Half shaft forging quality monitoring system based on machine vision

Abstract

The invention provides a half-shaft forging quality monitoring system based on machine vision, and the system comprises a data collection module which is used for collecting half-shaft data, and the half-shaft data comprises half-shaft image data, vibration data and half-shaft hardness data; the communication module is connected with the data acquisition module and is used for sending the half-shaft data to a server; the data preprocessing module is arranged in the server, is in communication connection with the communication module and is used for preprocessing the half-shaft data; the data analysis module is arranged in the server, is connected with the data preprocessing module, and is used for performing data analysis on the preprocessed half shaft data by using a data analysis algorithm to obtain the defect type of the half shaft; and the alarm module is connected with the data analysis module and is used for sending different alarm information based on the defect types. According to the invention, the forging quality of the half shaft can be monitored in real time, and the detection efficiency and accuracy are improved.

Description

Technical Field

[0001] The invention relates to the technical field of half-axle forging, and in particular to a half-axle forging quality monitoring system based on machine vision. Background Art

[0002] As a key component of the vehicle transmission system, the half-shaft connects the differential and the wheels. Its main function is to transfer the power of the engine to the wheels to propel the car forward. The quality and performance of the half-shaft are directly related to the car's handling, acceleration, fuel efficiency and driving safety. Therefore, strict monitoring of the quality of the half-shaft is an important part of ensuring the overall performance and safety of the car. The half-shaft not only has to withstand the torque from the engine, but also has to cope with various road conditions and changes in the driving environment. The reliability and durability of the half-shaft directly affect the driving stability and ride comfort of the car. If the half-shaft fails during driving, it may cause the vehicle to lose power, or even cause a serious accident, endangering the lives of the driver and passengers.

[0003] In the prior art, technicians usually perform quality inspection on forged axles by manual visual inspection. However, this method has many disadvantages, such as low inspection efficiency, large influence of human factors on inspection results, and difficulty in achieving comprehensive and real-time monitoring of forged axle quality. Summary of the invention

[0004] The present invention aims to at least solve the technical problem of low detection efficiency in the prior art, and in particular innovatively proposes a half-axle forging quality monitoring system based on machine vision.

[0005] In order to achieve the above-mentioned object of the present invention, the present invention provides a half-axle forging quality monitoring system based on machine vision, the system comprising:

[0006] A data acquisition module, used for acquiring semi-axis data, wherein the semi-axis data includes semi-axis image data, vibration data and semi-axis hardness data;

[0007] A communication module, connected to the data acquisition module, for sending the semi-axle data to a server;

[0008] A data preprocessing module, which is disposed in the server and is communicatively connected to the communication module, and is used for preprocessing the semi-axle data;

[0009] A data analysis module is provided in the server and connected to the data preprocessing module, and is used to analyze the preprocessed semi-axle data using a data analysis algorithm to obtain the defect type of the semi-axle;

[0010] An alarm module is connected to the data analysis module and is used to send different alarm information based on the defect type.

[0011] As an optional embodiment of the present invention, optionally, the acquisition module includes:

[0012] An image acquisition unit, arranged above the half-axle forging production line, for capturing image data of the half-axle;

[0013] A vibration collection unit is provided on the half-shaft forging equipment and is used to collect vibration data during the half-shaft forging process;

[0014] The hardness detection unit is installed on the half-axle forging production line and is used to measure the hardness data of the half-axle.

[0015] As an optional embodiment of the present invention, optionally, the data preprocessing module includes:

[0016] A feature extraction unit, used to extract semi-axis image features, vibration features and semi-axis hardness features from the semi-axis image data, vibration data and semi-axis hardness data respectively;

[0017] The data conversion unit is connected to the feature extraction unit and is used to convert the semi-axis image features, vibration features and semi-axis hardness features into the format required by the data analysis module.

[0018] As an optional embodiment of the present invention, optionally, the expression for extracting the semi-axis image feature is:

[0019]

[0020] Among them, F image The comprehensive index representing the semi-axis image characteristics, N pixels Represents the total number of pixels of the semi-axis image data, I i represents the gray value of pixel i, μ I Represents the mean gray value of the semi-axis image data, σ I represents the standard deviation of the grayscale value of the semi-axis image data, neighbors(i) represents the neighborhood pixel set of pixel i, I j represents the gray value of neighborhood pixel j, x j and j represents the coordinates of pixel j, x i and i Represents the coordinates of pixel i.

