Gear defect detection system and method based on gear pump machining

Abstract

The invention discloses a gear defect detection system and method based on gear pump machining, and relates to the technical field of nondestructive testing. The gear defect detection platform is in communication connection with a magnetization preprocessing module, a magnetic signal acquisition module, a magnetic signal processing module, a gear defect analysis module and a defect visual labeling module, and all the modules are in electric signal connection; and the magnetization preprocessing module is used for customizing a magnetization scheme according to the material and structural characteristics of the gear. According to the method, the signal separation and feature extraction technology is adopted, independent features of different defect types can be effectively extracted from magnetic signals, the defect classification model is constructed, then surface damage, geometric deformation and fatigue cracks are accurately distinguished, quantitative analysis is conducted on the severity of the defects, the defect classification accuracy is improved, and the defect classification efficiency is improved. And classification errors caused by overlapping of magnetic signals in a traditional method are avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of nondestructive testing, and in particular to a gear defect detection system and method based on gear pump processing. Background Art

[0002] Gear pumps are common mechanical components in hydraulic systems and are widely used in automobiles, engineering machinery, aerospace, ships, energy and other fields. Their main function is to transport and pressurize liquids through the meshing and rotation of gears. The quality and performance of gears have a vital impact on the overall performance of gear pumps. If there are defects in the gears, such as cracks, wear, broken teeth, etc., it will directly affect the gear pump's delivery efficiency, sealing performance and service life. Therefore, defect detection of gears is an important part of ensuring the quality of gear pumps.

[0003] For example, the intelligent gear defect detection method and system based on magnetic memory technology, disclosed in Chinese patent publication number CN119246666A, can detect, locate and identify stress defects in gears based on magnetic memory technology, thereby providing a reliable fault detection basis for gear maintenance and inspection, and improving the accuracy and sustainability of gear use and maintenance.

[0004] In the existing technology, the stress defects of gears are detected, located and identified based on magnetic memory technology, which solves the problem that it is difficult to capture the potential defects and types of gears through changes in magnetization intensity and magnetic domain structure. However, due to the different materials, structures and working environments of gears, their magnetization characteristics and defect manifestations vary greatly. In addition, when identifying the defect type by comparing the degree of consistency of the magnetization intensity change, the magnetic signals of different defects overlap, resulting in inaccurate defect classification. Therefore, a gear defect detection system and method based on gear pump processing are proposed to solve the above problems. Summary of the Invention

[0005] The purpose of the present invention is to provide a gear defect detection system and method based on gear pump processing to solve the problems raised in the above background technology.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is: In a first aspect, a gear defect detection system based on gear pump processing includes a gear defect detection platform, wherein the gear defect detection platform is communicatively connected to a magnetization preprocessing module, a magnetic signal acquisition module, a magnetic signal processing module, a gear defect analysis module, and a defect visualization annotation module, wherein the modules are electrically connected; The magnetization pre-processing module is used to customize the magnetization scheme according to the material and structural characteristics of the gear; The magnetic signal acquisition module is used to collect magnetic signals during gear detection using a high-resolution, high-sensitivity magnetic sensor array, including magnetic signal data on the gear surface and interior. By capturing magnetic signal changes caused by defects on the gear surface and interior, the detection sensitivity is improved. The magnetic signal processing module is used to apply signal processing technology to perform customized processing on different types of noise in the magnetic signal to solve the problem of magnetization signal overlap, separate different defect signals, improve the signal-to-noise ratio of the magnetic signal, reduce the impact of noise on the detection results, and improve the clarity and accuracy of the magnetic signal; The gear defect analysis module is used to perform feature analysis on the signal after magnetic signal processing, extract defect features, and separate different defect signals, thereby building a defect classification model, identifying the type of gear defect, and analyzing the severity of the defect to determine the availability of the gear; The defect visualization and annotation module is used to visualize the detected defect location, size, type and availability information, providing a user-friendly interface to facilitate users to view, analyze and annotate the detection results, making the detection results more intuitive and easy to understand, and improving the user experience.

[0007] A further improvement of the technical solution of the present invention is that: the gear defect analysis module includes a defect signal separation unit, a defect classification and identification unit and a defect severity analysis unit; The defect signal separation unit is used to extract defect features from the processed magnetic signals and distinguish the magnetic signals of different defect types through signal analysis, thereby separating different defect signals, effectively extracting independent features of different defects from overlapping magnetic signals, reducing classification errors caused by overlapping defect types, and improving the detection accuracy of the system; The defect classification and identification unit is used to combine the separated defect signals and defect features to build a defect classification model, identify the defect type of the gear, and accurately distinguish different types of defects; The defect severity analysis unit is used to further analyze the severity of the defect based on the characteristics and classification results of the defect, evaluate the impact of the defect on the gear of the gear pump, determine the availability of the gear, and identify defects that have a greater impact on the performance of the gear pump by quantitatively evaluating the defects of the gear.

[0008] A further improvement of the technical solution of the present invention is that the magnetization pretreatment module specifically includes: Determine the gear parameters based on the gear pump's operating characteristics (flow rate, pressure, speed, etc.) and the expected operating environment (temperature, humidity, corrosive media, etc.), input the gear parameters into the system, and analyze the gear's material and structural characteristics; Based on the material and structural characteristics of the gear, analyze its magnetization characteristics, determine the required magnetization intensity, magnetization direction and pre-treatment steps, and customize the magnetization solution; Use demagnetization equipment to demagnetize the gears, and after demagnetization, use magnetization equipment to pre-magnetize the gears to make them have a consistent magnetization state; The magnetization state of the gears after pre-magnetization treatment is verified to verify whether the magnetization state of the gears meets the expected requirements, ensure the comparability and stability of the magnetic signals, and record and store the magnetization treatment parameters (magnetization intensity, magnetization direction, equipment type, etc.) in the system database. At the same time, a unique magnetization state identification is generated for each gear to facilitate traceability and comparison during subsequent testing.

