Fiber grating sensing and machine learning-based fine blanking equipment monitoring method and system

Abstract

The application provides a fine blanking equipment monitoring method and system based on fiber grating sensing and machine learning, and relates to the field of fine blanking equipment monitoring.The method steps include laying fiber grating sensors on the slider of a fine blanking machine, and collecting strain data of the slider in real time;preprocessing the strain data to remove noise and nonlinear trends in the strain data;based on minimum value point detection, splitting the strain data from multi-stroke data into single-stroke data, dividing each single-stroke data into data of four stages of fast forward, detection, fine blanking, and fast reverse, and saving to an array;calculating time domain feature parameters of the data in each stage of the array, setting classification labels for the four stages respectively, and constructing a training data set;training a support vector machine classifier based on the training data set to obtain a stage recognition model;and building a visual interface to display real-time data monitoring data of the stage recognition model and perform fault early warning.

Description

Technical Field

[0001] This invention relates to the field of fine blanking equipment monitoring technology, and in particular to a fine blanking equipment monitoring method and system based on fiber optic grating sensing and machine learning. Background Technology

[0002] A fine blanking press is a processing equipment specifically designed for performing fine blanking processes. It is widely used in the manufacturing of precision parts for automobiles, electronics, and other industries. Real-time monitoring of the operating status of the fine blanking press and providing fault warnings can effectively prevent losses caused by major malfunctions.

[0003] The fine blanking process typically includes four stages: rapid advance, inspection, fine blanking, and rapid return. Strain data from the slide block can reflect the stage of operation of the fine blanking machine. Furthermore, data from these four stages can reflect the operating status of the fine blanking machine, and real-time monitoring of each stage is crucial for the intelligent development of fine blanking equipment.

[0004] However, traditional methods for identifying the stages of the fine blanking process mainly rely on human experience or pressure sensors. This approach is not only inefficient but also lacks accuracy, making it difficult to meet the needs of intelligent detection. In order to achieve high-precision real-time identification of the fine blanking process stages, this application proposes a fine blanking equipment monitoring method and system based on fiber optic grating sensing and machine learning. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for monitoring fine blanking equipment based on fiber optic grating sensing and machine learning, so as to solve the problems of low accuracy and poor real-time performance of traditional fine blanking process stage identification methods mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a monitoring method for fine blanking equipment based on fiber optic grating sensing and machine learning, comprising the following steps: deploying fiber optic grating sensors on the slider of the fine blanking machine to collect strain data of the slider in real time; preprocessing the strain data to remove noise and nonlinear trends; splitting the strain data from multi-stroke data into single-stroke data based on minimum point detection, further dividing each single-stroke data into four stages: fast advance, detection, fine blanking, and fast retraction, and saving them to an array; calculating the temporal feature parameters of each stage data in the array, and setting classification labels for each of the four stages to construct a training dataset; training a support vector machine classifier based on the training dataset to obtain a stage recognition model; and building a visualization interface to display the real-time data monitoring data of the stage recognition model and provide fault warnings.

[0007] Optionally, the preprocessing step of the strain data specifically includes: using the db4 wavelet basis function from the Daubechies wavelet family as the decomposition basis function to perform multi-scale wavelet decomposition on the strain data D1 to obtain a set of wavelet coefficients containing noise information; dividing the set of wavelet coefficients into multiple analysis windows on an average basis, and estimating the noise standard deviation using the median absolute deviation method; dynamically adjusting the weighting coefficients based on the stability coefficient of the window data to increase the weight of stable data in the window, and then calculating the weighted noise standard deviation; calculating the Donoho universal threshold based on the weighted noise standard deviation, and then performing soft thresholding on the wavelet coefficients, setting those below the threshold as noise to 0, and smoothing those above the threshold, and reconstructing the processed wavelet coefficients to obtain the denoised strain data D2.

[0008] Optionally, the step of preprocessing the strain data further includes: setting the window size of the strain data D2 to three stroke data lengths, using the local minimum values ​​within each window to form a trend term array, removing the trend term from the strain data D2, and obtaining strain data D3 with nonlinear trends removed.

