A method for analyzing the wear state of a micro drill tool of ultrafine grain cemented carbide

Abstract

The application provides a kind of super fine grain hard metal micro drill tool wear state analysis method, comprising: obtaining the original vibration signal in the process of super fine grain hard metal micro drill tool machining, the vibration frequency spectrum is obtained by frequency domain conversion to the original vibration signal by fast fourier transform, high frequency band energy and mid frequency band energy are extracted from the vibration frequency spectrum respectively;From the signal segment of the high-frequency surge stage, the time-domain amplitude distribution reflecting the severity of the blade impact is extracted, and the waveform kurtosis of the time-domain amplitude distribution is calculated;For the serious wear stage, the blade edge image of the micro drill tool is obtained, the edge detection algorithm is used to process the blade edge image to obtain the blade edge contour, and the number of discrete spots is extracted from the blade edge contour.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a method for analyzing the wear state of ultrafine-grained cemented carbide micro-drilling tools. Background Technology

[0002] Ultrafine-grained cemented carbide micro-drills play a crucial role in precision manufacturing. These tools typically have diameters ranging from 0.3 mm to 3 mm, and even minute changes in edge sharpness and geometric precision directly impact the accuracy of machined holes, surface roughness, and hole wall integrity. Accurate monitoring of their wear is a core element in ensuring machining quality. Current methods for monitoring tool wear largely rely on absolute threshold divisions of vibration signal frequency bands. This approach ignores the dynamic evolution of the tool-material contact state during machining, easily leading to misjudgments in complex cutting environments. In the wear process of micro-drills, there is a highly deceptive contradiction between the vibration spectrum energy distribution and the evolution of the edge wear morphology. Conventionally, when the proportion of high-frequency energy in the vibration signal gradually increases and exceeds the mid-frequency band, it indicates that the impact and collision between the tool edge and the workpiece is intensifying. This surge in high-frequency energy causes the wear area of ​​the edge to break down from a smooth, continuous distribution into a discrete, speckled distribution, which is then directly interpreted as severely accelerated tool wear. However, in actual drilling operations, this change in energy distribution does not always signify tool failure. For example, in the early stages of a tool's use, the initial break-in of the microstructure of the cutting edge will also cause a significant increase in the proportion of high-frequency energy. If, at this point, the tool is deemed severely worn and the machine is forcibly stopped simply because the high-frequency energy surpasses the mid-frequency energy, it will result in enormous waste. This method of assessment based solely on the proportion of frequency band energy masks the differences in the kurtosis value of the vibration waveform, which reflects the sharpness of the signal impact, at different stages of wear, making it impossible to accurately distinguish between early normal break-in and later severe wear. Therefore, how to comprehensively consider the changes in the proportion of high-frequency and mid-frequency energy during drilling, and combine this with the evolution trend of the vibration waveform kurtosis value to clarify the contradiction between energy distribution and wear morphology, and accurately identify the true stage of the wear area's evolution from continuous to discrete spots, has become a key issue in achieving accurate determination of the wear state of micro-drill tools. Summary of the Invention

[0003] This invention provides a method for analyzing the wear state of ultrafine-grained cemented carbide micro-drilling tools, the method comprising: The original vibration signal during the machining process of ultrafine-grained cemented carbide micro-drilling tool is acquired, and the vibration spectrum is obtained by frequency domain transformation of the original vibration signal through fast Fourier transform. High-frequency energy and mid-frequency energy are extracted from the vibration spectrum respectively. The high-frequency energy ratio is obtained based on the high-frequency energy and the mid-frequency energy. When the high-frequency energy ratio exceeds a preset threshold, it is determined that the micro-drilling tool has entered the high-frequency surge stage, and a signal segment containing the high-frequency surge stage is extracted. Extract the time-domain amplitude distribution reflecting the impact intensity of the cutting edge from the signal segment, and calculate the waveform kurtosis of the time-domain amplitude distribution; When the waveform kurtosis is lower than the preset kurtosis threshold, it is confirmed that the micro-drilling tool is in the early break-in stage; when the waveform kurtosis is higher than or reaches the preset kurtosis threshold, it is confirmed that the micro-drilling tool is in the severe wear stage. For the severe wear stage, the cutting edge image of the micro-drill tool is acquired, and the cutting edge image is processed by an edge detection algorithm to obtain the cutting edge contour. The number of discrete spots is extracted from the cutting edge contour. When the number of discrete spots exceeds a preset spot threshold, the edge wear area is determined to change from a continuous distribution to a discrete spot distribution, thereby obtaining the target wear state of the micro-drill tool.

[0004] Furthermore, acquiring the original vibration signal during the machining process of the ultrafine-grained cemented carbide micro-drill tool includes: An accelerometer is attached to the vibration pickup position at the spindle clamping end of the micro-drilling tool, and the original vibration signal of the drilling process is continuously collected according to a preset sampling rate. The original vibration signal is denoised by bandpass filtering to obtain a clean vibration signal. The signal is then segmented into frames according to a fixed time window to obtain time-domain waveform segments arranged frame by frame.

[0005] Furthermore, the step of performing frequency domain transformation on the original vibration signal using Fast Fourier Transform to obtain the vibration spectrum, and extracting high-frequency energy and mid-frequency energy from the vibration spectrum, includes: For the time-domain waveform segments obtained by framing the original vibration signal, a Hanning window is applied to each frame to weight and attenuate the edge samples, resulting in a windowed waveform. The windowed waveform is then converted to the frequency domain using a fast Fourier transform to obtain the vibration spectrum. The vibration spectrum is then divided into frequency bands according to the frequency boundary between the cutting resonance band and the impact response band, resulting in two continuous frequency intervals: the mid-frequency band and the high-frequency band. For the two continuous frequency intervals of the mid-frequency band and the high-frequency band, the square of the amplitude of the vibration spectrum is discretely accumulated and summed along the frequency axis within the interval to obtain two numerical indicators: the mid-frequency band energy and the high-frequency band energy. The mid-frequency band energy and the high-frequency band energy are aligned and arranged according to the same sampling time to obtain a frequency band energy sequence arranged frame by frame.

[0006] Furthermore, obtaining the high-frequency energy ratio based on the high-frequency energy and the mid-frequency energy includes: The high-frequency energy ratio is obtained by dividing the high-frequency energy by the sum of the high-frequency energy and the mid-frequency energy. A high-frequency energy ratio sequence is formed along the time axis. The high-frequency energy ratio sequence is then smoothed by a moving average to obtain a smoothed ratio sequence.

[0007] Furthermore, the interception of the signal segment containing the high-frequency surge phase includes: The high-frequency energy ratio is compared with the preset threshold. Sampling times that exceed the preset threshold are marked with an over-limit mark to form an over-limit mark sequence. Short-term isolated over-limit segments are screened out according to a preset minimum duration threshold. The first moment of continuous over-limit segments is retained as the surge start point and the last moment is retained as the surge end point. A preset transition margin is extended outward to form a complete time window. Segments are extracted from the original vibration signal along the corresponding position of the complete time window.

