Tire internal defect detection method and system based on tactile perception and ultrasonic fusion

Abstract

The invention relates to the technical field of tire detection, in particular to a tire internal defect detection method and system based on tactile perception and ultrasonic fusion. According to the method, through cooperative work of a touch sensor and an ultrasonic sensor, three-dimensional stress data of the surface of a tire and ultrasonic echo signals in the tire are collected respectively. Through preprocessing, weighted fusion, time synchronization and space alignment of tactile data and ultrasonic data, the consistency and integrity of the data are ensured. On this basis, possible defect areas are detected and identified by using abnormal values, and defects are classified through a machine learning algorithm, including types of cracks, bubbles, cavities and the like. Meanwhile, the severity of the defects is evaluated according to the types, positions and sizes of the defects, and a spatial distribution diagram and a severity evaluation report of the defects are generated. The method can improve the precision, robustness and efficiency of tire defect detection, has high application value, and is suitable for quality monitoring and maintenance of tires.

Description

Technical Field

[0001] The present invention relates to the technical field of tire detection, and in particular to a tire internal defect detection method and system based on tactile perception and ultrasonic fusion. Background Art

[0002] With the continuous development of modern industry, tires, as important components of transportation tools, have a direct impact on safety and service life due to their quality. Therefore, tire quality control and defect detection have become a vital part of automobile manufacturing and maintenance. Existing tire defect detection methods mainly rely on manual inspection, visual inspection, and non-destructive detection techniques based on vibration or acoustics. However, these methods still have some limitations in terms of detection accuracy, efficiency, and robustness.

[0003] Although traditional manual inspection methods can detect obvious defects, they are inefficient during large-scale production or maintenance, and are easily affected by the experience and subjective factors of the inspectors. In addition, manual inspection is usually unable to effectively identify minor internal defects, and the sampling inspection method is prone to miss some potential problems.

[0004] Visual inspection technology has also been used in tire defect detection, especially machine vision technology can automatically detect surface defects. However, visual inspection methods are limited by light source conditions and cannot effectively detect defects inside the tire. The cleanliness of the environment, lighting conditions and external interference may also affect the detection results. For some deep defects, such as cracks, bubbles or delamination inside the tire, it is difficult for existing visual inspection methods to effectively detect them.

[0005] With the development of non-destructive testing technology (NDT), ultrasonic testing has become a common means of internal defect detection. Ultrasonic waves can penetrate tire materials and detect internal defects such as cracks, bubbles and cavities through the characteristics of sound wave propagation. Ultrasonic testing technology has strong depth penetration capability, but because ultrasonic echo signals are easily interfered by factors such as tire surface roughness and material unevenness, its detection results are often affected by environmental conditions and cannot provide comprehensive surface defect information.

[0006] In recent years, tactile sensing technology, as an emerging intelligent sensing method, has been widely used in the fields of robots, intelligent equipment, etc. The Chinese invention patent application (publication number: CN116818172A, publication date: 2023-09-29) discloses a flexible interface detection method and device based on tactile perception, the method comprising the following steps: controlling the three-dimensional displacement stage to move downward along the Z-axis direction, obtaining the Z-direction force value of the three-axis force sensor in real time, stopping the movement when the force value reaches a preset threshold, and establishing a reference plane for the sensor to scan the flexible interface; on the reference plane, realizing the synchronous control of the three-axis force sensor data acquisition and the displacement stage based on the hard trigger method; scanning the flexible interface line by line based on the data acquisition method, collecting the force change data of the sensor and the flexible interface in the X, Y and Z directions as three-dimensional tactile information; fusing and calculating the three-dimensional force data to realize high-precision, high-resolution, multi-dimensional tactile scanning imaging of the flexible interface, and finally realizing flexible interface detection based on tactile perception. The tactile sensor can perceive the slight changes on the contact surface in real time and provide more accurate surface morphology information based on the force data. Combining tactile sensing with ultrasonic detection technology can effectively make up for the shortcomings of single sensor technology. Tactile sensors can obtain microscopic force changes on the surface in real time, while ultrasonic sensors can penetrate deep into the tire to detect potential defects. The combination of the two helps to improve the accuracy, coverage and robustness of defect detection.

[0007] However, in the field of tire defect detection, there is no relevant technology that can effectively combine tactile perception and ultrasonic detection to comprehensively identify the surface and internal defects of tires through multi-modal perception data fusion. Most of the existing technologies focus on the application of single sensor technology and cannot give full play to the advantages of tactile perception and ultrasonic technology. Therefore, there is a lot of room for improvement in accuracy, adaptability and intelligence. Summary of the invention

[0008] In order to solve the above-mentioned technical problems, the purpose of the present invention is to provide a tire internal defect detection method and system based on tactile perception and ultrasonic fusion, which can give full play to the synergistic effect of tactile perception and ultrasonic detection through multimodal data fusion technology to improve the accuracy, reliability and automation level of tire defect detection.

[0009] In order to achieve the above-mentioned purpose, the present invention adopts the following technical solutions:

[0010] A tire internal defect detection method based on tactile perception and ultrasonic fusion includes the following steps:

[0011] 1) Collecting three-dimensional force data of the tire surface through a tactile sensor, wherein the tactile sensor can detect force changes on the tire surface in the X, Y, and Z directions respectively;

[0012] 2) Collecting echo signals inside the tire through an ultrasonic sensor, wherein the ultrasonic sensor can detect the echo amplitude and echo delay time of the ultrasonic wave to reflect the defect information inside the tire;

[0013] 3) preprocessing the tactile data and the ultrasonic data, including standardizing the tactile data, converting the tactile data into standardized three-dimensional force data, and denoising and standardizing the ultrasonic data, so as to facilitate subsequent fusion and analysis;

[0014] 4) performing weighted fusion of the tactile data processed in step 3 and the ultrasonic data, wherein the weighted fusion adopts a weighted average method, wherein the weights of the tactile data and the ultrasonic data are dynamically determined by a preset rule or a machine learning algorithm;

[0015] 5) Perform time synchronization and spatial alignment on the fused data to ensure the consistency of the tactile data and the ultrasonic data in terms of sampling time and spatial position;

[0016] 6) Perform outlier detection based on the fused data to identify candidate areas that may have defects and compare them with normal areas;

[0017] 7) Classify the detected candidate areas through machine learning algorithms to identify the types of defects, including but not limited to cracks, voids, and bubbles;

[0018] 8) Calculate the severity of defects based on the type, location and size of the defects, and output the spatial distribution map of the defects and the severity assessment report.