[0021] As an optional embodiment of the present invention, optionally, the expression for extracting vibration features is:

[0022]

[0023] Among them, F vibrationA comprehensive indicator of vibration characteristics, t0 represents the start time of the vibration signal, t f represents the end time of the vibration signal, A(t) represents the vibration amplitude at time t, represents the third-order derivative of the vibration amplitude, f(t) represents the vibration frequency at time t, f0 represents the target vibration frequency, σ f represents the standard deviation of the frequency distribution, and max(A) represents the maximum value of the vibration amplitude.

[0024] As an optional embodiment of the present invention, optionally, the expression of the semi-axle hardness characteristic is:

[0025]

[0026] Among them, F hardness A comprehensive index representing the hardness characteristics of the semi-axle, N tests Indicates the number of hardness tests, H k represents the hardness value of the kth test, represents the mean value of hardness, σ H represents the standard deviation of the hardness value, ΔH represents the maximum deviation between the hardness value and the mean hardness value, and τ represents the characteristic time constant of the hardness value change.

[0027] As an optional embodiment of the present invention, optionally, the data analysis module includes:

[0028] A defect recognition unit, used for identifying the defect type of the semi-shaft by using a defect recognition algorithm according to the semi-shaft image feature, vibration feature and semi-shaft hardness feature;

[0029] The defect grading unit is used to grade the semi-axle defects.

[0030] As an optional embodiment of the present invention, optionally, the defect recognition algorithm is expressed as:

[0031] F=[F image,1 ,F image,2 ,...,F image,M ,F vibration,1 ,F vibration,2 ,...,F vibration,N ,F hardness,1 ]

[0032] F pre =Preprocess(F)

[0033] Model=Train(X train ,y train ,Algorithm)

[0034]

[0035] Among them, F represents the semi-axis feature vector after feature fusion, [,] represents the feature fusion operation, and F image,1 ,F image,2 ,…,F image,M represents the M semi-axis image features extracted from the semi-axis image features, F vibration,1 ,F vibration,2 ,…,F vibration,N represents the N vibration features extracted from the vibration features, F hardness,1 Indicates one semi-axis hardness feature extracted from the semi-axis hardness feature, F pre represents the feature vector after feature processing, Preprocess() represents the feature processing function, Model represents the trained defect recognition model, Train() represents the training process of the defect recognition model, X train represents the feature vector matrix of the training set, y train represents the label vector of the training set, i.e., the defect type, and Algorithm represents the selected machine learning algorithm. Indicates the predicted defect type, and Predict() indicates the prediction function.

[0036] As an optional embodiment of the present invention, optionally, the expression for grading the semi-axis defects is:

[0037]

[0038] Among them, G represents the classification result of the semi-axis defect, W image represents the weight of the semi-axis image feature, W vibration represents the weight of the vibration feature, W hardness Represents the weight of the hardness feature, T1, T2, T3, ..., T n-1 Indicates the classification threshold.

[0039] As an optional embodiment of the present invention, optionally, the alarm module triggers a corresponding alarm mechanism according to the grading result of the defect grading unit.

[0040] Beneficial effects of the present invention: The present invention can realize real-time monitoring of the half-axle forging quality and improve detection efficiency and accuracy. Through machine vision technology, the present invention can automatically collect, analyze and process half-axle data, effectively avoiding the subjectivity and inefficiency of manual visual inspection. At the same time, the present invention comprehensively considers the half-axle image characteristics, vibration characteristics and hardness characteristics, and can more comprehensively and accurately evaluate the quality status of the half-axle to ensure the safety and reliability of the automobile transmission system. In addition, the alarm module of the present invention can trigger the corresponding alarm mechanism according to the defect classification results, and promptly remind the technical personnel to handle it, further improving the intelligence level of half-axle forging quality monitoring.

[0041] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0043] Figure 1 It is a structural schematic diagram of a half-axle forging quality monitoring system based on machine vision of the present invention. DETAILED DESCRIPTION

[0044] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0045] like Figure 1 As shown, a half-axle forging quality monitoring system based on machine vision, the system comprises:

[0046] A data acquisition module, used for acquiring semi-axis data, wherein the semi-axis data includes semi-axis image data, vibration data and semi-axis hardness data;