[0009] A further improvement of the technical solution of the present invention is that the magnetic signal acquisition module specifically includes: A magnetic sensor array is configured based on the size and shape of the gear. Multiple magnetic sensors are arranged into an array to ensure full coverage of the gear surface and possible internal defect areas. Multi-axis detection capabilities are used to capture magnetic field changes in different directions. The magnetic sensors are tunnel magnetoresistive sensors or fluxgate sensors. The gear is magnetized based on a customized magnetization scheme and placed on a detection platform so that its relative position to the magnetic sensor array is fixed. A magnetic sensor array is used to capture the magnetic signals generated on the surface and inside of the gear. The magnetic signals include the magnetic field generated by the normal structure of the gear and the magnetic field changes caused by defects.

[0010] A further improvement of the technical solution of the present invention is that the magnetic signal processing module specifically includes: The collected original magnetic signal is amplified by a differential amplifier to improve the detectability of the signal, and the magnetic signal is frequency modulated and amplified using a lock-in amplifier to further enhance the signal strength and improve the detectability of the signal; The magnetic signal is processed to remove DC offset to eliminate baseline drift caused by equipment or environmental factors. DC offset removal can be achieved through a high-pass filter to ensure signal stability. The pre-processed signal is then subjected to spectrum analysis to identify the main noise type and frequency range in the signal. The characteristics of the noise are analyzed to determine whether it is white noise, periodic noise, or random noise. Based on the results of the noise analysis, corresponding filters are designed to process different types of noise. For periodic noise, a notch filter or a band-stop filter is used for suppression. For random noise, adaptive filtering or Kalman filtering is used for noise reduction. Apply the signal separation technology of independent component analysis to separate the overlapping magnetization signals. Utilize the statistical independence between the signals to separate different defect signals from the mixed signals. Then perform feature extraction on the separated signals and calculate the statistical characteristics of the peak value, mean value, variance and frequency of the signals. A 24-bit high-precision analog-to-digital converter (ADC) is used to convert the amplified and filtered analog signals into digital signals to ensure high data resolution. The collected digital signals are then transmitted to the host computer through a communication module for subsequent analysis. A unique detection identifier is generated for each gear, and the gear's magnetic signal data is stored with the gear's identification information, detection parameters (magnetization intensity and direction), and acquisition timestamp for subsequent traceability.

[0011] A further improvement of the technical solution of the present invention is that the defect signal separation unit specifically includes: According to the working characteristics of the gear pump and the expected use environment, the gear defect types are identified, including surface damage defects, geometric deformation defects and fatigue crack defects. For each type of defect, the corresponding defect characteristics are extracted, and the detection benchmark value of each defect characteristic is determined. The detection benchmark value is formulated according to the design requirements, use environment and industry standards of the gear, and is used to assess whether the defect exceeds the acceptable range; For surface damage defects, the defect characteristics are wear characteristics, pitting characteristics, and bonding characteristics; for geometric deformation defects, the defect characteristics are tooth profile error, axial / radial runout, and center distance deviation; for fatigue crack defects, the defect characteristics are crack length, crack depth, and crack number. Then, the detection benchmark values ​​of each defect characteristic are integrated to obtain a standard defect characteristic sequence; The defect signal separation unit receives the processed magnetic signal data and performs feature analysis on the magnetic signal data, identifying the frequency components in the signal and distinguishing different types of defect characteristics. It then uses independent component analysis technology to separate different defect signals. Based on the statistical independence between the signals, it decomposes the mixed signal into independent defect signals and extracts the independent characteristics of each defect from the separated signals. The separated defect features are compared with the standard defect feature sequence, classified into known defect types (surface damage, geometric deformation or fatigue cracks), and the separated defect signals are output to the defect classification and identification unit.

[0012] A further improvement of the technical solution of the present invention is that the defect classification and identification unit specifically includes: Collect relevant data on gear pump gear defect detection, including different types of defect signals and corresponding defect features. Then, extract gear defect sample data from the gear pump gear defect detection database and mark the defect features contained in the sample data, such as surface damage defects, geometric deformation defects, and fatigue crack defects, to form a defect dataset. The dataset is divided into training and test sets to ensure that the training and test sets contain various types of defect samples to ensure the generalization ability of the model. Based on the data characteristics of gear defect detection, a recurrent neural network is selected as the model architecture of the defect classification model. The neural network model is trained using labeled training set data. During the training process, the model parameters (learning rate, batch size, number of network layers, etc.) are adjusted based on the model's training performance and verification results. Regularization techniques (Dropout, L2 regularization, etc.) are used to prevent overfitting. Early stopping is used to accelerate the training process and improve model performance. The model is then tested on a test set to evaluate the model's accuracy and generalization ability. Based on the evaluation results, the model structure or parameters are adjusted to further optimize the model performance. The defect features contained in the currently separated defect signal are input into the trained defect classification model. The defect classification model identifies and classifies the defect type (surface damage, geometric deformation or fatigue crack) based on the extracted defect features, and then outputs the identified defect type and its location information to the defect severity analysis unit.

