A saw blade detection method based on multi-source data fusion

Abstract

This invention discloses a saw blade detection method based on multi-source data fusion, specifically relating to the field of saw blade detection technology. It involves real-time acquisition of image data of the rake face, side edge, and cutting edge, as well as spindle current data, during the saw blade's cutting process. By integrating image analysis and current characteristic analysis techniques, the method performs a joint evaluation of the saw blade's visual wear state and current change trends using multi-source data. This accurately detects the saw blade's wear degree and abnormal tooth loss. Furthermore, by combining this with quality indicators such as dimensional deviations, surface finish, and defect rate of the processed product, the method determines the saw blade's wear threshold and tooth loss threshold. This enables automated decision-making for saw blade scrapping or replacement, effectively improving the accuracy and real-time performance of saw blade lifespan monitoring and ensuring the stability of processed product quality.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of saw blade detection, and more particularly to a saw blade detection method based on multi-source data fusion. BACKGROUND

[0002] In the field of mechanical processing, the saw blade as a commonly used tool, its wear condition is closely related to the processing quality. At present, the monitoring method for the wear state of the saw blade mainly depends on artificial periodic inspection or only relies on a single image analysis method or current monitoring method. The image analysis method has the problem that the detection precision is unstable due to the influence of environmental light changes and lens definition, and the current monitoring method is easily affected by motor load fluctuation and processing material hardness difference, and cannot accurately identify abnormal tooth loss and other local damage conditions. The above single monitoring method cannot accurately and timely reflect the real wear state of the saw blade. SUMMARY

[0003] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide a saw blade detection method based on multi-source data fusion to solve the problems raised in the background art.

[0004] To achieve the above object, the present application provides the following technical scheme: A saw blade detection method based on multi-source data fusion, comprising the following steps: S1: collecting image data of the saw blade rake face, side edge and saw blade edge during sawing workpieces, and performing image definition and effectiveness evaluation, and outputting saw blade real-time image feature data; S2: based on the saw blade real-time image feature data, analyzing the projection line change of the saw blade rake face, the wear morphology change of the side edge and the retreat amount change of the edge position, and outputting the saw blade visual wear feature data; S3: real-time collecting the spindle current data during sawing workpieces, extracting the current change trend feature and current variability feature, and outputting the saw blade current wear feature data; S4: based on the saw blade visual wear feature data and the saw blade current wear feature data, respectively performing wear degradation trend analysis and abnormal tooth loss detection analysis, and outputting the comprehensive evaluation data of the saw blade visual and current wear features; S5: according to the comprehensive evaluation data of the saw blade visual and current wear features, respectively calculating the wear threshold and the tooth loss quantity threshold of the saw blade, and outputting the threshold feature data of the saw blade visual and current fusion determination; S6: according to the threshold feature data, combining the processing product quality index, judging whether the saw blade generates a saw blade scrap or replacement trigger signal.

[0005] In a preferred embodiment, S1, in particular: During the sawing process, the front cutting surface, side cutting edge and cutting edge of the saw blade exposed during the sawing process are continuously photographed. Based on image data of the saw blade's front face, side edge, and cutting edge, the image brightness, focus status, and motion blur are determined. Images that meet the imaging requirements are selected, and invalid images are removed to form real-time image feature data of the saw blade.

[0006] In a preferred embodiment, S2 specifically refers to: Based on real-time image feature data of the saw blade, grayscale normalization and edge extraction are performed on the saw tooth area to obtain the boundary line of the saw blade front face, the boundary line of the side edge, and the boundary line of the saw blade cutting edge. Straightness deviation is calculated by straight-line fitting of the saw blade front face boundary line, side edge profile wear is calculated based on the side edge boundary line, and edge backing is calculated based on the saw blade edge boundary line and the saw tooth reference point, thus forming visual wear characteristic data of the saw blade.

[0007] In a preferred embodiment, S3 specifically refers to: Real-time acquisition of spindle current data during saw blade cutting of workpiece, followed by noise reduction and baseline correction of the spindle current data; The saw blade rotation cycle is generated based on the saw blade speed. The spindle current data is segmented according to the saw blade rotation cycle, and the average value and fluctuation amplitude of each segment are calculated. Based on the segmented average value, the current change trend characteristics are formed; based on the segmented fluctuation amplitude, the current variability characteristics are formed; and the saw blade current wear characteristic data are output.

[0008] In a preferred embodiment, S4 specifically refers to: The visual wear feature data of the saw blade and the current wear feature data of the saw blade are time-aligned to form an aligned feature sequence; Based on the visual wear characteristic data of the saw blade, the rate of change of straightness deviation, the rate of change of side edge contour wear, and the rate of change of cutting edge retraction are calculated to form a visual degradation trend. Based on the saw blade current wear characteristic data, the slope of the current change trend characteristic and the increment of the current variability characteristic are calculated. Based on the alignment characteristic sequence, the abrupt segment of the current variability characteristic is determined and the abnormal tooth loss segment is marked. The comprehensive evaluation data of the saw blade visual and current wear characteristics is output.

[0009] In a preferred embodiment, S5 specifically refers to: Based on the comprehensive evaluation data of saw blade visual and electrical wear characteristics, the alignment feature sequences corresponding to the stable growth segment and the abnormal tooth loss segment of the visual degradation trend are extracted respectively. Within the stable growth phase of visual degradation trend, the statistical boundaries of the straightness deviation change rate, the side edge profile wear change rate, and the cutting edge retraction change rate are calculated to form the wear threshold. Within the abnormal tooth loss segment, the number of occurrences of the abrupt change segment based on the current variability characteristic is counted in correspondence with the saw blade rotation cycle to form a threshold for the number of tooth loss, and output threshold feature data for the saw blade visual and current fusion judgment.

[0010] In a preferred embodiment, S6 specifically refers to: Based on the threshold feature data determined by the fusion of visual and electrical characteristics of the saw blade, the processing product size deviation, surface finish data and processing product defect rate corresponding to the processing product quality indicators are collected, and the changing trend of the processing product quality indicators is calculated. The trend of changes in the quality indicators of the processed products is correlated and compared with the threshold feature data of the saw blade visual and current fusion judgment to determine whether to generate a trigger signal for the saw blade to be scrapped or replaced.