[0008] Furthermore, the step of extracting the time-domain amplitude distribution reflecting the impact intensity of the cutting edge from the signal segment and calculating the waveform kurtosis of the time-domain amplitude distribution includes: The instantaneous vibration amplitude of each sampling point is collected point by point along the time axis to obtain the time-domain amplitude distribution. The mean value of all sampling points of the time-domain amplitude distribution is subtracted and then divided by the standard deviation to obtain the normalized amplitude sequence. The fourth power mean of each value of the normalized amplitude sequence is calculated to obtain the waveform kurtosis.

[0009] Furthermore, when the waveform kurtosis is lower than a preset kurtosis threshold, the micro-drilling tool is confirmed to be in the early break-in stage; when the waveform kurtosis is higher than or reaches the preset kurtosis threshold, the micro-drilling tool is confirmed to be in the severe wear stage, including: For the typical specifications of the micro-drill tool and the workpiece material, the preset kurtosis threshold is calibrated based on the impact level reflected by the time-domain amplitude distribution when the cutting edge is in stable contact during historical drilling processes.

[0010] Furthermore, for the severe wear stage, the cutting edge image of the micro-drill tool is acquired, and an edge detection algorithm is used to process the cutting edge image to obtain the cutting edge contour. The number of discrete spots is extracted from the cutting edge contour, including: The cutting edge image is obtained by acquiring the cutting edge end face of the micro-drill using an industrial microscope camera. The cutting edge image is then subjected to grayscale conversion and Gaussian filtering smoothing to obtain a pure grayscale image. The edge detection algorithm is then used to identify the contour boundaries of pixel grayscale abrupt changes in the pure grayscale image to obtain the cutting edge contour. The internal region of the cutting edge contour is marked with connected regions according to pixel adjacency to obtain a set of connected regions composed of independent closed blocks. Morphological opening and closing operations are performed on the set of connected regions to obtain a modified set of connected regions. Artifact spots below the lower limit and continuous wear main blocks above the upper limit are filtered out according to a preset lower limit and upper limit, and the target connected regions between the two thresholds are retained. Each independent closed block in the target connected region is counted and accumulated one by one to obtain the number of discrete spots.

[0011] Furthermore, when the number of discrete spots exceeds a preset spot threshold, determining that the edge wear area shifts from a continuous distribution to a discrete spot distribution includes: For the cutting edge geometry of the micro-drill, a preset spot threshold is established in advance. The preset spot threshold is calibrated based on the statistical upper limit of the number of discrete spots in the microscopic field of view when the cutting edge wear area is in a continuous distribution pattern.

[0012] Furthermore, obtaining the target wear state of the micro-drill tool includes: The number of discrete spots is compared with the preset spot threshold to obtain the deviation direction of the number of discrete spots relative to the preset spot threshold. When the number of discrete spots exceeds the preset spot threshold, the target wear state is recorded as discrete spot wear state, with the number of discrete spots attached as a quantitative description of the target wear state.

[0013] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention discloses a method for analyzing the wear state of ultrafine-grained cemented carbide micro-drilling tools. Addressing the difficulty in accurately determining the wear state of micro-drilling tools during machining, this method achieves precise identification by combining frequency domain analysis of vibration signals with time domain feature extraction. First, the method performs a Fast Fourier Transform (FFT) on the original vibration signal to obtain the vibration spectrum. By calculating the proportion of high-frequency energy, it identifies whether the tool has entered a high-frequency surge stage, thus capturing the key turning point in the tool's state. Subsequently, it extracts the time-domain amplitude distribution from the high-frequency surge signal segment and calculates the waveform kurtosis. The kurtosis value is compared with a preset threshold to distinguish between the early break-in stage and the severe wear stage, solving the problem of traditional methods failing to accurately distinguish the degree of wear. After confirming severe wear, this invention further processes the cutting edge image using an edge detection algorithm to extract contour features and counts the number of discrete spots to determine whether the wear area has shifted from a continuous distribution to a discrete spot distribution, ultimately accurately determining the target wear state of the tool. This method organically integrates vibration signal analysis and visual inspection, achieving dynamic monitoring and precise determination of the wear state of micro-drilling tools throughout the entire process. Attached Figure Description

[0014] Figure 1This is a flowchart of a method for analyzing the wear state of ultrafine-grained cemented carbide micro-drilling tools according to the present invention.

[0015] Figure 2 This is a schematic diagram of a method for analyzing the wear state of ultrafine-grained cemented carbide micro-drilling tools according to the present invention.

[0016] Figure 3 This is another schematic diagram of the wear state analysis method for ultrafine-grained cemented carbide micro-drilling tools according to the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0018] like Figures 1-3 This embodiment of a method for analyzing the wear state of ultrafine-grained cemented carbide micro-drilling tools may specifically include: Step S101: Obtain the original vibration signal during the machining process of the ultrafine-grained cemented carbide micro-drilling tool, and perform frequency domain transformation on the original vibration signal through fast Fourier transform to obtain the vibration spectrum. Extract the high-frequency energy and mid-frequency energy from the vibration spectrum respectively.

[0019] An accelerometer is attached to the vibration pickup position at the spindle clamping end of the micro-drilling tool, and the original vibration signal during the drilling process is continuously acquired according to a preset sampling rate. To address the low-frequency drift components introduced by power frequency interference and machine tool base resonance, bandpass filtering is used to denoise the original vibration signal, resulting in a clean vibration signal free of background disturbances. The clean vibration signal is then segmented into frames according to a fixed time window length, resulting in time-domain waveform segments arranged frame by frame. For each time-domain waveform segment, a Hanning window is applied to weighted attenuate edge samples, resulting in a windowed waveform that suppresses spectral leakage. A fast Fourier transform is used to perform frequency domain transformation on the windowed waveform to obtain the vibration spectrum. The vibration spectrum describes the frequency composition of the cutting edge vibration energy in the form of a continuous distribution of amplitude with frequency. Based on the empirically defined frequency boundary between the cutting resonance band and the impact response band, the vibration spectrum is divided into frequency bands, resulting in two continuous frequency intervals: the mid-frequency band and the high-frequency band. For the two continuous frequency intervals of the mid-frequency band and the high-frequency band, the square of the amplitude of the vibration spectrum is discretely accumulated and summed along the frequency axis within each interval to obtain two numerical indicators: the mid-frequency band energy and the high-frequency band energy. The mid-frequency band energy reflects the cumulative vibration generated by continuous friction between the cutting edge and the substrate during stable cutting, and the high-frequency band energy reflects the cumulative vibration generated by transient impact collision between the cutting edge and the substrate. The mid-frequency band energy and the high-frequency band energy are aligned and arranged according to the same sampling time to obtain a frequency band energy sequence arranged frame by frame.

[0020] In the machining process of ultrafine-grained cemented carbide micro-drills, the acquisition of vibration signals is a fundamental step in subsequent wear condition determination. The diameter of the micro-drill is in the range of 0.3mm to 3mm, and the cutting edge geometry is extremely small. The vibration energy generated during drilling is mainly transmitted to the clamping end through the spindle system.

