A fault early warning method and system of a metal rod automatic processing equipment

Abstract

The application discloses a kind of metal rod automatic processing equipment's fault early warning method and system, it is related to fault diagnosis technical field, the method includes: obtaining the running data of internal component of processing equipment preset, constructs comprehensive running data set;Determine abnormal frequency interval, according to the frequency characteristic value of abnormal frequency interval, trace to source potential wear and tear anomaly, determine source component;Obtain the optical imaging data of source component, extract component surface feature, the wear degree of component surface feature is quantified, and the wear degree quantization description is obtained;Wear degree quantization description is mapped to space position, and position calibration is carried out to preliminary distribution area, and accurate coordinate information is determined;Feature extraction is carried out to accurate coordinate information, and the continuous monitoring sequence is obtained, and internal running state is tracked, and dynamic change trajectory is obtained;Trend analysis is carried out to dynamic change trajectory, and fault early warning information is generated.The application ensures the accuracy and timeliness of fault early warning.

Description

Technical Field

[0001] This application belongs to the field of fault diagnosis technology, specifically a fault early warning method and system for an automatic metal rod processing equipment. Background Technology

[0002] In modern industrial manufacturing systems, automated metal bar processing equipment, such as CNC lathes, centerless grinders, and automatic sawing machines, is the core production unit responsible for processing raw materials into precision parts. This type of equipment typically integrates a high-speed rotating spindle, a precision feed transmission mechanism, a tooling system, and complex hydraulic and cooling auxiliary devices. Its continuous, stable, and high-precision operation is crucial for ensuring downstream production cycle time, product quality consistency, and controlling overall manufacturing costs. Unplanned downtime, especially failures caused by key components such as burnt-out spindle bearings, chipped tools, or worn ball screws, will directly lead to the shutdown of the entire production line, resulting in significant capacity losses, raw material waste, and potentially serious safety accidents.

[0003] However, current methods for fault detection and early warning of such processing equipment still have significant shortcomings. Existing methods often rely on superficial monitoring of the equipment, failing to deeply reveal its internal operating status. Especially under complex operating conditions, the hidden and variable nature of faults makes it difficult to accurately determine the root cause of the problem based solely on experience or external observation. This limitation leads to a lag in the discovery of processing equipment faults, often resulting in measures being taken only after the problem has worsened, increasing production risks.

[0004] Specifically, existing methods typically employ the following approaches: Regular maintenance and experience-based judgment: Processing equipment is shut down for maintenance at fixed time intervals (e.g., weekly, monthly), with maintenance engineers conducting inspections using auditory, tactile, or simple handheld vibration meters. This method is not only inefficient and labor-intensive, but also heavily reliant on the engineer's personal experience. It struggles to detect latent faults in their early stages, often only discovering them when they have progressed to the middle or late stages, producing noticeable abnormal noises or exceeding accuracy tolerances, thus missing the optimal opportunity for preventative maintenance.

[0005] Alarm based on a single physical quantity threshold: This method involves installing a single type of sensor, such as a vibration sensor, on critical parts of the equipment (e.g., the spindle housing) and setting a fixed alarm threshold for the total vibration. An alarm is triggered when the monitored value exceeds the threshold. While this method achieves online monitoring, it has poor anti-interference capabilities and is prone to false alarms due to non-fault factors such as changes in processing load or external impacts. More importantly, a single vibration quantity cannot distinguish the fault type or locate the fault source. For example, it cannot accurately determine whether the abnormal vibration originates from the inner ring, outer ring, or cage of the bearing, nor can it detect changes in surface microstructure caused by initial wear.

[0006] Status monitoring based on a Programmable Logic Controller (PLC): This method utilizes the equipment's own PLC system to monitor logical states such as motor current and limit switch signals. It is primarily used to detect hard faults such as overload and overtravel, but its ability to monitor soft faults such as progressive mechanical wear and fatigue is very limited.

[0007] It should also be noted that due to the complex internal structure of processing equipment, the operation of key components such as transmission mechanisms or cutting tools is difficult to observe directly. Surface wear and crack propagation of sealed transmission components and cutting tools cannot be assessed intuitively and quantitatively without disassembling the machine. Maintenance personnel often cannot promptly understand their wear or abnormal conditions. This invisibility of the internal operating status of processing equipment further exacerbates the difficulty of fault location, making it difficult to accurately pinpoint the problem area in the early stages of a fault. For example, in the processing of metal bars, even minor wear on the cutting tool can cause deviations in machining accuracy. However, because the real-time status of the cutting tool cannot be directly observed, maintenance personnel can only detect the problem after obvious defects appear in the product, missing the optimal intervention opportunity.

[0008] Therefore, how to monitor the operating status of key internal components in real time and accurately locate the specific location of the fault without disassembling the processing equipment has become an important issue that the industry urgently needs to address. Summary of the Invention

[0009] To address the above issues, this application provides a fault early warning method for automatic metal rod processing equipment, which solves the problem that existing technologies cannot monitor the operating status of key internal components and accurately locate the specific location of faults in real time without disassembling the processing equipment.

[0010] To achieve the above objectives, the technical solution adopted in this application is as follows: In a first aspect, embodiments of this application provide a fault early warning method for an automatic metal rod processing equipment, including: The system acquires the operating data of the internal components of the processing equipment, preprocesses the operating data, and merges the preprocessed operating data to construct a comprehensive operating dataset. The operating data includes vibration signals, temperature distribution data, and sound wave data. Analyze the amplitude variation pattern of the vibration signal in the comprehensive operation data, determine all abnormal frequency ranges in the vibration signal, trace the source of potential wear anomalies based on the frequency characteristic values ​​of the abnormal frequency ranges, and identify the source components of potential wear anomalies; Acquire optical imaging data of the source component, extract the surface features of the component from the optical imaging data, quantify the wear degree of the surface features of the component, and obtain a quantitative description of the wear degree. The wear level is quantitatively described and mapped to the spatial location of the processing equipment to obtain a preliminary distribution area. The position of the preliminary distribution area is then calibrated to determine the precise coordinate information of potential wear anomalies. Feature extraction is performed on vibration signals and optical imaging data at precise coordinates. The extracted features are then fused and denoised to obtain a continuous monitoring sequence. The internal operating status of the processing equipment is tracked based on the continuous monitoring sequence to obtain a dynamic trajectory. Trend analysis is performed on the dynamic trajectory to generate fault early warning information.

[0011] In conjunction with the first aspect, in the first embodiment of the first aspect, the analysis and comprehensive analysis of the amplitude variation pattern of the vibration signal in the integrated operational data, determining all abnormal frequency ranges in the vibration signal, and tracing the source of potential wear anomalies based on the frequency characteristic values ​​of the abnormal frequency ranges to determine the source component of the potential wear anomaly, specifically includes: The preprocessed vibration signal is subjected to spectral decomposition to extract spectral features; Time-domain analysis of the amplitude variation patterns contained in the spectral features yields the amplitude fluctuation range; The amplitude fluctuation range is compared with the historical normal fluctuation range extracted based on historical operation data to determine all abnormal frequency intervals in the amplitude fluctuation range; All abnormal frequency intervals are further decomposed to obtain the frequency concentration intervals for each abnormal frequency interval, and the frequency feature values ​​of each frequency concentration interval are extracted. If the frequency characteristic value exceeds the preset threshold, the abnormal frequency range corresponding to the frequency characteristic value is marked as a potential wear anomaly. Based on the structural model of the processing equipment and the location distribution of the sensors used to acquire operational data, the propagation path of the abnormal frequency range marked as potential wear anomalies is traced to determine the source component of the potential wear anomaly.

[0012] In conjunction with the first embodiment of the first aspect, in the second embodiment of the first aspect, the step of acquiring optical imaging data of the source component, extracting surface features of the component from the optical imaging data, quantifying the wear degree of the surface features of the component, and obtaining a quantitative description of the wear degree of the source component specifically includes: Acquire optical imaging data from the source component; The optical imaging data is segmented to extract the surface feature information of the source component, resulting in a surface texture distribution map; Determine the vibration energy distribution and signal time domain window of the vibration signal, and based on the vibration energy distribution and signal time domain window, filter out the time points corresponding to the abnormal frequency intervals marked as potential wear anomalies; By temporally matching the time points with the optical imaging data, the optical imaging data of each frame under potential wear anomalies can be determined. Component surface features are extracted from optical imaging data of each frame under potential wear anomalies. Based on the component surface features and the intensity change of vibration energy distribution, the wear degree of the source component is quantified, and the wear quantification value of the source component is determined. The wear quantification value is compared with the preset wear level standard to determine the quantitative description of the wear degree of the source component.