[0047] It should be noted that the data acquisition module includes a camera, a vibration acquisition unit and a hardness tester. The camera is installed above the half-axle forging production line to capture the image data of the half-axle forging process in real time. The vibration acquisition unit is directly installed on the half-axle forging equipment, which can accurately sense the vibration information generated during the forging process. The hardness tester is set on the production line to accurately measure the hardness of the half-axle. These acquisition devices jointly ensure the comprehensiveness and accuracy of the half-axle data, providing a solid foundation for subsequent data analysis. For the collection of half-axle hardness data, the specific collection process is to perform an indentation test on the half-axle after the half-axle forging is completed by the hardness tester. The tester will record the depth or diameter of the indentation, and then calculate the hardness value of the half-axle based on the known hardness and indentation relationship curve. This hardness value is one of the important indicators for evaluating the performance and quality of the half-axle material. In order to ensure the accuracy of the data, the hardness tester will be calibrated and maintained regularly to reduce the test error. In this embodiment, the hardness tester is a sampling device, that is, after a certain number of half-axles are produced, a part is randomly selected for hardness testing. This method can not only ensure the representativeness of the data, but also effectively control the testing costs.

[0048] A communication module, connected to the data acquisition module, for sending the semi-axle data to a server;

[0049] It should be noted that in this embodiment, the communication module is a wireless communication module based on wireless communication technologies such as 4G / 5G or Wi-Fi, which can quickly and stably transmit the real-time collected half-axis data to the remote server, facilitating subsequent data analysis and processing. The application of the wireless communication module not only improves the efficiency and reliability of data transmission, but also enables the monitoring system to have the functions of remote monitoring and management, greatly improving the flexibility and convenience of the system.

[0050] A data preprocessing module, which is disposed in the server and is communicatively connected to the communication module, and is used for preprocessing the semi-axle data;

[0051] It should be noted that the data preprocessing module includes a feature extraction unit and a data conversion unit, which work together to ensure the accuracy and efficiency of subsequent data analysis. The feature extraction unit can accurately capture key features in the semi-axis image, such as edges, textures, etc., as well as features such as frequency and amplitude in vibration data, and features such as hardness value distribution in hardness data. These feature extraction processes are based on complex algorithms and models and can automatically adapt to different semi-axis types and forging conditions. The data conversion unit is responsible for converting these extracted features into a format that the data analysis module can understand and process, such as converting image features into vector representations, and converting vibration and hardness features into numerical matrices. Such a data preprocessing process provides a high-quality data foundation for subsequent data analysis and ensures the accuracy and reliability of the monitoring system.

[0052] A data analysis module is provided in the server and connected to the data preprocessing module, and is used to analyze the preprocessed semi-axle data using a data analysis algorithm to obtain the defect type of the semi-axle;

[0053] It should be noted that in this embodiment, the data analysis module includes a defect recognition unit and a defect grading unit. The defect recognition unit can accurately identify the defect types of the half-shaft, such as cracks, inclusions, deformations, etc., based on the half-shaft feature vector after feature fusion and through a trained defect recognition model. This process makes full use of the powerful capabilities of the machine learning algorithm to achieve intelligent identification of half-shaft defects. The defect grading unit further grades the identified defects and divides them into different levels, such as minor, moderate, severe, etc., according to the severity of the defects. Such grading results not only help technicians quickly understand the quality status of the half-shaft, but also provide important references for subsequent repairs and processing.

[0054] An alarm module is connected to the data analysis module and is used to send different alarm information based on the defect type.

[0055] It should be noted that in the present embodiment, the alarm module includes a plurality of alarm lights, each of which corresponds to a different defect level. After the data analysis module identifies the defect type of the half-axle and grades it, the corresponding alarm light will light up, indicating the severity of the defect with different colors or flashing frequencies. For example, minor defects correspond to yellow lights, medium defects correspond to orange lights, and serious defects correspond to red lights. This design enables technicians to understand the quality status of the half-axle at a glance and quickly take necessary repair and treatment measures. In addition, the alarm module can also be connected to the factory's monitoring system or SMS platform to send the alarm information to the mobile phones or computers of relevant personnel in real time, further improving the timeliness and effectiveness of the alarm.

[0056] In summary, in this embodiment, the image data, vibration data and hardness data of the half-axle are first collected by the camera, vibration collection unit and hardness tester respectively, ensuring the comprehensiveness and accuracy of the data. Then, these data are quickly transmitted to the remote server through the wireless communication module, realizing real-time data sharing and remote monitoring. On the server, the data preprocessing module preprocesses the received half-axle data, extracts key features and converts them into a format suitable for data analysis. Then, the data analysis module uses advanced machine learning algorithms to analyze these preprocessed data, accurately identify the defect types of the half-axle, and classify the defects. Finally, the alarm module triggers the corresponding alarm mechanism according to the defect classification results, and promptly reminds the technicians to handle it. The entire system realizes real-time monitoring and intelligent management of the forging quality of the half-axle, greatly improves the detection efficiency and accuracy, and ensures the safety and reliability of the automobile transmission system.