[0013] A further improvement of the technical solution of the present invention is that the process of identifying and classifying defect types includes: Integrate the extracted defect features to form a feature vector as the input of the defect classification model, and normalize the input feature vector to ensure that the feature values ​​are within the same dimension range. Check the integrity and consistency of the feature vector to ensure that there are no missing values ​​or outliers. Load the trained defect classification model, initialize the model's input interface, ensure the format of the input feature vector is consistent with that during model training, and input the constructed feature vector into the defect classification model. The model performs calculations based on the input feature vector, gradually extracting and integrating features through its internal neural network structure, and calculates the probability that the input feature belongs to different defect types, indicating the probability that the input feature belongs to surface damage, geometric deformation, or fatigue crack. Based on the probability value output by the model, the defect type with the highest probability is selected as the final identification result. The identified defect type and its related information are then organized into a structured output format and output to the defect severity analysis unit for subsequent defect severity analysis. The output information includes the defect type, the specific value of the defect feature, the defect location, and the probability value output by the model.

[0014] A further improvement of the technical solution of the present invention is that the defect severity analysis unit specifically includes: Obtaining the defect type, specific values ​​of the defect characteristics, the defect location, and the probability value output by the defect classification model from the defect classification and identification unit; The severity of defects is quantitatively assessed based on the specific values ​​of defect types and defect characteristics. For surface damage defects, the wear depth or area, the area or number of pitting, and the size or depth of the bonded area are calculated. For geometric deformation defects, the tooth profile deviation, runout deviation, and the deviation between the actual and designed center distance are measured. For fatigue crack defects, the crack extension length and depth are measured, and the number of cracks is counted. Compare the quantified defect characteristics with the pre-set defect tolerance to preliminarily judge the severity of the defect. Then, combine the ratio of all defect characteristics to the defect tolerance to calculate the defect severity coefficient and evaluate the comprehensive severity of the gear defect. Based on the gear defect severity analysis results and the defect size and location, the gear availability is judged and the defective gears are divided into three types: usable, needing repair and unusable. Among them, usable means that the defect has little impact on the gear performance and does not affect normal operation. Needing repair means that the defect has a certain impact on the gear performance and can be used after repair. Unusable means that the defect is serious and may cause gear failure.

[0015] A further improvement of the technical solution of the present invention is that the defect visualization annotation module specifically includes: Receive inspection result data from the defect severity analysis unit, including: defect location (tooth top, tooth root, gear end face, etc.), defect size (crack length, wear depth, pitting area, etc.), defect type (surface damage, geometric deformation, fatigue crack), defect severity coefficient, and gear usability judgment (usable, repair required, unusable); Create a visualization interface to display a 3D model of the gear, mark the defect locations on the 3D model, and use different colors to distinguish the defect type and severity. Red dots represent surface damage defects, blue rectangles represent geometric deformation defects, and green triangles represent fatigue crack defects. It provides zoom-in, zoom-out, and rotation tools to help users view various parts of the gear, and provides cross-sectional views to display the internal defects of the gear. It allows users to mark the inspection results, add comments, record the specific conditions of the defects or repair suggestions, and provides save and export functions to save the marked results as images or report files.

[0016] In a second aspect, a gear defect detection method based on gear pump processing is implemented based on the above-mentioned gear defect detection system based on gear pump processing, comprising the following steps: Step 1: Customize the magnetization scheme according to the gear material and structure, perform demagnetization and pre-magnetization to ensure consistent magnetization state; Step 2: Use a high-resolution magnetic sensor array to collect magnetic signals on the gear surface and inside, and capture signal changes caused by defects; Step 3: Apply signal processing technology to pre-process the collected magnetic signals and separate different defect signals to improve the signal-to-noise ratio and clarity; Step 4: Extract defect features from the processed magnetic signals, build a defect classification model, and identify defect types, including surface damage defects, geometric deformation defects, and fatigue crack defects; Step 5: Calculate the defect severity coefficient based on the defect characteristics and location, and determine the gear availability, which can be classified as usable, needing repair, or unusable. Step 6. Create a 3D model of the gear and mark the defect location, type and severity. Provide a user interface and support annotation and report export.

[0017] Due to the adoption of the above technical solution, the present invention has the following technical advancements compared to the prior art: 1. The present invention provides a gear defect detection system and method based on gear pump processing. By adopting signal separation and feature extraction technology, the system can effectively extract independent features of different defect types from magnetic signals, and construct a defect classification model, thereby accurately distinguishing surface damage, geometric deformation and fatigue cracks, and quantitatively analyzing the severity of defects. This not only improves the accuracy of defect classification, but also avoids classification errors caused by magnetic signal overlap in traditional methods.

[0018] 2. The present invention provides a gear defect detection system and method based on gear pump processing. Through a customized magnetization preprocessing solution, it can provide a consistent and stable magnetization state for gears of different materials and structures, significantly reducing the problem of uneven magnetization caused by material differences. Combined with a high-resolution, high-sensitivity magnetic sensor array, the system can accurately capture tiny defect signals on the surface and inside of the gear, greatly improving the accuracy and reliability of detection. In addition, the noise in the magnetic signal is customized to be removed through signal processing technology, further improving the clarity and accuracy of the magnetic signal.

[0019] 3. The present invention provides a gear defect detection system and method based on gear pump processing. According to the characteristics and classification results of the defects, the severity of the defects is quantitatively evaluated. By calculating the size, location and development trend of the defects, the availability of the gears can be quickly judged, and the gears can be divided into three types: usable, needing repair, and unusable. This not only improves the evaluation efficiency, but also provides a scientific basis for the processing and production of gear pumps, helping to ensure the service life and maintenance cost of the gear pumps. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0021] Figure 1 Schematic diagram of the system function modules of the present invention; Figure 2 Schematic diagram of the workflow of the defect severity analysis unit of the present invention; Figure 3 Schematic diagram of the method of the present invention. DETAILED DESCRIPTION

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. 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.