[0019] Preferably, the tactile sensor is a three-axis force sensor capable of collecting force changes on the tire surface in the X, Y and Z directions respectively; and / or, the ultrasonic sensor is used to collect echo signals inside the tire, the echo signals including echo amplitude and echo delay time, the echo amplitude is used to identify the reflection characteristics of the defect, and the echo delay time is used to infer the depth and position of the defect.

[0020] Preferably, the preprocessing in step 3) includes: standardizing the tactile data, normalizing the force value in each direction, and removing background noise from the ultrasonic data through a denoising algorithm to improve the validity of the data.

[0021] Preferably, the weighted fusion in step 4) adopts a weighted average method, and the weight value is dynamically adjusted according to the importance of the tactile data and ultrasonic data and their contribution to defect identification, and the weight is obtained by training a machine learning algorithm; and / or, the time synchronization and spatial alignment in step 5) are achieved by an interpolation algorithm, the timestamps of the tactile data and ultrasonic data are aligned, and the correspondence between the ultrasonic data and the tactile data on the tire surface and inside is ensured by a spatial mapping algorithm.

[0022] Preferably, the outlier detection in step 6) is based on threshold judgment, and when the changes in the force data and the ultrasonic echo signal exceed a preset threshold, it is marked as a defect candidate area.

[0023] Preferably, the machine learning algorithm in step 7) is a support vector machine (SVM), a random forest or a convolutional neural network (CNN), and the defect candidate areas are classified by the trained model to determine the type of defect.

[0024] Preferably, the defect severity assessment in step 8) is performed based on the type, location and size of the defect, and the assessment result is displayed in real time through a display module to provide a detailed report of the defect.

[0025] Furthermore, the present invention also discloses a tire internal defect detection system based on tactile perception and ultrasonic fusion, which implements the method described, including:

[0026] Tactile sensor: Tactile sensor is used to collect three-dimensional force data on the tire surface.

[0027] Ultrasonic sensor, which is used to collect ultrasonic echo data inside the tire.

[0028] The data fusion module is used to perform weighted fusion of tactile data and ultrasonic data.

[0029] Defect detection module, which is used to perform outlier detection, defect classification and severity assessment based on the fused data.

[0030] Display module,The display module is used to display the inspection results and provide a report on the spatial distribution and severity assessment of defects.

[0031] Furthermore, the present invention also discloses a computer-readable storage medium on which a computer program or instruction is stored. When the computer program or instruction is executed by a processor, step 3) to step 8) of the method are implemented.

[0032] Furthermore, the present invention also discloses a computer program product, including a computer program or an instruction, which implements step 3) to step 8) of the method when executed by a processor.

[0033] The present invention adopts the above technical solution and significantly improves the accuracy, efficiency and robustness of tire defect detection by fusing multimodal data of tactile sensors and ultrasonic sensors. The specific technical effects are as follows:

[0034] 1. Improve defect detection accuracy: By combining tactile perception and ultrasonic detection technology, the present invention can achieve all-round detection of tire surface and internal defects. Tactile sensors can accurately capture tiny force changes on the tire surface, while ultrasonic sensors can detect deep defects inside the tire, such as cracks, cavities, bubbles, etc. The combination of the two can effectively make up for the shortcomings of a single sensor and improve the accuracy of defect recognition, especially when detecting tiny and difficult-to-find internal defects, the present invention shows significant advantages.

[0035] 2. Enhance the robustness and adaptability of detection: Traditional visual and ultrasonic detection methods are easily affected by lighting, environmental interference and tire surface conditions, and the detection results have large fluctuations. However, the present invention fully utilizes the complementarity of two different sensors by integrating tactile perception and ultrasonic detection data, which can not only improve the reliability of detection, but also effectively cope with complex environmental conditions and diverse tire surface conditions, and improve the adaptability and robustness of the detection system.

[0036] 3. Improve detection efficiency and automation level: The present invention realizes automatic synchronization and fusion processing of tactile data and ultrasonic data through multimodal data fusion algorithm, avoids the complexity of manual intervention, and improves the automation level of the detection process. The multimodal fusion method makes defect identification more efficient, can quickly extract effective information from a large amount of data, reduces manual operation time, and greatly improves detection efficiency.

[0037] 4. Real-time defect location and severity assessment: The present invention can accurately locate the defect location, conduct severity assessment based on the defect type and size, and provide a comprehensive and detailed inspection report. Through the spatial distribution map and severity assessment of the defects, the operator can more clearly understand the quality status of the tire and avoid the occurrence of potential safety hazards.

[0038] 5. Improve the intelligence level of the detection system

[0039] The present invention uses a machine learning algorithm to classify and identify defects, and intelligently determines the type and location of defects through outlier detection, feature extraction, and defect type classification. Compared with the traditional experience-based manual judgment method, the present invention can achieve more intelligent, efficient, and objective defect detection, greatly reducing the impact of human factors and ensuring the accuracy and consistency of the detection results.

[0040] In summary, the present invention provides a high-precision, high-efficiency, and intelligent tire defect detection solution through an innovative method of fusion of tactile perception and ultrasound, which overcomes the limitations of the single sensor detection method in the prior art and has significant technical advantages and broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a flowchart of the invention process of the present invention.

