Excavator bucket tooth detection method, system and electronic equipment

Abstract

The present invention provides an excavator bucket tooth detection method, system and electronic equipment, which relate to the field of engineering machinery operation and maintenance. The method fuses data from three types of multimodal sensors to detect the bucket tooth life, alleviating the detection limitations of a single sensor. The method can also use the relationship curve obtained by data fusion processing to perform preventive detection and maintenance on the bucket teeth, thereby improving the reliability of the bucket teeth under complex working conditions.

Description

Technical Field

[0001] The present invention relates to the field of engineering machinery operation and maintenance, and in particular to an excavator bucket tooth detection method, system and electronic equipment. Background Art

[0002] Traditional methods for inspecting excavator bucket teeth rely primarily on manual visual inspection, which is prone to missed detections and lacks real-time monitoring. While existing technologies employ sensors for automated inspection, most rely on a single sensor type, which is susceptible to significant interference from ambient noise and results in a high false alarm rate. Furthermore, existing bucket tooth inspection techniques cannot predict tooth lifespan using historical data, leading to unplanned downtime and difficulty implementing preventative inspection and maintenance, thus compromising the reliability of bucket tooth inspection. Summary of the Invention

[0003] In view of this, the purpose of the present invention is to provide an excavator bucket tooth detection method, system and electronic equipment. This method fuses data from three types of multimodal sensors to realize the detection of bucket tooth life, alleviates the detection limitations of a single sensor, and can use the relationship curve obtained by data fusion processing to perform preventive detection and maintenance on bucket teeth, thereby improving the reliability of bucket teeth under complex working conditions.

[0004] In a first aspect, an embodiment of the present invention provides a method for detecting bucket teeth of an excavator, the method comprising:

[0005] Data collection steps: Based on the visual sensor, vibration sensor and stress sensor set up on the excavator, the visual data, vibration data and stress data corresponding to the bucket teeth in the excavator are respectively collected;

[0006] Feature extraction step: extracting the visual features, vibration features and stress features corresponding to the bucket teeth according to the spatial feature information of the visual data, the time-frequency feature information of the vibration data and the energy spectrum feature information corresponding to the stress data;

[0007] Feature fusion step: using the feature fusion results of visual features, vibration features and stress features to determine the relationship curve between the stress distribution entropy value and the life value corresponding to the bucket tooth;

[0008] Detection output steps: obtaining the stress distribution entropy value and life value corresponding to the number of working cycles of the bucket tooth according to the relationship curve, and determining the detection result of the bucket tooth using the stress distribution entropy value and life value.

[0009] Optional data collection steps include:

[0010] Obtain a vision unit installed on the side of the bucket of the excavator, and use a vision sensor provided in the vision unit to collect vision data corresponding to the bucket teeth at a first sampling rate;

[0011] Obtaining a vibration sensor array installed at the bucket of the excavator, and using vibration sensors provided in the vibration sensor array to collect vibration data corresponding to the bucket teeth at a second sampling rate;

[0012] A stress monitoring unit embedded in the root of the bucket tooth is obtained, and stress data corresponding to the bucket tooth is collected using a strain sensor in the stress monitoring unit at a third sampling rate.

[0013] Optionally, feature extraction steps include:

[0014] Calculate the spatial feature vector of the visual data, and determine the spatial feature information corresponding to the visual data using the temporal variation characteristics corresponding to the spatial feature vector; extract the defect area characteristics contained in the bucket tooth based on the spatial feature information, and determine the visual features based on the defect area characteristics;

[0015] Determine the time-frequency characteristics of the impact signal contained in the vibration data, use the time-frequency characteristics of the impact signal to determine the time-frequency characteristic information corresponding to the vibration data, and use the time-frequency characteristic information and time domain data corresponding to the kurtosis and impulse factor in the vibration data to determine the vibration characteristics;

[0016] The energy spectrum characteristics corresponding to the stress data are calculated using wavelet packet decomposition, the energy spectrum characteristics are used to determine the energy spectrum characteristic information, and the stress characteristics are determined using the statistics corresponding to the energy spectrum characteristic information and the mean square error value and peak factor in the stress data.

[0017] Optional feature fusion step, including:

[0018] Determining a first weight coefficient, a second weight coefficient, and a third weight coefficient based on abnormal data included in the visual feature, the vibration feature, and the stress feature, respectively;

[0019] The first weight coefficient, the second weight coefficient and the third weight coefficient are used to perform cascade splicing and weighted fusion calculation on the visual features, the vibration features and the stress features to obtain the feature fusion result corresponding to the bucket tooth;

[0020] Determine the stress distribution entropy value corresponding to the bucket tooth based on the feature fusion result, and determine the defect probability and defect type corresponding to the bucket tooth based on the stress distribution entropy value;

[0021] The life value of bucket teeth is determined by using the defect probability and defect type, and the relationship curve is determined by using the life value and the stress distribution entropy value.

[0022] Optionally, determining the first weight coefficient, the second weight coefficient, and the third weight coefficient based on abnormal data included in the visual feature, the vibration feature, and the stress feature, respectively, includes:

[0023] Determining features of a defective area contained in the bucket tooth using the visual features, and determining a first weight coefficient corresponding to the visual features based on the features of the defective area;

[0024] Determining an abnormal impact feature contained in the bucket tooth using the vibration feature, and determining a second weight coefficient corresponding to the vibration feature based on the abnormal impact feature;

[0025] The stress feature is used to determine a stress offset feature included in the bucket tooth, and a third weight coefficient corresponding to the stress feature is determined based on the stress offset feature.

[0026] Optionally, the output detection step includes:

[0027] Get the working cycle times of bucket teeth in real time;

[0028] The stress distribution entropy value corresponding to the number of working cycles is determined by using the three-stage segmented curve included in the relationship curve, and the life value corresponding to the number of working cycles is determined by using the two-stage segmented curve included in the relationship curve;

[0029] The health risk level of the bucket tooth is determined according to the stress distribution entropy value and the life value, and the health risk level is used to determine the test result.

[0030] Optionally, the three-stage segmented curve is:

[0031] ;

[0032] in, is the normalized entropy value corresponding to the stress distribution entropy value; is the number of working cycles; is the initial growth coefficient; is the exponential growth factor; is the platform constant, used to characterize the entropy benchmark of the second stage; is the fine-tuning coefficient; is the decay index; is the critical working cycle number; is the quadratic growth coefficient; is the secondary growth index;

[0033] The two-stage segmented curve is:

[0034] ;

[0035] in, is the remaining life ratio corresponding to the life value; is the linear damage rate coefficient; is the critical remaining life ratio; is the exponential decay coefficient; is the power law damage acceleration coefficient; is the power exponent.

