Nondestructive testing method and system for internal defects of precision metal parts

Abstract

The invention provides a nondestructive testing method and system for internal defects of precision metal parts, and relates to the technical field of nondestructive testing. The method comprises the following steps: firstly, collecting metal type information of a to-be-detected precise metal part so as to accurately screen out a defective target part, and preliminarily locking a detection key point; then, a multi-angle ultrasonic probe is used for scanning the target part in all directions, internal reflection data of the target part are collected, and original materials are provided for deep analysis; then combining the acquired metal type information and internal reflection data, inputting a defect detection model which is constructed by deep learning and has an accurate defect identification capability, and accurately judging defect data; constructing three-dimensional model information capable of intuitively presenting the internal condition of the part by using a three-dimensional reconstruction technology according to the defect data; and finally, pushing the three-dimensional model information to a visual end for display, thereby realizing visualization of a detection result. By implementing the method, the internal defect data of the precision metal part can be accurately judged, and the detection precision and efficiency are improved.

Description

Technical Field

[0001] This application relates to the field of non-destructive testing technology, and particularly to a non-destructive testing method and system for internal defects of precision metal parts. Background Art

[0002] In modern manufacturing, precision metal parts are widely used in many fields such as aerospace, automotive manufacturing, and electronic equipment. The quality of these parts is directly related to the performance and safety of the entire product. Therefore, it is crucial to perform efficient and accurate inspections on precision metal parts. Accurately obtaining the metal type information of the parts and identifying the defective parts can help enterprises promptly discover production problems, improve product quality, and reduce production costs.

[0003] Currently, in the field of precision metal part inspection, the existing technology usually uses a single-angle ultrasonic probe to inspect the parts. By emitting ultrasonic waves to the parts and then receiving the reflected signals, the presence of defects in the parts is preliminarily judged based on the characteristics of the reflected signals. For the judgment of the metal type, it generally relies on manual experience and is speculated based on the appearance, color, and previous production records of the parts. When determining whether a part has defects, it mainly depends on the simple analysis of the reflected signals by the inspectors, setting some fixed thresholds. If the reflected signals exceed or are lower than these thresholds, it is considered that the part may have defects.

[0004] However, this inspection method has many deficiencies. The single-angle ultrasonic probe scanning cannot comprehensively obtain the internal information of the parts and is very likely to miss some defects located in the non-scanning angle direction. Moreover, the manual judgment of the metal type based on experience has low accuracy and is greatly affected by subjective factors. Merely relying on simply setting thresholds to judge defects cannot adapt to precision metal parts with different complexities and materials. Therefore, it is difficult for the current traditional methods to accurately identify the target precision metal parts with defects. Summary of the Invention

[0005] This application provides a non-destructive testing method and system for internal defects of precision metal parts, which are used to efficiently and accurately identify the subtle and complex defects inside precision metal parts and effectively ensure the quality of the parts.

[0006] In a first aspect, the present application provides a non-destructive testing method for internal defects of precision metal parts, which is applied to a non-destructive testing system. The method includes: obtaining the metal type information of the precision metal part to be tested and determining the target precision metal part with defects; using a multi-angle ultrasonic probe to scan the target precision metal part to obtain internal reflection data; combining the metal type information and the internal reflection data, and determining the defect data through a defect detection model, which is constructed in advance through deep learning based on multiple metal type information and internal reflection data sets with defect data annotations; combining the defect data, and using three-dimensional reconstruction technology to determine the three-dimensional model information of the target precision metal part; and sending the three-dimensional model information to a visual terminal for display.

[0007] By adopting the above technical solution, obtaining the metal type information can preliminarily estimate the possible defect types according to the inherent characteristics of different metals. Then, determining the target precision metal part with defects can accurately focus on the detection object. Scanning with a multi-angle ultrasonic probe can capture internal reflection data from multiple directions and comprehensively reflect the internal condition of the part. Combining the metal type information and the reflection data, relying on the defect detection model constructed through deep learning, the model can accurately analyze the defect data based on the rules learned from a large number of annotated data, efficiently and accurately judge the internal defects of the part, and improve the detection accuracy and efficiency.

[0008] Combined with some embodiments of the first aspect, in some embodiments, the step of determining the target precision metal part with defects specifically includes: obtaining the usage times information of the precision metal part to be tested; combining the metal type information and the usage times information, and determining the target precision metal part with defects through a defect prediction model, which is constructed in advance through machine learning based on multiple metal type information and metal usage times information sets with defect status annotations.

[0009] By adopting the above technical solution, the metal type information of the part to be tested reflects its basic material characteristics, and different metals have different durability and vulnerable parts. The usage times information intuitively shows the wear accumulation of the part. Combining the two and inputting them into the defect prediction model constructed through machine learning, the model can quickly screen out the possible defective targets from many parts based on the experience summarized from a large number of past annotated data, comprehensively considering the material and loss factors, narrow the detection range in advance, avoid over-detecting a large number of defect-free parts, and save time, manpower and equipment loss.

[0010] Combined with some embodiments of the first aspect, in some embodiments, before the step of using a multi-angle ultrasonic probe to scan the target precision metal part to obtain internal reflection data, it further includes controlling a cleaning device to perform a preprocessing operation on the target precision metal part, and the preprocessing operation at least includes surface cleaning.