[0042] Figure 2 It is a system structure block diagram of the present invention. DETAILED DESCRIPTION

[0043] The following is a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with 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. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] Example 1

[0045] like Figure 1 As shown, this embodiment provides a method for detecting internal defects of a tire, and the main steps are as follows:

[0046] Step 1: Tactile Data Collection

[0047] A three-axis force sensor is used to collect real-time force data on the tire surface. The sensor can detect tiny force changes on the tire surface in the three directions of X, Y, and Z. The three-axis force sensor obtains detailed three-dimensional force data by accurately measuring the force changes at the contact point between the tire surface and the sensor.

[0048] The tactile sensor contacts the tire surface and records force data at each sampling point:

[0049] F X (t): represents the force in the X direction;

[0050] F Y (t): represents the force in the Y direction;

[0051] F Z (t): represents the force in the Z direction.

[0052] The formula means:

[0053]

[0054] Where F(t) represents the three-dimensional force data collected at time t. By scanning point by point, the sensor obtains the force information of the entire tire surface.

[0055] Furthermore, during the collection of tactile data, the sampling frequency and sensitivity of the sensor are adjusted according to the different surface conditions of the tire (such as roughness, curvature, etc.).

[0056] Step 2: Ultrasound data acquisition

[0057] Use ultrasonic sensors to detect defects inside tires. Ultrasonic sensors detect defects inside tires by emitting ultrasonic waves and receiving their echo signals. The amplitude and delay time of the echo signal provide information about potential defects such as cracks, bubbles or cavities inside the tire.

[0058] Ultrasonic sensors detect defects inside tires by emitting ultrasonic signals and receiving echo signals reflected back. The propagation characteristics of ultrasonic signals enable them to penetrate tire materials. When encountering defects (such as cracks or holes), the echo signal changes. The amplitude and delay time of the reflected signal can be used to analyze the location, size and type of the defect.

[0059] The amplitude and delay time of the echo signal are used to reflect the nature and location of the defect:

[0060] A(t): represents the amplitude of the echo signal; the amplitude of the echo reflects the reflection intensity of the defect. The larger or more serious the defect is, the larger the amplitude of the reflected echo signal is usually;

[0061] Δt(t): represents the delay time of the echo signal; the delay time of the echo reflects the time from the emission of the ultrasonic signal to its reflection. By calculating the delay time of the signal, the depth of the defect can be inferred.

[0062] The amplitude of the echo signal decays as the propagation distance increases. The attenuation of ultrasonic signals is usually exponential, which can be expressed by the following formula:

[0063] ;

[0064] Among them, A0 is the initial amplitude, α is the attenuation coefficient, and t is time. The size and depth of the defect can be inferred from the amplitude and attenuation characteristics of the echo.

[0065] The propagation speed of ultrasonic signals varies in different materials. Therefore, by calculating the delay time Δt(t) of the echo signal, the depth of the defect can be inferred. The relationship between the delay time of the echo signal and the depth D of the defect can be expressed by the following formula:

[0066] ;

[0067] Where: Δt(t) is the delay time of the echo signal; D is the depth of the defect; v is the propagation speed of the ultrasonic wave in the tire material (usually a constant when the material is known).

[0068] The depth of the defect can be inferred based on the echo delay time Δt(t). Through multiple sampling and analysis, a more accurate defect depth and position can be obtained.

[0069] Furthermore, in order to improve the accuracy of detection, ultrasonic sensors usually perform multiple sampling, especially when detecting large-area tires, to obtain echo signals through different sampling points. These multiple sampling echo signals can help eliminate errors caused by noise or environmental interference. The ultrasonic sensor will output the corresponding ultrasonic echo amplitude and echo delay time data to form a multidimensional data set. This data usually includes the following: A1(t), A2(t),…, A n (t) represents the time at different time points t1, t2, …, t n Echo amplitude data on; Δt1(t), Δt2(t),…, Δt n (t) represents different time points t1, t2, …, t n These data sets will be transmitted to the data acquisition and processing module for subsequent analysis, processing and fusion.

[0070] Step 3: Data Preprocessing

[0071] The tactile data is normalized to normalize the force values ​​in each direction for subsequent analysis. This process ensures the consistency of the tactile data and eliminates the impact of differences in tire surface materials.

[0072] ;

[0073] Among them, μ FX , μ FY , μ FZ are the means of the tactile data in the X, Y, and Z directions, σ FX , σ FY , σ FZ is the corresponding standard deviation.

[0074] The ultrasonic echo signal is denoised to remove environmental noise and equipment interference to improve the validity and signal-to-noise ratio of the data.

[0075] Step 4: Weighted Fusion of Tactile and Ultrasound Data

[0076] The processed tactile data is fused with the ultrasonic data using the weighted average method. The fusion weights of the tactile data and ultrasonic data can be dynamically adjusted through a machine learning algorithm to optimize the weight configuration according to the needs of the actual detection task.

[0077] The weighted average method achieves fusion by assigning different weights to tactile data and ultrasonic data. Tactile data can provide accurate surface information, while ultrasonic data provides defect information deep inside the tire. The configuration of weights will directly affect the performance of the fused data, so according to the actual inspection tasks and changes in the environment, the weights should be dynamically adjusted through appropriate algorithms.

[0078] Assume that the tactile data we have preprocessed is X touch (t), the ultrasound data is X ultrasound (t), the fused data X fused (t) can be fused by the following weighted average method:

[0079]

[0080] Where: X touch (t) represents the tactile data at time t; X ultrasound (t) represents the ultrasonic data at time t; ω1 and ω2 are the weights assigned to the tactile data and ultrasonic data respectively; ω1+ω2=1, that is, the sum of the two weights is 1, ensuring that the fused data can effectively combine the information of the two sensors.