[0013] In conjunction with the second embodiment of the first aspect, in the third embodiment of the first aspect, the step of quantifying and mapping the degree of wear to the spatial location of the processing equipment to obtain a preliminary distribution area, and performing position calibration on the preliminary distribution area to determine the precise coordinate information of potential wear anomalies, specifically includes: Based on the structural model of the processing equipment, establish the spatial coordinate system in which the processing equipment is located; The wear level is quantitatively described and mapped to the spatial coordinate system where the processing equipment is located, so as to obtain the preliminary distribution area of ​​potential wear anomalies in the spatial coordinate system; Based on the preset dominant frequency, extract the dominant frequency vibration characteristics in the abnormal frequency range marked as potential wear anomalies; The optical imaging data of each frame under potential wear anomalies are matched with the main frequency vibration characteristics to determine the offset between the vibration peak position and the wear center. The offset is superimposed onto the initial distribution area to obtain the calibration distribution area; The calibration distribution area is input into the constructed abnormal location and wear degree mapping model to obtain the accurate coordinate information output by the abnormal location and wear degree mapping model.

[0014] In conjunction with the first aspect, in the fourth embodiment of the first aspect, the step of extracting features from the vibration signal and optical imaging data at precise coordinate information, performing feature fusion and denoising on the extracted features to obtain a continuous monitoring sequence, and tracking the internal operating status of the processing equipment based on the continuous monitoring sequence to obtain a dynamic trajectory, specifically includes: Based on precise coordinate information, vibration signals and optical imaging data at the precise coordinate locations are acquired. Feature extraction was performed on the vibration signal and optical imaging data at the precise coordinate information, respectively, to obtain the vibration features of the vibration signal and the surface texture features of the optical imaging data; Based on the pre-assigned data weights, vibration features and surface texture features are fused to obtain abnormal fusion features; The abnormal fusion features are subjected to sliding smoothing to obtain a continuous monitoring sequence; The changing characteristics of the continuous monitoring sequence in the time dimension are determined, and the internal operating status of the processing equipment is tracked based on the changing characteristics to obtain the dynamic change trajectory.

[0015] In conjunction with the first aspect, in the fifth embodiment of the first aspect, the step of performing trend analysis on the dynamic trajectory to generate fault early warning information specifically includes: Interference signals in the dynamically changing trajectory are filtered out to obtain a denoised dynamically changing trajectory; Determine the rate of state change of the dynamic trajectory of the denoised system to obtain the rate change characteristics; The main frequency components and noise power ratio of the dynamic change trajectory of the denoising are determined to obtain the signal-to-noise ratio. When the rate change characteristics exceed a preset range and the signal-to-noise ratio does not exceed a preset ratio, a fault alarm message is generated.

[0016] In conjunction with the first aspect, in the sixth embodiment of the first aspect, the step of acquiring the operating data of preset internal components of the processing equipment, preprocessing the operating data, and fusing the preprocessed operating data to construct a comprehensive operating dataset specifically includes: Acquire the pre-set operating data of the internal components of the processing equipment; The vibration signal is resampled to unify the sampling frequency of all vibration signals. Then, the resampled vibration signal is subjected to signal noise reduction and signal drift suppression to obtain the preprocessed vibration signal. Temperature anomalies are extracted from the temperature distribution data. Abnormal temperature rise regions in the temperature distribution data are determined based on the temperature anomalies. The correlation between abnormal temperature rise regions and preprocessed vibration signals is established to obtain preprocessed temperature distribution data. Extract the spectral features of the acoustic wave data and identify abnormal frequency peaks in the spectral features to obtain preprocessed acoustic wave data; The preprocessed runtime data is fused to construct a comprehensive runtime dataset.

[0017] In conjunction with the sixth embodiment of the first aspect, in the seventh embodiment of the first aspect, the step of data fusion of the preprocessed operational data to construct a comprehensive operational dataset specifically includes: Feature extraction was performed on the preprocessed operational data to obtain vibration features, temperature distribution features, and sound wave features, respectively. Using the resampled vibration signal, feature alignment and splicing of vibration features, temperature distribution features, and acoustic wave features are performed to obtain a fused feature vector; The fused feature vector is processed to obtain a state feature vector that characterizes the internal operating state of the processing equipment. By integrating the preprocessed runtime data and state feature vectors, a comprehensive runtime dataset is constructed.

[0018] In conjunction with the first aspect, in the eighth embodiment of the first aspect, the method further includes the following steps: Using optical imaging data and the clean signal obtained after periodically filtering out vibration signals, a risk assessment is performed on the fault warning information to obtain the risk assessment results, and an intervention command execution sequence is generated based on the risk assessment results.

[0019] Secondly, embodiments of this application provide a fault early warning system for an automatic metal rod processing equipment, comprising: The data construction module is used to acquire the operating data of the internal components of the processing equipment, preprocess the operating data, and merge the preprocessed operating data to construct a comprehensive operating dataset; the operating data includes vibration signals, temperature distribution data, and sound wave data; The source tracing module is used to analyze the amplitude change pattern of the vibration signal in the comprehensive operation data, determine all abnormal frequency ranges in the vibration signal, trace the source of potential wear anomalies based on the frequency characteristic values ​​of the abnormal frequency ranges, and determine the source component of the potential wear anomaly. The wear quantization module is used to acquire optical imaging data of the source component, extract the surface features of the component from the optical imaging data, quantify the wear degree of the surface features of the component, and obtain a quantitative description of the wear degree. The coordinate correction module is used to map the quantitative description of wear degree to the spatial coordinate system where the processing equipment is located, obtain the preliminary distribution area, and perform position calibration on the preliminary distribution area to determine the precise coordinate information of potential wear anomalies; The trajectory prediction module is used to extract features from vibration signals and optical imaging data at precise coordinates, perform feature fusion and noise reduction on the extracted features to obtain a continuous monitoring sequence, and track the internal operating status of the processing equipment based on the continuous monitoring sequence to obtain a dynamically changing trajectory. The fault early warning module is used to perform trend analysis on dynamic change trajectories and generate fault early warning information.

[0020] The fault early warning method and system for automatic metal rod processing equipment of this application obtains the operating data of the internal components of the processing equipment, preprocesses the operating data, and integrates the preprocessed operating data to construct a comprehensive operating dataset. The operating data includes vibration signals, temperature distribution data, and sound wave data, and then cross-validates from different physical dimensions. This avoids the defects of the internal operating state determined by data obtained from a single sensor being one-sided and susceptible to interference. As a result, the comprehensive operating dataset constructed at the end can be a compact, robust, and information-rich digital representation of the internal operating state of the processing equipment. By continuously monitoring and tracking the internal operating status of the processing equipment, dynamic change trajectories are obtained. Then, trend analysis is performed on the dynamic change trajectories to generate fault early warning information. This can issue early warnings in the early stages of fault development, which is earlier than traditional threshold alarm methods, thus gaining valuable response time for maintenance personnel. By tracing the source of potential wear anomalies through frequency characteristic values ​​within abnormal frequency ranges, the originating components of potential wear anomalies are identified. The wear degree is then quantified and mapped onto the spatial coordinate system of the processing equipment to obtain a preliminary distribution area. The position of the preliminary distribution area is then calibrated to determine the precise coordinate information of potential wear anomalies. This allows users to perform visual internal operating status checks without disassembling the equipment, greatly improving fault diagnosis efficiency. Users can not only monitor the operating status of key internal components in real time and accurately locate the specific location of the fault, but also ensure the accuracy and timeliness of fault warnings, significantly improving equipment operating reliability and maintenance efficiency. Attached Figure Description

[0021] The features and advantages of this application will be more clearly understood by referring to the accompanying drawings, which are illustrative and should not be construed as limiting the application in any way. In the drawings: Figure 1 One of the flowcharts of the fault early warning method for the automatic metal rod processing equipment provided in this application is shown; Figure 2 The second flowchart illustrates the fault early warning method for the automatic metal rod processing equipment provided in this application. Figure 3 A schematic diagram of the fault early warning system of the automatic metal rod processing equipment provided in this application is shown. Detailed Implementation

[0022] To enable those skilled in the art to better understand the technical solution, the present application will be described in detail below with reference to the embodiments. The description in this section is only exemplary and explanatory, and should not be used to limit the scope of protection of the present application in any way.

