Deep learning based equipment structure nondestructive testing, terminal and storage medium

Abstract

The application discloses a deep learning-based equipment structure nondestructive detection method, a terminal and a storage medium, wherein a sound wave generator is controlled to emit a sound wave signal to a to-be-detected object, and a vibration velocimeter is controlled to collect vibration velocity signals of any one sampling point on the surface of the to-be-detected object within a preset time period; the collected vibration velocity signals are converted into time-frequency signal graphs and input into a first defect detection model to obtain a predicted detection result of the to-be-detected object; when the predicted detection result of the to-be-detected object is unrecognizable, the vibration velocimeter is controlled to collect vibration velocity signals of multiple preset sampling points on the surface of the to-be-detected object, time-frequency signal graphs converted from the vibration velocity signals of the preset sampling points are input into a second defect detection model, and it is determined whether defects exist in the preset sampling points, so that a defect region of the to-be-detected object is determined, nondestructive, non-contact and accurate and efficient defect detection of the to-be-detected object is realized, and the defect region of the equipment structure can be determined.

Description

Technical Field

[0001] This invention relates to the field of defect detection technology, and in particular to a method, terminal and storage medium for non-destructive testing of equipment structures based on deep learning. Background Technology

[0002] In daily life, some invisible defects can pose significant hidden dangers. Currently, equipment structures such as building structures, aerospace structures, vehicle engineering structures, additive manufacturing structures, and composite material structures may develop defects such as debonding, delamination, cracks, voids, and cavities during production and use, creating substantial risks. For example, the exterior walls of buildings are covered with hundreds or thousands of tiles, insulation boards, glass, and other protective structures. However, due to wind, sun, and even earthquakes, the adhesive layer between these materials and the building structure can age, creating cavities and eventually causing them to peel off and detach from the wall, posing a risk of falling objects. Similarly, defects such as debonding and cavities may occur between layers in composite material structures, also posing risks. Therefore, defect detection in equipment structures is essential. Furthermore, given the unknown extent of the defects, to avoid further damage, non-destructive testing methods must be used for defect detection in equipment structures.

[0003] Existing non-destructive testing technologies are mainly divided into two categories: contact testing and non-contact testing, including but not limited to mechanical methods, physical methods, electrical methods, chemical methods, and visual inspection. Contact testing requires direct contact between the testing equipment and the sample under test, thus requiring time for setup and sometimes even needing to apply adhesive or drill holes on the surface of the sample, making it unsuitable for testing equipment structures.

[0004] Currently, in non-contact inspection, visual inspection technology can only detect the surface of the sample and cannot detect the internal condition, making it unsuitable for detecting defects in equipment structures. Infrared detection technology requires the deployment of large thermal lamps to heat the sample, which is difficult to deploy and time-consuming, resulting in low detection efficiency for equipment structural defect detection. While radiation detection technologies such as X-rays and gamma rays offer high detection accuracy and depth, the transportation and storage of radiation sources require a high level of expertise, making deployment difficult and posing a risk of injury to workers during use, thus also unsuitable for equipment structural defect detection.

[0005] Therefore, how to provide a non-destructive, non-contact, efficient, and accurate defect detection solution suitable for equipment structures has become an urgent technical problem to be solved. Summary of the Invention

[0006] The main objective of this invention is to provide a terminal and storage medium for a non-destructive testing method for equipment structures based on deep learning, aiming to solve the problem that existing technologies cannot provide a non-destructive, non-contact, efficient, and accurate defect detection technology suitable for equipment structures.

[0007] To achieve the above objectives, the present invention provides a non-destructive testing method for equipment structures based on deep learning, the method comprising:

[0008] The sound wave generator is controlled to emit a sound wave signal to the object to be tested, and the vibration meter is controlled to collect the vibration velocity signal of any sampling point in the object to be tested within a preset time period.