[0011] By adopting the above technical solution, if there are impurities such as oil stains and dust adhering to the surface of the part, chaotic reflections and scatterings will occur when the ultrasonic wave contacts, interfering with signal transmission and resulting in large deviations in the collected internal reflection data. After cleaning, the ultrasonic wave emitted by the probe can smoothly penetrate the surface, and the received reflection signal is pure and real, accurately reflecting the internal structure of the part, laying a solid foundation for accurately judging defects based on the reflection data subsequently and ensuring the reliability of non-destructive testing.

[0012] Combined with some embodiments of the first aspect, in some embodiments, before the step of using a multi-angle ultrasonic probe to scan the target precision metal part to obtain internal reflection data, it further includes: obtaining the current ambient temperature data; combining the ambient temperature data, determining the optimal parameter data of the ultrasonic probe according to a preset ambient temperature probe parameter optimization strategy library; adjusting the parameters of the ultrasonic probe according to the optimal parameter data.

[0013] By adopting the above technical solution, obtaining the current ambient temperature data is the key starting step because temperature has a great influence on the propagation characteristics of ultrasonic waves, and temperature fluctuations will change the sound speed, wavelength, etc. According to the preset optimization strategy library, the optimal parameters of the ultrasonic probe are accurately matched in combination with the real-time temperature, enabling the probe to adapt to the environment and work efficiently. After adjusting the parameters, the transmitted and received ultrasonic signals are stable and accurate, and the collected internal reflection data is real and reliable. Subsequent defect analysis and judgment based on this data are more accurate, effectively reducing the interference of environmental factors.

[0014] Combined with some embodiments of the first aspect, in some embodiments, in the step of using three-dimensional reconstruction technology to determine the three-dimensional model information of the target precision metal part in combination with the defect data, it specifically includes: extracting the features of the defect data to determine the feature data of the defect, and the feature data at least includes the boundary, center point, and shape information of the defect; determining the acoustic characteristic parameters of the target precision metal part according to the metal type information, and the acoustic characteristic parameters at least include the sound speed and the sound attenuation coefficient; adopting a voxel-based three-dimensional reconstruction algorithm to map the internal reflection data and the feature data into three-dimensional space to construct an initial three-dimensional model; performing a defect annotation operation on the initial three-dimensional model to determine the final three-dimensional model information, and the defect annotation operation includes annotating the type, position, and size information of the defect on the initial three-dimensional model.

[0015] By adopting the above technical solutions, feature data such as boundaries, center points, and shapes are extracted from defect data. These are the "key coordinates" for accurate quantification of defects, making defect positioning clear. Acoustic characteristic parameters are determined according to the type of metal. Different metals have different acoustic performances, so it is more scientific to interpret reflection data based on this. Based on the voxel 3D reconstruction algorithm, the reflection and feature data are mapped into an initial 3D model to intuitively display the internal structure, and then the defect type, location, and size are annotated to fully present the overall picture of the defect, assisting inspection personnel to quickly understand the details of the defect and improve the accuracy and convenience of inspection.

[0016] In combination with some embodiments of the first aspect, in some embodiments, after the step of obtaining the metal type information of the precision metal parts to be inspected and determining the defective target precision metal parts, it also includes: using an electronic nose device to obtain the odor information of the target precision metal parts; combining the metal type information and the odor information, determining the defect probability data through an odor defect prediction model, and the odor defect prediction model is constructed in advance through deep learning using multiple sets of metal type information and odor information labeled with defect states.

[0017] By adopting the above technical solution, the electronic nose device is used to obtain the odor information of the target precision metal parts. When there are defects inside the metal, the odor components change due to stress and chemical reactions. Combined with the metal type information, the odor change patterns of different metals are different. The two are input into the odor defect prediction model after deep learning. The model mines the correlation between the two, predicts the defect probability data in advance, cross-validates multi-dimensional information, assists in precise detection, reduces the risk of misjudgment and missed judgment, and enhances detection accuracy.

[0018] In combination with some embodiments of the first aspect, in some embodiments, after the step of combining the metal type information and the odor information and determining the defect probability data through an odor defect prediction model, it also includes: if the defect probability data is greater than a set defect probability threshold, it is determined that the target precision metal part has a defect.

[0019] By adopting the above technical solution, after calculating the defect probability data by combining the metal type and odor information, it is of great significance to set a reasonable threshold. When the defect probability data is higher than the threshold, it is decisively determined that the target precision metal parts have defects, avoiding wasting too many inspection resources on high-risk parts, simplifying the process and improving efficiency.

[0020] In a second aspect, the present application provides a nondestructive testing system, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the nondestructive testing system to perform the method described in the first aspect and any possible implementation of the first aspect.

[0021] In a third aspect, the present application provides a computer-readable storage medium comprising instructions, which, when executed on a nondestructive testing system, causes the nondestructive testing system to perform the method described in the first aspect and any possible implementation of the first aspect.

[0022] In a fourth aspect, the present application provides a computer program product. When the computer program product is run on a nondestructive testing system, the nondestructive testing system executes the method described in the first aspect and any possible implementation manner of the first aspect.