[0021] In one embodiment, a piezoelectric accelerometer is attached to the spindle clamping end near the tool holder using a magnetic mount or adhesive. The vibration pickup direction is perpendicular to the spindle axis, and the original vibration signal during the drilling process is continuously acquired according to a preset sampling rate. For the drilling process, the upper limit of the frequency band of interest is typically set to 5000 Hz to 10000 Hz, and the sampling rate should be more than twice the upper limit of the frequency band of interest to satisfy the sampling theorem; typically, 20000 Hz to 25000 Hz is chosen.

[0022] It should be noted that the original vibration signal contains various interference components, including low-frequency vibrations generated by spindle rotation, machine tool bed resonance, power frequency interference from the electrical system, structural noise from the sensor mounting location, and random environmental noise. Signal preprocessing is required to extract effective drilling vibration characteristics. Power frequency interference introduced by the machine tool spindle motor drive is generally concentrated at 50Hz or 60Hz and its harmonics, exhibiting significant peaks at frequencies such as 50Hz, 100Hz, 150Hz, or 60Hz, 120Hz, 180Hz. The low-frequency drift components generated by the machine tool base and fixture structure under cutting force excitation are concentrated in the 0 to 50Hz frequency range. These components are unrelated to the vibration characteristics reflected by cutting edge wear.

[0023] Specifically, when using bandpass filtering to denoise the original vibration signal, the lower passband is set to 5 Hz to 20 Hz to filter out low-frequency drift components and environmental vibration interference below this frequency. The upper passband is set to 8000 Hz to 12000 Hz, corresponding to the upper limit of the effective response frequency range of the accelerometer. After filtering, the mid-to-high frequency vibration components related to the cutting edge contact state are retained, resulting in a pure vibration signal free of background disturbances. The pure vibration signal is then segmented into frames according to a fixed time window. The fixed time window is determined based on the vibration signal sampling frequency and the duration of the cutting edge contact process, specifically ranging from 0.05 seconds to 0.2 seconds, preferably 0.1 seconds. This ensures that a single frame of data can completely contain the vibration characteristics of the cutting edge contact without mixing in signals from multiple contact events due to an excessively large window length, resulting in time-domain waveform segments arranged frame by frame. Adjacent time-domain segments are overlapped by 50% to 75%, with an overlap length of 0.5L to 0.75L, where L is a time window length of 0.05 to 0.2 seconds, to avoid loss of transient information at frame boundaries. A Hanning window is applied to each time-domain segment to weighted attenuate edge samples and suppress spectral leakage. The window function w(n) of the Hanning window has a cosine square distribution, with a weight of 1 at the center of the window, gradually decreasing to 0 towards both ends. After multiplying point by point with the time-domain segment x(n), the windowed waveform y(n) = w(n)·x(n) is obtained. The edge samples of the windowed waveform are suppressed, and the sidelobe energy caused by truncation in the frequency domain is effectively suppressed. A fast Fourier transform is used to perform frequency domain transformation on the windowed waveform to obtain the vibration spectrum. The vibration spectrum describes the frequency composition of the blade vibration energy in the form of a continuous distribution of amplitude A with frequency f. The horizontal axis corresponds to discrete frequency points, and the vertical axis corresponds to the magnitude of the amplitude at each frequency point. The basis for frequency band division needs further explanation. When micro-drilling tools drill typical workpieces such as printed circuit board substrates and thin-walled aluminum alloy parts, the vibration energy generated by the continuous friction between the cutting edge and the substrate during the stable cutting stage exhibits an energy accumulation band in the frequency spectrum related to the tool's inherent cutting resonance, which is called the cutting resonance band. When the cutting edge suffers microscopic damage or undergoes transient impact collision with hard particles, the vibration energy exhibits a broadband dispersion pattern in a higher frequency range, which is called the impact response band.

[0024] Preferably, the frequency boundary between the cutting resonance band and the impact response band is determined based on the spindle speed. The value of the frequency boundary is 12 to 20 times the tooth passing frequency obtained by multiplying the spindle speed fundamental frequency (spindle speed ÷ 60) by the number of teeth. For example, when the spindle speed is 3000 rpm and the number of tool teeth is 4, the fundamental frequency is 50 Hz and the tooth passing frequency is 200 Hz. The frequency boundary value ranges from 2400 Hz to 4000 Hz. A specific value is selected within this range based on the tool specifications and workpiece material. The upper limit is taken when the tool diameter is large or the workpiece material is hard, and the lower limit is taken otherwise. The vibration spectrum is divided into frequency bands to obtain two continuous frequency intervals: the mid-frequency band and the high-frequency band. For the two continuous frequency intervals of the mid-frequency band and the high-frequency band, the squared amplitude of the vibration spectrum is discretely accumulated and summed along the frequency axis within each interval to obtain two numerical indicators: the mid-frequency band energy and the high-frequency band energy. The calculation formula is E=ΣA², where E represents the energy of the frequency band, A represents the amplitude at each discrete frequency point within the frequency band, and Σ represents the summation of all frequency points within the frequency band. Both the mid-frequency band and the high-frequency band are calculated using the same non-normalized method to ensure that the subsequent high-frequency energy ratio R=H / (H+M) is the true ratio of the energies of the two frequency bands.

[0025] For example, the energy integral is performed in the form E=ΣA², where E represents the energy value of the corresponding frequency band, A represents the amplitude at each discrete frequency point within the frequency band, the subscript of the summation symbol Σ is i=k1, and the superscript is k2, where k1 represents the index of the discrete frequency point corresponding to the starting frequency of the frequency band, and k2 represents the index of the discrete frequency point corresponding to the ending frequency of the frequency band. The summation range is limited to the frequency boundaries of the frequency band. The mid-frequency energy reflects the cumulative vibration generated by continuous friction between the cutting edge and the substrate during stable cutting, and the result of summing the squares of the amplitudes corresponds to the total energy contribution of the frequency band. The high-frequency energy reflects the cumulative vibration generated by transient impact collisions between the cutting edge and the substrate, and its value increases with the intensity of impact events. The mid-frequency energy and the high-frequency energy are aligned and arranged according to the same sampling time to obtain a frequency band energy sequence arranged frame by frame. The frequency band energy sequence records the synchronous evolution of energy in the two frequency intervals throughout the drilling process with time as the sequence axis, providing data support for the subsequent evaluation of the proportion of high-frequency energy.

[0026] Step S102: Obtain the high-frequency energy ratio based on the high-frequency energy and the mid-frequency energy, evaluate whether the high-frequency energy ratio exceeds a preset threshold, and determine that the micro-drilling tool has entered the high-frequency surge stage when the high-frequency energy ratio exceeds the preset threshold, and extract the signal segment containing the high-frequency surge stage.