[0009] The vibration velocity signal is converted into a time-frequency signal graph and input into a preset first defect detection model to obtain the predicted detection result of the object to be detected;

[0010] The predicted detection results include: normal, several preset defect types, and unidentifiable; the defect areas are different for each preset defect type.

[0011] When the predicted detection result of the object to be detected is that it cannot be identified, the vibration meter is controlled to collect vibration velocity signals from multiple preset sampling points on the surface of the object to be detected.

[0012] The vibration velocity signals of each preset sampling point are converted into time-frequency signal graphs and input into a preset second defect detection model to determine whether there are defects at each preset sampling point.

[0013] Based on whether defects exist at each of the preset sampling points, the defect area of ​​the object to be detected is determined.

[0014] Optionally, the preset first defect detection model is obtained by the following method:

[0015] The sound wave generator is controlled to emit sound wave signals to each sample object, and the vibration meter is controlled to collect vibration velocity signals at each preset sampling point on the surface of the sample object.

[0016] The sample object is of the same type as the object to be detected; the sample object is a pre-constructed group of objects with different defect types and normal objects.

[0017] The vibration velocity signals of the sample object are converted into time-frequency signal graphs and input into a first preset neural network model to obtain the predicted detection results of the sample object.

[0018] Based on the predicted detection results and the actual detection results of the sample object, the model parameters of the first preset neural network model are adjusted to obtain the preset first defect detection model.

[0019] Optionally, the preset second defect detection model is obtained by the following method:

[0020] The time-frequency signal graphs of each preset sampling point of the sample object are input into the second preset neural network model to obtain the prediction detection results of each preset sampling point;

[0021] The predicted detection result is used to indicate whether there is a defect at the corresponding preset sampling point;

[0022] Based on the predicted detection results and actual detection results of each preset sampling point, the model parameters of the second preset neural network model are adjusted to obtain the preset second defect detection model.

[0023] Optionally, after determining the defect region of the object to be detected based on whether defects exist at each of the preset sampling points, the method further includes:

[0024] The defect region of the object to be detected and each of the time-frequency signal maps are used as training samples;

[0025] The first defect detection model is trained and optimized based on the training samples to obtain an updated first defect detection model.

[0026] Optionally, before controlling the sound wave generator to emit a sound wave signal towards the object to be detected, the method further includes:

[0027] Obtain the feature information of the object to be detected;

[0028] The feature information includes at least: material information and size information;

[0029] Based on the feature information of the object to be detected, a first defect detection model and a second defect detection model corresponding to the object to be detected are determined from a preset model library.

[0030] Optionally, before controlling the sound wave generator to emit a sound wave signal towards the object to be detected, the method further includes:

[0031] Obtain the position information of the object to be detected;

[0032] Based on the position information of the object to be detected, the sound wave generator is controlled to move to the preset position corresponding to the object to be detected.

[0033] Optionally, the acoustic signal is a plurality of consecutive frequency-sweeping acoustic signals, and the frequency range of each frequency-sweeping acoustic signal is different.

[0034] Optionally, the object to be detected is any of the following: an object made of composite material; a mechanical structural component; or a tile or insulation board attached to a wall.

[0035] To achieve the above objectives, the present invention also provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores one or more programs, which can be executed by one or more processors to implement the steps in the deep learning-based non-destructive testing method for equipment structures as described in any of the above claims.

[0036] To achieve the above objectives, the present invention also provides a terminal, comprising: a processor and a memory; the memory storing a computer-readable program executable by the processor; and the processor executing the computer-readable program to implement the steps in the deep learning-based non-destructive testing method for equipment structures as described above.