[0011] The technical effects and advantages of the saw blade detection method based on multi-source data fusion proposed in this invention are as follows: By evaluating the clarity and effectiveness of image data of the saw blade's rake face, side edge, and cutting edge, and forming real-time image feature data of the saw blade, the impact of image quality fluctuations on the detection results is reduced. Based on the real-time image feature data of the saw blade, the changes in the rake face projection line, the changes in the side edge wear morphology, and the changes in the cutting edge retraction are extracted to form visual wear feature data of the saw blade, realizing quantitative characterization of key wear parts of the saw blade. Current change trend features and current variability features are extracted from the spindle current data to form saw blade current wear feature data, supplementing the information on sawing load changes. The visual wear feature data and the saw blade current wear feature data of the saw blade are comprehensively evaluated to complete the wear degradation trend analysis and abnormal tooth loss detection analysis, improving the completeness of wear identification and tooth loss identification. Based on the comprehensive evaluation data, wear thresholds and tooth loss number thresholds are formed respectively, and threshold feature data for the fusion judgment of saw blade vision and current is established to improve the repeatability of the judgment boundary. By combining the processing product dimensional deviation, surface finish, and defect rate and other processing product quality indicators, a saw blade scrapping or replacement trigger signal is generated, so that the saw blade status judgment and processing product quality control form a closed loop. Attached Figure Description

[0012] Figure 1 This is a schematic diagram of a saw blade detection method based on multi-source data fusion according to the present invention. Detailed Implementation

[0013] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention. Example

[0014] Figure 1 This invention presents a saw blade detection method based on multi-source data fusion, which includes the following steps: S1: Collect image data of the saw blade's front cutting face, side cutting edge, and cutting edge when the saw blade is sawing the workpiece, evaluate the image clarity and effectiveness, and output real-time image feature data of the saw blade; S2: Based on real-time image feature data of the saw blade, analyze the changes in the projection line of the saw blade's front face, the changes in the side edge wear pattern, and the changes in the amount of backlash at the edge position, and output the visual wear feature data of the saw blade. S3: Real-time acquisition of spindle current data when the saw blade is cutting the workpiece, extraction of current change trend characteristics and current variability characteristics, and output of saw blade current wear characteristic data; S4: Based on the visual wear characteristic data and the current wear characteristic data of the saw blade, the wear degradation trend analysis and abnormal tooth loss detection analysis are performed respectively, and the comprehensive evaluation data of the visual and current wear characteristics of the saw blade are output. S5: Based on the comprehensive evaluation data of saw blade visual and current wear characteristics, calculate the saw blade wear threshold and tooth loss threshold respectively, and output the threshold feature data of saw blade visual and current fusion judgment. S6: Based on the threshold feature data and the quality indicators of the processed products, determine whether the saw blade generates a trigger signal for scrapping or replacement.

[0015] S1: Acquire image data of the saw blade's front face, side edge, and cutting edge when the saw blade is sawing the workpiece, evaluate image clarity and effectiveness, and output real-time image feature data of the saw blade, including: During the sawing process, the front cutting surface, side cutting edge and cutting edge of the saw blade exposed during the sawing process are continuously photographed. During the sawing process, a high-speed camera is positioned on the front of the saw blade, with the camera lens perpendicular to the blade's surface. The front of the saw blade represents the planar position of the front face, side edges, and cutting edge of the saw teeth as the blade rotates. To ensure continuous capture of complete images of the saw blade's front face, side edges, and cutting edge, the high-speed camera's position is selected as follows: First, the diameter of the saw blade to be inspected is measured, and the center of rotation is determined. Then, using the center of rotation as a reference, the high-speed camera is installed in a frontal area that completely covers the movement trajectory of the saw blade's front face, side edges, and cutting edge. The installation position is determined by calculating the saw blade diameter and the high-speed camera's field of view. For example, when the saw blade diameter is 500 mm, after determining the field of view of the high-speed camera lens and considering a safety margin, the high-speed camera is positioned at an appropriate distance from the front of the saw blade, such as 600 mm.

[0016] The high-speed camera continuously captures images of the exposed front face, side edge, and cutting edge of the saw blade during the sawing process. The shooting frequency is set according to the saw blade rotation speed, requiring the shooting frequency to meet the requirements of continuous imaging. The method for determining the shooting frequency is as follows: First, obtain the rotation speed of the saw blade, for example, if the spindle speed is 3000 revolutions per minute, then it rotates 50 revolutions per second; then measure the number of circumferential saw teeth on the saw blade, for example, if the saw blade has 100 circumferential saw teeth, then the saw blade passes through the field of view position 5000 times per second; according to the principle that each saw tooth must obtain at least one complete image, the shooting frequency should be at least higher than the frequency at which the saw teeth pass through the field of view position, for example, the shooting frequency can be set to 8000 frames per second.

[0017] Based on the image data of the saw blade's front face, side edge, and cutting edge, the image brightness, focus status, and motion blur are determined. Images that meet the imaging requirements are selected and invalid images are removed to form real-time image feature data of the saw blade. The continuously captured image data consists of the front cutting face, side cutting edge, and blade edge of the saw blade, and is in grayscale image format. Image brightness is evaluated by calculating the average grayscale value of all pixels in each frame of the grayscale image and comparing it to a pre-set image brightness threshold to determine if the imaging requirements are met. The image brightness threshold is determined as follows: before the sawing operation begins, a set of high-quality reference images acquired under specific lighting conditions are selected, and the average grayscale value of the reference images is calculated as the image brightness threshold. For example, the average grayscale value threshold can be set between 100 and 200. If the average grayscale value is less than 100 or greater than 200, the image is considered invalid and discarded.

[0018] Focus status evaluation is performed on image data acquired by a high-speed camera: Sharpness parameters are calculated for the edge detail regions of each frame. The sharpness parameter is calculated using the Laplacian gradient operator method, i.e., convolution is performed on each frame using the Laplacian gradient operator to obtain the image edge sharpness. The focus status threshold is determined by using a sequence of sharp images with jagged edges taken before the sawing operation, and calculating the average sharpness obtained after convolution using the Laplacian gradient operator as a reference threshold. For example, the reference threshold can be set between 0.5 and 1.5. If the sharpness parameter of each frame does not reach the focus status threshold range, it is considered an invalid image and is discarded.