[0027] For the high-frequency energy and the mid-frequency energy, the high-frequency energy ratio is obtained by dividing the high-frequency energy by the sum of the two. A high-frequency energy ratio sequence is formed along the time axis. A moving average is used to smooth the high-frequency energy ratio sequence to obtain a smoothed ratio sequence. The smoothed ratio sequence is compared with a preset threshold. If the value at a certain sampling time exceeds the preset threshold, an over-limit mark is marked to form an over-limit mark sequence. For continuous over-limit segments in the over-limit mark sequence, short-term isolated over-limit segments are filtered out according to a preset minimum duration threshold. The first moment of the segment is retained as the surge start point and the last moment is retained as the surge end point to determine that the micro-drilling tool has entered the high-frequency surge stage. A continuous time interval is determined according to the surge start point and the surge end point. A preset transition margin is extended outward from both ends of the interval to form a complete time window. A segment is extracted from the original vibration signal along the corresponding position of the complete time window to obtain a signal segment containing the high-frequency surge stage.

[0028] Obtaining the proportion of high-frequency energy is the core entry point for determining the evolution of the cutting state of the micro-drill tool.

[0029] In one embodiment, based on the high-frequency energy and the mid-frequency energy obtained in the previous stage, normalization is performed by dividing the high-frequency energy by the sum of the two, i.e., the high-frequency energy proportion at the current sampling moment is obtained using the relationship R=H / (H+M), where R represents the high-frequency energy proportion, H represents the value of the high-frequency energy at the current sampling moment, and M represents the value of the mid-frequency energy at the current sampling moment. The sum of the two, as the denominator, ensures that the value of R falls within the range of 0 to 1. The high-frequency energy proportions obtained at each sampling moment are arranged in chronological order to obtain a high-frequency energy proportion sequence. The high-frequency energy proportion sequence describes the dynamic change trajectory of the contribution of high-frequency vibration energy throughout the drilling process.

[0030] It should be noted that the original high-frequency energy percentage sequence often exhibits transient high-frequency jitter due to random disturbances in the instantaneous contact state between the cutting edge and the substrate. For example, when drilling printed circuit board substrates, short high-frequency pulses may occur at the interface between the glass fiber cloth layer and the resin layer, causing local spikes in the percentage sequence. If the original percentage sequence is directly compared with a preset threshold, a large number of instantaneous false exceedances will be generated.

[0031] Specifically, a moving average is used to smooth the high-frequency energy proportion sequence. The moving average works by sliding a fixed-width time window along the time axis point by point, arithmetically averaging the proportion values ​​at several adjacent sampling times within the window, and using the mean as the smoothed value at the center of the window. After point-by-point moving average processing, a smoothed proportion sequence is obtained, which retains the overall trend of the proportion transitioning from a low, stable range to a high, rising range, while eliminating instantaneous jitter. The preset threshold is related to the specifications of the micro-drill tool and the workpiece material. For micro-drills with a diameter of (0.3, 1.0) mm, the threshold is continuously reduced by 0.02 for every 0.1 mm increase in tool diameter (0.5~1.0 mm corresponds to 0.50~0.55); the above benchmark value is determined statistically from the high-frequency proportion distribution of critical failure samples of the same type of tool, taking its lower quartile or mean. kσ. The smoothed proportion sequence is compared with the preset threshold at each sampling time. If the value at a certain sampling time exceeds the preset threshold, an over-limit mark is marked at that sampling time. All sampling times marked with over-limit are collected in chronological order to form an over-limit mark sequence.

[0032] It is understandable that not every consecutive limit exceedance in the limit exceedance sequence corresponds to a real high-frequency surge phase. During drilling, occasional factors such as hard point impact, instantaneous chip blockage resistance, and machine tool servo disturbances can trigger isolated limit exceedance segments lasting only a few milliseconds to tens of milliseconds. These short-term isolated limit exceedance segments are unrelated to the stable energy evolution corresponding to cutting edge break-in or wear.

[0033] Specifically, when filtering out consecutive out-of-limit segments in the out-of-limit marker sequence, a minimum duration threshold is introduced as a screening condition. The minimum duration threshold T is determined as follows: First, the feed rate v of the micro-drill tool is obtained in millimeters per minute, and the spindle speed n is obtained in revolutions per minute. The time required for a single spindle revolution is taken as the effective cutting cycle, denoted as T. c =60000 / n (in milliseconds, where n is the spindle speed in revolutions per minute); multiply this by the period multiplier k to get the threshold duration T = k × T c =k×60000 / n, where k ranges from 2 to 5, with 3 being commonly used. For example, when the spindle speed is 30000 rpm and k is 3, the single cutting cycle T c=2 milliseconds, threshold duration T is calculated to be 6 milliseconds. The time span of each continuous exceeding-limit segment is checked one by one. If the segment duration is lower than the threshold, the segment is removed from the exceeding-limit marker sequence; if the segment duration reaches or exceeds the threshold, it is retained as a stable exceeding-limit continuous segment. The remaining stable exceeding-limit continuous segments after threshold filtering reflect the actual continuous exceeding of the high-frequency energy proportion. The first sampling time of the stable exceeding-limit continuous segment is determined as the surge start point, and the last sampling time is determined as the surge end point, thus determining that the micro-drilling tool has entered the high-frequency surge stage. Furthermore, the continuous time interval between the surge start point and the surge end point defines the coverage range of the high-frequency surge stage on the time axis.

[0034] Preferably, a preset transition margin is extended outwards at both ends of the continuous time interval. The length of the transition margin is 1 / 5 to 1 / 3 of the duration of the surge. The forward extension is used to capture the vibration pattern of the critical transition before the surge, and the backward extension is used to capture the vibration pattern of the surge convergence process, thus obtaining a complete time window. Specifically, a 1 / 5 ratio is selected when the blade impact surge is rapid, and a 1 / 3 ratio is selected when the surge is slow. The transition margin length L = α × T, where T is the duration of the surge, and α is a proportionality coefficient 0.2 ≤ α ≤ 0.33. The boundary of the complete time window is jointly defined by the transition margin shifted forward from the surge start point and the transition margin shifted backward from the surge end point. A segment is extracted from the original vibration signal along the sampling position corresponding to the complete time window to obtain a signal segment containing the high-frequency surge stage. This signal segment completely preserves the vibration waveform in the time domain from the rise of the high-frequency energy proportion, through exceeding the limit, to the stable exceeding the limit. As time-domain material reflecting the blade impact intensity, it can provide undistorted time-domain evidence for subsequent differentiation between early break-in and severe wear.

[0035] Step S103: Extract the time-domain amplitude distribution reflecting the impact intensity of the cutting edge from the signal segment of the high-frequency surge phase, and calculate the waveform kurtosis of the time-domain amplitude distribution.

[0036] For the signal segment in the high-frequency surge phase, the instantaneous vibration amplitude of each sampling point is collected point by point along the time axis to obtain the time-domain amplitude distribution reflecting the impact intensity of the cutting edge. The time-domain amplitude distribution presents a waveform shape with the sampling number as the horizontal axis and the instantaneous amplitude as the vertical axis. For the time-domain amplitude distribution, the mean and standard deviation of the amplitude of all sampling points are calculated. The mean is subtracted from the amplitude of each sampling point and then divided by the standard deviation for centering and normalization processing to obtain a normalized amplitude sequence. The fourth power mean of each value in the normalized amplitude sequence is calculated to obtain the waveform kurtosis of the time-domain amplitude distribution.