[0023] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. The technology of obtaining metal type information to estimate defect types, using multi-angle ultrasonic probes to collect internal reflection data, and combining deep learning to build a defect detection model for accurate judgment has been adopted. Therefore, the problem of low accuracy and poor efficiency in the existing technology for detecting internal defects of precision metal parts has been effectively solved, thereby achieving the technical effect of efficiently and accurately judging the internal defects of parts and improving detection accuracy and efficiency.

[0024] 2. By combining the metal type information and usage frequency information of the parts to be inspected, and using the defect prediction model constructed by machine learning to screen the targets in advance, the problem of wide detection range and serious waste of resources in the existing technology is effectively solved, thereby achieving the technical effect of narrowing the detection range in advance and saving time, manpower and equipment loss.

[0025] 3. The technology of extracting features from defect data, determining acoustic characteristic parameters according to metal types, and constructing and annotating defect information using a voxel-based 3D reconstruction algorithm has been adopted. Therefore, the problem that the existing technology cannot intuitively and accurately present the overall picture of internal defects of precision metal parts has been effectively solved, thereby achieving the technical effect of assisting inspection personnel to quickly understand defect details and improve inspection accuracy and convenience. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a flow chart of a nondestructive detection method for internal defects of precision metal parts in an embodiment of the present application; Figure 2 It is another flow chart of the nondestructive detection method of internal defects of precision metal parts in the embodiment of the present application; Figure 3 It is a schematic diagram of the structure of a physical device of the nondestructive testing system in an embodiment of the present application. DETAILED DESCRIPTION

[0027] The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in the specification and appended claims of the present application, the singular forms "a", "an", "the", "above-mentioned", "said", "this" are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used in the present application refers to and includes any or all possible combinations of one or more of the listed items.

[0028] Hereinafter, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as implying or suggesting relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the embodiments of the present application, unless otherwise specified, the meaning of "a plurality" is two or more.

[0029] For ease of understanding, the method provided in this embodiment is described in terms of a process below. Please refer to Figure 1 , which is a schematic flowchart of a process for the non-destructive detection method of internal defects of precision metal parts in the embodiments of the present application.

[0030] S101. Obtain the metal type information of the precision metal part to be detected and determine the target precision metal part with defects; The non-destructive detection system first uses spectral analysis technology to obtain the metal type information of the precision metal part to be detected. The system is equipped with a high-precision spectral analyzer. By emitting light of a specific wavelength to irradiate the surface of the part, after the light interacts with the substances on the surface of the part, reflected light is generated. The spectral analyzer collects the reflected light and performs spectral analysis. Different metal elements have different absorption and reflection characteristics for light, thus presenting unique characteristic spectral lines on the spectrum. The system compares the obtained spectral data with the built-in metal spectral database, which contains a large amount of standard spectral information of known metal types. Using an advanced pattern matching algorithm, it can accurately identify the metal type of the part.

[0031] Defective target precision metal parts can be identified as follows: The non-destructive testing system can adopt different data acquisition methods according to the actual application scenarios of the parts. For parts installed on automated production equipment, the system can establish a data connection with the equipment's control system to directly read the data on the number of times the parts have been used recorded by the equipment. Taking the key metal components on an automobile engine production line as an example, the equipment control system will accurately record the number of working cycles of each component, and the non-destructive testing system can directly extract the corresponding data from this system. If the parts are applied to scenarios where the data on the number of times of use cannot be directly obtained, the non-destructive testing system resorts to sensor technology to achieve the goal. Sensors such as microswitches and counters are installed on the parts, and each time a use cycle is completed, the sensor will record once. For example, in some small mechanical equipment, since the equipment itself does not have the function of integrating the record of the number of times of use, a microswitch can be installed on the moving parts of the key metal parts. When the part moves once, the microswitch will generate a change in the electrical signal, and the system can accurately count the number of times the part has been used by counting these electrical signals. Different types of metals, due to differences in factors such as their internal crystal structures and atomic bonding forces, have very different mechanisms and development laws of internal defect generation during long-term use. For example, in an environment of high temperature, high pressure and alternating stress, microcracks are likely to appear inside titanium alloy parts, and the crack propagation speed is closely related to the number of stress cycles; while nickel-based alloys are prone to internal stress corrosion cracking under the combined action of specific corrosive media and stress. The non-destructive testing system will pre-construct a database covering rich metal material characteristic data and store relevant information on the generation of internal defects of various metals under different working conditions.

[0032] The defect prediction model is constructed based on machine learning technology, and its training data comes from a large number of metal type information and metal usage times information sets with defect status annotations. These data are accumulated through long-term experimental tests and actual production monitoring, covering the actual situations of internal defects of various metals under different usage conditions. Through in-depth learning of this vast amount of data, the model discovers the complex correlations and potential laws between metal types, usage times and the occurrence of internal defects.

[0033] During actual detection, the non-destructive testing system inputs the information on the metal type and usage times of the parts to be detected into the defect prediction model. The model uses deep learning algorithms to perform multi-level and multi-angle feature extraction and analysis on the input information, and calculates the probability of defects occurring inside each part. When the defect probability of a certain part exceeds the threshold preset by the system, the system determines it as a defective target precision metal part. For example, when detecting a batch of high-strength alloy steel parts for deep-sea exploration equipment, the model combines their usage times and the characteristics of the alloy steel being prone to internal defects, and predicts that there may be small cavities caused by fatigue inside some parts, and these parts are immediately marked as key detection objects. In this way, the system can efficiently screen out parts that may have internal defects, improve the detection efficiency and accuracy, reduce the ineffective detection of non-defective parts, and reasonably allocate detection resources.