[0009] Optionally, the step of splitting the strain data from multi-stroke data into single-stroke data based on minimum point detection specifically includes: setting a minimum interval between two minimum points in the strain data to ensure that a complete stroke can be split out, indexing the minimum points, and taking the data between two adjacent minimum points as single-stroke data.

[0010] Optionally, the step of dividing each single-stroke data into four stages—fast forward, detection, fine stroke, and fast retraction—specifically includes: dividing each single-stroke data into four stages according to strain change characteristics: fast forward, detection, fine stroke, and fast retraction; wherein, the fast forward stage is the data segment where the strain increases from zero to the first maximum point, the detection stage is the data segment where the strain decreases from the first maximum point to the first minimum point, the fine stroke stage is the data segment where the strain increases from the first minimum point to the global maximum point, and the fast retraction stage is the data segment where the strain decreases from the global maximum point to zero.

[0011] Optionally, the time-domain feature parameters include at least five of the following: maximum value, minimum value, average value, standard deviation, kurtosis, skewness, impact factor, and margin factor.

[0012] Optionally, the fault warning steps specifically include: setting time thresholds for the four stages of rapid advance, detection, fine blanking, and rapid return; determining a rapid advance system fault when the time of the rapid advance stage exceeds a preset proportion of the time threshold; determining a detection system fault or excessive waste on the die surface when the time of the detection stage exceeds a preset proportion of the time threshold; determining a main blanking system fault when the time of the fine blanking stage exceeds a preset proportion of the time threshold; and determining a rapid return system fault when the time of the rapid return stage exceeds a preset proportion of the time threshold.

[0013] On the other hand, the present invention also provides a fine blanking equipment monitoring system based on fiber optic grating sensing and machine learning, comprising: a data acquisition module for deploying fiber optic grating sensors on the slider of the fine blanking machine to acquire strain data of the slider in real time; a preprocessing module for preprocessing the strain data to remove noise and nonlinear trends; a stage division module for splitting the strain data from multi-stroke data into single-stroke data based on minimum point detection, dividing each single-stroke data into four stages: fast advance, detection, fine blanking, and fast retraction, and saving them to an array; a dataset construction module for calculating the temporal feature parameters of each stage data in the array, setting classification labels for each of the four stages, and constructing a training dataset; a model training module for training a support vector machine classifier based on the training dataset to obtain a fine blanking stage recognition model; and a visualization module for creating a visual graphical interface to visually display the real-time acquired data and the fine blanking stage recognition results.

[0014] On the other hand, the present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described fine stamping equipment monitoring method based on fiber optic grating sensing and machine learning.

[0015] On the other hand, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described monitoring method for fine stamping equipment based on fiber optic grating sensing and machine learning.

[0016] Compared with the prior art, the beneficial effects of the present invention are:

[0017] This application establishes a complete closed-loop monitoring system by acquiring real-time slider strain data, intelligently dividing process stages, visual monitoring, and fault early warning, solving the problems of low efficiency and insufficient accuracy in traditional manual inspection. By integrating fiber optic sensing with machine learning, and establishing a quantifiable mapping relationship between physical signal characteristics and equipment operating status, the real-time identification accuracy of the fine blanking process stage is greatly improved, realizing the transformation of fine blanking equipment from passive maintenance to predictive maintenance. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the method steps of the present invention.

[0019] Figure 2 This is a time-domain plot of the slider strain data D1 acquired by the fiber optic grating sensor according to the present invention.

[0020] Figure 3 This is the time-domain plot of the original strain data D1 after denoising.

[0021] Figure 4 This is the time-domain graph after removing the nonlinear trend using the sliding local minimum trend fitting method of this invention.

[0022] Figure 5 This is a time-domain diagram of a single stroke that incorporates the phase division principle of this invention.

[0023] Figure 6 This is a schematic diagram of the visualization GUI for stage detection of the fine punching machine built based on MATLAB according to the present invention.

[0024] Figure 7 This is a schematic diagram of the system structure of the present invention.

[0025] In the diagram: 10 - Acquisition module, 20 - Preprocessing module, 30 - Stage division module, 40 - Dataset construction module, 50 - Model training module, 60 - Visualization module. Detailed Implementation

[0026] The present invention will now be clearly and completely described in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0027] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be used interchangeably where appropriate for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0028] Those skilled in the art will understand that, unless explicitly stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in the specification of this application means the presence of features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0029] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0030] It should be understood that the sequence number and size of each step in this embodiment do not imply the order of execution. The execution order of each process is determined by its function and internal logic, and should not constitute any limitation on the implementation process of this application embodiment.