[0023] To address the aforementioned issues, this specification provides a fault early warning method for automated metal rod processing equipment. This method aims to monitor the real-time operating status of key internal components and accurately pinpoint the location of faults without disassembling the equipment, ensuring the accuracy and timeliness of fault warnings and significantly improving equipment reliability and maintenance efficiency. The fault early warning method for automated metal rod processing equipment provided in this specification can be applied to electronic devices, including laptops, desktop computers, smartphones, smart wearable devices, and tablets. Furthermore, the fault early warning method for automated metal rod processing equipment provided in this specification can also be applied to applications running on the aforementioned electronic devices. Figure 1 This is a flowchart illustrating a fault warning method for an automated metal rod processing equipment according to an embodiment of this application, as shown below. Figure 1 As shown, the method may include the following steps: S101. Obtain the operating data of the preset internal components of the processing equipment, preprocess the operating data, and merge the preprocessed operating data to construct a comprehensive operating dataset. The operating data includes vibration signals, temperature distribution data, and sound wave data. The preset internal components can be grinding wheel spindle, guide wheel spindle, grinding area, etc.

[0024] The aforementioned operational data is collected through a multimodal sensor array deployed around the processing equipment. This array includes several vibration sensors, several temperature sensors, several acoustic sensors, and several optical imaging devices. The vibration sensors can be mounted on the grinding wheel spindle bearing housing, guide wheel spindle bearing housing, etc., using magnetic bases or threaded installation. High-sensitivity, wide-frequency response integrated circuit piezoelectric sensors are preferred. Temperature sensors can be installed inside the protective cover of the processing equipment, aligned with the grinding wheel spindle, guide wheel spindle, grinding area, etc. Non-contact temperature sensors are preferred. Sensors: Acoustic sensors are installed near the operation panel of the processing equipment and close to the grinding area. The acoustic sensors can be pre-polarized electret microphone arrays. By arranging the array, the location of the acoustic emission source can be initially determined. Optical imaging devices are used to collect image information from various areas of the processing equipment. The optical imaging devices are high-resolution industrial cameras with microscope lenses. The industrial camera can be mounted on a two-dimensional precision guide rail driven by a micro servo motor, so that it can be moved in a controlled manner. The lens is aimed at potential fault areas (such as grinding wheel surface, guide wheel surface, dressing tool) that have been preliminarily located by vibration analysis. An LED light source can also be provided next to the industrial camera to ensure image quality.

[0025] Specifically, considering that characteristic frequencies such as bearing failures may reach several kilohertz, according to the sampling theorem, the sampling rate of the vibration signal can be set to 25600 Hz to capture high-frequency harmonics and sideband information. The corresponding sampling mode for the vibration signal is continuous acquisition. Temperature distribution data is acquired by transmitting a two-dimensional temperature matrix at a rate of 1 frame per second, with a temperature sensor data sampling rate of 10 Hz. The sampling rate for acoustic wave data acquisition is set to 44100 Hz to cover the audible range of the human ear and part of the ultrasonic frequency band, used to capture features such as tool wear and cutting noise. The optical imaging device performs continuous low-resolution (640×480) previews at a rate of 5 Hz to reduce computational load. High-resolution (1920×1200) imaging is only performed at the designated location upon receiving a specific trigger command.

[0026] In this embodiment, after obtaining the operating data of the preset internal components of the processing equipment through a multimodal sensor array, multi-dimensional physical quantity data of the preset internal components of the processing equipment, including vibration signals, temperature distribution data, and sound wave data, can be obtained simultaneously. Then, the operating data will be preprocessed. Specifically, the preprocessing of the vibration signal includes sampling rate adjustment and baseline calibration. The preprocessed vibration data is combined with the temperature distribution data to establish the correlation between the abnormal temperature rise area in the temperature distribution data and the preprocessed vibration signal, confirming the preliminary internal operating status. Then, the significant abnormal peak value in the sound wave data is combined to further verify the internal operating status of the processing equipment.

[0027] Finally, a data fusion model based on machine learning algorithms is used to perform feature-level fusion of the preprocessed multimodal data to form a comprehensive operating dataset. This comprehensive operating dataset serves as a digital representation of the internal operating status of the processing equipment and is used to quantify the stability of the internal operating status of the processing equipment.

[0028] In this way, in the initial stage, by fusing multimodal information such as vibration signals, temperature distribution data, and sound wave data, and cross-validating from different physical dimensions, the defects of the internal operating state determined by data obtained from a single sensor, which are one-sided and susceptible to interference, are avoided. This makes the final comprehensive operating dataset a compact, robust, and information-rich digital representation of the internal operating state of the processing equipment.

[0029] S102. Analyze the amplitude variation pattern of the vibration signal in the comprehensive operation data, determine all abnormal frequency ranges in the vibration signal, trace the source of potential wear anomalies based on the frequency characteristic values ​​of the abnormal frequency ranges, and determine the source component of the potential wear anomaly.

[0030] In this embodiment, signal processing technology is used to extract the frequency spectrum features and amplitude variation patterns of the vibration signal. The frequency spectrum features are then refined and decomposed with high precision to analyze the fluctuation range of the vibration signal's amplitude variation patterns and identify all abnormal frequency ranges within the vibration signal. If the extracted frequency feature values ​​exceed a preset threshold dynamically set based on historical operating data within an abnormal frequency range, it is determined that the processing equipment has potential wear anomalies. The source of these potential wear anomalies is then traced through signal features to preliminarily determine the originating component.

[0031] By performing in-depth analysis of highly sensitive vibration signals, an alarm can be quickly issued when abnormalities occur in the macroscopic internal operating state of the processing equipment, potential wear anomalies can be identified, and the direction of subsequent precise tracing can be guided to preliminarily determine the source component of the potential wear anomaly.

[0032] S103. Obtain optical imaging data of the source component, extract the surface features of the component from the optical imaging data, quantify the wear degree of the surface features of the component, and obtain a quantitative description of the wear degree.

[0033] In this embodiment, after determining the source component, an optical imaging device corresponding to the source component is activated to acquire high-resolution optical imaging data of the surface of the source component. This setup utilizes the spectral characteristics of the vibration signal to narrow down the fault range to a specific component (such as a bearing), and then triggers the optical imaging device in the corresponding area of ​​that component to acquire effective optical imaging data.

[0034] Next, image processing algorithms are used to extract features such as texture, edge, and color from the surface of the source component, thereby quantifying the wear level of the source component and obtaining a quantitative description of the wear level, which makes it easier for users to understand the specific wear level of the source component.

[0035] S104. Quantify and map the wear level to the spatial location of the processing equipment to obtain a preliminary distribution area, and perform position calibration on the preliminary distribution area to determine the precise coordinate information of potential wear anomalies.

[0036] In this embodiment, a three-dimensional spatial mapping technique is used to map the wear degree identified in the two-dimensional optical imaging data to the three-dimensional space where the processing equipment is located, thereby obtaining the preliminary distribution area of ​​potential wear anomalies in the spatial coordinate system. Then, based on the preset dominant frequency, the dominant frequency vibration features in the abnormal frequency range marked as potential wear anomalies are extracted. The optical imaging data of each frame under potential wear anomalies are matched with the dominant frequency vibration features to determine the offset between the vibration peak position and the wear center. The position of the preliminary distribution area is calibrated, and finally the precise coordinate information of potential wear anomalies is determined, thus realizing the positioning from which internal component to which point on the internal component.

[0037] By utilizing three-dimensional spatial mapping technology, the specific physical coordinates of potential wear anomalies such as wear or cracks can be accurately located. This setup of screening first and then accurately locating allows for visual internal inspection without disassembling the processing equipment, greatly improving the efficiency of fault diagnosis.

[0038] S105. Feature extraction is performed on the vibration signal and optical imaging data at the precise coordinate information. The extracted features are then fused and denoised to obtain a continuous monitoring sequence. The internal operating status of the processing equipment is tracked based on the continuous monitoring sequence to obtain a dynamic trajectory.

[0039] In this embodiment of the application, step S105 unfolds the state at a spatial point along the time axis to form a dynamic change trajectory, i.e., a dynamic change trend. By acquiring the dynamic change trend, it is possible to analyze whether there is a trend of accelerated deterioration in the internal operating state of the processing equipment. This allows for early warnings, earlier than traditional threshold alarm methods, at the early stage of fault development, thus buying valuable response time for maintenance personnel.