[0019] Motion blur determination is performed on image data acquired by high-speed cameras: Fourier transform is used to calculate the proportion of high-frequency components in the spectrum of each frame of the image. The proportion of high-frequency components in the image spectrum is defined as the ratio of the energy of high-frequency components to the total spectrum energy in the Fourier transform result of the image. The method for determining the motion blur threshold is as follows: select clear image samples taken before sawing, calculate the average proportion of high-frequency component energy, and set the threshold range of motion blur, for example, the high-frequency component proportion threshold is between 20% and 40%. If the proportion of high-frequency components in each frame of the image is less than 20%, the image is determined to have motion blur and should be removed.

[0020] Images that simultaneously meet the above three criteria are retained, and through screening, a set of valid images that meet the imaging requirements is formed. Feature extraction is performed on the valid image set to extract key edge information of the saw blade's front face, side edge, and cutting edge as feature data. The edge information extraction method is the Canny edge detection operator: first, the image is processed by Gaussian filtering to reduce the influence of noise, then the Canny edge detection operator is used to calculate the edge intensity and direction of the image, and the valid edges are determined according to the double threshold method; for example, the Gaussian filter kernel size of the Canny edge detection operator can be set to 3 to 5 pixels, and the double threshold can be set to a low threshold of 30 and a high threshold of 80.

[0021] Finally, the extracted saw blade front edge information, side edge information, and cutting edge information are stored as image feature data. The image feature data format is as follows: each valid image frame corresponds to one feature data file, including a set of two-dimensional coordinates for the front edge, side edge, and cutting edge. Each edge's two-dimensional coordinate set contains several coordinate points; for example, each coordinate point consists of horizontal and vertical pixel coordinates. The precision depends on the image resolution; for example, when the image resolution is 1280×1024 pixels, the coordinate precision is integer pixel coordinate values. The feature data set of all valid images obtained above constitutes the real-time image feature data of the saw blade.

[0022] S2: Based on real-time image feature data of the saw blade, analyze the changes in the projection line of the saw blade's rake face, the changes in the side edge wear pattern, and the changes in the amount of backlash at the cutting edge position, and output the visual wear feature data of the saw blade, including: Based on real-time image feature data of the saw blade, grayscale normalization and edge extraction are performed on the saw tooth area to obtain the boundary line of the saw blade front face, the boundary line of the side edge, and the boundary line of the saw blade cutting edge. Gray-level normalization operates on grayscale images within the real-time image feature data of the saw blade. These grayscale images encompass the complete area of ​​the saw blade's front face, side edges, and cutting edge at the time of capture. A linear transformation is applied to the pixel grayscale values ​​in each grayscale image, ensuring that the transformed image's grayscale values ​​are distributed within a uniform and standardized range. For example, grayscale values ​​are uniformly mapped to the range of 0 to 255. The grayscale value range of all grayscale images within the real-time image feature data of the saw blade is statistically analyzed. For instance, if the minimum grayscale value of all grayscale images is 50 and the maximum grayscale value is 200, then the starting point of the linear mapping is set to 50, and the ending point to 200. Each pixel grayscale value in the image is transformed using the linear mapping formula: the mapped grayscale value equals the original grayscale value minus the starting grayscale value, divided by the grayscale value range, and finally multiplied by the maximum value of the standard range, 255, to obtain the unified grayscale image data after gray-level normalization.

[0023] An edge extraction method is used to process the grayscale image data after grayscale normalization. Gaussian filtering is used to reduce noise interference in the grayscale image data. The size of the Gaussian filter kernel is selected according to the clarity and noise level of the grayscale image data, for example, a 5×5 filter kernel can be selected. The gradient intensity and gradient direction of the image pixels are calculated for the grayscale image data after Gaussian filtering. The gradient intensity is calculated by calculating the difference between the horizontal and vertical directions and then taking the square root of the sum of the squares. The gradient direction is calculated by the arctangent of the difference between the horizontal and vertical directions. Non-maximum suppression is used to process the image gradient intensity to remove non-edge pixels. A dual thresholding method is used to extract effective edges. The dual thresholds are determined by statistically analyzing the gradient intensity distribution of edge pixels in a certain number of grayscale image data. The part with higher intensity in the cumulative distribution is selected as the high threshold, and the part with medium intensity is selected as the low threshold. For example, the high threshold is set to 70% of the maximum gradient intensity, and the low threshold is set to 30% of the maximum gradient intensity. The effective edge lines obtained after the above processing are the saw blade front face boundary line, side edge boundary line and saw blade cutting edge boundary line in the real-time image feature data of the saw blade.

[0024] Straightness deviation is calculated by straight line fitting of the saw blade front face boundary line, side edge profile wear is calculated based on the side edge boundary line, and edge backing is calculated based on the saw blade edge boundary line and tooth reference point to form visual wear characteristic data of the saw blade. The straightness deviation is calculated by fitting a straight line to the boundary line of the saw blade's rake face: the boundary line of the saw blade's rake face is a set of continuous pixel coordinate data representing the rake face region within the effective edge line. First, a straight line is fitted to the continuous pixel coordinate data using the least squares method to find the straight line equation that minimizes the sum of the squared differences of all pixel coordinates. After fitting the straight line, the vertical distance from each pixel to the fitted straight line is calculated, and the average absolute value of all vertical distances is taken as the straightness deviation. The straightness deviation is used to measure the degree of shape change of the saw blade's rake face during actual wear. For example, when the boundary line of the saw blade's rake face is fitted to obtain the straight line equation y=kx+b, the vertical distance from each boundary line pixel to the straight line equation is calculated using the point-to-line distance formula, and the average value of all pixel distances reflects the straightness deviation characteristics of the saw blade's rake face.

[0025] Calculating Side Edge Wear Based on Side Edge Boundary Line: The side edge boundary line is the coordinate data of continuous pixels representing the side edge region within the effective edge lines of the saw blade's real-time image feature data. Before sawing begins, the side edge boundary line of a new, unworn saw blade is photographed, and initial side edge boundary line coordinate data is obtained using the same method described above. During actual sawing, the difference between the current side edge boundary line coordinates and the initial side edge boundary line coordinates is calculated. Specifically, the two boundary line coordinate data are matched point-by-point, and the absolute difference in the lateral or longitudinal coordinate offset between the matched pixels is calculated. This absolute difference is used as the measured value of the contour wear. For example, for each point in the side edge boundary line coordinate data, the lateral or longitudinal deviation from the corresponding initial side edge boundary line coordinate point is calculated, and the average of all deviations is taken as the side edge contour wear, used to characterize the actual degree of wear on the side edge.