[0037] The high-frequency surge stage and its signal segment are determined and extracted in step S102. This step directly processes this segment: the time-domain amplitude distribution is composed of the instantaneous amplitude values ​​of each sampling point extracted in chronological order within the segment. The i-th element A(i) of this sequence corresponds to the instantaneous impact amplitude generated by the contact between the micro-drill cutting edge and the substrate at the i-th sampling moment, where i increases from 1 to the total number of sampling points N in this stage. The sequence length N is determined by the product of the sampling rate and the stage duration. This amplitude sequence directly reflects the impact intensity change characteristics of the cutting edge in this stage. The peak points in the sequence correspond to the collision events between the cutting edge and hard inclusions, and the continuous low-amplitude segments correspond to the steady-state cutting process of the cutting edge in homogeneous materials.

[0038] It should be noted that waveform kurtosis is a standardized fourth-order moment index in statistics that describes the steepness of the distribution. The value directly reflects the prominence of sharp pulse components in the waveform relative to a flat background. A low kurtosis value means that the time-domain amplitude distribution is close to a smooth fluctuation, corresponding to a relatively uniform contact state of the cutting edge microstructure during the initial break-in stage; a high kurtosis value means that the waveform contains significant sharp pulses, corresponding to the severe instantaneous impact caused by localized detachment or microscopic damage to the cutting edge.

[0039] Specifically, the process of obtaining the waveform kurtosis of the time-domain amplitude distribution is implemented as follows: The mean of the amplitudes of all sampling points in the time-domain amplitude distribution is calculated, denoted as μ. The standard deviation is obtained by taking the square root of the squared average of the deviations between the amplitudes of each sampling point and the mean, denoted as σ. For each sampling point, the amplitude is subtracted from the mean and then divided by the standard deviation to perform centering and standardization processing, resulting in a standardized amplitude sequence. This standardized amplitude sequence eliminates the DC bias and amplitude magnitude differences of the waveform, retaining only the waveform morphology information. Further, the fourth power is calculated for each value in the standardized amplitude sequence, and then the arithmetic mean is calculated to obtain the fourth-order standardized central moment. Subtracting 3 from this fourth-order standardized central moment yields the waveform kurtosis of the time-domain amplitude distribution. The fourth power operation amplifies the contribution of the large-amplitude sampling points far from the mean to the result, allowing the presence or absence of sharp pulses in the waveform to be fully reflected by the kurtosis value. The waveform kurtosis is used as a quantitative indicator reflecting the sharpness of the blade impact.

[0040] Step S104: Analyze the waveform kurtosis relative to a preset kurtosis threshold. When the waveform kurtosis is lower than the preset kurtosis threshold, confirm that the micro-drill tool is in the early break-in stage. When the waveform kurtosis is higher than or reaches the preset kurtosis threshold, confirm that the micro-drill tool is in the severe wear stage.

[0041] For the typical specifications of the micro-drilling tool and the workpiece material, a preset kurtosis threshold is established in advance. This preset kurtosis threshold is calibrated based on the impact level reflected by the time-domain amplitude distribution when the cutting edge is in stable contact during historical drilling processes, serving as the kurtosis boundary benchmark between the early break-in stage and the severe wear stage. The waveform kurtosis is compared with the preset kurtosis threshold to obtain the deviation direction of the waveform kurtosis relative to the preset kurtosis threshold, serving as the basis for determining the wear stage of the micro-drilling tool. If the waveform kurtosis is lower than the preset kurtosis threshold, the micro-drilling tool is confirmed to be in the early break-in stage, corresponding to a stable contact state where the microstructure of the cutting edge has not yet shown localized detachment; if the waveform kurtosis is higher than or reaches the preset kurtosis threshold, the micro-drilling tool is confirmed to be in the severe wear stage, corresponding to a severe wear state where the cutting edge has shown sharp impact contact.

[0042] Properly calibrating the preset kurtosis threshold is a crucial prerequisite for distinguishing between the early break-in stage and the severe wear stage. The calibration process for the preset kurtosis threshold is completed before the micro-drill tool is put into formal drilling.

[0043] In one embodiment, for a typical combination of the micro-drill tool's specifications and the workpiece material, a sample of the same tool model that has undergone a stable cutting stage is selected, and corresponding vibration signal segments are collected under controlled drilling conditions. The controlled drilling conditions include a spindle speed maintained at 8000 rpm to 12000 rpm, a feed rate maintained at 0.5 mm / s to 1.5 mm / s, constant flow rate coolant spraying, and a workpiece clamping torque maintained within the range of 15 Nm to 20 Nm. Following the processing path formed by the aforementioned high-frequency surge extraction and time-domain amplitude distribution, waveform kurtosis samples corresponding to the time-domain amplitude distribution are obtained. These waveform kurtosis samples reflect the impact level when the cutting edge is in a stable contact state. Statistical analysis is performed on the collected N sets of waveform kurtosis samples, calculating the sample mean μ and standard deviation σ, and using μ + 3σ as the preset kurtosis threshold. Here, μ represents the arithmetic mean of the waveform kurtosis samples, and σ represents the standard deviation of the waveform kurtosis samples. This threshold covers 99.7% of normal stable cutting states.

[0044] It should be noted that the preset kurtosis threshold is strongly correlated with the tool specifications and workpiece material. This solution uses hyperkurtosis κ as the criterion, defined as κ equal to kurtosis minus 3, where 3 is the baseline value for kurtosis in a normal distribution. For the scenario of a 0.3 mm diameter carbide micro-drill drilling glass fiber reinforced substrate, the vibration signal is close to a normal distribution in the early break-in stage, with hyperkurtosis κ typically between 0.2 and 0.5. When the tool enters the severe wear stage, the increased impact vibration causes hyperkurtosis κ to rise to 2.0 to 3.0. The preset hyperkurtosis threshold κ0 is calibrated by μ + 3σ of the hyperkurtosis of stable cutting samples, with a typical value of approximately 1.5. If the measured hyperkurtosis κ is lower than κ0, the tool is confirmed to be in the early break-in stage, corresponding to a low hyperkurtosis characteristic of κ close to 0.2~0.5, reflecting a normal break-in state where the micro-protrusions of the cutting edge uniformly contact the substrate surface. If the waveform kurtosis K is greater than or equal to the preset kurtosis threshold of 3.5, the micro-drill tool is determined to have entered the severe wear stage. During this stage, the cutting edge experiences localized detachment or microscopic damage after prolonged cutting. Upon contact with the substrate, a violent impact occurs, resulting in sharp pulse peaks in the time-domain amplitude distribution that deviate significantly from the mean. Because the kurtosis calculation uses fourth-power operations, the contributions of these large-amplitude sampling points are significantly amplified, leading to a marked increase in the waveform kurtosis value. The vibration signal in the severe wear stage exhibits strong intermittent and non-stationary characteristics, with the superkurtosis κ typically rising to 2.0–3.0, significantly higher than the 0.2–0.5 in the early break-in stage.