[0034] S102. Use a multi-angle ultrasonic probe to scan the target precision metal part to obtain internal reflection data; When the non-destructive testing system executes this step, it first performs intelligent control on the multi-angle ultrasonic probe. Based on the three-dimensional model of the target precision metal part (which can be obtained from the previous design data or preliminary scanning), the system uses a path planning algorithm to determine the scanning trajectory of the probe. For example, for the complex-shaped blades of an aeroengine, the system will plan a scanning path that surrounds the blade contour and can cover all key parts, ensuring that ultrasonic waves can penetrate the inside of the part from multiple angles and obtain internal reflection data.

[0035] During the scanning process, the system uses phased array technology to drive the multi-angle ultrasonic probe. The phased array probe consists of multiple tiny piezoelectric wafers. The non-destructive testing system realizes the flexible deflection and focusing of the ultrasonic beam by precisely controlling the excitation time and voltage of each wafer. For example, when detecting defects near the surface of the part, the system controls the phased array probe to focus the ultrasonic beam in the shallow layer area to improve the detection resolution; for defects deep inside the part, it adjusts the angle and focal length of the ultrasonic beam so that it can penetrate to the corresponding depth and accurately locate the defects.

[0036] In some embodiments, the ambient temperature at the detection site will have a great impact on the propagation characteristics of ultrasonic waves in metal parts. In order to obtain accurate ambient temperature data, the non-destructive testing system uses high-precision temperature sensors to complete this task. These sensors are scientifically installed in a position close to the target part and can represent the detection ambient temperature. For example, in the aircraft engine parts detection scenario, the temperature sensor will be installed inside the detection equipment near the part placement area to avoid measurement errors caused by local thermal radiation or airflow. The temperature sensor has the ability to respond quickly and measure with high precision, can capture changes in ambient temperature in real time, and transmit temperature data to the control unit of the non-destructive testing system in the form of electrical signals or digital signals. After receiving the ambient temperature data from the temperature sensor, the control unit of the non-destructive testing system will immediately search for a matching optimization strategy from the preset ambient temperature probe parameter optimization strategy library. This strategy library is built based on a large amount of experimental data and theoretical analysis. The system compares and analyzes the real-time ambient temperature data with the data in the strategy library, and uses intelligent algorithms to quickly determine the optimal parameter data of the ultrasonic probe at the current temperature, including but not limited to key parameters such as probe frequency, gain, and pulse repetition frequency. The non-destructive testing system automatically adjusts the parameters of the ultrasonic probe based on the determined optimal parameter data. After the adjustment is completed, the ultrasonic probe will be able to work at the best performance under the current ambient temperature. The emitted ultrasonic signal can penetrate the interior of the part more stably and accurately, and receive clearer and more reliable internal reflection data, laying a solid foundation for the subsequent accurate judgment of internal defects of the part.

[0037] In some embodiments, during the production, transportation and storage of the target precision metal parts, various impurities such as oil, dust, metal debris, etc. will inevitably adhere to the surface. These impurities will seriously interfere with the propagation of ultrasound. When ultrasound encounters impurities on the surface of the parts, chaotic reflection and scattering will occur. For example, oil will aggravate the energy attenuation of ultrasound during propagation, resulting in the weakening of the reflected signal. These signals are confused with the reflected signals generated by the internal defects of the parts, causing the collected internal reflection data to deviate, making it difficult to accurately reflect the real structure and defects inside the parts, thereby affecting the subsequent judgment and analysis of defects. The non-destructive testing system will automatically select the appropriate cleaning equipment and cleaning process according to the material, shape and surface condition of the parts. For metal parts with more oil on the surface, high-pressure spray cleaning equipment may be selected. During the cleaning process, the system ensures the cleaning effect by controlling the flow, pressure and temperature of the cleaning fluid. During cleaning, the system controls the frequency and power of the ultrasound to make the cleaning fluid produce high-frequency vibrations to peel off dust and debris from the surface of the parts. During the entire cleaning process, the system will also monitor the operating status of the cleaning equipment in real time to ensure that the cleaning operation is carried out according to the predetermined procedure.

[0038] S103, combining the metal type information and the internal reflection data, determining defect data through a defect detection model, where the defect detection model is constructed in advance through deep learning based on a plurality of metal type information and internal reflection data sets annotated with defect data; The nondestructive testing system inputs the acquired metal type information and internal reflection data into the defect detection model. Before inputting the data, the system will pre-process the data and convert the metal type information into a feature vector suitable for model processing. For example, the chemical composition, crystal structure and other information of the metal will be digitally encoded, and the internal reflection data will be normalized to make the data collected in different batches and under different conditions comparable. When processing internal reflection data, in addition to normalization, wavelet transform technology is also used to decompose the data at multiple scales. Wavelet transform can analyze reflection data at different frequency scales and capture subtle features hidden in the data. For example, when detecting tiny cracks, the weak reflection signal characteristics at the crack edge in the low-frequency band data can be amplified by wavelet transform, making these features easier to identify in subsequent model processing.