[0026] The saw blade retraction is calculated based on the saw blade edge boundary line and the saw tooth reference point: The saw blade retraction represents the distance the saw blade edge retracts relative to its initial state during actual cutting and wear. The method for determining the saw tooth reference point is as follows: In the initial image data of a new, unworn saw blade, a stable marker point is selected based on the saw tooth shape characteristics, such as the vertex or root of the saw tooth, as the reference point. In real-time image data captured during the actual wear process, a feature point on the saw blade boundary line corresponding to the reference point is found. The distance difference between the feature point in the real-time image and the initial reference point is calculated using the Euclidean distance formula between the two points. That is, the saw blade retraction is equal to the straight-line distance between the feature point on the saw blade edge in the real-time image data and the initial saw tooth reference point. By calculating for all saw blade boundary lines, multiple saw blade retraction amounts are obtained, and the average value is taken as the final saw blade retraction amount.

[0027] The straightness deviation, side edge contour wear, and cutting edge retraction obtained above are integrated into the visual wear characteristic data of the saw blade.

[0028] S3: Real-time acquisition of spindle current data during saw blade cutting of the workpiece, extraction of current change trend characteristics and current variability characteristics, and output of saw blade current wear characteristic data, including: Real-time acquisition of spindle current data during saw blade cutting of workpiece, followed by noise reduction and baseline correction of the spindle current data; A current sensor, preferably a Hall effect type, is installed in the power supply line of the spindle motor to ensure the accuracy and real-time performance of the current data measurement. To ensure the validity of the spindle current data, the range of the current sensor is selected based on the rated operating current of the spindle motor. For example, if the rated current of the spindle motor is 20 amps, a current sensor with a range of 0 to 50 amps can be selected. The output signal of the current sensor is converted into a digital signal by an analog-to-digital converter (ADC). The resolution of the ADC is determined based on the precision of the current fluctuations under actual operating conditions of the spindle motor. For example, when a current measurement resolution of 0.01 amps is required, an ADC with a resolution of 12 bits or higher is selected.

[0029] A Butterworth low-pass filter is selected to denoise the acquired spindle current data. The parameter determination method for the Butterworth low-pass filter is as follows: First, determine the spectral distribution of the spindle current data, that is, collect the raw spindle current data for a certain period of time under actual sawing conditions, and perform Fourier transform analysis to obtain the current spectral distribution characteristics; based on the analysis results of the current spectrum, determine the upper limit of the effective signal frequency and the starting position of the noise frequency. For example, if the spectral analysis shows that the effective current signal frequency is concentrated between 0 and 500 Hz, the filter cutoff frequency can be set to 600 Hz; the filter order is determined by balancing the filtering requirements and processing performance. For example, a 4th-order Butterworth low-pass filter can be selected to effectively suppress high-frequency noise while retaining the effective signal components in the spindle current data.

[0030] Baseline correction is performed based on the denoised spindle current data: First, a segment of spindle current data under no-load operation is collected. The length of the no-load current data is set to cover at least several complete cycles of spindle rotation, such as collecting 10 consecutive seconds of no-load spindle current data. The average value of the no-load current data is calculated and determined as the current baseline correction value. The real-time spindle current data obtained during actual machining needs to be subtracted from the current baseline correction value to eliminate the influence of current baseline drift.

[0031] The saw blade rotation cycle is generated based on the saw blade speed. The spindle current data is segmented according to the saw blade rotation cycle, and the average value and fluctuation amplitude of each segment are calculated. The saw blade rotation cycle is generated based on the saw blade speed: The spindle motor speed is obtained, for example, by using a photoelectric speed sensor to measure the spindle rotation speed in real time. The photoelectric speed sensor is installed at the output shaft end of the spindle motor, and generates a pulse signal for each rotation of the spindle. The real-time spindle speed is determined by counting the pulse signals per unit time. For example, if the photoelectric speed sensor counts 50 pulse signals in 1 second, the real-time spindle speed is determined to be 50 revolutions per second. The saw blade rotation cycle is then calculated by reciprocating the real-time spindle speed, which is 1 second divided by 50, i.e., 0.02 seconds.

[0032] The denoised and baseline-corrected spindle current data is segmented according to the saw blade rotation cycle: Based on continuously acquired spindle current data, it is segmented into equal-length segments according to the saw blade rotation cycle, that is, the spindle current data is sequentially divided into multiple data segments, each with a length of one saw blade rotation cycle. For example, each data segment is a continuous current data with a length of 0.02 seconds. For each data segment, the segment mean and segment fluctuation amplitude are calculated. The segment mean is the arithmetic mean of all currents within the data segment, and the segment fluctuation amplitude is calculated as the difference between the maximum and minimum currents within the data segment. The segment mean and segment fluctuation amplitude of each data segment are obtained respectively.

[0033] Based on the segmented average value, the current change trend characteristics are formed; based on the segmented fluctuation amplitude, the current variability characteristics are formed; and the saw blade current wear characteristic data are output. The segmented averages of multiple consecutive data segments are arranged chronologically to form a trend curve of the average current changing with sawing time. Trend analysis is then performed on this curve, and the least squares method is used to calculate the linear relationship between the segmented average of each data segment and time. The slope of this linear relationship quantitatively describes the trend of current change. The slope characterizes the trend of saw blade wear over time, reflecting the severity of wear.

[0034] Current variability characteristics are formed based on the segmented fluctuation amplitude of each data segment: Statistical analysis is performed on the segmented fluctuation amplitude of multiple consecutive data segments, including moving window statistics of the segmented fluctuation amplitude to determine the short-term fluctuation characteristics of the spindle current. A moving window is formed by selecting several consecutive data segments, for example, a window length of 50 data segments. The window moves forward sequentially, moving one data segment at a time. The average and standard deviation of the fluctuation amplitude of all data segments within each window are calculated, and the average and standard deviation of the fluctuation amplitude within the window are used as quantitative characteristics of current variability to reflect the degree of spindle current fluctuation. For example, if the average fluctuation amplitude or standard deviation within the window increases, it is determined that the saw blade is abnormal, such as tooth loss.