[0045] Step S105: For the severe wear stage, acquire the cutting edge image of the micro-drill tool, process the cutting edge image using an edge detection algorithm to obtain the cutting edge contour, and extract the number of discrete spots from the cutting edge contour.

[0046] For the severe wear stage, an image of the cutting edge of the micro-drill tool is acquired using an industrial microscope camera. This image is then subjected to grayscale conversion and Gaussian filtering smoothing to obtain a clean grayscale image. An edge detection algorithm is used on this clean grayscale image to identify the contour boundaries of pixel grayscale abrupt changes, thus obtaining the cutting edge contour. Connected regions are marked within the cutting edge contour according to pixel adjacency relationships, resulting in a set of connected regions composed of independent closed blocks. Morphological opening and closing operations are used to repair fractures and peeling adhesions within this set of connected regions, resulting in a corrected set of connected regions. Artifact spots with too few pixels and continuous wear blocks with too many pixels are filtered out according to preset lower and upper area limits, retaining the target connected regions between these two thresholds. The number of discrete spots in the cutting edge contour is counted and accumulated for each independent closed block within the target connected regions, and this count is recorded along with the center coordinates of each independent closed block.

[0047] Once the micro-drill tool is determined to be in the severe wear stage, it is removed from the spindle and fixed to an image measurement platform. An industrial microscope camera, coupled with a ring-shaped cold light source, is used to capture a vertical top-down image of the cutting edge. For micro-drills with a diameter in the range of [0.3, 1) mm, an optical magnification of 200x to 300x is selected; for those in the range of [1, 3] mm, 100x to 150x is selected to cover the complete contour range of the main and secondary cutting edges. During imaging, it is ensured that the perpendicularity deviation between the cutting edge and the camera's optical axis is less than 2 degrees, the uniformity of the ring light source illumination is greater than 90%, and the imaging resolution is not less than 2048×2048 pixels to obtain the cutting edge image. The original image of the cutting edge is presented in a color three-channel format, and often exhibits uneven illumination and isolated pixel noise in the edge areas. The cutting edge image is subjected to grayscale conversion. The red channel R, green channel G, and blue channel B are fused using weight coefficients of 0.299, 0.587, and 0.114, respectively. The conversion formula is Gray = 0.299 × R + 0.587 × G + 0.114 × B, where Gray is the output grayscale value, ranging from 0 to 255, thus obtaining a single-channel cutting edge grayscale image. Gaussian filtering is then applied to the cutting edge grayscale image for neighborhood weighted smoothing. The standard deviation σ of the Gaussian function ranges from 0.5 to 2.0 pixels, preferably 1.0 pixel, and is adjusted according to the image noise level. A larger σ value is chosen when the noise is high to enhance the smoothing effect. Gaussian filtering employs a convolutional window with a Gaussian function as the weight template. The window size is 5×5 or 7×7 pixels, with the highest weight at the window center, decreasing outwards according to a Gaussian curve. The grayscale values ​​of each pixel within the window are multiplied by their corresponding weight coefficients and summed; the result is assigned to the center of the window. After traversing the entire image, a clean grayscale image is obtained. In this clean grayscale image, the grayscale values ​​of isolated noise pixels are weighted and smoothed to the weighted average of their neighborhood grayscale values, thus attenuating the main contour information of the cutting edge. The Canny edge detection algorithm is then used on this clean grayscale image to identify contour boundaries where pixel grayscale changes abruptly. A high threshold of 100 and a low threshold of 50 are set, and non-maximum suppression and dual-threshold concatenation are used to extract the complete cutting edge contour.

[0048] In one possible implementation, the Sobel operator based on the first-order gradient is used to perform gradient operations on the pure grayscale image along the horizontal and vertical directions to obtain the gradient magnitude at each pixel position. The Cannibal operator is used for edge detection; this operator, based on second-order derivative calculation and non-maximum suppression, can obtain continuous contour lines with a single pixel width. After edge detection processing, the cutting edge contour is obtained, which consists of a continuous closed chain of edge pixels, outlining the geometry of the outer periphery and internal wear blocks of the cutting edge. Multiple closed blocks formed by wear regions exist within the cutting edge contour. Connected regions within the cutting edge contour are marked according to pixel adjacency relationships, and the eight-neighbor adjacency rule is used to group interconnected foreground pixels with the same label. The eight-neighbor rule considers eight adjacent pixels in the top, bottom, left, right, and four diagonal directions of each pixel, which can identify diagonally connected wear regions and avoid misclassifying actually connected wear blocks as independent regions. The entire binary contour image is traversed, and each foreground pixel is scanned line by line. Each pixel is checked for a labeled pixel within its eight-neighborhood. If a labeled pixel exists, it inherits the label; otherwise, a new label is assigned. After traversal, a set of connected regions consisting of several independent closed blocks is obtained, each block corresponding to a unique label. Two types of morphological defects often exist in the original set of connected regions. The first type of defect manifests as tiny breaks on the closed boundaries, typically less than 3 pixels wide, causing pixels that should belong to the same wear patch to be incorrectly divided into two independent closed blocks. The second type of defect originates from the adhesion of edge pixels between adjacent wear patches, forming thin bridges, typically 1 to 5 pixels wide, causing two independent patches that should be separated to be incorrectly merged into one closed block. Morphological closing and opening operations are used to repair these two types of defects, respectively. The structuring element size for the closing operation is set to 5×5 pixels to bridge breaks, and the structuring element size for the opening operation is set to 3×3 pixels to cut thin bridges. The set of connected regions is then converted into a binary image, where the edge contour pixels are set to a foreground value of 255 and the background pixels are set to 0. Morphological opening and closing operations were used to repair and peel off the binary image. The morphological closing operation first dilates the binary image, causing the foreground pixels to expand outwards and fill the broken seams, then erodes them inwards to restore their original boundary sizes. The blocks on either side of the broken seams are reconnected into complete closed blocks. The morphological opening operation first erodes the image to peel off the adhering bridges, then dilates it to restore the size of the main closed blocks, separating two incorrectly merged closed blocks into independent blocks. After these sequential opening and closing operations, a set of corrected connected regions with complete boundaries and separated from each other is obtained.