[0036] Optionally, after the data collection step, the method further comprises a preprocessing step:

[0037] After downsampling and histogram equalization of the visual data using a preset first parameter, first time series data corresponding to the visual data is obtained;

[0038] After filtering the vibration data using a second parameter corresponding to a preset filter, second time series data corresponding to the vibration data is obtained;

[0039] After performing Kalman filtering on the stress data using a preset third parameter, third time series data corresponding to the stress data is obtained;

[0040] The sampling rates corresponding to the first time series data, the second time series data, and the third time series data are padded and calculated, and the visual data, the vibration data, and the stress data are respectively updated using the padded calculation results.

[0041] In a second aspect, the present invention provides an excavator bucket tooth detection system, the system comprising:

[0042] A data acquisition module is used to respectively collect visual data, vibration data and stress data corresponding to the bucket teeth in the excavator based on the visual sensor, vibration sensor and stress sensor set up in the excavator;

[0043] A feature extraction module is used to extract visual features, vibration features, and stress features corresponding to the bucket teeth based on spatial feature information of visual data, time-frequency feature information of vibration data, and energy spectrum feature information corresponding to stress data;

[0044] A feature fusion module is used to determine the relationship curve between the stress distribution entropy value and the life value corresponding to the bucket tooth using the feature fusion results of visual features, vibration features and stress features;

[0045] The detection output module is used to obtain the stress distribution entropy value and life value corresponding to the working cycle number of the bucket tooth according to the relationship curve, and use the stress distribution entropy value and life value to determine the detection result of the bucket tooth.

[0046] In a third aspect, an embodiment of the present invention further provides an electronic device comprising a processor and a memory, wherein the memory stores computer-executable instructions that can be executed by the processor, and the processor executes the computer-executable instructions to implement the steps of the excavator bucket tooth detection method provided in the first aspect.

[0047] In a fourth aspect, an embodiment of the present invention further provides a storage medium storing computer executable instructions. When the computer executable instructions are called and executed by a processor, the computer executable instructions prompt the processor to implement the steps of the excavator bucket tooth detection method provided in the first aspect.

[0048] The present invention provides a method, system, and electronic device for detecting bucket teeth for excavators. During the detection process of bucket teeth for excavators, the method first collects visual data, vibration data, and stress data corresponding to the bucket teeth in the excavator based on the visual sensor, vibration sensor, and stress sensor installed on the excavator, thereby completing data acquisition. Then, the visual features, vibration features, and stress features corresponding to the bucket teeth are extracted based on the spatial feature information of the visual data, the time-frequency feature information of the vibration data, and the energy spectrum feature information corresponding to the stress data, thereby completing the feature extraction process. Subsequently, the feature fusion results of the visual features, vibration features, and stress features are used to determine a relationship curve corresponding to the bucket teeth, including the stress distribution entropy value and life value, thereby completing the feature fusion process. Finally, the stress distribution entropy value and life value corresponding to the number of working cycles of the bucket teeth are obtained based on the relationship curve, and the stress distribution entropy value and life value are used to determine the detection result of the bucket teeth, thereby obtaining the detection output process of the excavator bucket teeth. This method fuses data from three types of multimodal sensors to detect bucket tooth life, alleviating the detection limitations of a single sensor. Furthermore, the relationship curve obtained through data fusion processing can be used to perform preventive detection and maintenance on the bucket teeth, thereby improving the reliability of the bucket teeth under complex working conditions.

[0049] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or understood by practicing the present invention. The purposes and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0050] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0052] Figure 1 A flowchart of a method for detecting bucket teeth of an excavator provided by an embodiment of the present invention;

[0053] Figure 2 This is a flow chart of step S101 in a method for detecting bucket teeth of an excavator provided by an embodiment of the present invention;

[0054] Figure 3 This is a flow chart of step S102 in a method for detecting bucket teeth of an excavator provided by an embodiment of the present invention;

[0055] Figure 4 This is a flow chart of step S103 of a method for detecting bucket teeth of an excavator provided by an embodiment of the present invention;

[0056] Figure 5 This is a flow chart of step S401 of a method for detecting bucket teeth of an excavator provided by an embodiment of the present invention;

[0057] Figure 6 This is a flow chart of step S104 of a method for detecting bucket teeth of an excavator provided by an embodiment of the present invention;

[0058] Figure 7 A flowchart of an excavator bucket tooth detection method including a preprocessing step provided by an embodiment of the present invention;

[0059] Figure 8 A flow chart of another excavator bucket tooth detection method provided by an embodiment of the present invention;

[0060] Figure 9 A schematic diagram of sensor deployment in an excavator bucket tooth detection method provided by an embodiment of the present invention;

[0061] Figure 10 A graph showing the relationship between stress distribution entropy and bucket tooth life in an excavator bucket tooth detection method provided by an embodiment of the present invention;

[0062] Figure 11 A schematic structural diagram of an excavator bucket tooth detection system provided by an embodiment of the present invention;

[0063] Figure 12 A schematic structural diagram of an electronic device provided by an embodiment of the present invention.

[0064] icon:

[0065] 1110 - data acquisition module; 1120 - feature extraction module; 1130 - feature fusion module; 1140 - detection output module;

[0066] 1-First visual sensor; 2-Second visual sensor; 3-Vibration sensor; 4-Stress sensor; 5-Edge computing module; 6-Decision-making and warning module;

[0067] 101 - processor; 102 - memory; 103 - bus; 104 - communication interface. DETAILED DESCRIPTION

[0068] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0069] In the detection of bucket teeth of excavators, the traditional method mainly relies on manual visual inspection, which is prone to missed detection and cannot be monitored in real time; although relevant sensors are used for automatic detection in the existing technology, most of them use a single type of sensor, which is greatly interfered by environmental noise and has a high false alarm rate (such as mistakenly identifying rock impact as bucket tooth defects). In addition, bucket tooth detection in the existing technology cannot predict the life of bucket teeth through historical data, which is prone to unplanned downtime and difficult to achieve preventive detection and maintenance, thereby affecting the reliability of bucket tooth detection. Based on this, the present invention implements a method, system and electronic equipment for detecting bucket teeth for excavators. The method fuses data from three types of multimodal sensors to detect bucket tooth life, alleviates the detection limitations of a single sensor, and can use the relationship curve obtained by data fusion processing to perform preventive detection and maintenance on bucket teeth, thereby improving the reliability of bucket teeth under complex working conditions.

[0070] To facilitate understanding of this embodiment, a method for detecting bucket teeth of an excavator disclosed in an embodiment of the present invention is first described in detail. Figure 1 As shown, the method includes:

[0071] Data collection step S101: Based on the visual sensor, vibration sensor and stress sensor provided on the excavator, visual data, vibration data and stress data corresponding to the bucket teeth in the excavator are respectively collected.

[0072] A high-resolution industrial camera should be appropriately positioned near the excavator's bucket teeth as a visual sensor. This camera should have a sufficient frame rate to capture the dynamic changes of the bucket teeth during operation. When collecting visual data, it's important not only to capture the overall appearance of the bucket teeth, but also to focus on capturing key areas of the tooth, such as the tooth tip and the tooth body. To improve data accuracy, multiple visual sensors at different angles can be set up to capture the bucket teeth from multiple perspectives, comprehensively capturing visual information such as the tooth's shape, degree of wear, and surface cracks. Furthermore, the captured image data must be preprocessed, including denoising, grayscaling, and contrast enhancement, to improve subsequent feature extraction.