[0031] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0032] Please refer to Figures 1-6 This invention discloses a monitoring method for fine blanking equipment based on fiber optic grating sensing and machine learning. The method includes the following steps: deploying fiber optic grating sensors on the slider of a fine blanking machine to collect strain data of the slider in real time; preprocessing the strain data to remove noise and nonlinear trends; splitting the strain data from multi-stroke data into single-stroke data based on minimum point detection, further dividing each single-stroke data into four stages: fast advance, detection, fine blanking, and fast retraction, and storing them in an array; calculating the temporal feature parameters of each stage in the array, and setting classification labels for each of the four stages to construct a training dataset; training a support vector machine classifier based on the training dataset to obtain a stage recognition model; and building a visualization interface to display the real-time data monitoring data of the stage recognition model and provide fault warnings.

[0033] Specifically, strain data D1 of the fine-brushing machine slider is acquired using a fiber Bragg grating data acquisition system. The fiber Bragg grating data acquisition system includes a fiber Bragg grating demodulator, a fiber Bragg grating sensor attached to the fine-brushing machine slider, and a computer. Data is acquired by the sensor and demodulator and transmitted to the computer via network cable for visualization. Since the acquired raw strain data contains a large amount of noise and DC components, denoising is necessary. Wavelet decomposition is performed on the strain data D1, and soft thresholding is applied to the decomposed data to obtain denoised strain data D2. Based on minimum point detection, the strain data is split from multi-stroke data into single-stroke data. Each single-stroke data is further divided into four stages: fast advance, detection, fine-brushing, and fast retraction. Classification labels are assigned to each stage to construct a training dataset. A support vector machine classifier is configured and trained based on this dataset, using a radial basis function as the kernel function. The trained classifier is saved in MATLAB. A graphical user interface (GUI) for fine-brushing stage recognition is created using MATLAB to visualize the real-time acquired data and recognition results.

[0034] This application establishes a complete closed-loop monitoring system by acquiring real-time slider strain data, intelligently dividing process stages, visual monitoring, and fault early warning, solving the problems of low efficiency and insufficient accuracy in traditional manual inspection. By integrating fiber optic sensing with machine learning, and establishing a quantifiable mapping relationship between physical signal characteristics and equipment operating status, the real-time identification accuracy of the fine blanking process stage is greatly improved, realizing the transformation of fine blanking equipment from passive maintenance to predictive maintenance.

[0035] In some embodiments, the preprocessing step of the strain data specifically includes: using the db4 wavelet basis function from the Daubechies wavelet family as the decomposition basis function to perform multi-scale wavelet decomposition on the strain data D1 to obtain a set of wavelet coefficients containing noise information; dividing the set of wavelet coefficients into multiple analysis windows on an average basis, and estimating the noise standard deviation using the median absolute deviation method; dynamically adjusting the weighting coefficients based on the stability coefficient of the window data to increase the weight of stable data in the window, and then calculating the weighted noise standard deviation; calculating the Donoho universal threshold based on the weighted noise standard deviation, and then performing soft thresholding on the wavelet coefficients, setting those below the threshold as noise to 0, and smoothing those above the threshold, and reconstructing the processed wavelet coefficients to obtain the denoised strain data D2.

[0036] Specifically, the db4 wavelet basis function from the Daubechies wavelet family is selected as the decomposition basis function to perform multi-scale wavelet decomposition on the strain data D1; the wavelet decomposition level is determined according to the signal sampling rate and frequency characteristics, and the decomposition is performed to obtain the wavelet coefficient set c; the standard deviation σ of the noise is estimated using the adaptive multi-window absolute bias method, which includes the following steps:

[0037] The high-frequency wavelet coefficients (cd) are averaged and divided into multiple windows. For each window wi, calculate the standard deviation of the noise based on the data. Where median is the median function;

[0038] According to the formula Calculate the stability coefficient of the data within the window, where The stability coefficient; This represents the variance of the data within the window; variance reflects the stability of the data. The maximum variance; This represents the window data corresponding to the maximum variance; ε is set to 10⁻⁶ to prevent the denominator from being 0.