[0049] Preferably, the structuring element used in the morphological opening and closing operation is a 3×3 square structuring element or a 5×5 cross structuring element. The size of the structuring element must be smaller than the minimum scale of the discrete spots to avoid accidentally swallowing real spots during the repair process. In addition to real wear spots, the corrected connected region set still contains two types of interference objects. One is a few pixel-counting imaging artifact spots, corresponding to isolated small blocks caused by dust adhesion, lens reflection, or edge detection residue; the other is a large number of pixel-counting continuous wear main body blocks, corresponding to continuous wear surfaces where the main cutting band of the cutting edge has not yet broken into discrete shapes. The corrected connected region set is filtered according to a preset lower and upper area limit. The preset lower area limit ranges from 5 to 50 pixels, corresponding to the minimum number of pixels that can be stably identified in the microscopic field of view for discrete spots; the preset upper area limit ranges from 3000 to 8000 pixels, corresponding to the common size of continuously distributed blocks of the main wear body of the cutting edge. Each independent closed region is compared with the preset lower and upper area limits. Artifact spots with fewer pixels than the preset lower area limit and continuous wear-prone areas with more pixels than the preset upper area limit are removed, retaining the target connected region between the two thresholds. Each independent closed region within the target connected region is counted, and the number of discrete spots in the cutting edge contour is obtained after traversal. Simultaneously, the center coordinates of each independent closed region are recorded, calculated by summing the x-coordinates of all pixels within the region and dividing by the total number of pixels to obtain the center x-coordinate. c The sum of the ordinates divided by the total number of pixels gives the center ordinate y. c , that is, x c =Σx i / n, yc=Σy i / n, where n is the total number of pixels contained in the block. For example, if a wear spot block contains 5 pixels with coordinates (10,20), (11,21), (12,20), (11,19), and (10,21), then the center coordinates of the spot are (10.8,20.2). The number of discrete spots and their corresponding center coordinates together characterize the discrete distribution of the edge wear area within the microscopic field of view.

[0050] Step S106: Evaluate the difference between the number of discrete spots and the preset spot threshold. When the number of discrete spots exceeds the preset spot threshold, determine that the edge wear area changes from continuous distribution to discrete spot distribution, and obtain the target wear state of the micro-drill tool.

[0051] For the cutting edge geometry of the micro-drill tool, a preset spot threshold is established in advance. This preset spot threshold is calibrated based on the statistical upper limit of the target connected region count within the microscopic field of view when the cutting edge wear area is in a continuous distribution pattern, serving as a boundary benchmark for the number of spots distinguishing between continuous and discrete spot distribution patterns. The number of discrete spots is compared with the preset spot threshold to obtain the deviation direction of the number of discrete spots relative to the preset spot threshold, serving as the basis for determining the morphology of the cutting edge wear area. If the number of discrete spots exceeds the preset spot threshold, the cutting edge wear area is determined to have shifted from a continuous distribution to a discrete spot distribution, and the target wear state of the micro-drill tool is recorded as a discrete spot wear state, accompanied by the number of discrete spots and the center coordinates of each independent closed block as a quantitative description of the target wear state.

[0052] A reasonable setting of the preset spot threshold is a prerequisite for determining the morphological classification of the edge wear area.

[0053] In one embodiment, a preset spot threshold is pre-established for the combination of the cutting edge geometry and the magnification of the microscopic field of view of the micro-drill tool.

[0054] Specifically, several samples of the same type of cutting tool in a continuous distribution pattern are selected. For each sample, the count value of the target connected region within the microscopic field of view is calculated along the processing path of the aforementioned edge detection and connected region extraction. The typical value of the preset spot threshold is estimated according to N0=D×K, where D is in millimeters, K is a material coefficient in units of spots / mm, and 33 is taken for glass fiber reinforced substrates and 20-25 for thin-walled aluminum alloy parts. For example, when D=0.8mm and K=20, N0=16. During calibration, N0 is further corrected by the 95th percentile of the spot count of several samples in a continuous distribution pattern, so that the threshold has both formulaic calculability and statistical representativeness. The preset spot threshold serves as the boundary benchmark for the number of spots in a continuous distribution pattern and a discrete spot distribution pattern. Choosing the 95th percentile can effectively eliminate the interference of extreme outliers while ensuring that the characteristics of most samples in a continuous distribution pattern are covered, so that the threshold setting is both statistically representative and practically operable.

[0055] For example, when collecting 30 consecutive distribution pattern samples, the spot count values ​​of each sample are sorted from smallest to largest, and the 29th value is taken as the 95th percentile. This value is used as the preset spot threshold for subsequent determination of the distribution pattern type of the tool under test.

[0056] It should be noted that the preset spot threshold is closely related to the tool specification. For scenarios where a 0.3 mm diameter carbide micro-drill is used to drill glass fiber reinforced substrates, the preset spot threshold is set to 10; for scenarios where a micro-drill with a diameter of 0.8 mm or more is used to drill thin-walled aluminum alloy parts, the preset spot threshold is set to 15 to 20.

[0057] Specifically, after obtaining the number of discrete spots, the number of discrete spots is compared with a preset spot threshold, and the morphology of the edge wear area is determined based on the comparison result. When the number of discrete spots is greater than the preset spot threshold, it is determined to be a discrete spot distribution pattern, in which case the wear exhibits discontinuous characteristics; when the number of discrete spots is less than or equal to the preset spot threshold, it is determined to be a continuous wear pattern, in which case the wear exhibits overall continuous characteristics. The morphology classification result serves as the basis for subsequent wear degree assessment and tool life prediction. If the number of discrete spots exceeds the preset spot threshold, it is determined that the edge wear area changes from a continuous distribution to a discrete spot distribution. The continuous distribution refers to the wear area existing in the form of a large area of ​​connected boundaries, while the discrete spot distribution refers to the wear area fragmenting into several small-scale spots that are separated from each other. At this time, the target wear state of the micro-drill tool is recorded as a discrete spot wear state, which serves as the final output of this discrimination process, along with the number of discrete spots and the center coordinates of each independent closed block.

[0058] Preferably, the archiving method for the target wear state includes the number of discrete spots and the center coordinate pair of each independent closed block. The center coordinate pair records the horizontal and vertical pixel positions of the spot in the cutting edge image. For scenarios where the number of discrete spots exceeds the preset spot threshold, the system determines that the cutting edge wear area has changed from a continuous distribution to a discrete distribution. At this time, the target wear state of the micro-drill tool is recorded as a discrete wear state. This state indicates that the wear on the cutting edge surface exhibits multi-point dispersion characteristics; the wear area no longer exists as a continuous patch, but is distributed at different positions on the cutting edge as independent wear spots. The system completes the morphology classification based on the comparison between the discrete spot count result N and the preset spot threshold N0. When N is greater than N0, the morphology change is confirmed, and the classification result is stored in the wear state database for subsequent tool management decisions.

[0059] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for analyzing the wear state of ultrafine-grained cemented carbide micro-drilling tools, characterized in that, The method includes: The original vibration signal during the machining process of ultrafine-grained cemented carbide micro-drilling tool is acquired, and the vibration spectrum is obtained by frequency domain transformation of the original vibration signal through fast Fourier transform. High-frequency energy and mid-frequency energy are extracted from the vibration spectrum respectively. The high-frequency energy ratio is obtained based on the high-frequency energy and the mid-frequency energy. When the high-frequency energy ratio exceeds a preset threshold, it is determined that the micro-drilling tool has entered the high-frequency surge stage, and a signal segment containing the high-frequency surge stage is extracted. Extract the time-domain amplitude distribution reflecting the impact intensity of the cutting edge from the signal segment, and calculate the waveform kurtosis of the time-domain amplitude distribution; When the waveform kurtosis is lower than the preset kurtosis threshold, it is confirmed that the micro-drilling tool is in the early break-in stage; when the waveform kurtosis is higher than or reaches the preset kurtosis threshold, it is confirmed that the micro-drilling tool is in the severe wear stage. For the severe wear stage, the cutting edge image of the micro-drill tool is acquired, and the cutting edge image is processed by an edge detection algorithm to obtain the cutting edge contour. The number of discrete spots is extracted from the cutting edge contour. When the number of discrete spots exceeds a preset spot threshold, the edge wear area is determined to change from a continuous distribution to a discrete spot distribution, thereby obtaining the target wear state of the micro-drill tool.