[0040] S106. Perform trend analysis on the dynamic trajectory and generate fault early warning information.

[0041] In this embodiment of the application, the rate of change of state and the signal-to-noise ratio are calculated based on the dynamic change trajectory. Then, trend analysis is performed on the rate of change and the noise ratio. If the analysis results show that the dynamic change trajectory shows signs of accelerated deterioration that meet preset conditions, the early warning signal generation process is triggered to generate targeted fault early warning information that includes fault location, severity and recommended maintenance measures.

[0042] The fault early warning method for automatic metal rod processing equipment disclosed in this application integrates multimodal information such as vibration signals, temperature distribution data, and acoustic wave data, and cross-validates from different physical dimensions. This avoids the shortcomings of determining the internal operating status based on data acquired from a single sensor, which is one-sided and susceptible to interference. The resulting comprehensive operating dataset is a compact, robust, and information-rich digital representation of the internal operating status of the processing equipment. By tracking the internal operating status of the processing equipment according to a continuous monitoring sequence, a dynamic change trajectory is obtained. Trend analysis of this dynamic trajectory reveals an accelerating deterioration trend, enabling early warning at the early stages of fault development, compared to traditional threshold alarm methods. The early warning system provides valuable response time for maintenance personnel. By utilizing the spectral characteristics of vibration signals, the fault range is narrowed down to specific components, thus identifying the source component. Then, by triggering optical imaging of the corresponding area of ​​that component, optical imaging data is obtained. Using three-dimensional spatial mapping technology, the specific physical coordinates of wear or cracks can be accurately located. This allows for visualized internal operating status checks without disassembling the equipment, greatly improving fault diagnosis efficiency. Therefore, this application can monitor the operating status of key internal components in real time without disassembling the processing equipment and accurately locate the specific location of the fault, ensuring the accuracy and timeliness of fault early warning and significantly improving equipment reliability and maintenance efficiency.

[0043] In this embodiment of the application, step S101 specifically includes: S1011. Obtain the preset operating data of the internal components of the processing equipment.

[0044] S1012. Resample the vibration signal, unify the sampling frequency of all vibration signals, and perform signal noise reduction and signal drift suppression on the resampled vibration signal to obtain the preprocessed vibration signal.

[0045] First, check if any vibration signals are lost, and compensate for the lost points using cubic spline interpolation. To ensure consistent subsequent processing, resample the vibration signals from different channels to a standard frequency, such as 20480 Hz.

[0046] During long-term operation of processing equipment, due to factors such as temperature drift, a slow DC component shift may occur in the vibration signal. Baseline calibration by calculating the moving average of the vibration signal (with a window size of 1000 sampling points) can eliminate long-term drift and obtain a stable vibration signal.

[0047] To address the high-frequency background noise mixed in the vibration signal, a wavelet thresholding denoising method is employed. A wavelet basis function matching the signal characteristics is selected, and the signal is subjected to multi-level wavelet decomposition. The wavelet basis is set to Daubechies level 4, and the number of decomposition levels is 5.

[0048] S1013. Extract temperature anomalies from the temperature distribution data, determine abnormal temperature rise areas in the temperature distribution data based on the temperature anomalies, and establish the correlation between the abnormal temperature rise areas and the preprocessed vibration signal to obtain the preprocessed temperature distribution data.

[0049] In this embodiment, a temperature threshold, such as 50°C, is preset. When the temperature distribution data exceeds the temperature threshold, it can be identified as a temperature anomaly. Based on all the temperature anomalies, the abnormal temperature rise area in the temperature distribution data can be determined, and the correlation between the abnormal temperature rise area and the preprocessed vibration signal can be calculated. For example, the correlation between the two can be calculated using the Pearson correlation coefficient. If the calculated correlation exceeds the preset correlation, the abnormal temperature rise area can be pre-marked as a potential wear anomaly area, and the preprocessed temperature distribution data can be obtained further.

[0050] S1014. Extract the spectral features of the acoustic wave data and identify abnormal frequency peaks in the spectral features to obtain the preprocessed acoustic wave data.

[0051] In this embodiment, the spectral features of the acoustic wave data are extracted using the Fast Fourier Transform algorithm. The analysis frequency band is set to 20Hz to 2000Hz to identify abnormal frequency peaks in the spectral features. For example, a significant peak is found at 500Hz, which can be further identified as potential wear anomalies.

[0052] S1015. Perform data fusion on the preprocessed running data to construct a comprehensive running dataset.

[0053] Considering the significant differences in the physical meaning and scale of different data, this application adopts a hybrid strategy based on feature-level fusion and deep autoencoder to fuse the data and construct a comprehensive running dataset.

[0054] More specifically, step S1015 includes: S10151. Perform feature extraction on the preprocessed running data to obtain vibration features, temperature distribution features, and sound wave features.

[0055] From the preprocessed vibration signal, time-domain features (such as mean, root mean square, peak value, kurtosis, waveform factor, etc.) and frequency-domain features (such as peak frequency, energy, centroid frequency, frequency variance, etc.) are extracted to form vibration characteristics.

[0056] Statistical features (such as maximum temperature difference, average temperature difference, area, equivalent diameter, etc.) are extracted from each abnormal temperature rise region in the preprocessed temperature distribution data to form temperature distribution features.

[0057] Mel frequency cepstral coefficients and their first and second order differences are extracted from the preprocessed acoustic signal to form acoustic features.

[0058] Through the above processing, vibration characteristics, temperature distribution characteristics, and sound wave characteristics are extracted respectively.

[0059] S10152. Using the resampled vibration signal, perform feature alignment and splicing of vibration features, temperature distribution features, and sound wave features to obtain a fused feature vector.

[0060] Because the sampling rates of each sensor are different, the feature vectors need to be aligned on the time axis. In this embodiment of the application, Since the sampling frequency of the preprocessed vibration signals is already consistent, an upsampling strategy based on the time axis of the vibration signals is adopted. For low sampling rate features such as temperature distribution features and sound wave features, interpolation is performed in time to keep the most recent value, aligning their time indices with the vibration features. After alignment, all feature vectors are concatenated into a high-dimensional fused feature vector.

[0061] S10153. Perform feature processing on the fused feature vector to obtain a state feature vector that characterizes the internal operating state of the processing equipment.

[0062] In this embodiment of the application, feature processing is performed based on a pre-trained autoencoder, which is used for nonlinear dimensionality reduction and deep feature fusion.

[0063] Because the fused feature vectors have high dimensionality and redundancy, the encoder part of the autoencoder consists of multiple fully connected layers, for example, the input layer has a dimension of... The layer then consists of hidden layers of 512, 256, 128, and 64 dimensions, with the final layer being the bottleneck layer, which outputs a 32-dimensional low-dimensional dense vector. The decoder part of the autoencoder is a mirror-symmetric structure of the encoder part.

[0064] The training process of a pre-trained autoencoder involves collecting a large amount of historical operational data from the processing equipment under normal operating conditions. This historical data is used as both input and output labels to train the autoencoder network. The training objective is to minimize the reconstruction error. In this way, the autoencoder learns to compress high-dimensional raw features into a low-dimensional space, automatically uncovering nonlinear relationships between multimodal data and filtering out noise and redundancy in the process.

[0065] During online monitoring, the fused feature vector calculated in real time is input into the trained autoencoder. The output of its bottleneck layer is the fused state feature vector, which can be regarded as the "health fingerprint" of the internal operating status of the processing equipment, that is, the digital representation of the internal operating status of the processing equipment.

[0066] S10154. By integrating the preprocessed running data and the state feature vector, a comprehensive running dataset is constructed.

[0067] By fusing the state feature vectors from a continuous time series with the corresponding preprocessed operational data, a comprehensive operational dataset can be constructed. This comprehensive operational dataset is understood to be a compact, robust, and information-rich digital representation of the internal operating state of the processing equipment.

[0068] In this embodiment of the application, step S102 specifically includes: S1021. Perform spectral decomposition on the preprocessed vibration signal and extract spectral features.

[0069] In step S102, the preprocessed vibration signal is first subjected to spectral decomposition using the Fast Fourier Transform algorithm to extract the main and secondary frequency components of the vibration signal, thereby obtaining spectral characteristics. For example, after spectral decomposition of the vibration signal, a significant frequency peak of 100Hz is found during the operation of the processing equipment, and the main frequency component of the vibration signal is also 100Hz. Furthermore, secondary frequency components can be obtained at 20Hz.