[0073] A high-precision accelerometer is selected as the vibration sensor and securely mounted at an appropriate location on the bucket tooth, such as the middle or root of the tooth body, to ensure accurate measurement of the tooth's vibration during operation. The vibration data collected by the vibration sensor includes information such as the tooth's vibration frequency and amplitude under different operating conditions. During data collection, data should be categorized according to the excavator's operating mode (such as digging, lifting, and slewing) to facilitate subsequent analysis of the bucket tooth's vibration characteristics under different operating conditions. Furthermore, to ensure data stability and reliability, multiple vibration sensors can be used for redundant data collection. The collected data is then filtered to remove high-frequency noise and interference signals.

[0074] Stress sensors, using strain gauges or piezoresistive stress sensors, are attached or mounted on bucket teeth at stress concentration points, such as the transition area between the tip and the body, and at the tooth root. These sensors measure the magnitude and direction of stress experienced by the bucket teeth in real time during operation. When collecting stress data, it's important to consider the varying stress patterns under different operating conditions, such as the impact of digging and impact forces on bucket tooth stress. To improve stress data accuracy, the sensors should be regularly calibrated, and appropriate signal amplification and acquisition circuits should be employed to ensure accurate stress data acquisition.

[0075] Feature extraction step S102: extracting visual features, vibration features and stress features corresponding to the bucket teeth according to the spatial feature information of the visual data, the time-frequency feature information of the vibration data and the energy spectrum feature information corresponding to the stress data.

[0076] The visual feature extraction process uses computer vision technology to extract the spatial features of the bucket teeth based on the collected visual data. First, edge detection algorithms (such as the Canny algorithm) are used to extract the tooth's outline edges, thereby determining the tooth's shape characteristics. Next, image segmentation techniques are used to separate the tooth from the background, further analyzing the tooth's surface texture characteristics, such as texture changes in worn areas and crack morphology. Furthermore, deep learning methods, such as convolutional neural networks (CNNs), can be used to train bucket tooth images to automatically extract more advanced visual features, such as the degree of tooth wear and the length and width of cracks.

[0077] When extracting vibration features, the collected vibration data is converted into time-domain feature information using time-frequency analysis methods (such as short-time Fourier transform and wavelet transform). By analyzing the time-frequency plot, the vibration energy distribution of the bucket tooth at different frequencies can be determined, thereby extracting vibration features. For example, characteristics such as the vibration amplitude and frequency component changes of the bucket tooth at a specific operating frequency can be determined. Statistical features of the vibration data, such as the mean, variance, and kurtosis, can also be calculated to reflect vibration stability and anomalies.

[0078] Based on the collected stress data, its energy spectrum characteristics are calculated. First, the stress data is Fourier transformed to convert it into a frequency domain signal. Then, the energy distribution at different frequencies is calculated. By analyzing the energy spectrum, the main frequency components and energy concentration areas of the stress data can be extracted, thereby obtaining stress characteristics. In addition, correlation coefficients and cross-correlation functions of the stress data can be calculated to analyze the relationship and variation patterns between stresses at different locations.

[0079] Feature fusion step S103: using the feature fusion results of the visual features, vibration features and stress features, a relationship curve including the stress distribution entropy value and the life value corresponding to the bucket tooth is determined.

[0080] The feature fusion process requires the use of an appropriate feature fusion method to fuse the extracted visual, vibration, and stress features. Common feature fusion methods include weighted fusion, serial fusion, and parallel fusion. When selecting a fusion method, consider the characteristics and importance of each feature. For example, visual features closely related to bucket tooth wear can be assigned a higher weight; vibration and stress features that reflect the working state of bucket teeth can be weighted based on their impact on tooth life.

[0081] It's worth noting that although visual data, vibration data, and stress data are not of the same dimensionality, they can be fused using weights associated with their corresponding data values. For example, specific weights can be assigned to different data types, and then multiplied by that weight with the data type to obtain the corresponding product. By setting different weights, the product can be kept within a specific range, thus resolving the fusion difficulties caused by different dimensionality.

[0082] After obtaining the fused features through feature fusion, machine learning or data mining methods can be used to establish a relationship model between the fused features and the stress distribution entropy and lifespan of the bucket teeth. Regression analysis methods (such as linear regression and nonlinear regression) or neural network methods (such as multilayer perceptrons and radial basis function networks) can be used to train the fused features, thereby generating a relationship curve between the stress distribution entropy and lifespan. When establishing this relationship model, extensive historical data should be used for training and validation to ensure its accuracy and reliability.

[0083] Detection output step S104: obtaining the stress distribution entropy value and life value corresponding to the number of working cycles of the bucket tooth according to the relationship curve, and determining the detection result of the bucket tooth using the stress distribution entropy value and the life value.

[0084] Obtain the number of bucket tooth operating cycles from the excavator's operating records or real-time monitoring system. Then, use the relationship curve to find the stress distribution entropy and lifespan corresponding to this number of operating cycles. During this search, interpolation methods (such as linear interpolation and spline interpolation) can be used to process the relationship curve to improve the accuracy of the acquired data.

[0085] The bucket tooth inspection results are determined based on the acquired stress distribution entropy and lifespan values, combined with pre-set thresholds and standards. For example, if the stress distribution entropy exceeds a certain threshold, it indicates uneven stress distribution in the bucket tooth, potentially posing a risk of failure. If the lifespan is less than a set value, the remaining life of the bucket tooth is short and requires prompt replacement. Furthermore, the inspection results can be output as intuitive charts or reports, providing a reference for excavator maintenance and management.

[0086] Optionally, the data collection step S101 is as follows: Figure 2 As shown, including:

[0087] Step S201: obtain a vision unit installed on the side of the bucket of the excavator, and use a vision sensor provided in the vision unit to collect vision data corresponding to the bucket teeth according to a first sampling rate.

[0088] The vision unit is an industrial-grade image acquisition unit that features high resolution, high frame rate, and good environmental adaptability. For example, in dusty excavation environments, a vision unit with dust and moisture-proof features must be selected. The vision unit should be installed in the cab area, with the shooting angle facing the bucket. The field of view should fully cover the bucket teeth and minimize obstructions. At the same time, the installation angle should be precisely adjusted so that the collected visual data can truly reflect the actual condition of the bucket teeth. The setting of the first sampling rate requires comprehensive consideration of the excavator's operating speed and the frequency of bucket tooth status changes. If the excavator is operating at a high speed and the bucket tooth status changes rapidly, a higher sampling rate is required to ensure that subtle changes in the bucket teeth can be captured. The visual sensor continuously collects visual data of the bucket teeth at the set first sampling rate. During the collection process, the integrity and accuracy of the data must be guaranteed to prevent frame loss or image blur.