[0057] As an optional embodiment of the present invention, optionally, the acquisition module includes:

[0058] An image acquisition unit, arranged above the half-axle forging production line, for capturing image data of the half-axle;

[0059] It should be noted that in this embodiment, the image acquisition unit is a camera, which has high resolution and high-speed shooting capabilities, and can capture subtle changes in the half-axle forging process in real time to ensure the clarity and accuracy of the image data. The camera installation position must ensure that it can fully cover the forging area of ​​the half-axle to avoid missing any possible defects. In addition, the camera is also equipped with automatic dimming and anti-shake functions to adapt to shooting needs under different light and vibration conditions, further improving the quality of image data.

[0060] A vibration collection unit is provided on the half-shaft forging equipment and is used to collect vibration data during the half-shaft forging process;

[0061] It should be noted that the vibration collection unit is a vibration sensor, which can sense the tiny vibrations generated during the forging process of the half-shaft with high precision, and convert these vibration information into electrical signals for transmission. The vibration sensor is usually installed at the key parts of the forging equipment, such as the mold or the workbench, to ensure that the vibration characteristics generated during the forging process can be accurately captured. These vibration characteristics are of great significance for identifying the internal defects of the half-shaft, such as cracks or inclusions. Through the analysis of the vibration data, the stress conditions and material properties of the half-shaft during the forging process can be further understood, providing an important basis for subsequent defect identification and quality assessment. The vibration data is collected by the vibration sensor. Specifically, in this embodiment, the vibration sensor is specifically installed on the contact surface between the mold and the workpiece of the half-shaft forging equipment. In this way, when the mold forges the half-shaft, the vibration sensor can sense the vibration signal generated by the interaction force between the mold and the half-shaft in real time. In order to ensure the accuracy and stability of the vibration data, the vibration sensor adopts a high-precision and high-sensitivity design, and is equipped with a signal processing circuit with strong anti-interference ability, which can effectively filter out environmental noise and other interference factors, and ensure that the collected vibration data truly reflects the vibration characteristics during the forging process of the half-shaft. In addition, the vibration sensor has long-term stability and reliability, and can continue to work in harsh industrial environments, providing a strong guarantee for the stable operation of the half-axle forging quality monitoring system.

[0062] Vibration data analysis methods include spectrum analysis, time domain analysis, and modal analysis. Spectral analysis can decompose vibration signals into different frequency components, and identify abnormal vibration characteristics by observing frequency distribution and amplitude changes; time domain analysis focuses on the change pattern of vibration signals over time, such as statistical characteristics such as peak value, mean value, and variance, to evaluate the stability of the half-shaft during the forging process; modal analysis reveals the dynamic response characteristics of the half-shaft under specific excitation by solving the modal parameters of the system, such as natural frequency and damping ratio. Combining these analysis methods, a comprehensive analysis of vibration data during the half-shaft forging process can be achieved, providing a more accurate basis for defect identification and quality assessment.

[0063] The hardness detection unit is installed on the half-axle forging production line and is used to measure the hardness data of the half-axle.

[0064] It should be noted that in this embodiment, the hardness detection unit is a hardness detector, which adopts the advanced indentation test principle and can accurately measure the hardness value of the half shaft. The hardness detector is usually set at the end of the production line. When the half shaft is forged, it will be automatically sent to the detection area for hardness testing. During the test, the hardness detector will apply a certain pressure to the half shaft, and record the depth or diameter of the indentation, and then calculate the hardness value of the half shaft according to the known hardness and indentation relationship curve. This hardness value is one of the important indicators for evaluating the performance and quality of the half shaft material, which is of great significance for ensuring the safety and reliability of the automobile transmission system. In order to ensure the accuracy and stability of the hardness test, the hardness detector will be calibrated and maintained regularly to reduce the test error. At the same time, in order to avoid interference with the normal operation of the production line, the hardness detector adopts an efficient and fast testing method, which can complete the hardness test of a large number of half shafts in a short time, meeting the efficient and automated production requirements of the production line.