[0039] The defect detection model is built based on a convolutional neural network (CNN) and a recurrent neural network (RNN) architecture. CNN has a strong ability to extract image features and is suitable for processing internal reflection data with spatial structural characteristics. It can effectively identify spatial features such as the shape and position of defects. RNN is good at processing data with time series characteristics. In this scenario, the reflection data obtained at different scanning angles can be analyzed according to the time series to explore the change rules of the data in the time dimension, such as the trend of defects changing with the scanning angle, so as to more accurately judge the nature of the defects.

[0040] In the model training phase, in addition to using a large amount of metal type information and internal reflection data sets with defect data annotations, the defect data tag can be an internal defect tag, and the transfer learning technology is introduced to learn from the model parameters that have been trained in other similar material detection fields, and these parameters are used as the initial parameters of this model for fine-tuning. For example, in the field of semiconductor material defect detection, a model with good recognition effect on tiny structural defects has been trained. Some of its convolutional layer parameters are transferred to this precision metal parts defect detection model, and fine-tuned in combination with the current metal parts detection annotation data, which can accelerate the model convergence speed and improve the detection accuracy of the model in this field.

[0041] In the actual detection process, the system inputs the fused metal type information and internal reflection data into the defect detection model. The model outputs detailed defect data such as defect type (such as cracks, pores, looseness, etc.), size (dimension data in millimeters or microns), and depth (depth from the part surface) through the operation of multiple convolutional layers and loop layers.

[0042] S104. Combine the defect data and use 3D reconstruction technology to determine the 3D model information of the target precision metal part; After the non-destructive testing system obtains the defect data, it first performs feature extraction on it. With an advanced edge detection algorithm as the core, the system will process the defect data in multiple rounds. For example, the Canny edge detection algorithm is adopted. It smooths the image through Gaussian filtering to reduce noise interference, then calculates the image gradient to determine the possible edge positions, and then refines the edges using non-maximum suppression to accurately outline the defect boundaries. When determining the center point of the defect, the system uses the geometric center calculation method or the centroid algorithm based on the boundary information. For defects with regular shapes, the geometric center calculation method can quickly calculate the center point; for defects with complex shapes, the centroid algorithm calculates the weighted average position of each point within the defect area to more accurately find the center point. For the shape information, the system uses shape descriptors to quantify. Such as the Fourier descriptor, which transforms the boundary contour of the defect shape into frequency domain information. By analyzing the spectral characteristics, shape features such as roundness and aspect ratio are extracted, providing accurate data support for subsequent modeling.

[0043] The acoustic properties vary significantly among different metal types, which has a great impact on the accuracy of 3D reconstruction. The system determines the acoustic property parameters through two methods: the material property database and experimental measurement. In the material property database, a large amount of data such as the sound speed and sound attenuation coefficient of different metals under standard conditions are stored. The system quickly retrieves the corresponding basic data according to the identified metal type. However, the actual detection environment is complex and the parameters will change, so the system also uses the ultrasonic measurement method for real-time calibration. Ultrasonic waves are emitted at different positions on the surface of the part, and the time and amplitude changes of the ultrasonic waves propagating inside the part are measured. The sound speed is calculated based on the propagation time, and the sound attenuation coefficient is determined according to the amplitude attenuation. For example, for an aluminum alloy part, the sound speed is about 5000 - 6000 m / s. In a high-temperature environment, the sound speed will decrease. The system obtains more accurate acoustic property parameters through real-time measurement, providing a reliable basis for 3D reconstruction.

[0044] The system maps the internal reflection data and feature data to a three-dimensional space to construct an initial three-dimensional model. First, the voxel size is set according to the part size and detection accuracy. For tiny parts with high-precision detection, the voxel size may be set to dozens of micrometers; for large mechanical parts, the voxel size will be appropriately increased. After determining the voxels, the system fills the internal reflection data and defect feature data into the corresponding voxels according to the spatial positions. The internal reflection data reflects the acoustic reflection of the internal structure of the part, and the defect feature data determines the position and shape of the defect in the voxel space. The system uses an interpolation algorithm to fill the sparse data areas to ensure the integrity and continuity of the model. For example, for the internal cavity defect of a part, the interpolation algorithm is used to make the voxel data around the cavity transition smoothly, constructing an initial three-dimensional model that conforms to the actual situation and preliminarily presenting the internal structure and defect distribution of the part. Defect annotation is performed on the initial three-dimensional model to generate the final three-dimensional model information. The system marks information such as the defect type, position, and size on the model. When annotating the defect type, different types of defects such as cracks, pores, and porosity are distinguished and displayed with different colors or textures according to the output results of the defect detection model. The position annotation is based on the center point coordinates and boundary information of the defect to accurately locate in the three-dimensional model space, and the inspectors can directly view the specific position of the defect inside the part through the interaction interface. For the defect size, the system calculates the size parameters such as the length, width, and depth of the defect according to the extracted shape feature data and presents them in the form of marked text or lines on the model.

[0045] S105. Send the three-dimensional model information to the visual terminal for display.