[0089] Step S202: obtaining a vibration sensor array installed at the bucket of the excavator, and using vibration sensors provided in the vibration sensor array to collect vibration data corresponding to the bucket teeth at a second sampling rate.

[0090] Choose a vibration sensor array with high sensitivity and a wide frequency response range. For example, piezoelectric vibration sensors offer excellent dynamic response characteristics, enabling accurate measurement of bucket tooth vibration under various operating conditions. When installing the vibration sensor array on the bucket, consider the vibration transmission path and vibration characteristics of the bucket teeth. Generally, the vibration sensor should be installed at the connection between the bucket and the teeth, providing a more direct measurement of the tooth vibration. Ensure the sensor is securely mounted to prevent looseness from affecting the measurement results.

[0091] The second sampling rate is set based on the bucket tooth vibration frequency range. Higher tooth vibration frequencies require a higher sampling rate to accurately capture details of the vibration signal. The vibration sensor collects bucket tooth vibration data at the set second sampling rate. During the data collection process, real-time monitoring is performed to ensure data stability and reliability.

[0092] Step S203: obtaining a stress monitoring unit embedded in the root of the bucket tooth, and using a strain sensor in the stress monitoring unit to collect stress data corresponding to the bucket tooth at a third sampling rate.

[0093] Choose a stress monitoring unit with high precision and good stability. For example, foil strain gauges are a commonly used stress measurement element, offering advantages such as high sensitivity and good linearity. When embedding the stress monitoring unit into the bucket tooth root, ensure that the strain sensor fits snugly and does not damage the tooth during installation. Before installation, prepare the tooth root surface to ensure it is smooth and clean. After installation, seal it to prevent moisture and dust from entering and affecting measurement results.

[0094] The third sampling rate is set based on the frequency and amplitude of bucket tooth stress changes. Bucket tooth stress changes are complex during operation, so the sampling rate needs to be appropriately set based on actual conditions. The strain sensor collects bucket tooth stress data at the set third sampling rate. During the acquisition process, the data is recorded and stored in real time for subsequent analysis and processing.

[0095] Optionally, the feature extraction step S102, such as Figure 3 As shown, including:

[0096] Step S301, calculate the spatial feature vector of the visual data, and determine the spatial feature information corresponding to the visual data using the temporal variation characteristics corresponding to the spatial feature vector; extract the defect area characteristics contained in the bucket tooth according to the spatial feature information, and determine the visual features based on the defect area characteristics.

[0097] When calculating the spatial feature vectors of visual data, a variety of image features can be used. For example, edge features, using the Canny edge detection algorithm, can accurately identify bucket tooth edges and outline the general outline of the bucket tooth. Corner features, using Harris corner detection, can identify corners in bucket tooth images with drastic grayscale changes. These corners are often key locations where the bucket tooth shape changes. Regarding texture features, the gray-level co-occurrence matrix can quantify the roughness or fineness of the bucket tooth surface texture, providing rich information for the spatial feature vector.

[0098] Spatial feature vectors change over time. Using a sliding window method, the feature vector sequence is divided into intervals. The Euclidean distance between vectors within adjacent windows is calculated. A large distance indicates significant changes in the bucket tooth's spatial features. These changes, such as the concentrated areas of feature change and the frequency of change, are then combined to determine the spatial feature information of the visual data.

[0099] To extract features of defect areas, a threshold segmentation algorithm (such as the Otsu algorithm) is used to separate the bucket tooth from the background. For suspected defect areas, morphological operations (such as dilation and erosion) are combined to remove noise. Next, features such as the area, perimeter, and shape factor of the defect area are extracted. The area reflects the size of the defect, the perimeter indicates the length of the defect boundary, and the shape factor distinguishes defects of different shapes.

[0100] In the process of determining visual features, the extracted defect area features are combined with spatial feature information to form a complete visual feature. For example, the defect area is combined with features such as the number of corner points and texture roughness in the spatial feature vector to form a feature vector that can fully describe the visual condition of the bucket tooth.

[0101] Step S302 , determining the time-frequency characteristics of the impact signal contained in the vibration data, using the time-frequency characteristics of the impact signal to determine the time-frequency feature information corresponding to the vibration data, and using the time-frequency feature information and the time domain data corresponding to the kurtosis and impulse factor in the vibration data to determine the vibration characteristics.

[0102] To determine the time-frequency characteristics of shock signals, vibration data analysis uses either the short-time Fourier transform (STFT) or wavelet transform. The STFT expands the signal in both the time and frequency domains, but the time and frequency resolutions are fixed. The wavelet transform, with its multi-resolution properties, is more suitable for analyzing shock signals. These transforms generate time-frequency plots, from which we can extract characteristics such as the shock signal's occurrence time, frequency components, and energy distribution.

[0103] When determining time-frequency characteristics, an energy threshold can be set. When the energy of the transform coefficient exceeds the threshold, it is identified as an impact signal. This is also screened and confirmed based on factors such as the frequency range and duration of the impact signal. Statistics are collected to determine the time-frequency characteristics of the vibration data, including the frequency of occurrence and average energy of the impact signal.

[0104] Determining vibration signatures requires calculating the kurtosis and impulse factor of the vibration data. Kurtosis reflects the degree to which the signal's peak value deviates from a normal distribution, while the impulse factor reflects the ratio of the signal's peak value to its effective value. Combining time-frequency characteristics with time-domain data such as kurtosis and impulse factor yields the final vibration signature.

[0105] Step S303: Calculate the energy spectrum characteristics corresponding to the stress data using wavelet packet decomposition, determine energy spectrum characteristic information using the energy spectrum characteristics, and determine stress characteristics using statistics corresponding to the energy spectrum characteristic information and the mean square error value and peak factor in the stress data.

[0106] To calculate the energy spectrum, wavelet packets are used to decompose stress data into frequency bands. An appropriate wavelet basis and number of decomposition levels are selected, for example, based on the stress signal's frequency range and the required analysis accuracy. The energy within each frequency band is calculated, and the energy spectrum is plotted with frequency band as the horizontal axis and energy as the vertical axis to obtain the energy spectrum characteristics of the stress data.

[0107] The process of determining energy spectrum characteristics involves extracting key information from the energy spectrum, such as the frequency bands where energy is concentrated and the standard deviation of the energy distribution. The energy-concentrated frequency bands likely correspond to specific force patterns on the bucket teeth, while the standard deviation of the energy distribution reflects the degree of energy dispersion across these frequency bands.

[0108] Determining stress characteristics requires calculating the mean square error (MSE) and peak factor of the stress data. The mean square error reflects the magnitude of stress fluctuations, while the peak factor reflects the stress peaks. Combining energy spectrum characteristics with statistics such as the MSE and peak factor yields a stress signature, which is used to comprehensively assess the stress state of the bucket tooth.

[0109] Optionally, the feature fusion step S103, such as Figure 4 Shown, including:

[0110] Step S401 : determining a first weight coefficient, a second weight coefficient, and a third weight coefficient based on abnormal data included in the visual feature, the vibration feature, and the stress feature, respectively.