[0023] Example 1, as Figure 1 As shown, the present invention provides a gear defect detection system based on gear pump processing, including a gear defect detection platform, which is communicatively connected to a magnetization preprocessing module, a magnetic signal acquisition module, a magnetic signal processing module, a gear defect analysis module, and a defect visualization annotation module, wherein electrical signals are connected between the modules; The magnetization pre-processing module is used to customize the magnetization scheme according to the material and structural characteristics of the gear, including selecting the appropriate magnetization equipment, magnetization intensity and magnetization direction. Through pre-processing steps such as demagnetization and pre-magnetization, the influence of the residual magnetic field inside the material on the test results is reduced, ensuring that the gear has a consistent magnetization state before testing, improving the comparability and stability of the magnetic signal, reducing the problem of uneven magnetization caused by material differences, and providing a reliable basis for subsequent testing. The gear parameters are determined according to the working characteristics of the gear pump (flow, pressure, speed, etc.) and the expected use environment (temperature, humidity, corrosive media, etc.), and the gear parameters are input into the system. The material and structural characteristics of the gear are analyzed. Among them, the gear material includes ferromagnetic materials and non-ferromagnetic materials. Material, gears of different materials respond differently to magnetization treatment. Structural characteristics include size, shape, number of teeth, tooth pitch and tooth thickness. Ferromagnetic materials (such as carbon steel and alloy steel) have high magnetic permeability and hysteresis characteristics and respond strongly to magnetic fields. Non-ferromagnetic materials (such as stainless steel and aluminum alloy) have weak magnetization ability. Based on the material and structural characteristics of the gear, its magnetization characteristics are analyzed to determine the required magnetization intensity, magnetization direction and pretreatment steps to customize the magnetization scheme. Among them, the selection of magnetization intensity and magnetization direction depends on the material of the gear and the required detection sensitivity. The greater the magnetization intensity, the higher the detection sensitivity, but the bearing capacity of the gear and the response to magnetization treatment also need to be considered. The magnetization direction should be perpendicular to the expected direction of the defect to improve the detection accuracy. To ensure accuracy, use a demagnetization device to demagnetize the gear, and after demagnetization, use a magnetization device to pre-magnetize the gear to make it have a consistent magnetization state. The demagnetization treatment is used to remove the residual magnetic field inside the gear to avoid its interference with the subsequent magnetization and detection process. Place the gear in the demagnetization device, starting from a high magnetic field intensity, and gradually reduce the magnetic field intensity to zero. At the same time, slowly rotate the gear to ensure that the magnetic field dissipates evenly from the gear. Repeat the above process many times to ensure that the residual magnetic field inside the gear is reduced to the lowest level. The demagnetization device is a coil device with adjustable magnetic field intensity. The pre-magnetization treatment is used to apply an initial, uniform magnetization state to the gear to provide a consistent background signal for subsequent defect detection. According to the material and structure of the gear Based on the structural characteristics, select the appropriate magnetization equipment (AC magnetization equipment, DC magnetization equipment or pulse magnetization equipment), set the magnetization intensity and direction, use circumferential magnetization for axial defect detection, and use longitudinal magnetization for radial defect detection, and then pre-magnetize the gear to ensure that the magnetization process is uniform and stable. Multiple magnetization and verification are required to ensure the consistency of the magnetization state. The magnetization state of the gear after pre-magnetization is verified to verify whether the magnetization state of the gear meets the expected requirements and ensure the comparability and stability of the magnetic signal. The magnetization parameters (magnetization intensity, magnetization direction, equipment type, etc.) are recorded and stored in the system database. At the same time, a unique magnetization state identifier is generated for each gear to facilitate traceability and comparison in the subsequent detection process; A magnetic signal acquisition module is used to use a high-resolution, high-sensitivity magnetic sensor array to collect magnetic signals during gear testing, including magnetic signal data from the gear surface and interior. By capturing magnetic signal changes caused by defects on the gear surface and interior, the detection sensitivity is improved. The magnetic sensor array is configured according to the size and shape of the gear, and multiple magnetic sensors are arranged into an array to ensure that the gear surface and possible internal defect areas can be fully covered. Multi-axis detection capabilities are used to capture magnetic field changes in different directions. The magnetic sensor is a tunnel magnetoresistive sensor or a fluxgate sensor. The gear is magnetized based on a customized magnetization scheme and placed on a testing platform so that its relative position to the magnetic sensor array is fixed. The magnetic sensor array is used to capture magnetic signals generated on the gear surface and interior. The magnetic signal includes the magnetic field generated by the normal structure of the gear and the magnetic field changes caused by defects. The magnetic signal processing module is used to apply signal processing technology to perform customized processing on different types of noise in the magnetic signal to solve the problem of magnetization signal overlap, separate different defect signals, improve the signal-to-noise ratio of the magnetic signal, reduce the impact of noise on the detection results, and improve the clarity and accuracy of the magnetic signal. The collected original magnetic signal is amplified by a differential amplifier to improve the detectability of the signal, and a phase-locked amplifier is used to perform frequency modulation amplification on the magnetic signal to further enhance the signal strength and improve the detectability of the signal. The magnetic signal is de-DC offset processed to eliminate the baseline drift caused by equipment or environmental factors. De-DC offset can be achieved through a high-pass filter to ensure signal stability. The pre-processed signal is subjected to spectrum analysis to identify the main noise type and frequency range in the signal, analyze the characteristics of the noise, and determine whether it is white noise, periodic noise or random noise. Then, according to the results of the noise analysis, the corresponding filter is designed. The filter processes different types of noise. For periodic noise, it uses notch filters or band-stop filters to suppress it. For random noise, it uses adaptive filtering or Kalman filtering methods to reduce noise. The signal separation technology of independent component analysis is applied to separate overlapping magnetization signals. The statistical independence between signals is used to separate different defect signals from the mixed signal. The separated signals are then feature extracted and the statistical characteristics of the peak value, mean value, variance and frequency of the signals are calculated. A 24-bit high-precision analog-to-digital converter (ADC) is used to convert the amplified and filtered analog signals into digital signals to ensure high data resolution. The collected digital signals are then transmitted to the host computer through the communication module for subsequent analysis. A unique detection identification is generated for each gear, and the magnetic signal data of the gear is stored with the gear identification information, detection parameters (magnetization intensity and direction), and acquisition timestamp for subsequent traceability. The gear defect analysis module is used to perform feature analysis on the signals after magnetic signal processing, extract defect features, and separate different defect signals. It then builds a defect classification model to identify the gear defect type and analyze the severity of the defect to determine the gear's usability. The defect visualization and annotation module is used to visualize the detected defect location, size, type and availability information, providing a user-friendly interface to facilitate users to view, analyze and annotate the detection results, making the detection results more intuitive and easy to understand, and improving the user experience.