[0070] S1022. Perform time-domain analysis on the amplitude variation patterns contained in the spectral characteristics to obtain the amplitude fluctuation range.

[0071] Next, the spectral characteristics, namely the amplitude variation patterns of the primary and secondary frequency components, are analyzed. More specifically, the amplitude fluctuation range of the spectral characteristics is calculated by performing time-domain analysis on the amplitude variation patterns contained in the spectral characteristics. For example, by performing time-domain analysis with a time window of 10 seconds, the amplitude fluctuation range of the primary frequency components is shown to be between 0.02 and 0.06 mm, with a mean of 0.04 mm and a standard deviation of 0.01 mm.

[0072] S1023. Compare the amplitude fluctuation range with the historical normal fluctuation range extracted based on historical operating data to determine all abnormal frequency intervals in the amplitude fluctuation range.

[0073] In this embodiment, historical operating data of the processing equipment is also acquired. Specifically, this historical operating data includes vibration signals, temperature distribution data, and sound wave data acquired when the processing equipment is in normal operating condition. The historical operating data is then processed based on the data processing methods of steps S1021 and S1022 to obtain the historical normal fluctuation range of the processing equipment's historical operating data. After obtaining the historical normal fluctuation range, the amplitude fluctuation range is compared with the historical normal fluctuation range to determine whether there are abnormal frequency intervals in the amplitude fluctuation range, and if so, the frequency intervals of each abnormal frequency interval in the vibration signal.

[0074] It should be noted that both the primary and secondary frequency components can be obtained with a corresponding amplitude fluctuation range.

[0075] For example, if the standard deviation of the amplitude fluctuation range of the main frequency component is 20% higher than that of the historical normal fluctuation range, it indicates that the frequency range corresponding to the main frequency component is an abnormal frequency range.

[0076] S1024. All abnormal frequency intervals are further decomposed to obtain the frequency concentration intervals of each abnormal frequency interval, and the frequency feature values ​​of each frequency concentration interval are extracted.

[0077] In this embodiment of the application, all abnormal frequency intervals are refined and decomposed using a wavelet transform algorithm. The decomposition level is set to 5 levels. It is found that the abnormal frequencies are concentrated in the 95 to 105 Hz interval, that is, the frequency concentration interval is 95 to 105 Hz interval. Correspondingly, the frequency characteristic value of the frequency concentration interval in the 95 to 105 Hz interval is 0.035 mm.

[0078] S1025. If it is determined that the frequency characteristic value exceeds the preset threshold, the abnormal frequency range corresponding to the frequency characteristic value is marked as a potential wear anomaly.

[0079] If the frequency characteristic value of the concentrated frequency range between 95 and 105 Hz is 0.035 mm, and exceeds the preset threshold of 0.025 mm, then the abnormal frequency range corresponding to the frequency characteristic value can be marked as a potential wear anomaly.

[0080] S1026. Based on the structural model of the processing equipment and the location distribution of the sensors used to acquire operating data, trace the propagation path of the abnormal frequency range marked as potential wear anomalies to determine the source component of the potential wear anomalies.

[0081] To eliminate misjudgments caused by complex transmission paths, in this embodiment of the application, the structural model of the processing equipment and sensors, i.e. vibration sensors, arranged at different positions of the processing equipment are also used for verification to trace the propagation path of abnormal frequency ranges marked as potential wear anomalies.

[0082] Assuming the source component is near the grinding wheel bearing housing, the signal strength measured by vibration sensors 1 and 2 mounted on the spindle should be significantly lower than that of vibration sensor 3 mounted on the grinding wheel bearing housing. By acquiring the signal strength of these vibration sensors separately, the propagation path can be traced based on the signal strength. For example, the ratio of the signal strength of vibration sensors 1 and 2 to that of vibration sensor 3 was found to be 0.667:1. Combined with triangulation, the source component of the potential wear anomaly was inferred to be the bearing housing, further supporting the correlation between bearing wear and excessive load, and providing data support for subsequent maintenance.

[0083] In this embodiment of the application, step S103 specifically includes: S1031. Acquire optical imaging data from the source component.

[0084] Once one or more internal components of the processing equipment are pre-marked as source components, the system will generate an image acquisition task instruction, which includes: Target components: such as the front bearing of the grinding wheel spindle, which is not visible due to the seal of the bearing itself, but its related exposed components, such as the journal, the sealed end cap, or the interior which can be observed through the reserved observation hole.

[0085] Target location: Based on the structural model of the equipment, the optimal camera coordinates required to capture the area are calculated, and the servo motor drives the optical imaging device to move to the designated position.

[0086] Shooting parameters: The image acquisition task instruction also includes the image acquisition frame rate and image acquisition frequency. This ensures that the optical imaging device's image acquisition frequency is within the corresponding acquisition time window. For example, the shooting parameters may instruct the optical imaging device to perform continuous high-resolution (1920×1200) shooting at the highest frame rate for 2 seconds.

[0087] In step S1031, after acquiring the optical imaging data of the source component, image preprocessing can be performed on the optical imaging data of the source component. Image preprocessing includes, but is not limited to: Frame selection: Select the highest-resolution frames from the optical imaging data. Sharpness is evaluated using the Tenengrad gradient method, which calculates the gradient sum after convolving the image with the Sobel operator. Image enhancement: The Contrast-Limited Adaptive Histogram Equalization (CLAHE) algorithm is applied to the selected optical imaging data to enhance the contrast of local textures and defects, with particularly significant effects on uneven reflections on metal surfaces.

[0088] S1032. Perform texture segmentation on the optical imaging data, extract the surface feature information of the source component, and obtain a surface texture distribution map.

[0089] In this embodiment, a pre-trained U-Net or DeepLabv3+ semantic segmentation network is used to perform pixel-level classification of optical imaging data. The output of the semantic segmentation network is surface feature information, which yields a surface texture distribution map, where pixels with a value of 1 represent "wear areas" or "defect areas," and pixels with a value of 0 represent "normal surfaces."

[0090] S1033. Determine the vibration energy distribution and signal time domain window of the vibration signal, and based on the vibration energy distribution and signal time domain window, select the time points corresponding to the abnormal frequency intervals marked as potential wear anomalies.

[0091] S1034. Match the time points with the optical imaging data in time to determine the optical imaging data of each frame under potential wear anomalies.

[0092] By setting the vibration energy threshold to 5.0 mJ and combining it with a signal segment with a signal time domain window width of 0.1 seconds, the surface texture distribution map is matched for fluctuations in the time dimension. This allows us to filter out the time points corresponding to the abnormal frequency ranges marked as potential wear anomalies. These time points are then matched with the timestamps of the optical imaging data to lock the optical imaging data of each frame under potential wear anomalies, thus forming a data association.

[0093] S1035. Extract the surface features of the component from the optical imaging data of each frame under potential wear anomalies, and quantify the wear degree of the source component based on the surface features of the component and the intensity change of vibration energy distribution, and determine the wear quantification value of the source component.

[0094] Subsequently, the surface features of the component were further analyzed using image processing technology. Edge detection algorithms such as the Canny operator were used to extract surface features such as surface texture and crack features. The low threshold of the Canny operator was set to 50 and the high threshold was set to 150. The proportion of crack area was calculated. For example, if the total area of ​​the detected crack was 2.5 square millimeters and the total area of ​​the source component was 100 square millimeters, the proportion was 2.5%.

[0095] Finally, the wear degree of the source component is quantitatively described. Based on the crack area ratio and vibration energy anomaly value, the wear quantification value is set as Crack ratio × 100 + Vibration energy threshold × 0.5. The wear quantification value is calculated to be 2.5 × 100 + 5.0 × 0.5 = 252.5.

[0096] S1036. Compare the wear quantification value with the preset wear level standard to determine the quantitative description of the wear degree of the source component.

[0097] By comparing the wear quantification value with a pre-established wear level standard, the wear level of the source component is determined, and a final wear status description is obtained. For example, if the wear quantification value is greater than 200, it is judged as severe wear, and the system automatically generates an early warning report and records the data to the database.

[0098] In this embodiment of the application, step S104 specifically includes: S1041. Based on the structural model of the processing equipment, establish the spatial coordinate system of the processing equipment.