[0111] When determining the first, second, and third weighting factors, identifying abnormal data is crucial. Abnormal data within visual features, such as features in the tooth defect area exceeding the normal range or drastic changes in spatial feature information, can be identified by setting a threshold. When the area of ​​the defective area in a visual feature suddenly increases to a certain percentage exceeding the historical average, an abnormality is considered present. The first weighting factor is determined by statistically analyzing the proportion of abnormal data within the visual features, combined with expert experience and historical data. A high proportion of abnormalities indicates that visual features are more important for determining the tooth's condition, and the first weighting factor can be appropriately increased.

[0112] Regarding vibration characteristics, if the impact signal's time-frequency characteristics or time-domain data (kurtosis, impulse factor) show abnormalities, such as a sudden increase in the impact signal frequency or a kurtosis value far exceeding the normal range, the threshold is also used to determine abnormalities. The second weighting factor is determined based on the proportion of abnormal data in the vibration characteristics. If the vibration characteristics are frequently abnormal, it indicates that the vibration is significantly affecting the bucket tooth condition, and the second weighting factor should be increased.

[0113] Abnormalities in the energy spectrum characteristics or statistics (mean square error, peak factor) within the stress signature, such as a sudden and significant increase in energy in a frequency band of the energy spectrum or a mean square error outside the normal fluctuation range, are considered abnormalities. The third weighting factor is determined based on the proportion of abnormal data in the stress signature. Significant stress abnormalities indicate a significant impact of stress on the bucket tooth condition, and the third weighting factor should be increased.

[0114] Step S402 , using the first weight coefficient, the second weight coefficient, and the third weight coefficient, cascade splicing and weighted fusion calculation are performed on the visual features, the vibration features, and the stress features to obtain a feature fusion result corresponding to the bucket tooth.

[0115] When performing cascade splicing and weighted fusion calculations, the visual features, vibration features, and stress features are first concatenated in sequence to form a longer feature vector. Then, the first, second, and third weight coefficients are used to weight the corresponding partial features after concatenation. Specifically, cascade splicing + attention weighted fusion can be used, as shown in the following formula:

[0116] F fusion =αF vision ⊕βF vibration ⊕γF stress ;

[0117] Where α, β, and γ are the first, second, and third weight coefficients respectively; F vision 、F vibration 、F stress are visual features, vibration features and stress features respectively, F fusion is the feature fusion result.

[0118] Step S403: determining the stress distribution entropy value corresponding to the bucket tooth according to the feature fusion result, and determining the defect probability and defect type corresponding to the bucket tooth based on the stress distribution entropy value.

[0119] To determine the stress distribution entropy based on feature fusion results, machine learning models such as support vector machines (SVMs) or neural networks can be used. The feature fusion results are used as input, and after training, the model outputs the stress distribution entropy. The stress distribution entropy reflects the degree of disorder in the bucket tooth stress distribution. A higher entropy value indicates a more uneven stress distribution.

[0120] Based on the stress distribution entropy value, the bucket tooth defect probability and defect type can be determined, and a mapping relationship table can be established. Through a large amount of experimental data and historical records, the bucket tooth defect probability and common defect types corresponding to different stress distribution entropy value ranges are statistically analyzed.

[0121] Step S404: determining the life value of the bucket tooth using the defect probability and the defect type, and determining a relationship curve using the life value and the stress distribution entropy value.

[0122] When determining bucket tooth life using defect probability and defect type, empirical formulas or life prediction models can be used. For different defect types, corresponding life decay models are developed. For example, for wear-type defects, the remaining useful operating time of the bucket tooth can be calculated based on the degree of wear (quantified by the defect probability) and the wear rate. For crack-type defects, the remaining number of operating cycles before failure is predicted as the life value, taking into account the crack growth rate and critical crack size.

[0123] After determining the life value, multiple sets of corresponding stress distribution entropy values ​​and life value data are plotted in a coordinate system. Using a fitting algorithm (such as linear fitting or polynomial fitting), a curve is found that best represents the distribution pattern of these data points. This curve is the relationship between life value and stress distribution entropy value. This curve can be used to subsequently predict the remaining life of the bucket tooth based on the real-time stress distribution entropy value.

[0124] Optionally, step S401 of determining the first weight coefficient, the second weight coefficient and the third weight coefficient respectively based on the abnormal data contained in the visual feature, the vibration feature and the stress feature, such as Figure 5 Shown, including:

[0125] Step S501, determining features of a defective area contained in a bucket tooth using visual features, and determining a first weight coefficient corresponding to the visual features based on the features of the defective area;

[0126] Step S502, determining an abnormal impact feature contained in the bucket tooth using the vibration feature, and determining a second weight coefficient corresponding to the vibration feature based on the abnormal impact feature;

[0127] Step S503: determining a stress offset feature included in the bucket tooth using the stress feature, and determining a third weight coefficient corresponding to the stress feature based on the stress offset feature.

[0128] In specific scenarios, the weight coefficient can be determined by constructing a cascade neural network architecture. For example, the visual branch in the network architecture uses EfficientNet-B4 to extract the features of the defect area (cracks, missing area), thereby determining the first weight coefficient; the vibration branch 1D-CNN analyzes the time-frequency characteristics of the impact signal and identifies abnormal impact patterns, thereby determining the second weight coefficient; the stress branch uses the LSTM network to predict the stress distribution offset trend, thereby determining the third weight coefficient, which specifically corresponds to the above-mentioned α, β, and γ.

[0129] Optionally, the detection output step S104 is as follows: Figure 6 Shown, including:

[0130] Step S601, obtaining the number of working cycles of the bucket teeth in real time.

[0131] Real-time information on the number of bucket tooth cycles can be obtained using sensors installed in key locations. For example, a pressure sensor can be installed in the bucket tooth. When the bucket tooth performs actions such as digging, lifting, and unloading, the pressure in the pressure sensor will change periodically. By analyzing the periodicity of the pressure change, the number of cycles can be accurately calculated.

[0132] Step S602 : using the three-stage segmented curve included in the relationship curve, determining the stress distribution entropy value corresponding to the number of working cycles, and using the two-stage segmented curve included in the relationship curve, determining the life value corresponding to the number of working cycles.

[0133] The three-stage segmented curve and the two-stage segmented curve in the relationship curve are obtained based on a large amount of experimental data and actual working condition analysis. Optionally, the three-stage segmented curve is:

[0134] ;

[0135] in, is the normalized entropy value corresponding to the stress distribution entropy value; is the number of working cycles; is the initial growth coefficient; is the exponential growth factor; is the platform constant, used to characterize the entropy benchmark of the second stage; is the fine-tuning coefficient; is the decay index; is the critical working cycle number; is the quadratic growth coefficient; is the secondary growth index;

[0136] The formula indicates that initial stress distribution is relatively dispersed and uneven, leading to a rapid increase in entropy. Initially, microcracks appear on the surface or within the material, and stress concentrates at these defects and is rapidly released, causing entropy (disorder) to grow in a power-law fashion. n represents the number of cycles (e.g., the number of cyclical effects such as mechanical loads and temperature fluctuations); a1 represents the initial growth coefficient, reflecting the sensitivity of the material's initial defects or microstructure to the increase in entropy; and b1 represents the exponential growth factor (b1 > 1), controlling the rate of accelerated entropy growth.