[0024] Example 2, as Figure 2 As shown, based on Example 1, the present invention provides a technical solution: preferably, the gear defect analysis module includes a defect signal separation unit, a defect classification and identification unit, and a defect severity analysis unit; The defect signal separation unit is used to extract defect features from the processed magnetic signal, and distinguish the magnetic signals of different defect types through signal analysis, thereby separating different defect signals, effectively extracting independent features of different defects from overlapping magnetic signals, reducing classification errors caused by overlapping defect types, and improving the detection accuracy of the system. According to the working characteristics of the gear pump and the expected use environment, the defect type of the gear is clarified, including surface damage defects, geometric deformation defects and fatigue crack defects, and for each type of defect, the corresponding defect features are extracted, and the detection reference value of each defect feature is determined. Among them, the detection reference value is formulated according to the design requirements, use environment and industry standards of the gear, and is used to evaluate whether the defect exceeds the acceptable range. For surface damage defects, the defect features are wear features, pitting features and bonding features. For geometric deformation defects, the defect features are tooth profile errors, axial / radial runout and center distance deviation. For fatigue crack defects, the defect features are crack length, crack depth and number of cracks. Then, the detection reference values ​​of each defect feature are integrated to obtain a standard defect feature sequence, wherein the wear feature reflects the gear surface material. The wear degree and pitting characteristics indicate small hole-like damage on the gear surface caused by fatigue or corrosion. The bonding characteristics reflect adhesion damage on the gear surface caused by poor lubrication or overload. The tooth profile error is manifested as the shape deviation of the gear tooth profile. The axial / radial runout is manifested as the position deviation of the gear in the axial or radial direction. The center distance deviation is manifested as the deviation between the actual value and the design value of the gear center distance. The crack length is manifested as the extension length of the crack. The crack depth is manifested as the depth of the crack. The crack number is manifested as the number of cracks on the gear. The defect signal separation unit receives the processed magnetic signal data and performs feature analysis on the magnetic signal data, identifies the frequency components in the signal, and distinguishes different types of defect characteristics. Then, the independent component analysis technology is used to separate different defect signals. The mixed signal is decomposed into independent defect signals through the statistical independence between the signals. The independent features of each defect are extracted from the separated signals to ensure that the features of different defect signals are clearly distinguishable. Specifically, the implementation steps of the independent component analysis technology are as follows: the collected magnetic signal is demeaned and whitened so that the signal has unit variance and the components are uncorrelated. FastICA (Fast The Independent Component Analysis (ICA) algorithm performs fixed-point iteration to maximize the non-Gaussianity of the signal, separate the independent components, and extract statistical features from the separated independent components, including peak value, mean value, variance, frequency characteristics, and kurtosis. The extracted features are compared with the standard defect feature sequence, classified into known defect types, and the separation results are output. Among them, the wear feature extracts the amplitude change of the signal to reflect the degree of wear of the surface material. The pitting feature extracts the high-frequency component in the signal to reflect the presence of small hole-like damage. The bonding feature extracts the low-frequency component in the signal to reflect the characteristics of adhesion damage. The tooth profile error extracts the shape deviation characteristics of the signal.Extract position deviation features of axial / radial runout signals, center distance deviation features of center distance deviation signals, and crack features of crack length, depth, and number signals, such as magnetic field distortion caused by cracks. Compare the separated defect features with the standard defect feature sequence, classify them into known defect types (surface damage, geometric deformation, or fatigue cracks), and output the separated defect signals to the defect classification and identification unit; The defect classification and recognition unit is used to combine the separated defect signals and defect features, build a defect classification model, identify the defect type of the gear, accurately distinguish different types of defects, and collect relevant data for gear pump gear defect detection, including different types of defect signals and corresponding defect features. Then, from the gear pump gear defect detection database, the gear defect sample data is extracted, and the defect features contained in the surface damage defects, geometric deformation defects and fatigue crack defects in the sample data are marked to form a defect data set, which is divided into a training set and a test set to ensure that the training set and the test set contain various types of defect samples to ensure the generalization ability of the model. According to the data characteristics of gear defect detection, a recurrent neural network is selected as the model architecture of the defect classification model, and the labeled training set data is used to train the neural network. The network model is trained. During the training process, the model parameters (learning rate, batch size, number of network layers, etc.) are adjusted according to the training performance and verification results of the model. Regularization techniques (Dropout, L2 regularization, etc.) are used to prevent overfitting. Early stopping is used to accelerate the training process and improve model performance. The model is then tested using a test set to evaluate the accuracy and generalization ability of the model. The model structure or parameters are adjusted based on the evaluation results to further optimize the model performance. The defect features contained in the currently separated defect signal are input into the trained defect classification model. The defect classification model identifies and classifies the defect type (surface damage, geometric deformation or fatigue crack) based on the extracted defect features, and then outputs the identified defect type and its location information to the defect severity analysis unit; Additionally, the process of identifying and categorizing defect types includes: The extracted defect features are integrated to form a feature vector as the input of the defect classification model, and the input feature vector is normalized to ensure that the feature values ​​are within the