[0065] As an optional embodiment of the present invention, optionally, the data preprocessing module includes:

[0066] A feature extraction unit, used to extract semi-axis image features, vibration features and semi-axis hardness features from the semi-axis image data, vibration data and semi-axis hardness data respectively;

[0067] It should be noted that in this embodiment, the feature extraction unit uses advanced image processing technology, signal processing technology and data analysis methods, which can accurately capture key features in the semi-axle image, such as edges, textures, shapes, etc., as well as frequency, amplitude and other features in vibration data, and hardness value, hardness gradient and other features in hardness data. These feature extraction processes are based on complex algorithms and models such as deep learning and machine learning, and can automatically adapt to different semi-axle types and forging conditions, ensuring the accuracy and robustness of feature extraction. The extracted feature data will be used in subsequent data analysis and defect identification processes, providing an important basis for monitoring and evaluating the quality of semi-axle forging.

[0068] The data conversion unit is connected to the feature extraction unit and is used to convert the semi-axis image features, vibration features and semi-axis hardness features into the format required by the data analysis module.

[0069] It should be noted that in this embodiment, the data conversion unit is responsible for format conversion and standardization of the extracted semi-axis image features, vibration features and semi-axis hardness features to ensure that they can be correctly understood and processed by the data analysis module. Specifically, image processing:

[0070] Grayscale conversion: Convert color images to grayscale images to reduce the amount of calculation.

[0071] Noise Removal: Apply a filter (such as a Gaussian filter) to remove image noise.

[0072] Image enhancement: Enhance image details using methods such as histogram equalization or contrast stretching.

[0073] Feature extraction: Use pre-trained convolutional neural network (CNN) models (such as VGG, ResNet) to extract image feature vectors.

[0074] Vibration data processing:

[0075] Signal denoising: Apply wavelet transform or Fourier transform to remove noise from vibration signals.

[0076] Feature extraction: Extract the statistical features (such as mean, variance, peak value, skewness, kurtosis) and frequency domain features (such as main frequency, secondary frequency, etc.) of the vibration signal.

[0077] Hardness data preprocessing:

[0078] Data cleaning: remove outliers or missing values.

[0079] Feature transformation: transform the hardness data into a form suitable for model input, such as standardization or normalization.

[0080] As an optional embodiment of the present invention, optionally, the expression for extracting the semi-axis image feature is:

[0081]

[0082] Among them, F image A comprehensive index representing the characteristics of the semi-axis image;

[0083] N pixels The total number of pixels representing the semi-axis image data;

[0084] I i Represents the gray value of pixel i;

[0085] μ I Represents the mean gray value of the semi-axis image data;

[0086] σ I Represents the standard deviation of the grayscale value of the semi-axis image data;

[0087] neighbors(i) represents the set of neighboring pixels of pixel i;

[0088] I j Represents the gray value of neighborhood pixel j;

[0089] x j and j represents the coordinates of pixel j;

[0090] x i and i Represents the coordinates of pixel i.

[0091] It should be noted that this expression combines the normalization of pixel values, the inverse tangent transformation of gradient information, and the difference of neighborhood pixels to capture the texture, edge, and contrast characteristics of the image.

[0092] As an optional embodiment of the present invention, optionally, the expression for extracting vibration features is:

[0093]

[0094] Among them, F vibration A comprehensive index that represents vibration characteristics;

[0095] t0 represents the start time of the vibration signal;

[0096] t f Indicates the end time of the vibration signal;

[0097] A(t) represents the vibration amplitude at time t;

[0098] The third derivative of the vibration amplitude represents the rate of change of acceleration;

[0099] f(t) represents the vibration frequency at time t;

[0100] f0 represents the target vibration frequency;

[0101] σ f represents the standard deviation of the frequency distribution;

[0102] max(A) indicates the maximum value of the vibration amplitude.

[0103] It should be noted that this expression combines the rate of change of acceleration, the weighting of frequency, and the periodic change of vibration amplitude to capture the dynamic characteristics of vibration.

[0104] As an optional embodiment of the present invention, optionally, the expression of the semi-axle hardness characteristic is:

[0105]

[0106] Among them, F hardness A comprehensive index that represents the hardness characteristics of the half shaft;

[0107] N tests Indicates the number of hardness tests;

[0108] H k represents the hardness value of the kth test;

[0109] Indicates the mean value of hardness value;

[0110] σ H Indicates the standard deviation of hardness values;

[0111] ΔH represents the maximum deviation between the hardness value and the mean hardness value;

[0112] τ represents the characteristic time constant of hardness value change.

[0113] It should be noted that this expression combines the dispersion of hardness values ​​(expressed by the fourth variance) and the exponential decay of hardness value changes to capture the uniformity and stability characteristics of hardness.