[0081] In practical applications, the quality and importance of tactile data and ultrasonic data may be affected by different conditions. In some cases, surface defects may be more important, and the weight ω1 of tactile data can be higher; while for deep defects, the weight ω2 of ultrasonic data may need to be higher. Therefore, the weights ω1 and ω2 should not be fixed, but can be dynamically adjusted according to the actual task. The method of dynamically adjusting weights can be achieved through machine learning algorithms. Common methods include:

[0082] Rules based on experience: A set of rules is set based on domain knowledge or experimental data to adjust weights according to environmental changes. For example, when surface defects are more obvious, the weight of tactile data is increased first.

[0083] Model-based adjustment: Use machine learning models (such as linear regression, support vector machine (SVM) or neural network) to train data and automatically learn the optimal weights of tactile data and ultrasonic data in different defect situations. Specifically, the following methods can be used:

[0084]

[0085] Among them, f1(.) is a function obtained by the training model, which automatically calculates the optimal weights ω1 and ω2 based on the input tactile data and ultrasonic data.

[0086] The key advantage of weighted fusion is that it can be adaptively adjusted under different detection tasks and environmental conditions to ensure that the data of each sensor can complement each other. When dealing with complex tire surface and internal defects, fused data can provide more accurate and comprehensive information. For example, tactile sensors can detect small cracks and dents on the tire surface, while ultrasonic sensors can deeply detect bubbles and cracks inside the tire. The weighted fusion of the two can effectively improve the overall perception of tire defects.

[0087] Step 5: Data synchronization and spatial alignment

[0088] The fused data is time synchronized and spatially aligned. There may be time lag or spatial displacement when sampling the tactile data and ultrasonic data, so it is necessary to time align the data of the two through an interpolation algorithm, and ensure the correspondence between the data on the tire surface and inside through a spatial mapping algorithm.

[0089] The specific implementation method is as follows:

[0090] 5.1 Time Synchronization

[0091] Time synchronization is a key step to ensure that tactile data and ultrasonic data are compared and analyzed at the same time point. Since the sampling frequencies of tactile sensors and ultrasonic sensors may be different, and there may be delays in the process of tactile sensor data collection, time alignment is required.

[0092] 5.1.1 Basic principles of time synchronization

[0093] Assume that the timestamp of the tactile data is t touch , the timestamp of the ultrasound data is t ultrasound Since the sampling times are inconsistent, the tactile data and ultrasonic data need to be interpolated or resampled to make their timestamps consistent.

[0094] 5.1.2 Linear interpolation

[0095] Linear interpolation is a simple and commonly used time synchronization method. It calculates new data values ​​through the linear relationship between existing data points to ensure that the two data are aligned in time. Assume that the timestamps of tactile data and ultrasound data are inconsistent. At time point t sync The interpolation formula is as follows:

[0096]

[0097] Where: t0 and t1 are time points tsync Two adjacent timestamps; X touch (t sync ) and X ultrasound (t sync ) is at time t sync Corresponding tactile data and ultrasonic data; t sync is the target synchronization time point. Through this interpolation method, at the target time point t sync The synchronized tactile and ultrasonic data are obtained.

[0098] 5.2 Spatial Alignment

[0099] Spatial alignment is an important step to ensure that the tactile data and the ultrasonic data reflect the same defect location. Since the tactile sensor and the ultrasonic sensor are located at different locations, the data needs to be aligned through a spatial mapping algorithm to ensure that the defect areas in the two data sets correspond.

[0100] 5.2.1 Basic principles of spatial alignment

[0101] Assuming that the relative positions of the tactile and ultrasonic sensors on and inside the tire are known, the spatial alignment process requires mapping the defect locations in the ultrasonic data into the surface coordinate system of the tactile data.

[0102] 5.2.2 Spatial Mapping Method

[0103] In order to perform spatial alignment, geometric transformation or interpolation methods are usually used. The following is one of the commonly used spatial mapping methods: Affine Transformation.

[0104] Assume that the spatial coordinates of the tactile data are (x touch ,y touch ), the spatial coordinates of the ultrasound data are (x ultrasound ,y ultrasound ), the affine transformation model can be expressed by the following formula:

[0105] ;

[0106] Among them: a, b, c, d are the parameters of the affine transformation, which determine the rotation, scaling and translation of the spatial coordinates; e, f are the translation parameters used to translate the coordinate system of the ultrasound data.

[0107] By optimizing these parameters, the defect location of the ultrasound data can be made consistent with the defect location of the tactile data, thus achieving spatial alignment.

[0108] 5.2.3 Optimizing spatial mapping parameters

[0109] The process of optimizing spatial mapping parameters is usually performed by minimizing the error, and the error metric can be based on the mean squared error (MSE):

[0110] ;

[0111] This formula can be used to adjust the parameters a, b, c, d, e, and f of the affine transformation to minimize the spatial error between the tactile data and the ultrasonic data.

[0112] 5.3 Comprehensive Implementation of Time Synchronization and Spatial Alignment

[0113] Combining the above time synchronization and space alignment methods, the specific implementation steps are as follows:

[0114] 1) Time synchronization of tactile data and ultrasonic data to ensure that the two data are consistent at the same time point;

[0115] 2) Use a spatial mapping algorithm to map the defect location in the ultrasound data to the coordinate system of the tactile data to ensure that the defect area in the two data is consistent.

[0116] Ultimately, the synchronized and aligned data can be fused and further analyzed to provide accurate tire defect detection results.

[0117] Step 6: Outlier Detection

[0118] Outlier detection is used to identify potential defect areas in the fused tactile and ultrasonic data. Outliers usually indicate abnormal conditions on the surface or inside of the tire, such as cracks, bubbles, holes, etc., which may be manifestations of defects. Outlier detection uses a preset threshold to identify abnormal points that are beyond the normal range and mark them as defect candidate areas.

[0119] 6.1 Basic Principles of Outlier Detection

[0120] Outlier detection is based on the following principles:

[0121] 1) Abnormal force in tactile data: Abnormal force changes may occur on the tire surface due to certain defects (such as cracks, dents, etc.). Therefore, abnormal points in tactile data usually appear as rapid changes or obvious fluctuations in force at a certain position.