same dimensional range. The integrity and consistency of the feature vector are checked to ensure that there are no missing values ​​or outliers. The trained defect classification model is loaded, the input interface of the model is initialized, and the format of the input feature vector is consistent with that during model training. The constructed feature vector is input into the defect classification model. The model performs calculations based on the input feature vector, gradually extracts and integrates features through its internal neural network structure, and calculates the probability that the input features belong to different defect types, indicating the probability that the input features belong to surface damage, geometric deformation or fatigue cracks. According to the probability value output by the model, the defect type with the highest probability is selected as the final recognition result. The identified defect type and its related information are then organized into a structured output format and output to the defect severity analysis unit for subsequent defect severity analysis. The output information includes the defect type, the specific value of the defect feature, the defect location, and the probability value output by the model. The expression for the probability that the input feature belongs to different defect types is: ; ; Where, is the probability that the input feature belongs to different defect types, is the input feature vector, containing the extracted defect features, , is the defect type (surface damage, geometric deformation or fatigue crack), is the index, which means it is used to traverse all defect types in the sum function. For the model The score function of the class defect represents the input feature vector With the The degree of matching of class defects, is the total number of defect types, For the The weight coefficient of the defect type indicates the importance of the defect type. For the The detection benchmark value of the feature, used to normalize the feature value , For the A bias term for class defects, used to adjust the baseline of the model, is the dimension of the input feature vector (number of features); ; Where, represents the probability value of each defect type output by the model , and select the defect type with the highest probability As a result of identification; The defect severity analysis unit is used to further analyze the severity of the defect based on the characteristics and classification results of the defect, evaluate the impact of the defect on the gear of the gear pump, determine the availability of the gear, and identify the defects that have a greater impact on the performance of the gear pump by quantitatively evaluating the defects of the gear. The defect type, specific values ​​of the defect characteristics, defect location and probability value output by the defect classification model are obtained from the defect classification and identification unit. According to the specific values ​​of the defect type and defect characteristics, the severity of the defect is quantitatively evaluated. For surface damage defects, the wear depth or area, the area or number of pitting and the size or depth of the bonding area are calculated. For geometric deformation defects, the tooth profile deviation value, the runout deviation value and the difference between the actual value and the design value of the center distance are measured. Deviation, for fatigue crack defects, measure the extension length of the crack, measure the depth of the crack and count the number of cracks, compare the quantified defect characteristics with the pre-set defect allowable value, preliminarily judge the severity of the defect, and then comprehensively calculate the ratio of all defect characteristics to the defect allowable value, calculate the defect severity coefficient, and evaluate the comprehensive severity of gear defects. According to the gear defect severity analysis results and the defect size and location, judge the availability of the gear, and divide the defective gears into three types: available, need to be repaired and unusable. Among them, available means that the defect has little effect on the gear performance and does not affect normal operation. Need to be repaired means that the defect has a certain impact on the gear performance and can be used after repair. Unusable means that the defect is serious and may cause gear failure. The expression of defect severity coefficient is: ; Where, is the defect severity coefficient, which is used to evaluate the comprehensive severity of gear defects. For the The quantitative value of the defect feature, For the The allowed value of defect characteristics, is the total number of defect features, is the adjustment coefficient, which is used to adjust the sensitivity of the defect severity coefficient. is a natural exponential function, which is used to map the weighted sum of defect features to the range of (0,1). The value range of is (0,1). When it is close to 0, it means that the severity of the defect is low and the gear condition is good. When it is close to 1, it means the severity of the defect is high and the gear condition is poor. are much smaller than the allowed value hour, Close to 0, resulting in Close to 0, when some Approaching or exceeding its allowed value hour, Increase, leading to Increase, adjustment coefficient The larger the value of , the more sensitive the defect severity coefficient is to the change of the defect characteristic value. , the defect location is not in the critical area (such as tooth root, tooth top, etc.), and the defect size is much smaller than the allowable value ( ), need to be repaired: the defect severity coefficient is medium, , the defect location is in a non-critical area, but the defect size is close to the allowable value ( ), or the defect is located in the critical area, but the defect size does not exceed the allowable value ( ), unavailable: the defect severity coefficient is high, , the defect is located in the key area (such as tooth root, tooth top, etc.), and the defect size exceeds the allowable value ( ); The defect visualization annotation module specifically includes: Receive inspection result data from the defect severity analysis unit, including: defect location (tooth top, tooth root, gear end face, etc.), defect size (crack length, wear depth, pitting area, etc.), defect type (surface damage, geometric deformation, fatigue crack), defect severity coefficient, and gear availability judgment (usable, repair required, unusable). Create a visualization interface to display the 3D model of the gear, mark the defect location on the 3D model, and use different colors to distinguish the defect type and severity. Among them, red dots are used to represent surface damage defects, blue rectangles are used to represent geometric deformation defects, and green triangles are used to represent fatigue crack defects. Provide zoom in, zoom out, and rotate tools so that users can view various parts of the gear. Provide a cross-sectional view to display the internal defects of the gear, allow users to mark the inspection results, add comments, record the specific conditions of the defects or repair suggestions, and provide save and export functions to save the marked results as images or report files.