[0037] This invention controls a sound wave generator to emit sound wave signals towards an object under test, and controls a vibration meter to collect vibration velocity signals from any sampling point on the surface of the object within a preset time period. The collected vibration velocity signals are converted into time-frequency signal graphs and input into a preset first defect detection model to obtain a predicted detection result for the object. If the predicted detection result for the object is unidentifiable, the vibration meter collects vibration velocity signals from multiple preset sampling points on the object, converts the vibration velocity signals from each preset sampling point into time-frequency signal graphs, and inputs them into a preset second defect detection model to determine whether defects exist at each preset sampling point, thereby identifying the defect area of ​​the object. On the one hand, this enables non-destructive and non-contact equipment structural defect detection; on the other hand, by performing step-by-step defect detection through the first and second defect detection models, the efficiency of equipment structural defect detection can be improved, and the location and shape of defect areas can also be detected, thus improving the accuracy and precision of equipment structural defect detection. Attached Figure Description

[0038] Figure 1 A flowchart of a deep learning-based nondestructive testing method for equipment structures provided in an embodiment of the present invention;

[0039] Figure 2 A schematic diagram of a vibration velocity signal provided in an embodiment of the present invention;

[0040] Figure 3 This is a schematic diagram illustrating a defect-free object to be inspected, provided as an embodiment of the present invention.

[0041] Figure 4 This is a schematic diagram of a square defect type of an object to be detected, provided in an embodiment of the present invention;

[0042] Figure 5 This is a schematic diagram of an equilateral triangular defect type of an object to be detected, provided in an embodiment of the present invention;

[0043] Figure 6 This is a schematic diagram illustrating a type of edge corner defect in an object to be detected, provided in an embodiment of the present invention.

[0044] Figure 7 This is a schematic diagram of a type of half-debonding defect of an object to be tested, provided by an embodiment of the present invention;

[0045] Figure 8 This is a schematic diagram of a circular defect type of an object to be detected, provided as an embodiment of the present invention;

[0046] Figure 9 This is an example diagram of a time-frequency signal diagram provided in an embodiment of the present invention;

[0047] Figure 10 An example diagram of preset sampling points provided in an embodiment of the present invention;

[0048] Figure 11 This is an illustration of the effect of non-destructive testing of ceramic tiles attached to a wall, provided by an embodiment of the present invention.

[0049] Figure 12 The training method for the first defect detection model provided in the embodiments of the present invention;

[0050] Figure 13 The training method for the second defect detection model provided in this embodiment of the invention;

[0051] Figure 14 A model structure diagram of a deep learning model based on a convolutional neural network provided in an embodiment of the present invention;

[0052] Figure 15 This is a schematic diagram of a classification result of the first defect detection model provided in an embodiment of the present invention;

[0053] Figure 16 This is a schematic diagram of a classification result of the second defect detection model provided in an embodiment of the present invention;

[0054] Figure 17 This is a schematic diagram of the structure of a terminal provided in an embodiment of the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0056] Defective regions (i.e., regions with defects such as debonding, delamination, cracks, pores, and cavities) exhibit different vibration characteristics under acoustic excitation compared to normal regions. Due to the presence of cavities, defective regions have reduced structural strength, and their natural frequencies differ from those of normal objects. Furthermore, different defect types and shapes result in different vibration models. Based on this, this invention provides a deep learning-based non-destructive testing (NDT) system for equipment structures, including a terminal and storage medium, to offer a non-destructive, non-contact, and efficient defect detection solution for equipment structures.

[0057] First, this invention provides a non-destructive testing method for equipment structures based on deep learning, such as... Figure 1 As shown, this deep learning-based non-destructive testing method for equipment structures can be implemented through at least the following steps:

[0058] S101, acquire the feature information of the object to be detected, and select the first defect detection model and the second defect detection model corresponding to the object to be detected from the preset model library according to the feature information of the object to be detected.

[0059] The feature information includes at least: material information and size information.

[0060] In this embodiment of the invention, the aforementioned preset model library stores multiple first defect detection models and second defect detection models. Each first defect detection model and second defect detection model is designed for different types of objects to be detected. That is, the corresponding first defect detection model and second defect detection model are different for different objects to be detected. For example, the first defect detection model and second defect detection model used to detect composite boards and ceramic tiles are different.