[0035] Finally, the current change trend characteristics and current variability characteristics obtained above are integrated to generate saw blade current wear characteristic data, including the segment mean and segment fluctuation amplitude corresponding to each data segment, as well as the slope of the current change trend characteristics obtained by linear fitting, and the mean and standard deviation of the current variability characteristics obtained by moving window statistics.

[0036] S4: Based on the visual wear characteristic data and the current wear characteristic data of the saw blade, wear degradation trend analysis and abnormal tooth loss detection analysis are performed respectively, and comprehensive evaluation data of the visual and current wear characteristics of the saw blade are output, including: The visual wear feature data of the saw blade and the current wear feature data of the saw blade are time-aligned to form an aligned feature sequence; Time alignment of saw blade visual wear feature data and saw blade current wear feature data: Both types of saw blade visual wear feature data and saw blade current wear feature data are time-stamped separately. The time-stamping method for the saw blade visual wear feature data is based on the timestamps of images captured by a high-speed camera. Each frame of real-time image feature data of the saw blade has a unique timestamp. The timestamp accuracy is determined according to the frame rate of the high-speed camera. For example, when the frame rate of the high-speed camera is 8000 frames per second, the timestamp accuracy of each frame is 0.000125 seconds. The time-stamping method for the saw blade current wear feature data uses a high-precision clock built into the current data acquisition system. The data acquisition frequency is an integer multiple of the spindle speed to ensure the timestamp accuracy of the spindle current data. For example, when the spindle speed is 50 revolutions per second, the current data acquisition frequency can be set to 5000 times per second, then the timestamp accuracy of the current data is 0.0002 seconds.

[0037] Using the timestamps of the saw blade's visual wear characteristic data as a benchmark, the timestamps of the saw blade's current wear characteristic data are interpolated or sampled to ensure that the two types of wear characteristic data accurately correspond to the same moment, achieving feature data alignment. The interpolation or sampling process uses linear interpolation. When the visual data timestamp is located between two adjacent current data timestamps, the distance ratio between the visual data timestamp and the two adjacent current data timestamps is calculated. The values ​​of the two adjacent current wear characteristic data are then weighted using this distance ratio to generate current wear characteristic data that is strictly aligned with the visual wear characteristic data timestamps.

[0038] After time alignment is completed, a unified alignment feature sequence is formed. The alignment feature sequence is a synchronous combination of the saw blade visual wear feature data and the saw blade current wear feature data at each moment, including: straightness deviation, side edge contour wear amount, cutting edge back amount, current change trend features and current variability features. Each data unit corresponds to a unified high-precision timestamp.

[0039] Based on the visual wear characteristic data of the saw blade, the rate of change of straightness deviation, the rate of change of side edge contour wear, and the rate of change of cutting edge retraction are calculated to form a visual degradation trend. Visual degradation trend features are calculated based on the visual wear characteristics of saw blades. These features include the rate of change of straightness deviation, the rate of change of side edge contour wear, and the rate of change of cutting edge retraction. A finite difference method is uniformly used, which involves dividing the difference between consecutive visual wear characteristic data points across multiple time stamps by the corresponding time interval to obtain the rate of change for each feature. For example, the calculation process for the rate of change of straightness deviation is as follows: the difference between two consecutive adjacent straightness deviations within the aligned feature sequence is taken as the change, and then the change is divided by the time difference between the corresponding two time stamps to obtain the rate of change of straightness deviation within the time period. The rates of change of side edge contour wear and cutting edge retraction are also calculated using the same finite difference method to form the visual degradation trend features.

[0040] Based on the saw blade current wear characteristic data, the slope of the current change trend characteristic and the increment of the current variability characteristic are calculated. Based on the alignment characteristic sequence, the abrupt segment of the current variability characteristic is determined and the abnormal tooth loss segment is marked. The comprehensive evaluation data of the saw blade visual and current wear characteristics is output. The method for calculating the slope of the current change trend feature is as follows: the piecewise mean of multiple consecutive current change trend features in the aligned feature sequence is used for linear fitting. The least squares method is used for linear fitting, and the slope of the trend feature of the feature data changing with time is used as a quantitative index to reflect the strength of the current change trend.

[0041] The method for calculating the increment of current variability characteristics is as follows: subtract the current variability characteristics of two consecutive adjacent time points to obtain the increment of current variability characteristics, thereby quantifying the degree to which the amplitude of the main shaft current fluctuation changes over time.

[0042] Abrupt segment detection and abnormal tooth loss segments are performed on current variability features based on a unified alignment feature sequence. The threshold for determining abrupt segments in current variability features is established by conducting long-term statistical analysis of current variability feature data under normal saw blade cutting conditions, calculating the average and standard deviation of the current variability features within the normal fluctuation range, and using the average plus a certain multiple of the standard deviation as the abrupt segment determination threshold (e.g., the average plus three times the standard deviation). Each increment of current variability feature in the alignment feature sequence is evaluated; when the increment of current variability feature at a certain moment exceeds the determination threshold, it is determined to be an abrupt segment of the current variability feature.

[0043] For detected abrupt changes in current variability characteristics, abnormal tooth-drop segments are marked: when the increment of current variability characteristics at a certain timestamp continuously exceeds the judgment threshold for multiple data segments, it indicates that abnormal tooth drop has occurred at the timestamp and multiple adjacent timestamps. These multiple data segments are collectively referred to as abnormal tooth-drop segments. For example, if the increment of current variability characteristics in three or more consecutive data segments all exceed the judgment threshold, the starting timestamp of this segment is marked as the start time of the abnormal tooth-drop segment, and correspondingly, the ending timestamp of the consecutive data segments exceeding the judgment threshold is marked as the end time of the abnormal tooth-drop segment.

[0044] Finally, all the information obtained above, including visual degradation trend features, slope of current change trend features, increment of current variability features, and marked abnormal tooth loss segments, is integrated and output to form comprehensive evaluation data of saw blade visual and current wear characteristics. The comprehensive evaluation data includes the rate of change of straightness deviation, the rate of change of side edge contour wear, and the rate of change of cutting edge retraction for visual degradation trend features corresponding to each time stamp, as well as the slope of current change trend features and the increment of current variability features; simultaneously, the location and duration of abrupt changes in current variability features are marked to identify and track the timing and duration of abnormal tooth loss events.