[0025] Example 3, as Figure 3 As shown, based on Examples 1-2, the present invention further provides a gear defect detection method based on gear pump processing, which is implemented based on the above-mentioned gear defect detection system based on gear pump processing, and includes the following steps: Step 1: Customize the magnetization scheme according to the gear material and structure, perform demagnetization and pre-magnetization to ensure consistent magnetization state; Step 2: Use a high-resolution magnetic sensor array to collect magnetic signals on the gear surface and inside, and capture signal changes caused by defects; Step 3: Apply signal processing technology to pre-process the collected magnetic signals and separate different defect signals to improve the signal-to-noise ratio and clarity; Step 4: Extract defect features from the processed magnetic signals, build a defect classification model, and identify defect types, including surface damage defects, geometric deformation defects, and fatigue crack defects; Step 5: Calculate the defect severity coefficient based on the defect characteristics and location, and determine the gear availability, which can be classified as usable, needing repair, or unusable. Step 6. Create a 3D model of the gear and mark the defect location, type and severity. Provide a user interface and support annotation and report export.

[0026] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A gear defect detection system based on gear pump processing, including a gear defect detection platform, characterized by: The gear defect detection platform is communicatively connected to a magnetization pre-processing module, a magnetic signal acquisition module, a magnetic signal processing module, a gear defect analysis module, and a defect visualization annotation module, wherein electrical signals are connected between the modules; The magnetization pre-processing module is used to customize the magnetization scheme according to the material and structural characteristics of the gear; The magnetic signal acquisition module is used to use a magnetic sensor array to collect magnetic signals during gear detection, including magnetic signal data on the gear surface and inside; The magnetic signal processing module is used to apply signal processing technology to perform customized processing on different types of noise in the magnetic signal. The gear defect analysis module is used to perform feature analysis on the signal after magnetic signal processing, extract defect features, and separate different defect signals, thereby building a defect classification model, identifying the type of gear defect, and analyzing the severity of the defect to determine the availability of the gear; The defect visualization annotation module is used to visualize the detected defect location, size, type and availability information.

2. A gear defect detection system based on gear pump processing according to claim 1, characterized in that: The gear defect analysis module includes a defect signal separation unit, a defect classification and identification unit, and a defect severity analysis unit; The defect signal separation unit is used to extract defect features from the processed magnetic signal, and distinguish magnetic signals of different defect types through signal analysis, thereby separating different defect signals; The defect classification and identification unit is used to combine the separated defect signals and defect features to build a defect classification model and identify the defect type of the gear; The defect severity analysis unit is used to further analyze the severity of the defect based on the defect characteristics and classification results, evaluate the impact of the defect on the gear of the gear pump, and determine the availability of the gear.

3. The gear defect detection system based on gear pump processing according to claim 2, characterized in that: The magnetization preprocessing module specifically includes: Determine the gear parameters based on the working characteristics of the gear pump and the expected use environment, input the gear parameters into the system, and analyze the material and structural characteristics of the gear; Based on the material and structural characteristics of the gear, analyze its magnetization characteristics, determine the required magnetization intensity, magnetization direction and pre-treatment steps, and customize the magnetization solution; Use demagnetization equipment to demagnetize the gears, and after demagnetization, use magnetization equipment to pre-magnetize the gears to make them have a consistent magnetization state; The magnetization state of the gear after pre-magnetization treatment is verified to verify whether the magnetization state of the gear meets the expected requirements, and the magnetization treatment parameter records are stored in the system database, and a unique magnetization state identification is generated for each gear.

4. The gear defect detection system based on gear pump processing according to claim 3 is characterized in that: The magnetic signal acquisition module specifically includes: According to the size and shape of the gear, a magnetic sensor array is configured. Multiple magnetic sensors are arranged into an array and a multi-axis detection function is used to capture magnetic field changes in different directions. The magnetic sensors are tunnel magnetoresistive sensors or fluxgate sensors. The gear is magnetized based on a customized magnetization scheme and placed on a detection platform so that its relative position to the magnetic sensor array is fixed. A magnetic sensor array is used to capture the magnetic signals generated on the surface and inside of the gear. The magnetic signals include the magnetic field generated by the normal structure of the gear and the magnetic field changes caused by defects.