[0114] As an optional embodiment of the present invention, optionally, the data analysis module includes:

[0115] A defect recognition unit, used for identifying the defect type of the semi-shaft by using a defect recognition algorithm according to the semi-shaft image feature, vibration feature and semi-shaft hardness feature;

[0116] It should be noted that the defect recognition unit combines the semi-axis image features, vibration features and semi-axis hardness features to identify the defect type through a method combining multi-feature fusion and machine learning. First, this embodiment needs to combine different types of features into a feature vector. Before defect recognition, the features usually need to be preprocessed, such as standardization or normalization. Next, this embodiment uses a machine learning algorithm to train a defect recognition model and uses the model for prediction. The selection of specific machine learning algorithms may include support vector machines (SVM), random forests, neural networks, etc. These algorithms can learn the complex relationship between different features, so as to achieve accurate recognition of the semi-axis defect type. In order to improve the accuracy and robustness of recognition, this embodiment can also use ensemble learning methods, such as AdaBoost, Bagging, etc., to fuse the prediction results of multiple machine learning models to obtain more reliable recognition results. In addition, the defect recognition unit will continue to collect new semi-axis data for continuous training and optimization of the model to adapt to different production conditions and changes in semi-axis types, and ensure the accuracy and timeliness of defect recognition.

[0117] The defect grading unit is used to grade the semi-axle defects.

[0118] It should be noted that in this embodiment, the defect grading unit mainly grades the identified defects based on factors such as the severity, size, position of the semi-axle image features, vibration features and semi-axle hardness features, as well as possible impacts on the semi-axle performance. The grading standards are usually formulated based on industry standards, historical data and actual customer needs. Through defect grading, the quality status of the semi-axle can be evaluated more accurately, providing a basis for subsequent decision-making and processing. The defect grading unit adopts advanced classification algorithms and rule engines, combined with the knowledge and experience of experts in the field, to achieve automatic and intelligent grading of semi-axle defects. At the same time, the unit will also dynamically adjust and optimize the grading standards and methods according to the actual situation in the production process and changes in customer needs to ensure the accuracy and practicality of the grading.

[0119] As an optional embodiment of the present invention, optionally, the defect recognition algorithm is expressed as:

[0120] F=[F image,1 ,F image,2 ,…,F image,M ,F vibration,1 ,F vibration,2 ,…,F vibration,N ,F hardness,1 ]

[0121] F pre =Preprocess(F)

[0122] Model=Train(X train ,y train ,Algorithm)

[0123]

[0124] Among them, F represents the semi-axis feature vector after feature fusion;

[0125] [,] indicates feature fusion operation;

[0126] F image,1 ,F image,2 ,…,F image,M represents M semi-axis image features extracted from the semi-axis image features;

[0127] F vibration,1 ,F vibration,2 ,…,F vibration,N represents N vibration features extracted from the vibration features;

[0128] F hardness,1 represents a semi-axis hardness feature extracted from the semi-axis hardness feature;

[0129] F preRepresents the feature vector after feature processing

[0130] Preprocess() represents the feature processing function;

[0131] Model represents the trained defect recognition model;

[0132] Train() represents the training process of the defect recognition model;

[0133] X train represents the feature vector matrix representing the training set;

[0134] y train The label vector representing the training set, i.e., the defect type;

[0135] Algorithm represents the selected machine learning algorithm;

[0136] Indicates the predicted defect type;

[0137] Predict() represents the prediction function.

[0138] It should be noted that the defect recognition algorithm first fuses the semi-axle image features, vibration features and semi-axle hardness features to form a comprehensive feature vector. The feature fusion operation can make full use of the complementarity between different features and improve the accuracy of defect recognition. Then, the fused feature vector is subjected to feature processing, such as standardization or normalization, to ensure the stability and consistency of the model input. Next, the processed feature vector is predicted using the trained defect recognition model, and the predicted defect type is output. The training process of the defect recognition model includes steps such as preparing a training set, selecting a machine learning algorithm, and training a model. The training set consists of a large amount of semi-axle data, including a feature vector matrix and a corresponding label vector (i.e., defect type). By selecting a suitable machine learning algorithm and training it on the training set, a model that can accurately identify the defect type of the semi-axle can be obtained. Finally, the prediction function is used to predict the defect type of the new semi-axle data to achieve real-time monitoring and evaluation of the semi-axle forging quality. The specific steps for training the defect recognition model are as follows:

[0139] S1. Collect historical half-axis data, which should include half-axis images, vibration signals, hardness test values ​​and corresponding defect type labels.

