying the fault identification coefficient by U and taking the integer to obtain W, and randomly selecting and configuring W sideband fault identification paths as the fault identification strategy; Inputting the sideband parameters into the W sideband fault identification paths, identifying W sideband failure rates, and calculating the mean value to obtain the first failure rate; Among them, combining the first failure rate to calculate the device failure rate as the device fault analysis and prediction result includes: Performing weighted calculation on the first failure rate and the second failure rate to obtain the device failure rate; Taking the device failure rate as the device fault analysis and prediction result.

2. The device fault analysis and prediction system based on frequency domain characteristics according to claim 1, wherein Acquiring vibration signals during the operation of the target device, performing frequency-domain processing to obtain sideband parameters, and performing sideband recurrence analysis in the historical device frequency-domain data to obtain the sideband recurrence coefficient, including: Acquiring vibration signals during the operation of the target device; Performing frequency-domain transformation on the vibration signals to obtain frequency-domain signals; Performing sideband extraction in the frequency-domain signals to obtain sideband parameters, where the sideband parameters include sideband spacing; Performing sideband recurrence analysis in the historical device frequency-domain data according to the sideband spacing to obtain the sideband recurrence coefficient.

3. The device fault analysis and prediction system based on frequency domain characteristics according to claim 2, characterized in that, Performing sideband recurrence analysis in the historical device frequency-domain data according to the sideband spacing to obtain the sideband recurrence coefficient, including: Extracting the historical sideband spacing set when the device fails in the historical device frequency-domain data; Obtaining the number of occurrences of the sideband spacing in the historical sideband spacing set to obtain the recurrence ratio; Obtaining the maximum ratio of different historical sideband spacings in the historical sideband spacing set as the maximum recurrence ratio; Calculating the ratio of the recurrence ratio to the maximum recurrence ratio as the sideband recurrence coefficient.

4. The device fault analysis and prediction system based on frequency domain characteristics according to claim 1, characterized in that Collect workpiece processing images of workpieces during the processing of the target device, perform workpiece anomaly analysis, and obtain workpiece anomaly parameters, including: According to the processing data of similar workpieces, collect a set of sample workpiece processing images, and collect the deformation amplitudes of workpieces in different sample workpiece processing images, which are marked as a set of sample workpiece anomaly parameters; Based on a convolutional neural network, construct a workpiece image recognizer; Use the set of sample workpiece processing images and the set of sample workpiece anomaly parameters to perform supervised training on the workpiece image recognizer until the accuracy of the workpiece image recognizer meets the requirements; Input the workpiece processing image into the workpiece image recognizer, and recognize and output workpiece anomaly parameters.

5. The device fault analysis and prediction system based on frequency domain characteristics according to claim 1, characterized in that Construct a sideband frequency fault recognizer, including: According to the fault diagnosis data of the device in historical time, collect a set of sample sideband frequency parameters, and collect the probability of the device failing when different sample sideband frequency parameters appear, which is marked as a set of sample sideband frequency failure rates; Randomly draw U training samples with replacement from the set of sample sideband frequency parameters and the set of sample sideband frequency failure rates; Use a feedforward neural network to construct U sideband frequency fault recognition paths, and use the U training samples to perform supervised training and testing on the U sideband frequency fault recognition paths until the accuracy meets the requirements; Combine the U sideband frequency fault recognition paths to obtain a sideband frequency fault recognizer.

6. The device fault analysis and prediction system based on frequency domain characteristics according to claim 3, characterized in that Calculate a second failure rate according to the sideband frequency parameters, including: Obtain the average ratio of different historical sideband frequency spacings appearing in the historical sideband frequency spacing set as the average recurrence ratio; When the recurrence ratio is less than the average recurrence ratio, calculate the ratio of the difference between the recurrence ratio and the average recurrence ratio to the average recurrence ratio as the second failure probability; When the recurrence ratio is greater than or equal to the average recurrence ratio, output the second failure probability as 0.

7. A device fault analysis and prediction method based on frequency domain characteristics, characterized in that, The method is applied to a device fault analysis and prediction system based on frequency domain characteristics according to any one of claims 1-6. The method includes: Collect vibration signals during the operation of the target device, perform frequency domain processing to obtain sideband frequency parameters, and perform sideband frequency recurrence analysis on historical device frequency domain data to obtain a sideband frequency recurrence coefficient; Collect workpiece processing images of workpieces during the processing of the target device, perform workpiece anomaly analysis, and obtain workpiece anomaly parameters; Configure a fault recognition strategy according to the sideband frequency recurrence coefficient and workpiece anomaly parameters, perform fault recognition on the sideband frequency parameters, and obtain a first failure rate; Calculate a second failure rate according to the sideband frequency parameters, and combine the first failure rate to calculate the device failure rate as the device fault analysis and prediction result.

Images