[0045] S5: Based on the comprehensive evaluation data of saw blade visual and current wear characteristics, calculate the saw blade wear threshold and tooth loss threshold respectively, and output the threshold feature data of saw blade visual and current fusion judgment, including: Based on the comprehensive evaluation data of saw blade visual and electrical wear characteristics, the alignment feature sequences corresponding to the stable growth segment and the abnormal tooth loss segment of the visual degradation trend are extracted respectively. The comprehensive evaluation data of visual and electrical wear characteristics of saw blades are used for trend classification and abnormal segment identification to divide the stable growth segment of visual degradation trend and the abnormal tooth loss segment. The stable growth segment of visual degradation trend is the period when the visual wear characteristic data continues to increase steadily throughout the entire processing process; the abnormal tooth loss segment is the period when the current variability characteristic marked in the aligned feature sequence shows abrupt changes and continuously exceeds the abrupt change judgment threshold.

[0046] The method for extracting the aligned feature sequence corresponding to the stable growth segment of the visual degradation trend is as follows: Visual degradation trend curves are plotted based on the rate of change of straightness deviation, the rate of change of side edge contour wear, and the rate of change of cutting edge retraction. A sliding window method is used to detect the fluctuation of the curves over a certain period. The length of the sliding window can be determined according to the processing characteristics and the variation law of the saw blade wear rate; for example, the sliding window length can be set to 50 consecutive timestamps of data points. Standard deviation analysis is performed on the rate of change of straightness deviation, the rate of change of side edge contour wear, and the rate of change of cutting edge retraction within each sliding window to quantitatively assess the degree of curve fluctuation. If the standard deviation of each rate of change within the window is less than a pre-set stability threshold, the data segment within the window is determined to belong to the stable growth segment of the visual degradation trend. The stability threshold is determined as follows: Under normal and stable cutting conditions of the saw blade, the standard deviations of the three rates of change are statistically analyzed over a relatively long time period. The upper limit of the statistical standard deviation is taken as the stability threshold. For example, the stability threshold can be set as the average of the statistical standard deviations plus two standard deviations.

[0047] The method for extracting the alignment feature sequence corresponding to the abnormal tooth loss segment is as follows: Based on the marked current variability feature mutation segment positions in the comprehensive evaluation data, all data between the start and end timestamps of each abnormal tooth loss segment is extracted to form the alignment feature sequence of the abnormal tooth loss segment. The start and end timestamps of the abnormal tooth loss segment are obtained through the current variability feature mutation judgment method. When multiple consecutive data segments of the current variability feature exceed the current variability feature mutation segment judgment threshold, the timestamp of the first data segment exceeding the threshold is recorded as the start timestamp, and the timestamp of the last consecutive data segment exceeding the threshold is recorded as the end timestamp.

[0048] Within the stable growth phase of visual degradation trend, the statistical boundaries of the straightness deviation change rate, the side edge profile wear change rate, and the cutting edge retraction change rate are calculated to form the wear threshold. Within the alignment feature sequence of the extracted stable growth segment of visual degradation trend, the data distribution of the rate of change in straightness deviation, the rate of change in side edge contour wear, and the rate of change in edge retreat are statistically analyzed. The statistical distribution characteristics of each rate of change are calculated throughout the entire stable growth segment. A normal distribution hypothesis test is used to determine the data distribution characteristics, and the Shapiro-Wilk method is used to determine whether the data for each rate of change conforms to a normal distribution. If the data conforms to a normal distribution, the statistical boundaries for each rate of change are determined based on the mean and standard deviation of the statistical data. The method for determining the statistical boundaries is to use the mean plus or minus a certain multiple of the standard deviation as the upper and lower limits of the statistical boundaries; for example, setting the upper and lower limits of the statistical boundaries to the mean plus or minus three times the standard deviation, respectively. If the data does not conform to a normal distribution, a nonparametric statistical method is used to determine the statistical boundaries. Specifically, the kernel density estimation method is used to calculate the probability density function of the rate of change, and the statistical boundaries are determined using a specific confidence interval of the probability density function; for example, using a 95% confidence level as the upper and lower limits of the statistical boundaries.

[0049] For the statistical boundaries of the three obtained rates of change, the boundary closer to the actual application requirements is selected as the wear threshold for each. For example, an increase in the rate of change of straightness deviation indicates a greater degree of saw blade wear, so the upper limit of the statistical boundary is selected as the wear threshold for the rate of change of straightness deviation. Similarly, for the rate of change of side edge contour wear and the rate of change of cutting edge retraction, the statistical boundary on the side with the greater rate of change is used as the wear threshold for each. The wear thresholds of the final obtained rate of change of straightness deviation, rate of change of side edge contour wear, and rate of change of cutting edge retraction together constitute the complete threshold feature data for visual wear state determination.

[0050] Within the abnormal tooth loss segment, the number of occurrences of the abrupt segment based on the current variability characteristic is counted in correspondence with the saw blade rotation cycle to form a threshold for the number of tooth loss, and output threshold feature data for the saw blade visual and current fusion judgment. Within the abnormal tooth loss segment, each instance of abrupt changes in current variability is marked, and the timestamp of each event is recorded. Based on the saw blade rotation cycle, the abnormal tooth loss segment is subdivided into multiple periodic windows, each with a rotation cycle length of 0.02 seconds. For example, if the saw blade rotation cycle is 0.02 seconds, the entire abnormal tooth loss segment is divided into periodic windows of 0.02 seconds each. The number of abrupt changes in current variability within each periodic window is counted separately to obtain the total number of abrupt changes in each periodic window.

[0051] Based on the abrupt change pattern exhibited by the spindle current variation characteristics when saw blade teeth fall out, a relationship model is established between the number of tooth-falling events and the number of abrupt change events. Methods for establishing this model include, but are not limited to, linear regression models. Experimental methods are used to obtain the average number of abrupt change events corresponding to a known number of tooth falls on the saw blade. The least squares method is then used to fit the linear relationship equation between the number of abrupt change events and the actual number of tooth falls, obtaining the correlation coefficient for the actual number of tooth falls corresponding to each abrupt change event. This represents the correspondence between the number of tooth falls and the number of abrupt change events.