[0122] 2) Abnormal echo in ultrasonic data: When the ultrasonic signal passes through the tire, if it encounters a defect, the amplitude and delay time of the echo signal will be abnormal. For example, if the amplitude of the echo changes too much or the echo delay time is abnormal, it may indicate that there is a defect inside the tire.

[0123] Abnormal changes in tactile and ultrasonic data are detected by setting thresholds, and points exceeding these thresholds are considered as candidate areas for potential defects.

[0124] 6.2 Tactile Data Outlier Detection

[0125] Abnormal values ​​of tactile data are usually manifested as a sudden increase or decrease in force changes. In order to detect abnormal changes in tactile data, a threshold-based detection method can be used. For example, assume that the normal range of force changes in each direction (X, Y, Z direction) of tactile data is [L, U], where L is the lower limit and U is the upper limit. If the force value at a certain time t exceeds this range, the data point is considered to be an abnormal value.

[0126] The formula means:

[0127]

[0128] in:

[0129] F X ′(t), F Y ′(t), F Z ′(t) represents the force value of tactile data in the X, Y, and Z directions respectively; θ F is the threshold for detecting tactile data anomalies. Force data exceeding this threshold is considered anomaly. F When the tire surface is marked as abnormal, it indicates that there is a potential defect on the tire surface and the area is marked as abnormal.

[0130] 6.3 Ultrasound Data Outlier Detection

[0131] Abnormal values ​​of ultrasound data are usually manifested as abnormal changes in echo signals. Common anomalies include a sudden increase or decrease in the amplitude of the echo signal, or abnormal fluctuations in the echo delay time. In order to detect these anomalies, a threshold for the echo signal can be set, and signals outside this range are considered abnormal.

[0132] The outlier detection formula of the echo signal is as follows:

[0133]

[0134] in:

[0135] A′(t) represents the amplitude of the echo signal; Δt′(t) represents the delay time of the echo signal; θ A and θ Δt are the thresholds for echo amplitude and echo delay time, respectively, and echo signals exceeding these thresholds are considered abnormal.

[0136] When the amplitude or delay time of the echo signal exceeds the set threshold, it indicates that there may be defects inside the tire (such as bubbles, cracks, etc.), and the area is marked as abnormal.

[0137] 6.4 Joint Outlier Detection of Tactile and Ultrasonic Data

[0138] In practical applications, tactile data and ultrasonic data often need to be considered jointly because of their different data types. Tactile data may show anomalies in the surface area, while ultrasonic data may show internal defects. In order to improve the accuracy of detection, abnormal values ​​in tactile data and ultrasonic data can be detected jointly.

[0139] The basic idea of ​​joint outlier detection is that when outliers in tactile data and ultrasonic data appear at the same time, it is more likely to indicate a defect. Therefore, joint detection can be performed using the following formula:

[0140]

[0141] Among them: Anomaly F Anomaly is an outlier in the tactile data; A Anomaly is an outlier in ultrasound data; joint is a joint outlier, indicating that when both tactile and ultrasound data are abnormal, the area is considered to be defective.

[0142] 6.5 Further processing of outliers

[0143] When an outlier is detected, it can be further processed based on the type and location of the defect, such as:

[0144] Mark candidate areas: Mark the abnormal points detected in the tactile and ultrasonic data as defect candidate areas;

[0145] Subsequent analysis: Further analysis is performed on the marked defect candidate areas, such as calculating the depth, width, length and other parameters of the defect to provide detailed information of the defect.

[0146] 6.6 Optimizing Outlier Detection

[0147] Furthermore, in order to improve the accuracy and robustness of detection, the following optimization methods can be used:

[0148] Adaptive threshold: Dynamically adjust the threshold θ through statistical analysis of the data (such as mean, variance, etc.) F ,θ A and θ Δt .

[0149] Machine learning algorithms: Use model-based outlier detection methods, such as Isolation Forest and Local Outlier Factor (LOF), to automatically identify potential abnormal areas.

[0150] Step 7: Defect Classification and Identification

[0151] The candidate areas are classified using a machine learning algorithm to determine the defect type. This step can classify and identify the specific defect location inside the tire according to different defect types (such as cracks, bubbles, cavities, etc.). In the present invention, a fusion adaptive weighted support vector machine (AWSVM) is used, which combines the traditional support vector machine (SVM) and the adaptive weighting mechanism, and is particularly suitable for the task of tire defect detection. This algorithm can dynamically adjust the weights to optimize the detection accuracy when dealing with different types of defects, and can efficiently handle the fusion of heterogeneous data.

[0152] 7.1 Algorithm Overview

[0153] The core idea of ​​the algorithm is to dynamically adjust the importance of input features in the support vector machine through an adaptive weighting mechanism to cope with the different characteristic manifestations of different defect types (such as cracks, bubbles, voids, etc.). Adaptive weighted support vector machine (AWSVM) adds adaptive weights to features on the basis of traditional SVM, and continuously adjusts the weights through an optimization algorithm, so that the model can efficiently classify different input data (such as tactile data and ultrasonic data).

[0154] 7.2 Algorithm Principle

[0155] 7.2.1 Adaptive Weighting Mechanism

[0156] In traditional SVM, all input features usually have equal weights, but in practical applications, different defect types have different sensitivities to different features. To adapt to this difference, this algorithm introduces an adaptive weighting mechanism to adjust the weight of each feature according to the defect type and the contribution of the input feature. Let the input feature vector be x=[f1,f2,…,f n ], where f i is extracted from the tactile data and ultrasound data. i Through the adaptive weighting mechanism, each feature is assigned a weight w i The weights are adjusted dynamically during the training process. The weighted input feature vector is: weighted =[w1f1,w2f2,…,w n f n ]; these weights w i Adjusting the feature based on its contribution to classification accuracy can effectively reflect the differences in feature performance of different defect types.