[0140] S2. Preprocess the collected data, including image processing operations such as image grayscale conversion, noise removal, image enhancement, vibration signal denoising and feature extraction, and hardness data cleaning and feature conversion.

[0141] S3. Divide the preprocessed data into a training set and a test set. The training set is used to train the defect recognition model, and the test set is used to evaluate the performance of the model.

[0142] S4. Select an appropriate machine learning algorithm, such as support vector machine (SVM), random forest or neural network, as the basis of the defect recognition model.

[0143] S5. Use the training set to train the selected machine learning algorithm to obtain a preliminary defect recognition model. During the training process, the parameters of the model can be optimized through methods such as cross-validation to improve the recognition accuracy of the model.

[0144] S6. Evaluate the trained model and use the test set data to verify the performance of the model. Evaluation indicators can include accuracy, recall, F1 score, etc.

[0145] S7. Adjust and optimize the model based on the evaluation results, such as adjusting feature selection, improving feature fusion methods, or trying other machine learning algorithms to improve the recognition performance of the model.

[0146] S8. When the model performance meets the requirements, it is deployed to the actual half-axle forging quality monitoring system to monitor and evaluate the forging quality of the half-axle in real time. At the same time, new half-axle data is continuously collected to update and optimize the model to adapt to changes in production conditions and half-axle types.

[0147] As an optional embodiment of the present invention, optionally, the expression for grading the semi-axis defects is:

[0148]

[0149] Among them, G represents the classification result of the semi-axis defect;

[0150] W image Represents the weight of the semi-axis image feature;

[0151] W vibration represents the weight of the vibration feature;

[0152] W hardness represents the weight of the hardness feature;

[0153] T1, T2, T3, ..., T n-1 Indicates the classification threshold.

[0154] It should be noted that this expression comprehensively considers the importance of the semi-axle image features, vibration features and hardness features in defect grading. By assigning weights to different features and combining grading thresholds, accurate grading of semi-axle defects can be achieved. The weights can be determined based on historical data, industry standards, and the knowledge and experience of experts in the field. The grading threshold is set according to factors such as the severity of the defect and the impact on the performance of the semi-axle. Through this expression, the quality status of the semi-axle can be evaluated more scientifically, providing strong support for subsequent processing and decision-making. At the same time, the expression also has certain flexibility and scalability. The weights and grading thresholds can be adjusted and optimized according to actual needs and changes in production conditions to ensure the accuracy and practicality of the grading.

[0155] As an optional embodiment of the present invention, optionally, the alarm module triggers a corresponding alarm mechanism according to the grading result of the defect grading unit.

[0156] It should be noted that when the defect grading unit completes the grading of the half-axle defects, the alarm module will automatically trigger the corresponding alarm mechanism according to the grading results. For defects of different levels, the alarm module can take different response measures. For example, for severe defects, the alarm module will immediately trigger an emergency shutdown signal to stop the operation of the half-axle forging production line to prevent defective products from flowing into the next process, and at the same time send an audio-visual alarm signal to the operator to remind him to check and handle it immediately. For minor defects, the alarm module will trigger an early warning signal to remind the operator to pay attention to the potential quality problems of the half-axle and recommend further testing and analysis. In addition, the alarm module can also upload the defect grading results and alarm information to the company's quality management system in real time, so that managers can monitor and analyze the quality of half-axle forging in real time, and promptly discover and solve potential quality problems. Through the intelligent response of the alarm module, the efficiency and accuracy of the half-axle forging quality monitoring system can be greatly improved to ensure the quality and safety of half-axle products.

[0157] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.

Claims

1. A half-axle forging quality monitoring system based on machine vision, characterized in that: The system comprises: A data acquisition module, used for acquiring semi-axis data, wherein the semi-axis data includes semi-axis image data, vibration data and semi-axis hardness data; A communication module, connected to the data acquisition module, for sending the semi-axle data to a server; A data preprocessing module, which is disposed in the server and is communicatively connected to the communication module, and is used for preprocessing the semi-axle data; A data analysis module is provided in the server and connected to the data preprocessing module, and is used to analyze the preprocessed semi-axle data using a data analysis algorithm to obtain the defect type of the semi-axle; An alarm module is connected to the data analysis module and is used to send different alarm information based on the defect type.