[0061] In this embodiment of the invention, each first defect detection model and second defect detection model can be generated from the training samples corresponding to the object to be detected. Then, the association between the feature information of the object to be detected and the first and second defect detection models can be generated and stored. Thus, after obtaining the feature information of the object to be detected, the first and second defect detection models used for defect detection of the current object to be detected can be obtained through the pre-stored association.

[0062] For example, the first defect detection model and the second defect detection model are used to detect defects in tile B (i.e. the object to be detected) attached to wall A. Wall A is a reinforced concrete structure with a thickness of 240mm, and tile B attached to wall A is a 100mm×100mm tile. The first defect detection model and the second defect detection model can be selected from the preset model library according to the pre-stored association relationship.

[0063] In real-world scenarios, different objects to be inspected vary in type, which can manifest as differences in material and size. Therefore, based on the characteristic information of the object to be inspected, a first and second defect detection model suitable for the current object can be selected from a pre-set model library, ensuring the accuracy of defect detection. When a different object is selected, its corresponding first and second defect detection models can be retrieved from the pre-set model library based on the new object's characteristic information, making it applicable to different objects and improving the versatility of non-destructive testing of equipment structures.

[0064] In some embodiments of the present invention, the object to be detected is any one of the following: an object made of composite material (e.g., carbon fiber composite material used to manufacture aircraft shells); a mechanical structural component; or a tile or insulation board attached to a wall.

[0065] S102, obtain the position information of the object to be detected, and control the sound wave generator to move to the preset position corresponding to the object to be detected based on the position information of the object to be detected.

[0066] In this embodiment of the invention, the acoustic wave generator may be, for example, a directional acoustic wave actuator to achieve remote detection.

[0067] In practical applications, the objects to be inspected may exist in different locations. Taking a building structure as an example, multiple objects to be inspected (such as tiles, insulation boards, etc.) are attached to the surface of wall A. Defect detection needs to be performed on these multiple objects attached to wall A. Therefore, the position information of each object to be inspected can be obtained first. Based on the position information, the sound wave generator and vibration meter can be moved to perform defect detection on multiple objects.

[0068] In addition, to further ensure the accuracy of equipment structural defect detection, it is necessary to control the sound wave generator to move to a preset position of the object to be detected, so as to ensure that the object to be detected can receive a preset sound wave signal intensity, which needs to be the same as the sound wave signal intensity corresponding to the training samples used for training the first defect detection model and the second defect detection model.

[0069] Specifically, the relative position of the sound wave generator and the object to be detected can be controlled to be the same as the relative position of the sound wave generator and the sample object when collecting training samples. For example, the detection of tiles requires controlling the sound wave generator to maintain a distance of 1m from the tile.

[0070] Furthermore, the acoustic generator and vibration meter can be mounted on the drone to adjust their relative positions to the object being tested, thus making them suitable for objects in different locations.

[0071] In some embodiments of the present invention, the vibration meter can be a laser Doppler vibration meter. The laser Doppler vibration meter has a large spatial range for collecting data and is almost unaffected by factors such as angle. Therefore, the vibration meter does not need to set a preset position as long as it can collect the vibration velocity signal of the object to be detected.

[0072] It should be noted that step S102 can be executed first, followed by step S103; or step S103 can be executed first, followed by step S102; or steps S103 and S102 can be executed simultaneously. No specific limitation is made in this embodiment of the invention.

[0073] S103 controls the sound wave generator to emit sound wave signals to the object to be tested, and controls the vibration meter to collect the vibration velocity signal of any sampling point on the surface of the object to be tested within a preset time period.

[0074] After performing the above steps S101-S102, the sound wave generator can be controlled to emit a sound wave signal towards the object to be tested, and the vibration meter can be controlled to collect the vibration velocity signal of any sampling point on the surface of the object to be tested within a preset time period, such as... Figure 2 As shown.