[0046] Before sending the three-dimensional model information to the visual terminal for display, the non-destructive testing system will perform adaptive optimization processing according to the device type and display capabilities of the visual terminal. After receiving the three-dimensional model information, the visual terminal uses advanced visualization rendering technology for display. The ray tracing technology is used to simulate the light propagation in the real world, making the three-dimensional model present a realistic light and shadow effect on the visual terminal and more intuitively displaying the internal structure and defect position of the part.

[0047] In the embodiment of the present application, the target part is determined by obtaining the metal type information of the precision metal part to be detected, the internal reflection data is obtained by scanning with a multi-angle ultrasonic probe, the defect data is accurately determined by combining the metal type information and the reflection data through a defect detection model, and then the three-dimensional reconstruction technology is used to construct the three-dimensional model information and display it, realizing the efficient and accurate detection of the internal defects of the precision metal part. It not only effectively solves the problems of low detection accuracy and poor efficiency in the prior art, but also realizes pre-reducing the detection range, saving time, manpower, and equipment loss, and at the same time assisting the inspectors to quickly understand the details of the defects and improving the detection accuracy and convenience.

[0048] After combining the above content, the following is a further and more specific process description of the method provided in this embodiment. Please refer to Figure 2 , which is another process schematic diagram of the non-destructive testing method for internal defects of precision metal parts in the embodiments of the present application.

[0049] S201. Obtain the odor information of the target precision metal part by using an electronic nose device; During the production, processing, and use of metal parts, internal defects may cause the generation of volatile gases, which is mainly due to physical and chemical changes at the defect sites. For example, internal pores are cavities formed in metal parts during casting or processing. These pores may trap some gases during part manufacturing, such as air and water vapor entrained during the solidification of the metal melt in the casting process. When the part is in subsequent use and the temperature and pressure change, these trapped gases may undergo physical or chemical changes. For example, at high temperatures, water vapor may react with the metal. Taking an aluminum alloy part as an example, water vapor in the pores reacts with aluminum at high temperatures to generate hydrogen and aluminum oxide. In addition, if there are impurities in the metal around the pores, under certain conditions, chemical reactions may also occur between the impurities to produce gases. Internal cracks in metal parts also disrupt the continuity of the metal, and the metal atoms at the crack sites are in a high-energy unstable state. In a high-temperature environment, such as the high-temperature components of a gas turbine, the activity of metal atoms at the internal crack sites increases, and chemical reactions occur with the residual gases (such as oxygen and nitrogen) inside the part. Taking steel materials as an example, iron atoms at the crack sites react with oxygen at high temperatures to form iron oxide, and in this process, carbon monoxide or carbon dioxide gases may be generated. This is because steel contains a certain amount of carbon element, and during the oxidation reaction, the carbon element combines with oxygen to form carbon oxide gases. In addition, when a metal part is subjected to alternating stress, the crack will gradually expand. During the expansion process, some impurities inside the crack will be squeezed and rubbed, which may also trigger chemical reactions to produce volatile gases. For example, if the crack contains sulfur element, under the action of stress, it may react with iron to generate hydrogen sulfide gas.

[0050] In this case, the non-destructive testing system can detect odors through an electronic nose device, which consists of a gas sensor array, a signal preprocessing module, and a data acquisition module. During the detection process, the gas sensor array of the electronic nose device plays a key role. This array contains a variety of gas sensors that have specific responses to different gas components, and the sensitivity and selectivity of each sensor to a specific gas are carefully calibrated. For example, for the volatile gases that may be generated by metal parts due to defects, the system selects sensors that are sensitive to hydrogen, hydrocarbons, nitrogen oxides, etc. These sensors are based on different working principles. For instance, metal oxide semiconductor sensors detect gas concentration by the change in conductivity caused by the surface adsorption of gases, while electrochemical sensors detect based on the current signal generated by the redox reaction of gases on the electrodes.

[0051] When obtaining odor information, the electronic nose device is accurately positioned near the target precision metal part to ensure that the gas samples emitted by the part can be effectively collected. To ensure the representativeness of the collected gases, the system will set multiple sampling points around the part and adopt a combination of dynamic sampling and static sampling. During dynamic sampling, a micro air pump is used to continuously flow air over the surface of the part, quickly transporting the gases emitted by the part to the electronic nose device; static sampling is to allow the gases to naturally diffuse to the gas sensor array at specific time intervals to obtain more stable odor information.

[0052] S202. Combine the metal type information and the odor information, and determine the defect probability data through the odor defect prediction model. The odor defect prediction model is constructed in advance through multiple sets of metal type information and odor information with defect status annotations through deep learning; After the non-destructive testing system obtains the metal type information and odor information of the target precision metal parts, it will perform in-depth fusion processing on these two types of information to provide high-quality data input for the odor defect prediction model. For the metal type information, the system will further explore its deep features. Taking aluminum alloy as an example, different series of aluminum alloys (such as 2 series, 5 series, 6 series, etc.), due to the differences in their alloy compositions (different contents of elements such as copper, magnesium, silicon, etc.), when internal defects occur, their physical and chemical change processes will be different, which will in turn affect the generation of volatile gases. The system calls the built-in material property database to obtain the detailed composition information, crystal structure data, etc. of this metal type, and digitally encodes this information to convert it into a feature vector suitable for model input. When processing odor information, the system will perform noise reduction and feature extraction operations on the original data collected by the electronic nose. Due to various interference factors in the actual detection environment, such as other volatile substances in the environment and the background noise of the detection equipment, the original odor data may contain a large amount of noise. The system uses the wavelet denoising algorithm, which can analyze the odor data at different frequency scales, effectively remove high-frequency noise, and retain the useful features of the signal.