[0099] By importing the structural model of the processing equipment and selecting several feature points on the structural model that are easy to measure in reality, such as screw holes and edge intersections, a one-to-one correspondence between the spatial coordinate system of the processing equipment and the world coordinate system is established by measuring its actual world coordinates.

[0100] For example, a high-precision scan of the processing equipment can be performed using a 3D laser scanning device to obtain point cloud data, and the aforementioned structural model can be obtained based on the point cloud data.

[0101] S1042. Quantify and map the degree of wear to the spatial coordinate system where the processing equipment is located to obtain the preliminary distribution area of ​​potential wear anomalies in the spatial coordinate system.

[0102] Using point cloud processing algorithms such as point cloud filtering and surface fitting, spatial geometric features that quantify the wear degree are extracted. For example, the average wear depth is calculated to be 2.5 mm with a standard deviation of 0.3 mm. The spatial geometric features that quantify the wear degree are then mapped to the spatial coordinate system of the processing equipment. The preliminary distribution area of ​​potential wear anomalies in the spatial coordinate system is obtained as 400-600 mm on the X-axis, 300-500 mm on the Y-axis, and 50-100 mm on the Z-axis.

[0103] S1043. Based on the preset main frequency, extract the main frequency vibration characteristics in the abnormal frequency range marked as potential wear abnormality.

[0104] Next, based on a preset main frequency, such as 50Hz, Fourier transform is used to analyze the vibration signal and extract the vibration characteristics of the main frequency of 50Hz in the abnormal frequency range.

[0105] S1044. Match the optical imaging data of each frame under potential wear anomalies with the main frequency vibration characteristics to determine the offset between the vibration peak position and the wear center.

[0106] S1045. The offset is superimposed onto the initial distribution area to obtain the calibration distribution area.

[0107] By matching the dominant frequency vibration characteristics with the optical imaging data of each frame under potential wear anomalies, the offset between the vibration peak position and the wear center in the optical imaging data is calculated to be 5 mm on the X-axis and 3 mm on the Y-axis. Then, the coordinates of the abnormal position are calibrated to be 405-605 mm on the X-axis, 303-503 mm on the Y-axis, and 50-100 mm on the Z-axis.

[0108] S1046. Input the calibration distribution area into the constructed abnormal location and wear degree mapping model to obtain the accurate coordinate information output by the abnormal location and wear degree mapping model.

[0109] Using the calibrated distribution area and machine learning algorithms such as support vector machines to classify the wear area, the training dataset of the constructed abnormal location and wear degree mapping model contains 1000 wear samples. The features include depth, area and vibration intensity. The constructed abnormal location and wear degree mapping model finally outputs the precise coordinates of the abnormal parts as 410 mm on the X-axis, 310 mm on the Y-axis and 60 mm on the Z-axis, and generates a visualization report, marking the wear degree and location distribution for reference in subsequent maintenance decisions.

[0110] In this embodiment of the application, step S105 specifically includes: S1051. Based on precise coordinate information, acquire vibration signals and optical imaging data at the precise coordinate information location.

[0111] Assuming the precise coordinate information is (X=12.5, Y=8.3, Z=4.2), the vibration signal and optical imaging data at this precise coordinate information are obtained accordingly.

[0112] S1052. Extract features from the vibration signal and optical imaging data at the precise coordinate information to obtain the vibration features of the vibration signal and the surface texture features of the optical imaging data.

[0113] The surface texture distribution map above contains surface texture features.

[0114] S1053. Based on the pre-assigned data weights, feature fusion is performed on vibration features and surface texture features to obtain abnormal fusion features.

[0115] In the pre-allocated data weights, the vibration signal weight is set to 0.6 and the surface texture feature weight is set to 0.4. Based on these two data weights, the vibration features and surface texture features can be fused to obtain the abnormal fusion features, that is, the abnormal fusion features are the weighted sum of the real features and the surface texture features.

[0116] It should be noted that the allocation of data weights in this embodiment is dynamic. In the early stages of a fault, image features (such as microcracks) may not yet be apparent or may change slowly. At this time, vibration features are more sensitive to the fault, so the weight of surface texture features is less than that of vibration features. As wear intensifies, the changes in surface texture features of the same type as image features become significant and intuitive, and the weight of surface texture features can be adaptively increased at this time.

[0117] S1054. Perform sliding smoothing on the abnormal fusion features to obtain a continuous monitoring sequence.

[0118] Anomalies in fusion features may be caused by measurement noise or instantaneous fluctuations in operating conditions. In this embodiment, a locally weighted scatterplot smoothing (LOWESS) method is used for smoothing. For each data point, a weighted multinomial regression is performed within its neighborhood, the size of which is controlled by the bandwidth parameter span (e.g., 0.2). This non-parametric method effectively preserves the overall trend of the data while filtering out local short-term fluctuations, thereby reducing noise interference and generating a smoothed continuous monitoring sequence.

[0119] The smoothed continuous monitoring sequence is continuous in time, and fully records the evolution of the health status at that specific location over time.

[0120] S1055. Determine the change characteristics of the continuous monitoring sequence in the time dimension, and track the internal operating status of the processing equipment based on the change characteristics to obtain the dynamic change trajectory.

[0121] Finally, through time series analysis, data changes in the time dimension are recorded. If abnormal fluctuations occur in the data changes in the time dimension, i.e., exceeding the preset amount, it is determined that the internal operating state has changed significantly. Combined with the Kalman filter algorithm, the dynamic change trajectory of the state is predicted, the internal operating state of the processing equipment is tracked, the dynamic trajectory information of the continuous monitoring sequence is updated, and the final internal state change record is obtained, i.e., the dynamic change trajectory is obtained.

[0122] The Kalman filter algorithm recursively estimates the optimal internal state value using a noisy observation sequence through a predictive and updating cycle. The output of the Kalman filter algorithm is not merely a smooth value, but also includes an estimate of its rate of change, resulting in a smooth, dynamic trajectory containing trend information. Assuming the state value of the continuously monitored sequence is 0.31 at t=6s and the predicted value at t=7s is 0.32, with an error range of ±0.02, combined with historical trends, the internal operating state is judged to be slowly deteriorating, and there is a positive correlation between vibration frequency and crack propagation, providing a basis for subsequent maintenance decisions.

[0123] In this embodiment of the application, step S106 specifically includes: S1061. Filter out interference signals in the dynamically changing trajectory, filter out irrelevant interference information, and obtain the denoised dynamically changing trajectory. S1062. Determine the rate of change of the state of the dynamic trajectory of the denoised system to obtain the rate change characteristics.

[0124] S1063. Determine the main frequency components and noise power ratio of the dynamic change trajectory of the denoising process to obtain the signal-to-noise ratio.

[0125] S1064. If the rate change characteristics exceed a preset range and the signal-to-noise ratio does not exceed a preset ratio, generate a fault alarm message. A comprehensive evaluation is performed based on the rate change characteristics and the signal-to-noise ratio. If the rate change characteristics exceed a preset range and the signal-to-noise ratio does not exceed a preset ratio, it is determined that there are signs of accelerated deterioration, triggering the warning signal generation process to generate targeted fault warning information including the fault location, severity, and recommended maintenance measures.

[0126] The fault location is the precise coordinate information, and the severity is the quantitative description of the wear degree. Based on the precise coordinate information and severity, the system can match and output suggested intervention measures from the maintenance knowledge base, such as "It is recommended to check and replace the grinding wheel in the next shift" or "It is recommended to reduce the feed rate to 80%".

[0127] Please see Figure 2 In this embodiment of the application, the method further includes the following steps: S107. Using optical imaging data and the pure signal obtained after periodically filtering out vibration signals, a risk assessment is performed on the fault warning information to obtain the risk assessment result, and an intervention command execution sequence is generated based on the risk assessment result.

[0128] Step S107 is responsible for converting the fault warning information into a specific sequence of intervention instructions.

[0129] The generated targeted fault warning information is encapsulated into a JSON format data packet. First, the relevant data is sent to the control system via a transmission protocol. Assuming the industry-standard Modbus protocol is used, the data packet contains the fault code 0x001A and the timestamp 2026-03-15 14:30:00. The data is encapsulated in 18-bit unsigned integer format, and the transmission rate is set to 9600 baud rate to ensure real-time performance. After receiving the data, the system verifies its integrity using a CRC check algorithm. If the verification passes, the system proceeds to the next step of analysis; otherwise, a retransmission mechanism is triggered.