[0039] The data from the blanking press slide is similar to an impact signal; therefore, more stable data should have a higher weight, while unstable data may contain real signals and should therefore have a lower weight. The weighting coefficients are calculated using the formula. It can adaptively adjust the weighting coefficients based on the stability of wavelet coefficients, thereby more accurately estimating the standard deviation of noise; according to the formula Calculate the standard deviation σ of the noise after weighted fusion.

[0040] Based on the weighted noise standard deviation σ, the Donoho universal threshold formula is used. Calculate the threshold, where: The threshold is calculated, and N is the length of the original data D1. Then, the wavelet coefficients are subjected to soft thresholding.

[0041] The wavelet coefficients are reconstructed to obtain the denoised data D2.

[0042] This application decomposes a signal into multiple frequency bands using the db4 wavelet. The decomposition level is determined based on the sampling frequency and the frequency characteristics of the fine-bearing machine. Specifically, decomposition is performed based on the low-frequency components of the signal (i.e., the operating frequency of the fine-bearing machine), and the final decomposition level completely includes the operating frequency. Too high a decomposition level may lead to signal feature loss, while too low a level results in poor denoising. To further prevent signal distortion due to excessively high decomposition levels and poor denoising due to excessively low levels, a stability coefficient is introduced as a weight. More stable data receives a larger weight, while unstable data may contain the true signal and therefore should have a smaller weight. The noise standard deviation is then estimated based on the weights. Soft processing is applied to the decomposed wavelet coefficients to achieve denoising. First, a threshold is calculated based on the signal length and the noise standard deviation. Values ​​below the threshold are treated as noise and set to 0, while values ​​above the threshold are smoothed. After threshold processing, wavelet coefficients are reconstructed, ultimately achieving smooth denoising. By adaptively adjusting the noise threshold through multi-window stability assessment, the abrupt change characteristics of the strain signal can still be effectively preserved even in a strong electromagnetic interference environment. Compared with the traditional fixed threshold method, the signal-to-noise ratio is significantly improved, laying a data foundation for the subsequent accurate division of process stages.

[0043] In some embodiments, the step of preprocessing the strain data further includes: setting the window size of the strain data D2 to three stroke data lengths, using the local minimum values ​​within each window as sliding indices to form a trend term array, removing the trend term from the strain data D2, and obtaining strain data D3 with nonlinear trends removed.

[0044] Specifically, the window size is set according to the sampling frequency and data characteristics. The window size is set to W=3×s, which means the window contains data from 3 strokes, where s is the number of data points in a single stroke. Setting the window too large will result in the inability to completely remove nonlinear trends, while setting the window too small will increase the computation time.

[0045] The index starts with the local minimum value within the initial window, then slides to the next group of adjacent data until the end of the data, obtaining the trend term to be eliminated, I=min(W), and saving the obtained trend term as an array T.

[0046] According to the formula D3=D2-T, the data D3 after removing the non-linear trend is obtained.

[0047] The sliding window trend elimination method designed in this application overcomes the problem of data drift in long-term monitoring. Based on the local minimum fitting of the 3-stroke window, it eliminates nonlinear baseline interference while maintaining the original waveform characteristics, making the data collected at different time periods comparable. This technology is particularly suitable for dealing with signal distortion caused by pressure fluctuations in hydraulic systems.

[0048] In some embodiments, the step of splitting the strain data from multi-stroke data into single-stroke data based on minimum point detection specifically includes: setting a minimum interval between two minimum points in the strain data to ensure that a complete stroke can be split out, indexing the minimum points, and taking the data between two adjacent minimum points as single-stroke data.

[0049] Specifically, the extreme point splitting mechanism proposed in this application makes full use of the periodic motion characteristics of the fine-stamping machine, ensures the integrity of the stroke by constraining the minimum interval, transforms the mechanical motion characteristics into data segmentation rules, realizes the deep coupling of physical process and data processing, and provides a standardized data foundation for single-stroke analysis.