[0053] After processing the metal type information and odor information, the non-destructive testing system inputs the fused information into the odor defect prediction model. This model is built based on a deep learning framework and adopts an architecture that combines a convolutional neural network (CNN) and a long short-term memory network (LSTM). CNN has a powerful local feature extraction ability and can effectively capture local feature patterns in odor information. For example, when detecting specific volatile gases generated by internal defects in aluminum alloy parts, CNN can identify local features in the gas component combination, such as the concentration change trend of certain gas components and the relative proportion relationship between different gases. LSTM, on the other hand, is good at processing data with time series characteristics and can remember long-term information dependencies. In this scenario, since the development of internal defects in metal parts varies at different usage stages and the generated odors also change, LSTM can analyze the odor information collected at different time points, explore the long-term trend of odor changes, and thus more accurately judge the degree of defect development. During the model training stage, the non-destructive testing system will use a large number of metal type information and odor information sets with defect status annotations. These data come from detection samples in the actual production process, laboratory simulation experiment samples, etc. The system will strictly clean and preprocess the data, remove outliers and duplicate data, and ensure the quality of the training data. At the same time, the cross-validation method is adopted to divide the training data into multiple subsets, and training and validation are carried out on different subsets to improve the generalization ability of the model. During the training process, by adjusting the parameters of the model (such as the size of the convolutional kernel, the number of layers, the learning rate, etc.), the model can continuously learn the complex relationship between metal type information, odor information, and defect status, and finally achieve the purpose of accurately predicting the defect probability. When the system inputs the fused information into the trained odor defect prediction model, the model will output a numerical value representing the defect probability of the target precision metal part through the operations of multiple convolutional layers and LSTM layers, that is, the defect probability data. This data reflects the likelihood of the existence of internal defects in the part based on the currently obtained metal type information and odor information.

[0054] S203. If the defect probability data is greater than the set defect probability threshold, it is determined that the target precision metal part has a defect.

[0055] After the non-destructive testing system obtains the defect probability data, it will compare it with a pre-set defect probability threshold. The system determines the appropriate threshold by analyzing a large amount of historical detection data and combining the actual usage requirements and safety standards of different types of metal parts. For example, for critical metal parts used in the aerospace field, due to their extremely high safety requirements, even minor defects may cause serious consequences. Therefore, a relatively low defect probability threshold is set to ensure that as many potentially defective parts as possible are detected. For metal parts in some civilian products with relatively lower safety requirements, a relatively higher threshold may be set to improve the detection efficiency while ensuring a certain detection accuracy.

[0056] During the comparison process, the system uses a logic judgment module for rapid calculation. When the defect probability data is greater than the set defect probability threshold, the system will mark the part and include it in the list of key attention objects. The marking information includes detailed information such as the part number, detection time, and defect probability data, which is convenient for subsequent tracing and management of the detection process and handling situation of the part. At the same time, the system will send an alarm message to the operator, and the alarm method can be various forms such as audible and visual alarms, SMS notifications, and system pop-up prompts to ensure that the operator can promptly know the existence of defective parts.

[0057] In the embodiments of the present application, by using an electronic nose device to obtain the odor information of the target precision metal part and combining the metal type information, the defect probability data is determined through an odor defect pre-judgment model. If the defect probability data is greater than the set threshold, it is determined that the part is defective, realizing the efficient screening and accurate judgment of potential defects of precision metal parts. It not only effectively solves the problems of single detection means, easy missed judgment and misjudgment in the prior art, but also further improves the detection accuracy and efficiency, and enhances the reliability of quality control of precision metal parts under different working conditions.

[0058] The following describes the non-destructive testing system in the embodiments of the present invention application from the perspective of hardware processing. Please refer to Figure 3 , which is a schematic structural diagram of an entity device of the non-destructive testing system in the embodiments of the present application.

[0059] It should be noted that Figure 3 The structure of the non-destructive testing system shown is only an example and should not bring any limitations to the functions and usage scopes of the embodiments of the present invention.

[0060] As Figure 3As shown, the non-destructive testing system includes a Central Processing Unit (CPU) 301, which can perform various appropriate actions and processes according to the program stored in the Read-Only Memory (ROM) 302 or the program loaded from the storage section 308 into the Random Access Memory (RAM) 303, such as executing the methods described in the above embodiments. In the RAM 303, various programs and data required for system operation are also stored. The CPU 301, ROM 302, and RAM 303 are connected to each other via a bus 304. An Input / Output (I / O) interface 305 is also connected to the bus 304.

[0061] The following components are connected to the I / O interface 305: an input section 306 including an audio input device, a button switch, etc.; an output section 307 including a Liquid Crystal Display (LCD), an audio output device, an indicator light, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as needed. A removable medium 311, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 310 as needed so that a computer program read from it can be installed into the storage section 308 as needed.

[0062] Specifically, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network via the communication section 309, and / or installed from the removable medium 311. When the computer program is executed by the Central Processing Unit (CPU) 301, various functions defined in the present invention are executed.