[0052] Based on the relationship coefficient, the number of abrupt events within each periodic window is converted into the number of tooth losses. For example, if the linear relationship model states that the number of tooth losses equals the number of abrupt events multiplied by 0.5, then when four current variability characteristic abrupt events are recorded within a certain periodic window, it is determined that the actual number of tooth losses within that periodic window is two. The distribution of the actual number of tooth losses obtained across all periodic windows is statistically analyzed, and the statistical mean and standard deviation of the number of tooth losses are calculated. This statistical measure is used to determine the threshold for the number of tooth losses. The method for determining the threshold for the number of tooth losses is as follows: the statistical mean plus a certain multiple of the statistical standard deviation is used as the threshold for the number of tooth losses. For example, the statistical mean plus three times the standard deviation is used. This threshold is used to determine the severity of abnormal tooth loss on the saw blade.

[0053] Finally, the threshold values ​​for the rate of change of straightness deviation, the rate of change of side edge contour wear, and the rate of change of cutting edge retreat within the stable growth segment of visual degradation trend, determined above, are integrated with the threshold value for the number of missing teeth within the abnormal tooth loss segment to form the threshold feature data for the visual and electrical fusion judgment of the saw blade.

[0054] S6: Based on threshold feature data and combined with processed product quality indicators, determine whether the saw blade generates a trigger signal for scrapping or replacement, including: Based on the threshold feature data determined by the fusion of visual and electrical characteristics of the saw blade, the processing product size deviation, surface finish data and processing product defect rate corresponding to the processing product quality indicators are collected, and the changing trend of the processing product quality indicators is calculated. The dimensional deviation data of the processed product refers to the absolute deviation between the actual size and the design size generated by the saw blade during the sawing process. The method for collecting this data is as follows: a coordinate measuring machine (CMM) is used to inspect the dimensions of the processed workpiece to obtain its actual size data. The CMM's measurement accuracy is set according to the dimensional tolerance requirements of the processed product. For example, if the tolerance requirement is ±0.01 mm, the CMM's measurement accuracy is set to 0.005 mm or higher to ensure the accuracy of the measurement data. Multiple measurements are performed on each processed workpiece. The number of measurement points on each workpiece is determined based on the complexity of the workpiece's geometry. For example, for workpieces with regular geometric shapes, at least 10 measurement points are taken per workpiece. The absolute deviation between the actual size and the design size of all measurement points is calculated, and the arithmetic mean of the deviation values ​​is taken as the dimensional deviation data of the processed product.

[0055] Surface finish data of processed products is a quantitative indicator of workpiece surface roughness. The method for collecting surface finish data is as follows: a surface roughness measuring instrument is used, specifically a pin-type profilometer. The probe moves at a constant speed along the workpiece surface to measure the surface profile changes, outputting the surface roughness Ra value. The probe speed and measurement range of the surface roughness measuring instrument are determined based on the surface quality requirements of the processed workpiece. For example, when the required surface roughness Ra is 0.8 micrometers, the probe speed is set to 0.1 mm / s to 0.5 mm / s, and the measurement range is 4 mm to 5 mm to ensure measurement accuracy. Multiple typical surface locations are selected for measurement for each processed workpiece; for example, at least five different areas are selected for measurement on each workpiece. The arithmetic mean of the Ra values ​​at each measurement location is taken as the surface finish data of the processed product.

[0056] The product defect rate refers to the percentage of processed workpieces that fail to meet pre-set quality requirements out of a certain number of processed workpieces. The statistical method for the defect rate is as follows: A certain number of processed workpieces are continuously produced within a specified time period, for example, 100 workpieces. Each workpiece is inspected according to pre-set quality inspection standards, including multiple indicators such as dimensional tolerances, surface finish, and appearance defects. If any indicator of a processed workpiece fails to meet the pre-set quality requirements, the workpiece is deemed a defective product. The number of defective workpieces divided by the total number of workpieces produced, multiplied by 100%, is the product defect rate.

[0057] Based on the obtained data on dimensional deviations, surface finish, and defect rates of processed products, the trends in these quality indicators are calculated. The calculation method is as follows: Trend analysis is performed on the dimensional deviation, surface finish, and defect rate data separately. The moving average method is used to smooth short-term data fluctuations and highlight long-term trends. A fixed-length moving window is selected; for example, the most recent 20 consecutive data points are used as one moving window. The arithmetic mean of the data within each window is calculated sequentially to obtain a moving average sequence. Then, a linear fitting method is used to perform trend fitting analysis on the moving average sequence. The slope of the linear trend fitting line is calculated using the least squares method. The slope of the linear trend fitting line quantifies the trend of the processed product quality indicators. A positive slope indicates a deteriorating trend in the quality indicators; a larger slope indicates a more pronounced deterioration.

[0058] The trend of changes in the quality indicators of processed products is correlated and compared with the threshold feature data of the saw blade visual and current fusion judgment to determine whether to generate a trigger signal for saw blade scrapping or replacement. This study establishes the correspondence between the changing trends of processed product quality indicators and the threshold feature data used in the visual and electrical fusion judgment of the saw blade. Using the thresholds for the rate of change of straightness deviation, side edge contour wear, cutting edge retraction, and tooth loss in the visual wear state of the saw blade as benchmarks, correlation analysis is performed with the trend slopes of dimensional deviation, surface finish, and defect rate of processed products. The correlation analysis method employs Pearson correlation coefficient analysis to calculate the correlation coefficient between the trend slope of each quality indicator and each threshold feature data. Data pairs are constructed by simultaneously obtaining the changing trend data of quality indicators over multiple consecutive processing cycles and the threshold feature data used in the visual and electrical fusion judgment of the saw blade, and the Pearson correlation coefficient for each data pair is calculated. The calculated Pearson correlation coefficient values ​​determine the correlation strength between the changing trends of processed product quality indicators and each threshold feature data; for example, a correlation coefficient greater than 0.8 indicates a strong correlation, less than 0.3 indicates a weak correlation, and between 0.3 and 0.8 indicates a moderate correlation.