[0050] In some embodiments, the step of dividing each single-stroke data into four stages—fast forward, detection, fine stroke, and fast return—specifically includes: dividing each single-stroke data into four stages according to strain change characteristics: fast forward, detection, fine stroke, and fast return; wherein, the fast forward stage is the data segment where the strain increases from zero to a first maximum point, the detection stage is the data segment where the strain decreases from the first maximum point to a first minimum point, the fine stroke stage is the data segment where the strain increases from the first minimum point to a global maximum point, and the fast return stage is the data segment where the strain decreases from the global maximum point to zero.

[0051] Specifically, each single-stroke data in array C1 is split into four stages: rapid advance, detection, fine stroke, and rapid retraction, and then stored in array C2. The specific splitting rules based on sensor data characteristics and the working characteristics of the fine stroke machine are as follows:

[0052] ;

[0053] In the rapid traverse phase s1(t), the slider moves rapidly upward due to the pressure of the rapid cylinder, and the strain measured by the sensor increases rapidly from 0 to the first strain maximum point s1max. Therefore, the data between the starting point (0, 0) and the strain maximum point (t1, s1max) corresponds to the rapid traverse phase of fine blanking. In the detection phase s2(t), the fine blanking machine automatically performs a system check. During this phase, the slider hardly moves upward, and the strain in the grating area decreases rapidly to the minimum point s2min. Therefore, the data between (t1, s1max) and (t2, s2min) corresponds to the detection phase of fine blanking. In the stroke stage s3(t), the slider is subjected to a large pressure by the main cylinder for fine stroke, which causes the grid strain to increase rapidly to the global strain maximum value s3max. Therefore, the data from (t2, s2min) to the global strain maximum point (t3, smax) corresponds to the fine stroke stage of fine stroke. In the fast return stage s4(t), the grid strain decreases slowly due to the pre-pressure relief effect. Then, the slider is also subjected to a large speed by the fast cylinder, which causes the grid strain to decrease rapidly to 0. Therefore, the data from (t3, smax1) to the end of the stroke (0, 0) corresponds to the fast return stage of fine stroke.

[0054] This application uses mathematical modeling to analyze the dynamic strain characteristics of the fine blanking machine's operation, precisely dividing single-stroke data into four physically distinct stages: rapid advance, inspection, fine blanking, and rapid retraction. Based on a composite criterion of strain change rate (ds / dt) and extreme points (s1max, s2min, smax), a data decomposition rule strictly corresponding to the equipment's mechanical motion characteristics is constructed. This not only enables quantifiable identification of process stages but also provides characteristic data with clear engineering semantics for subsequent fault diagnosis. The acceleration characteristics of the rapid advance stage, the system stability of the inspection stage, the pressure peak of the fine blanking stage, and the pressure relief characteristics of the rapid retraction stage are all intuitively represented through standardized data intervals, thus establishing a mapping relationship between sensor signals and equipment operating status, laying a data foundation for intelligent monitoring of the fine blanking process.

[0055] In some embodiments, the time-domain feature parameters include at least five of the following: maximum value, minimum value, average value, standard deviation, kurtosis, skewness, impact factor, and margin factor.

[0056] Specifically, the signal's characteristics are fully extracted from four aspects: basic statistical features of the signal (maximum, minimum, and average), signal fluctuation features (standard deviation), distribution morphology features (kurtosis and skewness), and impact sensitivity features (impact factor and margin factor) to ensure accurate identification of the fine-tuning stage and improve the identification accuracy.

[0057] In some embodiments, the fault warning steps specifically include: setting time thresholds for four stages: rapid advance, detection, fine blanking, and rapid return; determining a rapid advance system fault when the time of the rapid advance stage exceeds a preset proportion of the time threshold; determining a detection system fault or excessive waste on the die surface when the time of the detection stage exceeds a preset proportion of the time threshold; determining a main blanking system fault when the time of the fine blanking stage exceeds a preset proportion of the time threshold; and determining a rapid return system fault when the time of the rapid return stage exceeds a preset proportion of the time threshold.

[0058] Specifically, a threshold is set for the time of each stage. When the time of the fast advance, inspection, fine blanking, and fast return stages is greater than or less than 10% of the normal time, it indicates that the fast advance system is faulty, the inspection system is faulty, or there is too much scrap on the mold surface, the main blanking system is faulty, and the fast return system is faulty, respectively.