2. The method for analyzing the wear state of ultrafine-grained cemented carbide micro-drilling tools according to claim 1, characterized in that, The acquisition of the original vibration signal during the machining process of the ultrafine-grained cemented carbide micro-drill tool includes: An accelerometer is attached to the vibration pickup position at the spindle clamping end of the micro-drilling tool, and the original vibration signal of the drilling process is continuously collected according to a preset sampling rate. The original vibration signal is denoised by bandpass filtering to obtain a clean vibration signal. The signal is then segmented into frames according to a fixed time window to obtain time-domain waveform segments arranged frame by frame.

3. The method for analyzing the wear state of ultrafine-grained cemented carbide micro-drilling tools according to claim 1, characterized in that, The step of performing frequency domain transformation on the original vibration signal using Fast Fourier Transform to obtain the vibration spectrum, and extracting high-frequency energy and mid-frequency energy from the vibration spectrum, includes: For the time-domain waveform segments obtained by framing the original vibration signal, a Hanning window is applied to each frame to weight and attenuate the edge samples, resulting in a windowed waveform. The windowed waveform is then converted to the frequency domain using a fast Fourier transform to obtain the vibration spectrum. The vibration spectrum is then divided into frequency bands according to the frequency boundary between the cutting resonance band and the impact response band, resulting in two continuous frequency intervals: the mid-frequency band and the high-frequency band. For the two continuous frequency intervals of the mid-frequency band and the high-frequency band, the square of the amplitude of the vibration spectrum is discretely accumulated and summed along the frequency axis within the interval to obtain two numerical indicators: the mid-frequency band energy and the high-frequency band energy. The mid-frequency band energy and the high-frequency band energy are aligned and arranged according to the same sampling time to obtain a frequency band energy sequence arranged frame by frame.

4. The method for analyzing the wear state of ultrafine-grained cemented carbide micro-drilling tools according to claim 1, characterized in that, The step of obtaining the high-frequency energy ratio based on the high-frequency energy and the mid-frequency energy includes: The high-frequency energy ratio is obtained by dividing the high-frequency energy by the sum of the high-frequency energy and the mid-frequency energy. A high-frequency energy ratio sequence is formed along the time axis. The high-frequency energy ratio sequence is then smoothed by a moving average to obtain a smoothed ratio sequence.

5. The method for analyzing the wear state of ultrafine-grained cemented carbide micro-drilling tools according to claim 1, characterized in that, The extraction of the signal segment containing the high-frequency surge phase includes: The high-frequency energy ratio is compared with the preset threshold. Sampling times that exceed the preset threshold are marked with an over-limit mark to form an over-limit mark sequence. Short-term isolated over-limit segments are screened out according to a preset minimum duration threshold. The first moment of continuous over-limit segments is retained as the surge start point and the last moment is retained as the surge end point. A preset transition margin is extended outward to form a complete time window. Segments are extracted from the original vibration signal along the corresponding position of the complete time window.

6. The method for analyzing the wear state of ultrafine-grained cemented carbide micro-drilling tools according to claim 1, characterized in that, Extracting the time-domain amplitude distribution reflecting the impact intensity of the cutting edge from the signal segment and calculating the waveform kurtosis of the time-domain amplitude distribution includes: The instantaneous vibration amplitude of each sampling point is collected point by point along the time axis to obtain the time-domain amplitude distribution. The mean value of all sampling points of the time-domain amplitude distribution is subtracted and then divided by the standard deviation to obtain the normalized amplitude sequence. The fourth power mean of each value of the normalized amplitude sequence is calculated to obtain the waveform kurtosis.

7. The method for analyzing the wear state of ultrafine-grained cemented carbide micro-drilling tools according to claim 1, characterized in that, When the waveform kurtosis is lower than a preset kurtosis threshold, the micro-drilling tool is confirmed to be in the early break-in stage; when the waveform kurtosis is higher than or reaches the preset kurtosis threshold, the micro-drilling tool is confirmed to be in the severe wear stage, including: For the typical specifications of the micro-drill tool and the workpiece material, the preset kurtosis threshold is calibrated based on the impact level reflected by the time-domain amplitude distribution when the cutting edge is in stable contact during historical drilling processes.

8. The method for analyzing the wear state of ultrafine-grained cemented carbide micro-drilling tools according to claim 1, characterized in that, For the severe wear stage, the cutting edge image of the micro-drill tool is acquired, and an edge detection algorithm is used to process the cutting edge image to obtain the cutting edge contour. The number of discrete spots is extracted from the cutting edge contour, including: The cutting edge image is obtained by acquiring the cutting edge end face of the micro-drill using an industrial microscope camera. The cutting edge image is then subjected to grayscale conversion and Gaussian filtering smoothing to obtain a pure grayscale image. The edge detection algorithm is then used to identify the contour boundaries of pixel grayscale abrupt changes in the pure grayscale image to obtain the cutting edge contour. The internal region of the cutting edge contour is marked with connected regions according to pixel adjacency to obtain a set of connected regions composed of independent closed blocks. Morphological opening and closing operations are performed on the set of connected regions to obtain a modified set of connected regions. Artifact spots below the lower limit and continuous wear main blocks above the upper limit are filtered out according to a preset lower limit and upper limit, and the target connected regions between the two thresholds are retained. Each independent closed block in the target connected region is counted and accumulated one by one to obtain the number of discrete spots.

9. The method for analyzing the wear state of ultrafine-grained cemented carbide micro-drilling tools according to claim 1, characterized in that, When the number of discrete spots exceeds a preset spot threshold, determining that the edge wear area changes from a continuous distribution to a discrete spot distribution includes: For the cutting edge geometry of the micro-drill, a preset spot threshold is established in advance. The preset spot threshold is calibrated based on the statistical upper limit of the number of discrete spots in the microscopic field of view when the cutting edge wear area is in a continuous distribution pattern.

10. The method for analyzing the wear state of ultrafine-grained cemented carbide micro-drilling tools according to claim 1, characterized in that, Obtaining the target wear state of the micro-drill tool includes: The number of discrete spots is compared with the preset spot threshold to obtain the deviation direction of the number of discrete spots relative to the preset spot threshold. When the number of discrete spots exceeds the preset spot threshold, the target wear state is recorded as discrete spot wear state, with the number of discrete spots attached as a quantitative description of the target wear state.

Images