[0157] 7.2.2 SVM model training

[0158] The adaptively weighted feature vector x weighted , we use support vector machines for classification. The goal of SVM is to find an optimal hyperplane to separate data points of different categories. Specifically, SVM maximizes the classification boundary (interval) by solving the following optimization problem:

[0159]

[0160] Among them: w is the normal vector of the hyperplane, which determines the boundary of the classification; ξ i is a slack variable used to deal with noise and misclassification in the data; C is a regularization parameter used to control the trade-off between error and model complexity.

[0161] The goal is to find the optimal w and b, that is, the position and direction of the hyperplane, through optimization problems, so that the defect data of different categories are separated as much as possible.

[0162] 7.2.3 Adaptive Weight Adjustment

[0163] The adjustment of the adaptive weighting mechanism is based on the impact of each feature on the classification effect. In order to make the weight of each feature match the characteristics of the defect type, the weight can be dynamically adjusted through the error feedback mechanism during the training process. The specific steps are as follows:

[0164] 1) Initial weight setting: The initial weights of all features are set equal, that is, w i =1.

[0165] 2) Weight adjustment during training: After each classification, adjust the weight w of the misclassified features based on the misclassified samples i ; For the misclassified samples, the corresponding feature weight w i Increase, indicating that the importance of this feature for classification increases; for correctly classified samples, its weight w i reduce.

[0166] The formula means:

[0167]

[0168] Where: η is the learning rate, which controls the speed of weight update; error(f i ) is the feature f i The error representation in classification is usually calculated based on the classification results.

[0169] In this way, the algorithm can dynamically adjust the weights of each feature to improve classification performance.

[0170] 7.2.4 Classification Decision

[0171] The trained SVM model can be used to new Classify the defect type. The classification decision is given by the following formula:

[0172]

[0173] Where: x new is the new input feature vector; w is the optimized normal vector; b is the bias term.

[0174] Based on the output of the decision function, the sign function returns the corresponding defect category (such as cracks, bubbles, voids, etc.).

[0175] 7.3 Training and Optimization

[0176] Furthermore, during the training process, the present invention can use cross-validation to evaluate the performance of the model, and use methods such as grid search or random search to optimize the hyperparameters of the model.

[0177] 7.3.1 Cross-Validation

[0178] Cross-validation verifies the generalization ability of the model by dividing the dataset into multiple subsets. A common cross-validation method is K-fold cross-validation, in which the model is trained and tested on different training and validation sets in each fold.

[0179] 7.3.2 Grid Search

[0180] Grid search is a hyperparameter optimization method that systematically searches for the best parameter combination in a specified parameter space. For example, when training AWSVM, you can perform a grid search on the regularization parameter C, the learning rate η, and other hyperparameters to select the best parameters.

[0181] 7.4 Defect type identification

[0182] With the trained model, we can classify each candidate defect area. AWSVM will output the probability distribution of different defect types. Finally, the system will determine the defect type based on these classification results and provide a spatial distribution map of the defect and a severity assessment.

[0183] Step 8: Defect Severity Assessment

[0184] The severity of defects is evaluated based on the type, location and size of the defects. The evaluation results are displayed in real time through the display module, providing a spatial distribution map of the defects and outputting a detailed report of the defects to help operators take timely repair or replacement measures based on the test results.

[0185] 8.1 Basic Principles of Defect Severity Assessment

[0186] The severity of a defect is usually determined by three factors:

[0187] 1) Defect type: Different types of defects (such as cracks, bubbles, holes, etc.) have different degrees of impact on the tire. For example, cracks are usually more serious than bubbles or holes.

[0188] 2) Defect location: The location of the defect in the tire affects its severity. Surface defects and internal defects have different hazards. Surface cracks may more easily cause tire rupture, while internal bubbles or cavities may cause long-term structural damage.

[0189] 3) Defect size: The size of the defect is an important factor in assessing its severity. Larger defects usually have a greater impact on the safety and service life of the tire.

[0190] 8.2 Mathematical Model for Defect Severity Assessment

[0191] To accurately assess the severity of defects, a weighted model based on defect type, location, and size can be established. Specifically, the severity of a defect can be expressed as the following mathematical formula:

[0192]

[0193] Among them: S is the severity score of the defect; T is the weight value of the defect type, which is set based on the degree of harm of the defect type; P is the weight value of the defect position, considering the impact of the defect position in the tire on the severity; A is the size of the defect, usually expressed by the area or volume of the defect area; α, β, γ are the weight coefficients of each factor, which are set according to actual conditions.

[0194] 8.3 Weight assessment of defect types

[0195] Different types of defects have different impacts on the tire. For example, a crack may be more serious than a bubble, so we can assign different weights to each defect type. Suppose the score of defect type T is:

[0196] Crack: T crack =1.0

[0197] Bubble: T bubble= 0.6T

[0198] Cavity: T cavity =0.8

[0199] Depending on the defect type, the resulting T value can be used in the severity scoring formula.

[0200] 8.4 Weight assessment of defect locations

[0201] The weight P of the defect location is mainly scored based on which part of the tire the defect is located. For example, the sidewall area of ​​the tire is more susceptible to external pressure than the tread area, so a defect in the sidewall may be more serious than a defect in the tread. Assume that the score of location P is as follows:

[0202] Tread defect: P tread =0.5;

[0203] Tire sidewall defect (Sidewall): P sidewall =1.0;

[0204] Internal defect (Inner): P inner =0.8.

[0205] 8.5 Evaluation of defect size

[0206] The size of the defect A is usually expressed as the area or volume of the defect. The larger the defect, the greater the impact on the performance of the tire. The area or volume of the defect can be obtained through image processing or sensor measurement. Assuming A is expressed in area (unit: square millimeter), the evaluation formula is as follows:

[0207] A=Area of ​​Defect

[0208] In order to normalize the defect size, the defect area A can be compared with the maximum allowable defect area A of the tire. max Compare and get the standardized score A of defect size norm :

[0209] .