[0059] Finally, a trigger signal for saw blade scrapping or replacement is generated. The judgment criteria are as follows: when the correlation coefficient between the trend of the processed product quality indicators and the threshold feature data determined by the fusion of visual and current measurements of the saw blade reaches a strong correlation level, and the slope of the trend of the processed product quality indicators exceeds the preset quality control threshold, the signal for saw blade scrapping or replacement is triggered. The preset quality control threshold is determined as follows: data is collected at the beginning of the production process or when the saw blade is new, and historical data on the trend of the processed product quality indicators under normal production conditions are recorded. The maximum value of the trend slope of the historical data is used as the benchmark value, and the quality control threshold is set as a multiple of the benchmark value, for example, the quality control threshold is set between 1.5 and 2 times the benchmark value.

[0060] Once the judgment conditions are met, a trigger signal for saw blade scrapping or replacement is automatically generated. The trigger signal includes, but is not limited to, an automatic alarm signal or an automatic shutdown and replacement signal, which is used to prompt equipment maintenance personnel to take corresponding maintenance measures in a timely manner to ensure the normal operation of the processing equipment and the quality stability of the processed products.

[0061] By using the above methods, the actual changes in the quality of processed products can be correlated with the real-time wear status of the saw blade in a timely and accurate manner. This effectively avoids the decline in the quality of processed products or scrap due to saw blade wear or tooth loss, ensuring the safe and efficient operation of processing equipment and the continuous and stable control of product quality.

[0062] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0063] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0064] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0065] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0066] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0067] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0068] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion 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 described in the various embodiments of this application. 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.

[0069] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0070] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A saw blade detection method based on multi-source data fusion, characterized in that, Includes the following steps: S1: Collect image data of the saw blade's front cutting face, side cutting edge, and cutting edge when the saw blade is sawing the workpiece, evaluate the image clarity and effectiveness, and output real-time image feature data of the saw blade; S2: Based on real-time image feature data of the saw blade, analyze the changes in the projection line of the saw blade's front face, the changes in the side edge wear pattern, and the changes in the amount of backlash at the edge position, and output the visual wear feature data of the saw blade. S3: Real-time acquisition of spindle current data when the saw blade is cutting the workpiece, extraction of current change trend characteristics and current variability characteristics, and output of saw blade current wear characteristic data; S4: Based on the visual wear characteristic data and the current wear characteristic data of the saw blade, the wear degradation trend analysis and abnormal tooth loss detection analysis are performed respectively, and the comprehensive evaluation data of the visual and current wear characteristics of the saw blade are output. S5: Based on the comprehensive evaluation data of saw blade visual and current wear characteristics, calculate the saw blade wear threshold and tooth loss threshold respectively, and output the threshold feature data of saw blade visual and current fusion judgment. S6: Based on the threshold feature data and the quality indicators of the processed products, determine whether the saw blade generates a trigger signal for scrapping or replacement.

2. The saw blade detection method based on multi-source data fusion according to claim 1, characterized in that, S1, specifically: During the sawing process, the front cutting surface, side cutting edge and cutting edge of the saw blade exposed during the sawing process are continuously photographed. Based on image data of the saw blade's front face, side edge, and cutting edge, the image brightness, focus status, and motion blur are determined. Images that meet the imaging requirements are selected, and invalid images are removed to form real-time image feature data of the saw blade.

3. The saw blade detection method based on multi-source data fusion according to claim 2, characterized in that, S2, specifically: Based on real-time image feature data of the saw blade, grayscale normalization and edge extraction are performed on the saw tooth area to obtain the boundary line of the saw blade front face, the boundary line of the side edge, and the boundary line of the saw blade cutting edge. Straightness deviation is calculated by straight-line fitting of the saw blade front face boundary line, side edge profile wear is calculated based on the side edge boundary line, and edge backing is calculated based on the saw blade edge boundary line and the saw tooth reference point, thus forming visual wear characteristic data of the saw blade.

4. The saw blade detection method based on multi-source data fusion according to claim 3, characterized in that, S3, specifically: Real-time acquisition of spindle current data during saw blade cutting of workpiece, followed by noise reduction and baseline correction of the spindle current data; The saw blade rotation cycle is generated based on the saw blade speed. The spindle current data is segmented according to the saw blade rotation cycle, and the average value and fluctuation amplitude of each segment are calculated. Based on the segmented average value, the current change trend characteristics are formed; based on the segmented fluctuation amplitude, the current variability characteristics are formed; and the saw blade current wear characteristic data are output.

5. The saw blade detection method based on multi-source data fusion according to claim 4, characterized in that, S4, specifically: The visual wear feature data of the saw blade and the current wear feature data of the saw blade are time-aligned to form an aligned feature sequence; Based on the visual wear characteristic data of the saw blade, the rate of change of straightness deviation, the rate of change of side edge contour wear, and the rate of change of cutting edge retraction are calculated to form a visual degradation trend. Based on the saw blade current wear characteristic data, the slope of the current change trend characteristic and the increment of the current variability characteristic are calculated. Based on the alignment characteristic sequence, the abrupt segment of the current variability characteristic is determined and the abnormal tooth loss segment is marked. The comprehensive evaluation data of the saw blade visual and current wear characteristics is output.

6. The saw blade detection method based on multi-source data fusion according to claim 5, characterized in that, S5, specifically: Based on the comprehensive evaluation data of saw blade visual and electrical wear characteristics, the alignment feature sequences corresponding to the stable growth segment and the abnormal tooth loss segment of the visual degradation trend are extracted respectively. Within the stable growth phase of visual degradation trend, the statistical boundaries of the straightness deviation change rate, the side edge profile wear change rate, and the cutting edge retraction change rate are calculated to form the wear threshold. Within the abnormal tooth loss segment, the number of occurrences of the abrupt change segment based on the current variability characteristic is counted in correspondence with the saw blade rotation cycle to form a threshold for the number of tooth loss, and output threshold feature data for the saw blade visual and current fusion judgment.

7. The saw blade detection method based on multi-source data fusion according to claim 6, characterized in that, S6, specifically: Based on the threshold feature data determined by the fusion of visual and electrical characteristics of the saw blade, the processing product size deviation, surface finish data and processing product defect rate corresponding to the processing product quality indicators are collected, and the changing trend of the processing product quality indicators is calculated. The trend of changes in the quality indicators of the processed products is correlated and compared with the threshold feature data of the saw blade visual and current fusion judgment to determine whether to generate a trigger signal for the saw blade to be scrapped or replaced.

Images