[0137] The formula indicates that the stress distribution and wear enter a stable period, the entropy value slows down, the wear enters a dynamic equilibrium stage, the new cracks and the material self-repair (such as the formation of an oxide layer) reach a temporary balance, and the entropy value tends to be stable. , which reflects small fluctuations (such as changes in ambient temperature and humidity). c2 is the platform constant, representing the baseline level of entropy value in the wear stability stage; a2 is the fine-tuning coefficient, usually a2≪a1, which is used to describe the impact of weak environmental disturbances (such as random vibration) on entropy value. b2 is the attenuation exponent (b2≈0), when b2→0, n b2 →1, the formula degenerates to the constant c2.

[0138] The formula shows that crack or damage expansion leads to stress distribution. When the crack changes from stable expansion to unstable expansion (such as reaching a critical size), the stress distribution becomes extremely chaotic and the entropy value rises again in a power law. crit ) reflects the hysteresis effect (accelerated failure is triggered only after the damage accumulates to the threshold). Specifically, n crit is the critical number of cycles, corresponding to the starting point of a rapid decrease in the remaining life of the material (such as the threshold value of a sudden change in the crack growth rate); a3 is the secondary growth coefficient, reflecting the sensitivity of the entropy value in the crack growth stage; b3 is the secondary growth exponent (b3>1), which controls the secondary accelerated increase of the entropy value.

[0139] The two-stage segmented curve is:

[0140] ;

[0141] in, is the remaining life ratio corresponding to the life value; is the linear damage rate coefficient; is the critical remaining life ratio; is the exponential decay coefficient; is the power law damage acceleration coefficient; is the power exponent.

[0142] The formula represents the first stage, where damage is approximately linear with the number of cycles. In the initial stage, internal material damage (such as microcracks and plastic deformation) accumulates at a nearly constant rate, and the remaining life decreases linearly with the number of cycles. For example, if k1 = 0.0005, the remaining life decreases by 50% after every 1000 cycles, that is, L(1000) = 1 - 0.0005 × 1000 = 0.5. Specifically, L(n) is the remaining life ratio, which represents the percentage of the material or component's remaining life as a percentage of the initial life. For example, L(n) = 0.6 indicates a remaining life of 60% of the initial life. n is the number of cycles (the number of cyclical effects such as mechanical loads and temperature cycling). k1 is the linear damage rate coefficient (k1 > 0 and is relatively small), which controls the rate of linear decline of the remaining life.

[0143] The formula represents the second stage, where the damage accelerates (exponential or power law). When the damage accumulates to a critical value (such as the crack reaches a critical size), the material enters the unstable failure stage. If k2=0.002, then in stage 2, for every 100 cycles, the remaining life decays to the original e -0.002×100 ≈81.9%. If k3=0.001 and b4=2, then for every 100 cycles in stage 2, the remaining life decreases by 0.001×1002=10%. Specifically, L crit is the critical remaining life ratio, i.e. the remaining life at the end of stage 1 (e.g. L crit ≈40%=0.4);n crit is the critical number of cycles, i.e., the end point of stage 1 (1200 cycles in the figure); k2 is the exponential decay coefficient (k2>0), which controls the exponential decrease rate of the remaining life; k3 is the power-law damage acceleration coefficient (k3>0), which controls the power-law decrease amplitude of the remaining life; b4 is the power exponent (b4>1), which determines the steepness of the curve (the larger the value, the faster the decrease).

[0144] Based on the real-time acquired working cycle number, a search is performed on the three-stage segmented curve and the two-stage segmented curve. If the working cycle number corresponds exactly to a known point on the curve, the corresponding stress distribution entropy and life value can be directly read; if it is between two known points, the corresponding value can be estimated using linear interpolation.

[0145] Step S603: determining the health risk level of the bucket tooth according to the stress distribution entropy value and the life value, and determining the detection result using the health risk level.

[0146] Different stress distribution entropy and life span intervals are pre-set and associated with corresponding health risk levels. For example, when the stress distribution entropy is low and the life span is high, the bucket tooth is considered healthy and has a low health risk level. When the stress distribution entropy increases and the life span drops to a certain level, it indicates that the bucket tooth may have minor damage and has a medium health risk level. If the stress distribution entropy is very high and the life span is extremely low, the bucket tooth is at risk of serious damage and has a high health risk level.

[0147] Further clarification of the test results is based on the determined health risk level. For low-risk levels, it is recommended to continue using the bucket teeth normally, but regular monitoring is required. For medium-risk levels, a detailed inspection of the bucket teeth is recommended to assess whether repairs or replacements are necessary. For high-risk levels, the bucket teeth should be immediately discontinued and replaced to avoid safety accidents.

[0148] Optional, such as Figure 7 As shown, after the data acquisition step, the method further includes a preprocessing step:

[0149] Step S701, downsampling and histogram equalization are performed on the visual data using a preset first parameter to obtain first time series data corresponding to the visual data;

[0150] Step S702, filtering the vibration data using a second parameter corresponding to a preset filter to obtain second time series data corresponding to the vibration data;

[0151] Step S703, performing Kalman filtering on the stress data using a preset third parameter to obtain third time series data corresponding to the stress data;

[0152] Step S704 , performing a padding calculation on the sampling rates corresponding to the first time series data, the second time series data, and the third time series data, and using the padding calculation results to update the visual data, the vibration data, and the stress data respectively.

[0153] During preprocessing, the visual data uses adaptive histogram equalization (CLAHE) to enhance low-light images. Mask R-CNN is then used to segment the bucket tooth contours. Finally, the visual data is downsampled (224×224) and histogram equalized. Vibration data is processed using a wavelet transform to extract impact features in the 0.5-5kHz frequency band and filter out engine noise. A Butterworth filter (cutoff frequency 20-500Hz) is used for denoising. A Kalman filter is used to eliminate dynamic load fluctuations in the stress data, and the entropy of the bucket tooth stress distribution is calculated. Finally, the sampling rates of the first, second, and third time series data are padded. Dynamic time warping (DTW) is used to compensate for the sampling rate differences between the three sensors.

[0154] like Figure 8 The flowchart of another excavator bucket tooth detection method is shown. During the feature extraction process, ResNet-18 is used to extract spatial features (2048 dimensions) for visual features, and 3D-CNN is used to extract temporal variation features. An LSTM network is used to extract time-frequency features (128 dimensions) for vibration features, and time-domain metrics such as kurtosis and impulse factor are simultaneously calculated. Wavelet packet decomposition (5 layers) is used to extract energy spectrum features for stress features, and statistics such as RMS and peak factor are calculated. The feature fusion process uses cascaded concatenation and attention-weighted fusion, calculated using the following formula:

[0155] F fusion =αF vision ⊕βF vibration ⊕γF stress ;

[0156] Where α, β, and γ are the first, second, and third weight coefficients respectively; F vision 、F vibration 、F stress are visual features, vibration features and stress features respectively, F fusion is the feature fusion result.