[0210] 8.6 Comprehensive severity assessment

[0211] By combining the weights of the above factors, we can get the final severity score of each defect. According to the severity score S, defects can be divided into different levels, for example:

[0212] Minor defects (minor cracks or bubbles): S≤0.5;

[0213] Moderate defects (large cracks or voids): 0.5 <S≤0.8;

[0214] Severe defects (crack extension or large voids): S>0.8.

[0215] 8.7 Spatial distribution map and report output

[0216] Furthermore, the defect severity score can also generate a spatial distribution map based on the location of the defect. By associating the severity score of each defect with the spatial location of the tire, an intuitive defect spatial distribution map can be generated to help operators quickly identify the weak areas of the tire. The map can show the severity distribution of each defect area, using different colors or sizes to indicate the severity of the defect. Finally, the system will output a detailed defect report, which includes: the type, location, size and severity score of each defect; the spatial distribution map of the defect; and repair or replacement recommendations.

[0217] Example 2

[0218] like Figure 2 As shown, the tire defect detection system based on tactile perception and ultrasonic fusion of the present invention, its tactile perception hardware adopts the hardware part involved in the flexible interface detection method of the Chinese invention patent application (publication number: CN116818172A, publication date: 2023-09-29); the system includes the following components:

[0219] 1. Tactile Sensor

[0220] Used to collect three-dimensional force data on the tire surface. The tactile sensor can be a three-axis force sensor that can collect surface force data in the X, Y, and Z directions respectively and convert it into electrical signal output.

[0221] 2. Ultrasonic sensor

[0222] Used to collect ultrasonic echo data inside the tire. The ultrasonic sensor can emit ultrasonic waves and receive echo signals to obtain echo amplitude and echo delay time data.

[0223] 3. Data fusion module

[0224] Used to perform weighted fusion of tactile data and ultrasonic data. The data fusion module uses weighted average method to dynamically adjust the weights of tactile data and ultrasonic data. The weighted fused data can fully combine the advantages of the two sensors and enhance detection accuracy.

[0225] 4. Defect Detection Module

[0226] It is used to detect outliers, classify defects and evaluate severity of fused data. The defect detection module determines the defect candidate area based on the outliers, identifies the defect type through the classification algorithm, and outputs the defect type, location, size and severity.

[0227] 5. Display module

[0228] It is used to display the test results and provide the spatial distribution map of defects and severity assessment reports. The display module can present the entire defect detection process in real time, helping operators to understand the tire quality status in a timely manner.

[0229] Experimental example

[0230] In order to verify the effect of the tire defect detection method based on tactile perception and ultrasonic fusion of the present invention, the following experiment was conducted and actual detection data was collected. This experiment aims to test the accuracy, robustness and detection efficiency of the system under different defect types, locations and sizes.

[0231] 1. Experimental Environment Setup

[0232] Experimental equipment: as shown in Example 2.

[0233] Tire samples: 10 tire samples with artificial defects were selected. The defect types included cracks, bubbles, cavities and dents, which were arranged in different tire areas (tread, sidewall and interior).

[0234] 2. Defect type and location

[0235] Crack: Located on the sidewall of the tire, length 30 mm.

[0236] Air bubble: Located inside the tire, with a maximum diameter of 10 mm.

[0237] Cavity: Located inside the tire, with a diameter of 15 mm.

[0238] Depression: Located on the tire tread, with a depth of 5 mm and an area of ​​50 square millimeters.

[0239] 3. Experimental Data and Analysis

[0240] 3.1 Experimental steps: as shown in Example 1.

[0241] 3.2 Defect Detection Results

[0242] Detection accuracy: For 10 test samples, the system was able to correctly detect all defects with a detection accuracy of 100%.

[0243] Cracks: The system successfully detected all sidewall cracks with 100% classification accuracy.

[0244] Bubbles: All internal bubbles were successfully detected with a classification accuracy of 98%.

[0245] Holes: All holes are detected and the classification accuracy is 100%.

[0246] Depression: All tread depressions were detected and the classification accuracy was 95% (slightly lower, but still able to be accurately located, due to shallow depressions and sometimes less obvious force changes).

[0247] Time synchronization and spatial alignment: The time synchronization error of all tactile data and ultrasonic data is less than 1 millisecond, and the spatial alignment error is less than 0.5 mm, ensuring accurate mapping of defect locations.

[0248] Defect classification:

[0249] Cracks: The size and location of the detected cracks are consistent with those marked manually.

[0250] Bubbles: The size and position of the detected bubbles are consistent with the artificial marks, and the amplitude and delay time changes of the echo signal match the reflection characteristics of the bubbles.

[0251] Void: The position and size of the detected void are consistent with the artificial mark, the amplitude of the echo signal is greatly reduced, and the delay time is increased.

[0252] Depression: Successfully identified and classified through force changes in tactile data and attenuation characteristics of ultrasonic echoes.

[0253] 3.3 Defect severity assessment results

[0254] The detected defects were evaluated using the defect severity evaluation method of the present invention, and the scoring results are as follows:

[0255] Crack (side wall, length 30mm):

[0256] Type weight T crack =1.0

[0257] Position weight P sidewall =1.0

[0258] Area weight A norm =0.15

[0259] Severity score: Scrack = 0.95 (a serious defect);

[0260] Bubble (inside, 10mm diameter):

[0261] Type weight T bubble =0.6

[0262] Position weight P inne r=0.8

[0263] Area weight A norm =0.05

[0264] Severity Rating: Sbubble =0.65 (moderate defect);

[0265] Cavity (inside, 15mm diameter):

[0266] Type weight T cavity =0.8

[0267] Position weight P inner =0.8

[0268] Area weight A norm =0.0755

[0269] Severity Rating: S cavity =0.78 (moderate defect);

[0270] Depression (tread, area 50mm²):

[0271] Type weight T dent =0.4

[0272] Position weight P tread =0.5

[0273] Area weight A norm =0.25

[0274] Severity Rating: S dent =0.65 (moderate defect).