[0063] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fibers, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above. In the present invention, a computer-readable storage medium can be any tangible medium that contains or stores a program, and this program can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0064] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present invention. Among them, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code, and the above-mentioned module, program segment, or part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order from that marked in the accompanying drawings.

[0065] Specifically, the non-destructive testing system of this embodiment includes a processor and a memory. A computer program is stored on the memory. When the computer program is executed by the processor, it implements the non-destructive testing method for internal defects of precision metal parts provided in the above embodiment.

[0066] On the other hand, the present invention also provides a computer-readable storage medium, which may be included in the non-destructive testing system described in the above embodiment; or it may exist alone and not be assembled into the non-destructive testing system. The above storage medium carries one or more computer programs. When the above one or more computer programs are executed by a processor of the non-destructive testing system, the non-destructive testing system implements the non-destructive testing method for internal defects of precision metal parts provided in the above embodiment.

[0067] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the various embodiments of the present application.

[0068] As used in the foregoing embodiments, depending on the context, the term "when" may be construed to mean "if" or "after" or "in response to determining" or "in response to detecting". Similarly, depending on the context, the phrase "when determining" or "if (the stated condition or event) is detected" may be construed to mean "if determining" or "in response to determining" or "when (the stated condition or event) is detected" or "in response to detecting (the stated condition or event)".

[0069] Those of ordinary skill in the art can understand that all or part of the process in implementing the methods of the foregoing embodiments can be completed by relevant hardware instructed by a computer program, and the program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the foregoing method embodiments. The foregoing storage media include: various media such as ROM or random access memory RAM, magnetic disks, or optical discs that can store program codes.

Claims

1. A non-destructive testing method for internal defects of precision metal parts, applied to a non-destructive testing system, characterized in that, The method includes: Obtaining the metal type information of the precision metal part to be detected, and determining the defective target precision metal part; Scanning the target precision metal part with a multi-angle ultrasonic probe to obtain internal reflection data; Combining the metal type information and the internal reflection data, and determining the defect data through a defect detection model, which is constructed in advance by deep learning based on multiple metal type information and internal reflection data sets with defect data annotations; Combining the defect data, and using three-dimensional reconstruction technology to determine the three-dimensional model information of the target precision metal part; Sending the three-dimensional model information to the visual end for display.

2. The method according to claim 1, wherein In the step of determining the defective target precision metal part, it specifically includes: Obtaining the usage times information of the precision metal part to be detected; Combining the metal type information and the usage times information, and determining the defective target precision metal part through a defect prediction model, which is constructed in advance by machine learning based on multiple metal type information and metal usage times information sets with defect status annotations.

3. The method according to claim 1, wherein Before the step of scanning the target precision metal part with a multi-angle ultrasonic probe to obtain internal reflection data, it further includes controlling a cleaning device to perform a pretreatment operation on the target precision metal part, and the pretreatment operation at least includes surface cleaning.

4. The method according to claim 1, wherein Before the step of scanning the target precision metal part with a multi-angle ultrasonic probe to obtain internal reflection data, it further includes: Obtaining the current ambient temperature data; Combining the ambient temperature data, and determining the optimal parameter data of the ultrasonic probe according to a preset ambient temperature probe parameter optimization strategy library; Adjusting the parameters of the ultrasonic probe according to the optimal parameter data.

5. The method according to claim 1, wherein In the step of combining the defect data and using three-dimensional reconstruction technology to determine the three-dimensional model information of the target precision metal part, it specifically includes: Performing feature extraction on the defect data to determine the feature data of the defect, and the feature data at least includes the boundary, center point, and shape information of the defect; Determining the acoustic characteristic parameters of the target precision metal part according to the metal type information, and the acoustic characteristic parameters at least include the sound velocity and the sound attenuation coefficient; Adopting a voxel-based three-dimensional reconstruction algorithm to map the internal reflection data and the feature data into a three-dimensional space to construct an initial three-dimensional model; Performing a defect annotation operation on the initial three-dimensional model to determine the final three-dimensional model information, and the defect annotation operation includes annotating the type, position, and size information of the defect on the initial three-dimensional model.

6. The method according to claim 1, characterized in that, After the step of obtaining the metal type information of the precision metal part to be detected and determining the defective target precision metal part, it further includes: Using an electronic nose device to obtain the odor information of the target precision metal part; Combining the metal type information and the odor information, and determining the defect probability data through an odor defect prediction model, which is constructed in advance by deep learning based on multiple metal type information and odor information sets with defect status annotations.

7. The method according to claim 6, characterized in that, After the step of determining defect probability data through the odor defect prediction model by combining the metal type information and the odor information, the following steps are further included: If the defect probability data is greater than the set defect probability threshold, it is determined that the target precision metal part has a defect.

8. A non-destructive testing system, characterized in that, The non-destructive testing system includes: one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is used to store computer program code, and the computer program code includes computer instructions. The one or more processors call the computer instructions to cause the non-destructive testing system to execute the method according to any one of claims 1-7.

9. A computer-readable storage medium, comprising instructions, characterized in that, When the instructions run on the non-destructive testing system, the non-destructive testing system is caused to execute the method according to any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product runs on the non-destructive testing system, the non-destructive testing system is caused to execute the method according to any one of claims 1-7.

Images