5. The gear defect detection system based on gear pump processing according to claim 4 is characterized in that: The magnetic signal processing module specifically includes: The collected original magnetic signal is amplified by a differential amplifier, and the magnetic signal is frequency modulated and amplified using a lock-in amplifier to further enhance the signal strength. Perform DC offset processing on magnetic signals to eliminate baseline drift caused by equipment or environmental factors. Perform spectrum analysis on the pre-processed signals to identify the main noise types and frequency ranges in the signals. Analyze the characteristics of the noise to determine whether it is white noise, periodic noise, or random noise. Based on the results of the noise analysis, design appropriate filters to process different types of noise. Apply the signal separation technology of independent component analysis to separate the overlapping magnetization signals. Utilize the statistical independence between the signals to separate different defect signals from the mixed signals. Then perform feature extraction on the separated signals and calculate the statistical characteristics of the peak value, mean value, variance and frequency of the signals. A 24-bit high-precision analog-to-digital converter is used to convert the amplified and filtered analog signal into a digital signal, and the collected digital signal is transmitted to the host computer through the communication module, thereby generating a unique detection identification for each gear and storing the gear's magnetic signal data with the gear's identification information, detection parameters, and acquisition timestamp.

6. The gear defect detection system based on gear pump processing according to claim 2, characterized in that: The defect signal separation unit specifically includes: According to the working characteristics of the gear pump and the expected use environment, the gear defect types are clearly identified, including surface damage defects, geometric deformation defects and fatigue crack defects. For each type of defect, the corresponding defect characteristics are extracted, and the detection benchmark value of each defect characteristic is determined; For surface damage defects, the defect characteristics are wear characteristics, pitting characteristics, and bonding characteristics; for geometric deformation defects, the defect characteristics are tooth profile error, axial / radial runout, and center distance deviation; for fatigue crack defects, the defect characteristics are crack length, crack depth, and crack number. Then, the detection benchmark values ​​of each defect characteristic are integrated to obtain a standard defect characteristic sequence; The defect signal separation unit receives the processed magnetic signal data and performs feature analysis on the magnetic signal data, identifying the frequency components in the signal and distinguishing different types of defect characteristics. It then uses independent component analysis technology to separate different defect signals, decomposing the mixed signal into independent defect signals and extracting the independent characteristics of each defect from the separated signals. The separated defect features are compared with the standard defect feature sequence, classified into known defect types, and the separated defect signals are output to the defect classification and identification unit.

7. The gear defect detection system based on gear pump processing according to claim 6, characterized in that: The defect classification and identification unit specifically includes: Collect relevant data on gear pump gear defect detection, including different types of defect signals and corresponding defect features. Then, extract gear defect sample data from the gear pump gear defect detection database and mark the defect features contained in the sample data, such as surface damage defects, geometric deformation defects, and fatigue crack defects, to form a defect dataset, which is divided into a training set and a test set. Based on the data characteristics of gear defect detection, a recurrent neural network was selected as the model architecture of the defect classification model. The neural network model was trained using labeled training data. During the training process, the model parameters were adjusted based on the training performance and verification results. The model was then tested using a test set to further optimize the model performance. The defect features contained in the currently separated defect signal are input into the trained defect classification model. The defect classification model identifies and classifies the defect type based on the extracted defect features, and then outputs the identified defect type and its location information to the defect severity analysis unit.

8. The gear defect detection system based on gear pump processing according to claim 7, characterized in that: The process of identifying and categorizing defect types includes: Integrate the extracted defect features to form a feature vector as the input of the defect classification model, and normalize the input feature vector; Load the trained defect classification model, initialize the model's input interface, and input the constructed feature vector into the defect classification model. The model performs calculations based on the input feature vector, gradually extracting and integrating features through its internal neural network structure, and calculating the probability that the input feature belongs to different defect types, indicating the probability that the input feature belongs to surface damage, geometric deformation, or fatigue crack. Based on the probability value output by the model, the defect type with the highest probability is selected as the final identification result. The identified defect type and its related information are then organized into a structured output format and output to the defect severity analysis unit for subsequent defect severity analysis. The output information includes the defect type, the specific value of the defect feature, the defect location, and the probability value output by the model.

9. The gear defect detection system based on gear pump processing according to claim 8, characterized in that: The defect severity analysis unit specifically includes: Obtaining the defect type, specific values ​​of the defect characteristics, the defect location, and the probability value output by the defect classification model from the defect classification and identification unit; The severity of defects is quantitatively assessed based on the specific values ​​of defect types and defect characteristics. For surface damage defects, the wear depth or area, the area or number of pitting, and the size or depth of the bonded area are calculated. For geometric deformation defects, the tooth profile deviation, runout deviation, and the deviation between the actual and designed center distance are measured. For fatigue crack defects, the crack extension length and depth are measured, and the number of cracks is counted. Compare the quantified defect characteristics with the pre-set defect tolerance to preliminarily judge the severity of the defect. Then, combine the ratio of all defect characteristics to the defect tolerance to calculate the defect severity coefficient and evaluate the comprehensive severity of the gear defect. Based on the gear defect severity analysis results and the defect size and location, the gear availability is judged and the defective gears are divided into three types: usable, needing repair and unusable.

10. A gear defect detection method based on gear pump processing, implemented based on the gear defect detection system based on gear pump processing according to any one of claims 1 to 9, characterized in that: The following steps are involved: Step 1: Customize the magnetization scheme according to the gear material and structure, and perform demagnetization and pre-magnetization; Step 2: Use a high-resolution magnetic sensor array to collect magnetic signals on the gear surface and inside, and capture signal changes caused by defects; Step 3: Apply signal processing technology to pre-process the collected magnetic signals and separate different defect signals; Step 4: Extract defect features from the processed magnetic signals, build a defect classification model, and identify defect types, including surface damage defects, geometric deformation defects, and fatigue crack defects; Step 5: Calculate the defect severity coefficient based on the defect characteristics and location, and determine the gear availability, which can be classified as usable, needing repair, or unusable. Step 6. Create a 3D model of the gear and mark the defect location, type and severity. Provide a user interface and support annotation and report export.

Images