[0075] It is understandable that the vibration meter can be controlled to collect data at the same time as the sound wave generator emits the sound wave signal; or the vibration meter can be controlled to collect data before the sound wave generator emits the sound wave signal, thereby ensuring the integrity of the collected data and further improving the accuracy of equipment structural defect detection.

[0076] S104, convert the vibration velocity signal into a time-frequency signal graph and input it into the preset first defect detection model to obtain the predicted detection result of the object to be detected.

[0077] The predicted detection results include: normal, several preset defect types, and unidentifiable.

[0078] In the above-mentioned predicted test results, "normal" means that the object under test does not exhibit defects such as delamination, separation, cracks, pores, or cavities. For example, if the object under test is a ceramic tile attached to a wall, such as... Figure 3 As shown, the object under test is properly bonded to the wall, and no defects are found.

[0079] The aforementioned predicted defect types refer to the fact that the defect areas for each type are different. These differences can include variations in the location and / or shape of the defect areas. For example, taking a ceramic tile attached to a wall as the object to be inspected... Figure 4 The defect area shown is square, as... Figure 5 The defect area shown is an equilateral triangle, as... Figure 6 The defect area shown is a missing corner at the edge, such as... Figure 7 The defect area shown is half debonded, as... Figure 8 The defect area shown is a circular defect portion.

[0080] The "unidentifiable" result in the above-mentioned predictive detection results means that the predictive detection results cannot be identified through the time-frequency signal graph, and it is impossible to determine whether the object to be detected is normal or a preset defect type.

[0081] In some embodiments of the present invention, continuous wavelet transform can be used to convert the vibration velocity signal into a corresponding time-frequency signal graph, such as... Figure 9 As shown.

[0082] In this embodiment of the invention, the first defect detection model can be used to determine whether the object to be detected has a preset defect type.

[0083] S105, when the predicted detection result of the object to be detected is unrecognizable, the vibration meter is controlled to collect vibration velocity signals from multiple preset sampling points in the object to be detected.

[0084] For different objects to be detected, different numbers and locations of preset sampling points can be set. For example, Figure 10 As shown, for a 100×100mm tile attached to a wall, 36 test points can be set as preset sampling points, with a distance of 20mm between each preset sampling point, so as to collect vibration velocity signals from 36 preset sampling points.

[0085] In some embodiments of the present invention, after the vibration velocity signal is acquired, a filter can be used to eliminate noise in the signal and eliminate zero drift, so as to further improve the accuracy of equipment structural defect detection.

[0086] Understandably, during steps S101-S105, the acoustic generator continuously emits acoustic signals so that the vibration meter can collect the vibration velocity signal of the object under acoustic excitation.

[0087] S106, convert the vibration velocity signal of each preset sampling point into a time-frequency signal graph and input it into the preset second defect detection model to determine whether there is a defect at each preset sampling point.

[0088] The second defect detection model is used to determine whether a defect exists at a preset sampling point based on the time-frequency signal diagram of the preset sampling point.

[0089] S107, Based on whether there are defects at each preset sampling point, determine the defect area of ​​the object to be detected.

[0090] Specifically, by connecting the preset sampling points that have defects, the defect area of ​​the object to be detected can be drawn.

[0091] For example, taking a ceramic tile attached to a wall with a partial detachment defect as an example, the experiment is conducted. The first defect detection model detects the defect, and its predicted detection result is "unidentifiable." Then, the time-frequency signal maps corresponding to 36 preset sampling points on the surface of the object are input into the second defect detection model to determine whether defects exist at each preset sampling point, thus obtaining the predicted defect area of ​​the object. Figure 11 As shown.