[0157] like Figure 9 The sensor deployment diagram shown in the figure shows a multimodal sensing module consisting of the following three modules:

[0158] The high-dynamic vision unit includes a first vision sensor 1 and a second vision sensor 2. An anti-shake wide-angle camera (resolution ≥ 4K, frame rate 60fps) installed on the side of the bucket has an integrated polarization filter to eliminate reflection interference and capture bucket tooth images in real time.

[0159] The vibration sensing array includes vibration sensors 3, specifically triaxial acceleration sensors (sampling rate ≥ 10kHz) deployed at key locations in the bucket to capture the spectral characteristics of bucket tooth impact vibration;

[0160] The stress monitoring unit includes a stress sensor 4: a micro strain gauge (accuracy ±0.1%) is embedded in the root of the bucket tooth to measure the working load distribution in real time.

[0161] The Edge Computing Module 5 is equipped with the NVIDIA Jetson AGX Orin chip, which features a built-in lightweight multi-task deep learning model and supports real-time inference. It also integrates a 5G communication module for cloud-based data synchronization and remote control.

[0162] The decision and warning module 6 displays the defect location and severity through the HMI (human-machine interface) and triggers an audible and visual alarm.

[0163] The three-modal features are dynamically weighted and fused through the attention mechanism (Transformer-based Fusion), and the following results are output: defect probability (0-1), defect type (crack, fracture, wear), and remaining life prediction (based on the Weibull distribution model). On this basis, we can get the following Figure 10 The relationship curve between the stress distribution entropy value and the bucket tooth life is shown. There are three parameters in this coordinate system. The X-axis is the number of working cycles of the bucket tooth, and the Y-axis has two, namely the normalized entropy value of stress and the remaining life percentage of the bucket tooth. The normalized entropy value of stress has three stages, namely rapid increase, stability and rapid increase again. The remaining life has two stages, slow decline and rapid decline. The starting point of the rapid decline of the remaining life coincides with the rapid increase of the normalized entropy value of stress.

[0164] The curve characteristics are analyzed as follows:

[0165] In the initial stage (0–900 cycles), the entropy value increases rapidly (0.5→0.7), and the initial running-in stress is unevenly distributed and locally concentrated;

[0166] In the stable stage (900-1800 times), the entropy value fluctuated at 0.7 ± 0.05, and the stress release reached equilibrium after a period of use;

[0167] In the failure stage (>1800 times), the entropy value suddenly increases to above 0.95, micro cracks turn into macro cracks, resulting in stress concentration, and immediate replacement is required when the remaining life is <20%.

[0168] The hardware implementation example is as follows: the camera is installed on the side bracket of the bucket, adopts IP67 protection grade, and has a built-in heating film to prevent the mirror from fogging; the vibration sensor is fixed with a magnetic base and supports quick disassembly and assembly; the edge computing module integrates a power management unit and supports 24V DC power supply and solar backup power supply.

[0169] An example of algorithm implementation is as follows: 100,000 bucket tooth images under different lighting and dust conditions were collected for training data, and the defect types and locations were annotated. The model used knowledge distillation technology to compress the model (ResNet-101) into EfficientNet-B4, which can be run on edge devices. Online learning uses a federated learning framework to continuously optimize the model using field data from multiple excavators.

[0170] From the excavator bucket tooth detection method mentioned in the above embodiment, it can be seen that this method fuses data from three types of multimodal sensors to realize the detection of bucket tooth life, alleviates the detection limitations of a single sensor, and can use the relationship curve obtained by data fusion processing to perform preventive detection and maintenance on the bucket teeth, thereby improving the reliability of the bucket teeth under complex working conditions.

[0171] Corresponding to the excavator bucket tooth detection method provided in the above embodiment, the embodiment of the present invention provides an excavator bucket tooth detection system, such as Figure 11 As shown, the system includes:

[0172] The data acquisition module 1110 is used to respectively collect visual data, vibration data and stress data corresponding to the bucket teeth in the excavator based on the visual sensor, vibration sensor and stress sensor provided on the excavator;

[0173] A feature extraction module 1120 is configured to extract visual features, vibration features, and stress features corresponding to the bucket teeth based on spatial feature information of the visual data, time-frequency feature information of the vibration data, and energy spectrum feature information corresponding to the stress data.

[0174] A feature fusion module 1130 is used to determine a relationship curve between stress distribution entropy and life value corresponding to the bucket tooth using the feature fusion results of visual features, vibration features, and stress features;

[0175] The detection output module 1140 is used to obtain the stress distribution entropy value and life value corresponding to the working cycle number of the bucket tooth according to the relationship curve, and determine the detection result of the bucket tooth using the stress distribution entropy value and life value.

[0176] From the excavator bucket tooth detection system mentioned in the above embodiment, it can be seen that the system fuses data from three types of multimodal sensors to realize the detection of bucket tooth life, alleviates the detection limitations of a single sensor, and can use the relationship curve obtained by data fusion processing to perform preventive detection and maintenance on the bucket teeth, thereby improving the reliability of the bucket teeth under complex working conditions.

[0177] The excavator bucket tooth detection system provided in the embodiment of the present invention has the same implementation principle and technical effects as the aforementioned excavator bucket tooth detection method embodiment. For the sake of brief description, for matters not mentioned in the system embodiment, reference can be made to the corresponding content in the aforementioned excavator bucket tooth detection method embodiment.

[0178] This embodiment also provides an electronic device. The structural diagram of the electronic device is as follows: Figure 12 As shown, the excavator equipment is provided with a processing unit, which includes a processor 101 and a memory 102; wherein the memory 102 is used to store one or more computer instructions, and the one or more computer instructions are executed by the processor to implement the steps of the above-mentioned excavator bucket tooth detection method.

[0179] Figure 12 The processing unit in the electronic device shown further includes a bus 103 and a communication interface 104 , and the processor 101 , the communication interface 104 and the memory 102 are connected via the bus 103 .

[0180] The memory 102 may include a high-speed random access memory (RAM) and may also include a non-volatile memory, such as at least one disk storage. The bus 103 may be an ISA bus, a PCI bus, or an EISA bus. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 12 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0181] The communication interface 104 is used to connect to at least one user terminal and other network units through a network interface, and send the encapsulated IPv4 message or IPv4 message to the user terminal through the network interface.

[0182] The processor 101 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in the processor 101 or the instruction in the form of software. The above-mentioned processor 101 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The storage medium is located in the memory 102, and the processor 101 reads the information in the memory 102 and completes the steps of the method of the above embodiment in combination with its hardware.