[0059] The tiered early warning mechanism set up in this application links the time threshold with the health status of the equipment. By setting a reasonable range of ±10% to balance sensitivity and false alarm rate, a quantifiable fault judgment standard is established, so that different types of problems such as fast-forward system failure and mold waste accumulation can receive differentiated early warnings.

[0060] Furthermore, Figure 6 The upper center plotting area dynamically displays the slider strain curve, with the horizontal axis representing time (each data point representing 0.01 seconds) and the vertical axis representing micro-strain. This graphics window displays the slider's trajectory in real time, allowing observation of the stage changes during the fine blanking process. "Load Data" is used to input the slider strain data acquired in real time by the fiber Bragg grating sensor. After clicking "Start Recognition," the corresponding stage button turns gray after recognition. The running time is displayed below each stage after completion, and the time for each stage is updated after the next stroke. When the running time of any stage exceeds the set threshold, "YES" is displayed next to "Fault?", otherwise "NO" is displayed.

[0061] Please refer to Figure 7On the other hand, the present invention also provides a fine-bending equipment monitoring system based on fiber optic grating sensing and machine learning, comprising: a data acquisition module for deploying fiber optic grating sensors on the slider of the fine-bending machine to acquire strain data of the slider in real time; a preprocessing module for preprocessing the strain data to remove noise and nonlinear trends; a stage division module for splitting the strain data from multi-stroke data into single-stroke data based on minimum point detection, dividing each single-stroke data into four stages: fast advance, detection, fine-bending, and fast retraction, and saving them to an array; a dataset construction module for calculating the temporal feature parameters of each stage data in the array, setting classification labels for each of the four stages, and constructing a training dataset; a model training module for training a support vector machine classifier based on the training dataset to obtain a fine-bending stage recognition model; and a visualization module for creating a visual graphical interface to visually display the real-time acquired data and the fine-bending stage recognition results.

[0062] On the other hand, the present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described fine stamping equipment monitoring method based on fiber optic grating sensing and machine learning.

[0063] On the other hand, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described monitoring method for fine stamping equipment based on fiber optic grating sensing and machine learning.

[0064] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0065] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, database, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0066] The above are merely embodiments of the present invention and do not limit the patent scope of the present invention. Any equivalent modifications made based on the content of the present invention's specification and drawings, or direct or indirect applications in related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A monitoring method for fine stamping equipment based on fiber optic grating sensing and machine learning, characterized in that the steps include... include: A fiber optic grating sensor is installed on the slider of the fine blanking machine to collect the strain data of the slider in real time. The strain data is preprocessed, specifically including: The db4 wavelet basis function in the Daubechies wavelet family was selected as the decomposition basis function to perform multi-scale wavelet decomposition on the strain data D1, and a set of wavelet coefficients containing noise information was obtained. The wavelet coefficient set is divided into multiple analysis windows on average, and the noise standard deviation is estimated by the median absolute deviation method. Calculate the stability coefficient of the window data, and dynamically adjust the weighting coefficient based on the stability coefficient of the window data to increase the weight of stable data in the window and give less weight to unstable data, and then calculate the weighted noise standard deviation. The Donoho universal threshold is calculated based on the weighted noise standard deviation, and then the wavelet coefficients are subjected to soft thresholding. Those below the threshold are set to 0 as noise, and those above the threshold are smoothed. The processed wavelet coefficients are reconstructed to obtain the denoised strain data D2. The window size of strain data D2 is set to the length of 3 strokes. The local minimum values ​​in each window are used to form a trend term array. The trend term is removed from strain data D2 to obtain strain data D3 with nonlinear trend removed, thereby removing noise and nonlinear trend from the strain data. Based on the detection of the minimum point, the strain data is split from multi-stroke data into single-stroke data. Each single-stroke data is divided into four stages: fast advance, detection, fine stroke, and fast retraction, and saved to an array. Calculate the temporal feature parameters of the data in each stage of the array, and set classification labels for the four stages to construct a training dataset; A support vector machine classifier is trained based on the training dataset to obtain a stage recognition model; A visual interface is built to display real-time data monitoring data of the identification model during the stage, and to provide fault warnings for the fast forward system, detection system, main punching system, and fast retraction system.