[0130] Before generating the final sequence of intervention instructions, a final risk assessment is conducted by combining the image texture features and interference filtering of the optical imaging data to confirm the necessity and urgency of the intervention.

[0131] In this embodiment, a historical fault case library for storing several fault modes is also pre-set. After a fault warning message is generated, the historical fault case library is searched to find the fault mode that is most similar to, and thus has the highest matching degree, the surface texture distribution map extracted from the optical imaging data in step S103. If the fault warning message can be matched with the corresponding fault mode, the priority of the fault warning message is increased. Matching the fault warning message with the fault mode through the surface texture distribution map allows for an initial risk assessment of the current fault warning message, ensuring that the system automatically responds to the fault warning and reduces risk.

[0132] It is understandable that each fault mode has its own corresponding characteristic information, and the characteristic information corresponding to different fault modes is different. Users can set different characteristic information to form preset fault modes.

[0133] By pre-setting specific frequency, fault-independent periodic vibration sources, and targeting certain specific frequency, fault-independent periodic vibration sources (such as vibrations from other nearby equipment), the correlation function is calculated to determine the correlation between the current vibration signal and these known interference sources. Interference from fault-independent periodic vibration sources at specific frequencies is then filtered out to obtain a clean signal. If the correlation is high, it indicates that the vibration signal is likely affected by the periodic vibration source, thus reducing the confidence level of the warning and lowering the priority of the fault warning information to prevent false triggering. A second risk assessment is then performed using the clean signal obtained after filtering out periodic interference to further reduce risk.

[0134] Based on these two risk assessments, a risk assessment result for the fault warning information can be generated. This result characterizes the risk assessment level of the corresponding fault warning information. Using optical imaging data and the clean signal obtained after periodically filtering out vibration interference, the fault warning information undergoes two risk assessments to determine its final risk assessment level. The system then automatically generates or recommends an intervention command sequence to the operator. For example: First priority intervention command execution sequence (immediate execution): If the surface texture distribution map based on optical imaging data increases the priority of the fault warning information and the clean signal does not indicate a reduction in priority, it means that the risk assessment level of the fault warning information is extremely high. The system can directly write commands to the equipment PLC, such as "immediately execute emergency stop" or "spindle speed reduced to 100 RPM".

[0135] Second-priority intervention sequence (recommended): If the surface texture distribution map based on optical imaging data does not increase the priority of the fault warning information, and the clean signal does not indicate a decrease in priority; or if the surface texture distribution map based on optical imaging data increases the priority of the fault warning information, and the clean signal indicates a decrease in priority, the system will recommend an operation sequence for most warnings in these two cases. For example: "1. Please have the operator confirm the warning information. 2. Click the Parameter Adjustment button on the control panel; the system will automatically reduce the feed rate to 70% of the current level. 3. After completing this batch of processing, schedule planned maintenance." The third priority instruction intervention instruction execution sequence (information recording): If the surface texture distribution map based on optical imaging data does not increase the priority of the fault warning information, and the clean signal indicates a reduction in priority, all information such as the complete warning event, relevant data snapshots, and operator response records will be stored in the maintenance log database for subsequent data optimization and knowledge accumulation.

[0136] To address the aforementioned issues, this specification provides a fault early warning system for an automatic metal rod processing equipment. This system aims to monitor the operational status of key internal components in real time without disassembling the processing equipment, accurately pinpointing the location of any faults. This ensures the precision and timeliness of fault warnings, significantly improving equipment reliability and maintenance efficiency. Figure 3 This is a structural schematic diagram of a fault early warning system for an automatic metal rod processing equipment according to an embodiment of this application, as shown below. Figure 3 As shown, the system may include: The data construction module 10 is used to acquire the operating data of the preset internal components of the processing equipment, preprocess the operating data, and merge the preprocessed operating data to construct a comprehensive operating dataset. The operating data includes vibration signals, temperature distribution data, and sound wave data. The preset internal components can be grinding wheel spindle, guide wheel spindle, grinding area, etc.

[0137] The source tracing module 20 is used to analyze the amplitude change pattern of the vibration signal in the comprehensive operation data, determine all abnormal frequency ranges in the vibration signal, trace the source of potential wear anomalies based on the frequency characteristic values ​​of the abnormal frequency ranges, and determine the source component of the potential wear anomaly.

[0138] The wear quantization module 30 is used to acquire optical imaging data of the source component, extract the surface features of the component from the optical imaging data, quantify the wear degree of the surface features of the component, and obtain a quantitative description of the wear degree.

[0139] The coordinate correction module 40 is used to map the quantitative description of the wear degree to the spatial location of the processing equipment to obtain the preliminary distribution area, and to perform position calibration on the preliminary distribution area to determine the precise coordinate information of potential wear anomalies.

[0140] The trajectory prediction module 50 is used to extract features from vibration signals and optical imaging data at precise coordinate information, perform feature fusion and noise reduction on the extracted features to obtain a continuous monitoring sequence, and track the internal operating status of the processing equipment according to the continuous monitoring sequence to obtain a dynamically changing trajectory.

[0141] The fault early warning module 60 is used to perform trend analysis on the dynamic change trajectory and generate fault early warning information.

[0142] The fault early warning system for the automatic metal rod processing equipment of this application integrates multimodal information such as vibration signals, temperature distribution data, and acoustic wave data, and cross-verifies it from different physical dimensions. This avoids the shortcomings of determining the internal operating status based on data acquired by a single sensor, which is one-sided and susceptible to interference. The resulting comprehensive operating dataset is a compact, robust, and information-rich digital representation of the internal operating status of the processing equipment. By tracking the internal operating status of the processing equipment according to a continuous monitoring sequence, the system obtains a dynamic trajectory of change. Trend analysis of this dynamic trajectory reveals an accelerating deterioration trend, enabling early warning at the early stages of fault development, compared to traditional threshold alarm methods. The early warning system provides valuable response time for maintenance personnel. By utilizing the spectral characteristics of vibration signals, the fault range is narrowed down to specific components, thus identifying the source component. Then, by triggering optical imaging of the corresponding area of ​​that component, optical imaging data is obtained. Using three-dimensional spatial mapping technology, the specific physical coordinates of wear or cracks can be accurately located. This allows for visualized internal operating status checks without disassembling the equipment, greatly improving fault diagnosis efficiency. Therefore, this application can monitor the operating status of key internal components in real time without disassembling the processing equipment and accurately locate the specific location of the fault, ensuring the accuracy and timeliness of fault early warning and significantly improving equipment reliability and maintenance efficiency.

[0143] It should be noted that, in this document, the terms "comprising," "including," and any other variations are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Specific examples have been used in this document to illustrate the principles and implementation methods of the technical solutions of this application. The above examples are only for the purpose of helping to understand the methods and core ideas of this application. The above descriptions are merely preferred embodiments of this application. It should be pointed out that, due to the limitations of written expression and the objective existence of infinite specific structures, those skilled in the art can make several improvements, modifications, or changes without departing from the principles of this application, and can also combine the above technical features in an appropriate manner; these improvements, modifications, changes, or combinations, or the direct application of the concept and technical solutions of this application to other situations without modification, should all be considered within the scope of protection of this application.

Claims

1. A fault early warning method for an automatic metal rod processing equipment, characterized in that, include: The system acquires the operating data of the internal components of the processing equipment, preprocesses the operating data, and merges the preprocessed operating data to construct a comprehensive operating dataset. The operating data includes vibration signals, temperature distribution data, and sound wave data. Analyze the amplitude variation pattern of the vibration signal in the comprehensive operation data, determine all abnormal frequency ranges in the vibration signal, trace the source of potential wear anomalies based on the frequency characteristic values ​​of the abnormal frequency ranges, and identify the source components of potential wear anomalies; Acquire optical imaging data of the source component, extract the surface features of the component from the optical imaging data, quantify the wear degree of the surface features of the component, and obtain a quantitative description of the wear degree. The wear level is quantitatively described and mapped to the spatial location of the processing equipment to obtain a preliminary distribution area. The position of the preliminary distribution area is then calibrated to determine the precise coordinate information of potential wear anomalies. Feature extraction is performed on vibration signals and optical imaging data at precise coordinates. The extracted features are then fused and denoised to obtain a continuous monitoring sequence. The internal operating status of the processing equipment is tracked based on the continuous monitoring sequence to obtain a dynamic trajectory. Trend analysis is performed on the dynamic trajectory to generate fault early warning information.