[0183] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that can be executed by a processor. Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the aforementioned embodiments, those of ordinary skill in the art should understand that any person familiar with this technical field can still modify the technical solutions described in the aforementioned embodiments within the technical scope disclosed by the present invention, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for detecting bucket teeth of an excavator, characterized in that: The method comprises: Data collection step: collecting visual data, vibration data and stress data corresponding to the bucket teeth in the excavator based on the visual sensor, vibration sensor and stress sensor provided on the excavator; Feature extraction step: extracting visual features, vibration features, and stress features corresponding to the bucket tooth according to the spatial feature information of the visual data, the time-frequency feature information of the vibration data, and the energy spectrum feature information corresponding to the stress data; Feature fusion step: determining a relationship curve between a stress distribution entropy value and a life value corresponding to the bucket tooth using a feature fusion result of the visual feature, the vibration feature, and the stress feature; Detection and output step: obtaining the stress distribution entropy value and the life value corresponding to the number of working cycles of the bucket tooth according to the relationship curve, and determining the detection result of the bucket tooth using the stress distribution entropy value and the life value.

2. The excavator bucket tooth detection method according to claim 1, characterized in that: The data collection step includes: Obtaining a vision unit installed on the side of the bucket of the excavator, and using the vision sensor provided in the vision unit to collect the vision data corresponding to the bucket teeth at a first sampling rate; Obtaining a vibration sensor array installed at the bucket of the excavator, and using the vibration sensors provided in the vibration sensor array to collect vibration data corresponding to the bucket teeth at a second sampling rate; A stress monitoring unit embedded in the root of the bucket tooth is obtained, and the stress data corresponding to the bucket tooth is collected according to a third sampling rate using a strain sensor in the stress monitoring unit.

3. The excavator bucket tooth detection method according to claim 1, characterized in that: The feature extraction step comprises: Calculating a spatial feature vector of the visual data, and determining the spatial feature information corresponding to the visual data using a temporal variation feature corresponding to the spatial feature vector; extracting a defective region feature of the bucket tooth according to the spatial feature information, and determining the visual feature based on the defective region feature; Determining a time-frequency feature of an impact signal contained in the vibration data, determining the time-frequency feature information corresponding to the vibration data using the time-frequency feature of the impact signal, and determining the vibration feature using the time-frequency feature information and time domain data corresponding to the kurtosis and impulse factor in the vibration data; The energy spectrum characteristics corresponding to the stress data are calculated using wavelet packet decomposition, the energy spectrum characteristic information is determined using the energy spectrum characteristics, and the stress characteristics are determined using statistics corresponding to the energy spectrum characteristic information and the mean square error value and peak factor in the stress data.

4. The excavator bucket tooth detection method according to claim 1, characterized in that: The feature fusion step includes: determining a first weight coefficient, a second weight coefficient, and a third weight coefficient based on abnormal data included in the visual feature, the vibration feature, and the stress feature, respectively; Performing cascade splicing and weighted fusion calculation on the visual feature, the vibration feature, and the stress feature using the first weight coefficient, the second weight coefficient, and the third weight coefficient to obtain the feature fusion result corresponding to the bucket tooth; Determining the stress distribution entropy value corresponding to the bucket tooth according to the feature fusion result, and determining the defect probability and defect type corresponding to the bucket tooth based on the stress distribution entropy value; The life value of the bucket tooth is determined using the defect probability and the defect type, and the relationship curve is determined using the life value and the stress distribution entropy value.

5. The excavator bucket tooth detection method according to claim 4, characterized in that: Determining a first weight coefficient, a second weight coefficient, and a third weight coefficient based on abnormal data included in the visual feature, the vibration feature, and the stress feature, respectively, includes: Determining a defect region feature contained in the bucket tooth using the visual feature, and determining a first weight coefficient corresponding to the visual feature based on the defect region feature; Determining an abnormal impact feature contained in the bucket tooth using the vibration feature, and determining a second weight coefficient corresponding to the vibration feature based on the abnormal impact feature; The stress feature is used to determine a stress offset feature included in the bucket tooth, and a third weight coefficient corresponding to the stress feature is determined based on the stress offset feature.

6. The excavator bucket tooth detection method according to claim 1, characterized in that: The detection and output step includes: Acquire the working cycle number of the bucket tooth in real time; Determine the stress distribution entropy value corresponding to the number of working cycles using a three-stage segmented curve included in the relationship curve, and determine the life value corresponding to the number of working cycles using a two-stage segmented curve included in the relationship curve; determining a health risk level of the bucket tooth according to the stress distribution entropy value and the life value, and determining the detection result using the health risk level; The three-stage segmented curve is: ； in, is the normalized entropy value corresponding to the stress distribution entropy value; is the number of working cycles; is the initial growth coefficient; is the exponential growth factor; is the platform constant, used to characterize the entropy benchmark of the second stage; is the fine-tuning coefficient; is the decay index; is the critical working cycle number; is the quadratic growth coefficient; is the secondary growth index; The two-stage segmented curve is: ； in, is the remaining life ratio corresponding to the life value; is the linear damage rate coefficient; is the critical remaining life ratio; is the exponential decay coefficient; is the power law damage acceleration coefficient; is the power exponent.

7. The excavator bucket tooth detection method according to claim 1, characterized in that: After the data collection step, the method further comprises a pre-processing step: downsampling and histogram equalization are performed on the visual data using a preset first parameter to obtain first time series data corresponding to the visual data; After filtering the vibration data using a second parameter corresponding to a preset filter, second time series data corresponding to the vibration data is obtained; After performing Kalman filtering on the stress data using a preset third parameter, third time series data corresponding to the stress data is obtained; A padding calculation is performed on the sampling rates corresponding to the first time series data, the second time series data, and the third time series data, and the visual data, the vibration data, and the stress data are updated respectively using the padding calculation results.

8. An excavator bucket tooth detection system, characterized in that: The system comprises: A data acquisition module, configured to respectively acquire visual data, vibration data, and stress data corresponding to the bucket teeth of the excavator based on a visual sensor, a vibration sensor, and a stress sensor provided on the excavator; a feature extraction module, configured to extract visual features, vibration features, and stress features corresponding to the bucket tooth based on the spatial feature information of the visual data, the time-frequency feature information of the vibration data, and the energy spectrum feature information corresponding to the stress data; a feature fusion module, configured to determine a relationship curve between a stress distribution entropy value and a life value corresponding to the bucket tooth using a feature fusion result of the visual feature, the vibration feature, and the stress feature; A detection output module is used to obtain the stress distribution entropy value and the life value corresponding to the number of working cycles of the bucket tooth according to the relationship curve, and determine the detection result of the bucket tooth using the stress distribution entropy value and the life value.

9. An electronic device, characterized in that: The invention comprises a processor and a memory, wherein the memory stores computer-executable instructions that can be executed by the processor, and the processor executes the computer-executable instructions to implement the steps of the excavator bucket tooth detection method according to any one of claims 1 to 7.

Images