2. The monitoring method for fine stamping equipment based on fiber optic grating sensing and machine learning according to claim 1, characterized in that, The step of splitting the strain data from multi-stroke data into single-stroke data based on minimum point detection specifically includes: In the strain data, a minimum interval between two minimum points is set to ensure that the complete stroke can be separated. The minimum points are indexed, and the data between two adjacent minimum points is taken as single-stroke data.

3. The monitoring method for fine stamping equipment based on fiber optic grating sensing and machine learning according to claim 1, characterized in that, The step of dividing each single-stroke data into four stages—rapid advance, detection, fine stroke, and rapid rewind—specifically includes: Each single-stroke data point is divided into four stages based on strain variation characteristics: rapid advance, detection, fine stroke, and rapid retraction. The fast forward phase consists of a data segment in which the strain increases from zero to the first maximum point; the detection phase consists of a data segment in which the strain decreases from the first maximum point to the first minimum point; the fine pressing phase consists of a data segment in which the strain increases from the first minimum point to the global maximum point; and the fast retraction phase consists of a data segment in which the strain decreases from the global maximum point to zero.

4. The monitoring method for fine stamping equipment based on fiber optic grating sensing and machine learning according to claim 1, characterized in that, The time-domain feature parameters include at least five of the following: maximum value, minimum value, average value, standard deviation, kurtosis, skewness, impact factor, and margin factor.

5. The monitoring method for fine stamping equipment based on fiber optic grating sensing and machine learning according to claim 1, characterized in that, The specific steps of the fault early warning include: Set time thresholds for the four stages: fast forward, detection, fine punching, and fast rewind; When the time of the fast-forward phase exceeds a preset proportion of the time threshold, the fast-forward system is deemed to be malfunctioning. When the detection time exceeds the preset proportion of the time threshold, it is determined that the detection system is faulty or there is too much waste on the mold surface. When the time of the fine blanking stage exceeds the preset proportion of the time threshold, the main blanking system is judged to be faulty. When the time of the fast rewind phase exceeds a preset proportion of the time threshold, the fast rewind system is judged to be faulty.

6. A monitoring system for fine stamping equipment based on fiber optic grating sensing and machine learning, characterized in that, include: The acquisition module is used to deploy fiber optic grating sensors on the slider of the fine punching machine to acquire the strain data of the slider in real time. The preprocessing module is used to preprocess the strain data, specifically including: The db4 wavelet basis function in the Daubechies wavelet family was selected as the decomposition basis function to perform multi-scale wavelet decomposition on the strain data D1, and a set of wavelet coefficients containing noise information was obtained. The wavelet coefficient set is divided into multiple analysis windows on average, and the noise standard deviation is estimated by the median absolute deviation method. Calculate the stability coefficient of the window data, and dynamically adjust the weighting coefficient based on the stability coefficient of the window data to increase the weight of stable data in the window and give less weight to unstable data, and then calculate the weighted noise standard deviation. The Donoho universal threshold is calculated based on the weighted noise standard deviation, and then the wavelet coefficients are subjected to soft thresholding. Those below the threshold are set to 0 as noise, and those above the threshold are smoothed. The processed wavelet coefficients are reconstructed to obtain the denoised strain data D2. The window size of strain data D2 is set to the length of 3 strokes. The local minimum values ​​in each window are used to form a trend term array. The trend term is removed from strain data D2 to obtain strain data D3 with nonlinear trend removed, thereby removing noise and nonlinear trend from the strain data. The stage division module is used to split the strain data from multi-stroke data into single-stroke data based on the detection of the minimum value point, and to divide each single-stroke data into four stages of data: fast advance, detection, fine stroke and fast retreat, and save them to an array. The dataset construction module is used to calculate the temporal feature parameters of the data at each stage in the array, and to set classification labels for the four stages to build the training dataset. The model training module is used to train a support vector machine classifier based on the training dataset to obtain a fine-bending stage recognition model. The visualization module is used to create a visual graphical interface to display real-time collected data and the identification results of the fine blanking stage, and to provide early warnings for faults in the fast forward system, detection system, main blanking system, and fast retraction system.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the fine stamping equipment monitoring method based on fiber optic grating sensing and machine learning as described in any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the fine stamping equipment monitoring method based on fiber optic grating sensing and machine learning as described in any one of claims 1 to 5.

Images