2. The fault early warning method for the automatic metal rod processing equipment according to claim 1, characterized in that, The analysis integrates the amplitude variation patterns of vibration signals in the comprehensive operational data set, identifies all abnormal frequency ranges in the vibration signals, and traces the source of potential wear anomalies based on the frequency characteristic values ​​of these abnormal frequency ranges, specifically identifying the source components of the potential wear anomalies, including: The preprocessed vibration signal is subjected to spectral decomposition to extract spectral features; Time-domain analysis of the amplitude variation patterns contained in the spectral features yields the amplitude fluctuation range; The amplitude fluctuation range is compared with the historical normal fluctuation range extracted based on historical operation data to determine all abnormal frequency intervals in the amplitude fluctuation range; All abnormal frequency intervals are further decomposed to obtain the frequency concentration intervals for each abnormal frequency interval, and the frequency feature values ​​of each frequency concentration interval are extracted. If the frequency characteristic value exceeds the preset threshold, the abnormal frequency range corresponding to the frequency characteristic value is marked as a potential wear anomaly. Based on the structural model of the processing equipment and the location distribution of the sensors used to acquire operational data, the propagation path of the abnormal frequency range marked as potential wear anomalies is traced to determine the source component of the potential wear anomaly.

3. The fault early warning method for the automatic metal rod processing equipment according to claim 2, characterized in that, The process of acquiring optical imaging data of the source component, extracting surface features from the optical imaging data, quantifying the wear degree of the surface features, and obtaining a quantitative description of the wear degree of the source component specifically includes: Acquire optical imaging data from the source component; The optical imaging data is segmented to extract the surface feature information of the source component, resulting in a surface texture distribution map; Determine the vibration energy distribution and signal time domain window of the vibration signal, and based on the vibration energy distribution and signal time domain window, filter out the time points corresponding to the abnormal frequency intervals marked as potential wear anomalies; By temporally matching the time points with the optical imaging data, the optical imaging data of each frame under potential wear anomalies can be determined. Component surface features are extracted from optical imaging data of each frame under potential wear anomalies. Based on the component surface features and the intensity change of vibration energy distribution, the wear degree of the source component is quantified, and the wear quantification value of the source component is determined. The wear quantification value is compared with the preset wear level standard to determine the quantitative description of the wear degree of the source component.

4. The fault early warning method for the automatic metal rod processing equipment according to claim 3, characterized in that, The process of quantifying and mapping the degree of wear to the spatial location of the processing equipment to obtain a preliminary distribution area, and then calibrating the position of the preliminary distribution area to determine the precise coordinate information of potential wear anomalies, specifically includes: Based on the structural model of the processing equipment, establish the spatial coordinate system in which the processing equipment is located; The wear level is quantitatively described and mapped to the spatial coordinate system where the processing equipment is located, so as to obtain the preliminary distribution area of ​​potential wear anomalies in the spatial coordinate system; Based on the preset dominant frequency, extract the dominant frequency vibration characteristics in the abnormal frequency range marked as potential wear anomalies; The optical imaging data of each frame under potential wear anomalies are matched with the main frequency vibration characteristics to determine the offset between the vibration peak position and the wear center. The offset is superimposed onto the initial distribution area to obtain the calibration distribution area; The calibration distribution area is input into the constructed abnormal location and wear degree mapping model to obtain the accurate coordinate information output by the abnormal location and wear degree mapping model.

5. The fault early warning method for the automatic metal rod processing equipment according to claim 1, characterized in that, The process involves extracting features from vibration signals and optical imaging data at precise coordinate locations, fusing and denoising the extracted features to obtain a continuous monitoring sequence, and tracking the internal operating status of the processing equipment based on the continuous monitoring sequence to obtain a dynamic trajectory. Specifically, this includes: Based on precise coordinate information, vibration signals and optical imaging data at the precise coordinate locations are acquired. Feature extraction was performed on the vibration signal and optical imaging data at the precise coordinate information, respectively, to obtain the vibration features of the vibration signal and the surface texture features of the optical imaging data; Based on the pre-assigned data weights, vibration features and surface texture features are fused to obtain abnormal fusion features; The abnormal fusion features are subjected to sliding smoothing to obtain a continuous monitoring sequence; The changing characteristics of the continuous monitoring sequence in the time dimension are determined, and the internal operating status of the processing equipment is tracked based on the changing characteristics to obtain the dynamic change trajectory.

6. The fault early warning method for the automatic metal rod processing equipment according to claim 1, characterized in that, The step of performing trend analysis on the dynamic trajectory to generate fault early warning information specifically includes: Interference signals in the dynamically changing trajectory are filtered out to obtain a denoised dynamically changing trajectory; Determine the rate of state change of the dynamic trajectory of the denoised system to obtain the rate change characteristics; The main frequency components and noise power ratio of the dynamic change trajectory of the denoising are determined to obtain the signal-to-noise ratio. When the rate change characteristics exceed a preset range and the signal-to-noise ratio does not exceed a preset ratio, a fault alarm message is generated.

7. The fault early warning method for the automatic metal rod processing equipment according to claim 1, characterized in that, The process of acquiring pre-set operating data of internal components of the processing equipment, preprocessing the operating data, and fusing the preprocessed operating data to construct a comprehensive operating dataset specifically includes: Acquire the pre-set operating data of the internal components of the processing equipment; The vibration signal is resampled to unify the sampling frequency of all vibration signals. Then, the resampled vibration signal is subjected to signal noise reduction and signal drift suppression to obtain the preprocessed vibration signal. Temperature anomalies are extracted from the temperature distribution data. Abnormal temperature rise regions in the temperature distribution data are determined based on the temperature anomalies. The correlation between abnormal temperature rise regions and preprocessed vibration signals is established to obtain preprocessed temperature distribution data. Extract the spectral features of the acoustic wave data and identify abnormal frequency peaks in the spectral features to obtain preprocessed acoustic wave data; The preprocessed runtime data is fused to construct a comprehensive runtime dataset.

8. The fault early warning method for the automatic metal rod processing equipment according to claim 7, characterized in that, The process of fusing the preprocessed runtime data to construct a comprehensive runtime dataset specifically includes: Feature extraction was performed on the preprocessed operational data to obtain vibration features, temperature distribution features, and sound wave features, respectively. Using the resampled vibration signal, feature alignment and splicing of vibration features, temperature distribution features, and acoustic wave features are performed to obtain a fused feature vector; The fused feature vector is processed to obtain a state feature vector that characterizes the internal operating state of the processing equipment. By integrating the preprocessed runtime data and state feature vectors, a comprehensive runtime dataset is constructed.

9. The fault early warning method for the automatic metal rod processing equipment according to claim 1, characterized in that, The method also includes the following steps: Using optical imaging data and the clean signal obtained after periodically filtering out vibration signals, a risk assessment is performed on the fault warning information to obtain the risk assessment results, and an intervention command execution sequence is generated based on the risk assessment results.

10. A fault early warning system for an automatic metal rod processing equipment, characterized in that, include: The data construction module is used to acquire the operating data of the internal components of the processing equipment, preprocess the operating data, and merge the preprocessed operating data to construct a comprehensive operating dataset; the operating data includes vibration signals, temperature distribution data, and sound wave data; The source tracing module is used to analyze the amplitude change pattern of the vibration signal in the comprehensive operation data, determine all abnormal frequency ranges in the vibration signal, trace the source of potential wear anomalies based on the frequency characteristic values ​​of the abnormal frequency ranges, and determine the source component of the potential wear anomaly. The wear quantization module is used to acquire optical imaging data of the source component, extract the surface features of the component from the optical imaging data, quantify the wear degree of the surface features of the component, and obtain a quantitative description of the wear degree. The coordinate correction module is used to map the quantitative description of wear degree to the spatial coordinate system where the processing equipment is located, obtain the preliminary distribution area, and perform position calibration on the preliminary distribution area to determine the precise coordinate information of potential wear anomalies; The trajectory prediction module is used to extract features from vibration signals and optical imaging data at precise coordinates, perform feature fusion and noise reduction on the extracted features to obtain a continuous monitoring sequence, and track the internal operating status of the processing equipment based on the continuous monitoring sequence to obtain a dynamically changing trajectory. The fault early warning module is used to perform trend analysis on dynamic change trajectories and generate fault early warning information.

Images