[0275] 4. Results display

[0276] Defect Report: The type, location, size, severity score and other information of each defect are generated in a detailed report and displayed on the monitoring screen in real time to help operators make decisions quickly.

[0277] 5. The experimental data are shown in Table 1.

[0278] Table 1

[0279]

[0280] Classification accuracy: The system achieved 100% detection accuracy in all defect categories, with an overall detection accuracy of 98%.

[0281] Severity Assessment: Severity scores for all defects are accurately assessed, providing operators with clear repair and replacement recommendations.

[0282] 6. Conclusion

[0283] Through the above experiments, the tire defect detection system based on tactile perception and ultrasonic fusion of the present invention has shown excellent performance in accuracy, robustness and detection efficiency. The system can effectively detect various defects of tires and provide detailed reports through the defect severity assessment module to help operators take appropriate repair or replacement measures in time to ensure the safety and performance of tires.

[0284] Furthermore, the present invention also provides a computer-readable storage medium, comprising one or more programs for execution by one or more processors of an electronic device, wherein the one or more programs include instructions for executing the method described in this embodiment 1.

[0285] It should be noted that computer-readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include temporary computer-readable media such as modulated data signals and carrier waves.

[0286] The present invention may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present invention may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0287] The above is a description of the embodiments of the present invention. Through the above description of the disclosed embodiments, professionals and technicians in the field can implement or use the present invention. Various modifications to these embodiments will be apparent to professionals and technicians in the field. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown in this article, but will conform to the widest range consistent with the principles and novelties disclosed herein.

Claims

1. A tire internal defect detection method based on tactile perception and ultrasonic fusion, characterized in that: The following steps are involved: 1) Collecting three-dimensional force data of the tire surface through a tactile sensor, wherein the tactile sensor can detect force changes on the tire surface in the X, Y, and Z directions respectively; 2) Collecting the echo signal inside the tire through an ultrasonic sensor, wherein the ultrasonic sensor can detect the echo amplitude and echo delay time of the ultrasonic wave to reflect the defect information inside the tire; 3) preprocessing the tactile data and the ultrasonic data, including standardizing the tactile data, converting the tactile data into standardized three-dimensional force data, and denoising and standardizing the ultrasonic data, so as to facilitate subsequent fusion and analysis; 4) performing weighted fusion of the tactile data processed in step 3 and the ultrasonic data, wherein the weighted fusion adopts a weighted average method, wherein the weights of the tactile data and the ultrasonic data are dynamically determined by a preset rule or a machine learning algorithm; 5) Perform time synchronization and spatial alignment on the fused data to ensure the consistency of the tactile data and the ultrasonic data in terms of sampling time and spatial position; 6) Perform outlier detection based on the fused data to identify candidate areas that may have defects and compare them with normal areas; 7) Classify the detected candidate areas through machine learning algorithms to identify the types of defects, including but not limited to cracks, voids, and bubbles; 8) Calculate the severity of defects based on the type, location and size of the defects, and output the spatial distribution map of the defects and the severity assessment report.

2. The method according to claim 1, characterized in that The tactile sensor is a three-axis force sensor that can collect force changes on the tire surface in the X, Y, and Z directions respectively; and / or, the ultrasonic sensor is used to collect echo signals inside the tire, and the echo signals include echo amplitude and echo delay time. The echo amplitude is used to identify the reflection characteristics of the defect, and the echo delay time is used to infer the depth and position of the defect.

3. The method according to claim 1, characterized in that: The preprocessing in step 3) includes: standardizing the tactile data, normalizing the force value in each direction, and removing background noise from the ultrasonic data through a denoising algorithm to improve the validity of the data.

4. The method according to claim 1, characterized in that: The weighted fusion in step 4) adopts a weighted average method, and the weight value is dynamically adjusted according to the importance of the tactile data and the ultrasonic data and their contribution to defect identification. The weight is obtained by training the machine learning algorithm; And / or, the time synchronization and space alignment in step 5) are achieved through an interpolation algorithm, the timestamps of the tactile data and the ultrasonic data are aligned, and the correspondence between the ultrasonic data and the tactile data on the tire surface and inside is ensured through a spatial mapping algorithm.

5. The method according to claim 1, characterized in that The outlier detection in step 6) is based on threshold judgment. When the changes in the force data and ultrasonic echo signal exceed the preset threshold, it is marked as a defect candidate area.

6. The method according to claim 1, characterized in that The machine learning algorithm in step 7) is a support vector machine (SVM), a random forest or a convolutional neural network (CNN). The defect candidate area is classified through the trained model to determine the type of defect.

7. The method according to claim 1, characterized in that The defect severity assessment described in step 8) is performed based on the type, location and size of the defect, and the assessment result is displayed in real time through the display module, providing a detailed report of the defect.

8. A tire internal defect detection system based on tactile perception and ultrasonic fusion, characterized in that: The system implements the method described in any one of claims 1 to 7, including: Tactile sensor: Tactile sensor is used to collect three-dimensional force data on the tire surface. Ultrasonic sensor, which is used to collect ultrasonic echo data inside the tire. The data fusion module is used to perform weighted fusion of tactile data and ultrasonic data. Defect detection module, which is used to perform outlier detection, defect classification and severity assessment based on the fused data. Display module,The display module is used to display the inspection results and provide a report on the spatial distribution and severity assessment of defects.

9. A computer-readable storage medium having a computer program or instruction stored thereon, characterized in that: When the computer program or instruction is executed by a processor, step 3) to step 8) of the method according to any one of claims 1 to 7 are implemented.

10. A computer program product comprising a computer program or instructions, characterized in that When the computer program or instruction is executed by a processor, step 3) to step 8) of the method according to any one of claims 1 to 7 are implemented.

Images