[0092] This invention provides a deep learning-based non-destructive testing method for equipment structures. It controls a sound wave generator to emit sound wave signals towards the object under test, and controls a vibration meter to collect vibration velocity signals from any sampling point on the surface of the object within a preset time period. The collected vibration velocity signals are converted into time-frequency signal graphs and input into a preset first defect detection model to obtain a predicted detection result for the object. If the predicted detection result for the object is unrecognizable, the vibration meter collects vibration velocity signals from multiple preset sampling points on the object, converts the vibration velocity signals of each preset sampling point into time-frequency signal graphs, and inputs them into a preset second defect detection model to determine whether defects exist at each preset sampling point, thereby identifying the defect area of ​​the object. On the one hand, it enables non-destructive and non-contact defect detection of equipment structures; on the other hand, by performing step-by-step defect detection through the first and second defect detection models, it can improve the efficiency of equipment structure defect detection and also detect the location and shape of defect areas, thus improving the accuracy and precision of equipment structure defect detection.

[0093] like Figure 12 As shown, the aforementioned preset first defect detection model can be obtained through at least the following steps:

[0094] S1201 controls the sound wave generator to emit sound wave signals to each sample object, and controls the vibration meter to collect vibration velocity signals at each preset sampling point on the surface of the sample object.

[0095] The object to be tested is of the same type as the sample object. The sample objects consist of several pre-constructed objects with different defect types, as well as normal objects.

[0096] For example, using a 100×100mm tile attached to a wall as a sample object, five sample objects with different defect types can be manufactured. The defects are located in the cement between the tile and the wall, and each defect portion has a side length of 70mm, and is a square (e.g., Figure 4 As shown), equilateral triangle (as shown) Figure 5 As shown), edge missing corner (such as) Figure 6 As shown), half of the adhesive has come off (as shown) Figure 7 As shown), circular (as shown) Figure 8(As shown); at the same time, it can also manufacture normal objects, such as Figure 3 The image shows a normal adhesive layer. Thirty-six test points were set on the aforementioned tile, serving as preset sampling points. Each preset sampling point was 20mm apart. The sound wave generator was aimed at the sample object, and its position was adjusted so that the distance between the generator and the sample object was 1m. The vibration meter was aimed at each preset sampling point on the sample surface. The sound wave generator applied a sound wave signal to the sample object, and the vibration meter recorded the vibration velocity signal of each preset sampling point under sound wave excitation.

[0097] Furthermore, the term "same type" here refers to the same feature information.

[0098] Understandably, the sound wave generator also needs to be adjusted to a fixed preset position before step S1201.

[0099] In some embodiments of the present invention, the above-mentioned acoustic wave signal can be a fixed frequency sweep acoustic wave signal or multiple consecutive frequency sweep acoustic wave signals, and the frequency range of each frequency sweep acoustic wave signal is different.

[0100] Since vibration velocities are too small to be accurately distinguished in actual testing, multiple continuous sweep frequency acoustic signals are used to find the resonant frequency in order to increase the amplitude and generate different resonances, thereby further improving the accuracy of equipment structural defect detection.

[0101] S1202, convert each vibration velocity signal of the sample object into a time-frequency signal graph and input it into the first preset neural network model to obtain the predicted detection result of the sample object.

[0102] S1203, Based on the predicted detection results and the actual detection results of the sample object, adjust the model parameters of the first preset neural network model to obtain the preset first defect detection model.

[0103] The aforementioned actual test results refer to normal results or several different defect types.

[0104] Through the above steps S1201-S1203, a trained first defect detection model can be obtained. This first defect detection model is used to identify the defect type of the object to be detected, including normal, preset defect type, and unidentifiable defect type.

[0105] like Figure 13 As shown, the aforementioned pre-defined second defect detection model can be obtained through at least the following steps:

[0106] S1301, input the time-frequency signal graphs of each preset sampling point of the sample object into the second preset neural network model to obtain the prediction detection results of each preset sampling point.

[0107] In this embodiment of the invention, the time-frequency signal diagrams of the vibration velocity signals of the sample object obtained in step S1202 above can be used directly.