2. A half-axle forging quality monitoring system based on machine vision as claimed in claim 1, characterized in that: The acquisition module comprises: An image acquisition unit, arranged above the half-axle forging production line, for capturing image data of the half-axle; A vibration collection unit is provided on the half-shaft forging equipment and is used to collect vibration data during the half-shaft forging process; The hardness detection unit is installed on the half-axle forging production line and is used to measure the hardness data of the half-axle.

3. A half-axle forging quality monitoring system based on machine vision as claimed in claim 1, characterized in that: The data preprocessing module comprises: A feature extraction unit, used to extract semi-axis image features, vibration features and semi-axis hardness features from the semi-axis image data, vibration data and semi-axis hardness data respectively; The data conversion unit is connected to the feature extraction unit and is used to convert the semi-axis image features, vibration features and semi-axis hardness features into the format required by the data analysis module.

4. A half-axle forging quality monitoring system based on machine vision as claimed in claim 3, characterized in that: The expression for extracting the semi-axis image features is: Among them, F image The comprehensive index representing the semi-axis image characteristics, N pixels Represents the total number of pixels of the semi-axis image data, I i represents the gray value of pixel i, μ I Represents the mean gray value of the semi-axis image data, σ I represents the standard deviation of the grayscale value of the semi-axis image data, neighbors(i) represents the neighborhood pixel set of pixel i, I j represents the gray value of neighborhood pixel j, x j and j represents the coordinates of pixel j, x i and i Represents the coordinates of pixel i.

5. A half-axle forging quality monitoring system based on machine vision as claimed in claim 3, characterized in that: The expression for extracting vibration features is: Among them, F vibration A comprehensive indicator of vibration characteristics, t0 represents the start time of the vibration signal, t f represents the end time of the vibration signal, A(t) represents the vibration amplitude at time t, represents the third-order derivative of the vibration amplitude, f(t) represents the vibration frequency at time t, f0 represents the target vibration frequency, σ f represents the standard deviation of the frequency distribution, and max(A) represents the maximum value of the vibration amplitude.

6. A half-axle forging quality monitoring system based on machine vision as claimed in claim 3, characterized in that: The expression of the semi-axle hardness characteristic is: Among them, F hardness A comprehensive index representing the hardness characteristics of the semi-axle, N tests Indicates the number of hardness tests, H k represents the hardness value of the kth test, represents the mean value of hardness, σ H represents the standard deviation of the hardness value, ΔH represents the maximum deviation between the hardness value and the mean hardness value, and τ represents the characteristic time constant of the hardness value change.

7. A half-axle forging quality monitoring system based on machine vision as claimed in claim 3, characterized in that: The data analysis module includes: A defect recognition unit, used for identifying the defect type of the semi-shaft by using a defect recognition algorithm according to the semi-shaft image feature, vibration feature and semi-shaft hardness feature; The defect grading unit is used to grade the semi-axle defects.

8. A half-axle forging quality monitoring system based on machine vision as claimed in claim 7, characterized in that: The expression of the defect recognition algorithm is: F=[F image,1 ,F image,2 ,…,F image,M ,F vibration,1 ,F vibration,2 ,...,F vibration,N ,F hardness,1 ] F pre =Preprocess(F) Model=Train(X train ,y train ,Algorithm) Among them, F represents the semi-axis feature vector after feature fusion, [,] represents the feature fusion operation, and F image,1 ,F image,2 ,...,F image,M represents the M semi-axis image features extracted from the semi-axis image features, F vibration,1 ,F vibration,2 ,...,F vibration,N represents the N vibration features extracted from the vibration features, F hardness,1 Indicates one semi-axis hardness feature extracted from the semi-axis hardness feature, F pre represents the feature vector after feature processing, Preprocess() represents the feature processing function, Model represents the trained defect recognition model, Train() represents the training process of the defect recognition model, X train represents the feature vector matrix of the training set, y train represents the label vector of the training set, i.e., the defect type, and Algorithm represents the selected machine learning algorithm. Indicates the predicted defect type, and Predict() indicates the prediction function.

9. A half-axle forging quality monitoring system based on machine vision as claimed in claim 7, characterized in that: The expression for grading the semi-axle defects is: Among them, G represents the classification result of the semi-axis defect, W image represents the weight of the semi-axis image feature, W vibration represents the weight of the vibration feature, W hardness Represents the weight of the hardness feature, T1, T2, T3, ..., T n-1 Indicates the classification threshold.

10. A half-axle forging quality monitoring system based on machine vision as claimed in claim 7 or 9, characterized in that: The alarm module triggers a corresponding alarm mechanism according to the classification result of the defect classification unit.

Images