[0108] The predicted detection result is used to indicate whether there are defects at the corresponding preset sampling points.

[0109] S1302, Based on the predicted detection results and the actual detection results of each preset sampling point, adjust the model parameters of the second preset neural network model to obtain the preset second defect detection model.

[0110] Through the above steps S1301-S1302, a trained second defect detection model is obtained. The second defect detection model is used to identify whether there are defects at each sampling point on the surface of the object to be detected.

[0111] Both the first and second preset neural network models can be deep learning models based on convolutional neural networks, and their structures are as follows: Figure 14 As shown.

[0112] In some embodiments of the present invention, the defect region of the object to be detected and each time-frequency signal map can be used as training samples; and the first defect detection model can be trained and optimized based on the training samples to obtain an updated first defect detection model.

[0113] In this embodiment of the invention, the first defect detection model can classify vibration velocity signals into a variety of known preset defect types. The defect region of the object to be detected and the corresponding time-frequency signal diagrams obtained by the second defect detection model are used as training samples to optimize the first defect detection model, thereby enabling the optimized first defect detection model to be applicable to the detection of more different defect types.

[0114] As can be seen from the above embodiments, the first defect detection model can classify vibration velocity signals into several known preset defect types. Through training and optimization, the number of preset defect types that the first defect detection model can classify is continuously expanded. Figure 15 The diagram shows the classification results of the first defect detection model, with tiles attached to the wall as the object to be detected. Figure 16 The image shows the classification results of the second defect detection model, with the ceramic tile attached to the wall as the object to be detected. The second defect detection model can only classify vibration velocity signals into two categories: normal and defective. Figure 16 The damaged area in the image indicates the presence of a defect.

[0115] Based on the above-described deep learning-based non-destructive testing method for equipment structures, the present invention also provides a computer-readable storage medium storing one or more programs, which can be executed by one or more processors to implement the steps in the deep learning-based non-destructive testing method for equipment structures described in the above embodiments.

[0116] Based on the aforementioned deep learning-based nondestructive testing method for equipment structures, this invention also provides a terminal, such as... Figure 17 As shown, it includes at least one processor 30; a display screen 31; and a memory 32, and may also include a communications interface 33 and a bus 34. The processor 30, display screen 31, memory 32, and communications interface 33 can communicate with each other via the bus 34. The display screen 31 is configured to display a preset user guide interface in the initial setup mode. The communications interface 33 can transmit information. The processor 30 can call logical instructions in the memory 32 to execute the deep learning-based non-destructive testing method for equipment structures described in the above embodiments.

[0117] Furthermore, the logic instructions in the aforementioned memory 32 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium.

[0118] The memory 32, as a computer-readable storage medium, can be configured to store software programs, computer-executable programs, such as program instructions or modules corresponding to the methods in the embodiments of this disclosure. The processor 30 executes functional applications and data processing by running the software programs, instructions, or modules stored in the memory 32, thereby implementing the methods in the above embodiments.

[0119] The memory 32 may include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a given function; the data storage area may store data created based on the use of the terminal. Furthermore, the memory 32 may include high-speed random access memory (RAM) and non-volatile memory. Examples include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks; these can also be transient storage media.

[0120] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device and medium embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the description of the method embodiments.

[0121] The terminal and storage medium provided in this application are one-to-one with the method. Therefore, the terminal and storage medium also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the terminal and storage medium will not be repeated here.

[0122] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0123] Of course, those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.). The program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The computer-readable storage medium can be a memory, magnetic disk, optical disk, etc.

[0124] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A non-destructive testing method for equipment structures based on deep learning, characterized in that, The method includes: The sound wave generator is controlled to emit a sound wave signal to the object to be tested, and the vibration meter is controlled to collect the vibration velocity signal of any sampling point in the object to be tested within a preset time period. Before controlling the sound wave generator to emit a sound wave signal toward the object to be detected, the method further includes: Obtain the feature information of the object to be detected; The feature information includes at least: material information and size information; Based on the feature information of the object to be detected, a first defect detection model and a second defect detection model corresponding to the object to be detected are determined from a preset model library; The preset model library stores multiple first defect detection models and second defect detection models, each for different types of objects to be detected; By using pre-stored associations, obtain the first and second defect detection models for defect detection of the current object to be detected; The vibration velocity signal is converted into a time-frequency signal graph and input into a preset first defect detection model to obtain the predicted detection result of the object to be detected; The predicted detection results include: normal, several preset defect types, and unidentifiable; the defect areas are different for each preset defect type. When the predicted detection result of the object to be detected is that it cannot be identified, the vibration meter is controlled to collect vibration velocity signals from multiple preset sampling points on the surface of the object to be detected. The vibration velocity signals of each preset sampling point are converted into time-frequency signal graphs and input into a preset second defect detection model to determine whether there are defects at each preset sampling point. Based on whether there are defects at each of the preset sampling points, the defect area of ​​the object to be detected is determined; Connect the preset sampling points with defects to obtain the defect area of ​​the object to be detected; The preset first defect detection model is obtained through the following method: The sound wave generator is controlled to emit sound wave signals to each sample object, and the vibration meter is controlled to collect vibration velocity signals at each preset sampling point on the surface of the sample object. The sample object is of the same type as the object to be detected; the sample object is a pre-constructed group of objects with different defect types and normal objects. The vibration velocity signals of the sample object are converted into time-frequency signal graphs and input into a first preset neural network model to obtain the predicted detection results of the sample object. Based on the predicted detection results and the actual detection results of the sample object, the model parameters of the first preset neural network model are adjusted to obtain the preset first defect detection model. The preset second defect detection model is obtained through the following method: The time-frequency signal graphs of each preset sampling point of the sample object are input into the second preset neural network model to obtain the prediction detection results of each preset sampling point; The predicted detection result is used to indicate whether there is a defect at the corresponding preset sampling point; Based on the predicted detection results and actual detection results of each preset sampling point, the model parameters of the second preset neural network model are adjusted to obtain the preset second defect detection model.

2. The deep learning-based nondestructive testing method for equipment structures according to claim 1, characterized in that, After determining the defect region of the object to be detected based on whether defects exist at each of the preset sampling points, the method further includes: The defect region of the object to be detected and each of the time-frequency signal maps are used as training samples; The first defect detection model is trained and optimized based on the training samples to obtain an updated first defect detection model.

3. The deep learning-based nondestructive testing method for equipment structures according to claim 1, characterized in that, Before controlling the sound wave generator to emit a sound wave signal toward the object to be detected, the method further includes: Obtain the position information of the object to be detected; Based on the position information of the object to be detected, the sound wave generator is controlled to move to the preset position corresponding to the object to be detected.

4. The deep learning-based nondestructive testing method for equipment structures according to claim 1, characterized in that, The acoustic signal is a series of consecutive frequency-sweeping acoustic signals, and the frequency range of each frequency-sweeping acoustic signal is different.

5. The method for non-destructive testing of equipment structures based on deep learning according to claim 1, characterized in that, The object to be detected is an object made of composite materials.

6. The deep learning-based nondestructive testing method for equipment structures according to claim 1, characterized in that, The object to be detected is a mechanical structural component.

7. The deep learning-based nondestructive testing method for equipment structures according to claim 1, characterized in that, The object to be tested is: ceramic tile or insulation board attached to the wall.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores one or more programs, which can be executed by one or more processors to implement the steps in the deep learning-based non-destructive testing method for equipment structures as described in any one of claims 1-7.

9. A terminal, characterized in that, include: Processor and memory; The memory stores a computer-readable program that can be executed by the processor; when the processor executes the computer-readable program, it implements the steps in the deep learning-based non-destructive testing method for equipment structures as described in any one of claims 1-7.

Images