Wing structure damage detection method and device based on artificial intelligence

Abstract

The present invention discloses a wing structure damage detection method and device based on artificial intelligence, belonging to the field of artificial intelligence technology. The method comprises: obtaining N groups of strain data respectively detected by M sensors on the wings to be detected; the N groups of strain data correspond to N signal features one by one; according to the N groups of strain data, using a classification model to determine the damage type of the wing structure; the damage type includes structural fracture and connection problems; using a trained first detection model to predict the N groups of strain data and the damage type, and obtain a damage position prediction result. The present invention can improve the prediction accuracy of the damage position of the wing structure.

Description

Technical Field

[0001] The present invention relates to the technical field of damage detection, and in particular to a wing structure damage detection method and device based on artificial intelligence. Background Art

[0002] Damage to the wing structure will affect the flight safety of the aircraft. It is necessary to detect the damaged position of the wing structure. In the related art, manual inspection is used to determine the damage position. However, manual inspection can only find obvious damage on the surface of the wing, but the damage position of the internal wing cannot be accurately detected. Summary of the invention

[0003] The present invention provides a wing structure damage detection method and device based on artificial intelligence, which can accurately detect the damage location of the wing structure. The technical solution is as follows:

[0004] In one aspect, a wing structure damage detection method based on artificial intelligence is provided, the method comprising:

[0005] N groups of strain data are obtained by respectively detecting the wing to be detected by M sensors; the N groups of strain data correspond to N signal features one by one; the signal generator and the sensor are placed at the set position of the wing to be detected, and after the signal generator transmits a signal in sequence according to each signal feature, the M sensors detect each transmitted signal to obtain M strain data under the corresponding signal feature; M and N are both integers not less than 2;

[0006] According to the N groups of strain data, a classification model is used to determine the damage type of the wing structure; the damage type includes structural fracture and connection problem;

[0007] The trained first detection model is used to predict N groups of strain data and damage types to obtain damage location prediction results.

[0008] On the other hand, a wing structure damage detection device based on artificial intelligence is provided, the device comprising:

[0009] The acquisition unit is used to acquire N groups of strain data respectively detected by M sensors on the wing to be detected; the N groups of strain data correspond to N signal features one by one; the signal generator and the sensor are both placed at the set position of the wing to be detected, and after the signal generator transmits a signal in sequence according to each signal feature, the M sensors detect each transmitted signal to obtain M strain data under the corresponding signal feature; M and N are both integers not less than 2;

[0010] A classification unit, for determining the damage type of the wing structure using a classification model according to the N groups of strain data; the damage type includes structural fracture and connection problem;

[0011] The detection unit is used to predict N groups of strain data and damage types using the trained first detection model to obtain a damage position prediction result.

[0012] On the other hand, a computer device is provided, which includes a memory and a processor, the memory is used to store computer programs, and the processor is used to execute the computer programs stored in the memory to implement the steps of the above-mentioned artificial intelligence-based wing structure damage detection method.

[0013] On the other hand, a computer-readable storage medium is provided, wherein a computer program is stored in the storage medium, and when the computer program is executed by a processor, the steps of the above-mentioned artificial intelligence-based wing structure damage detection method are implemented.

[0014] On the other hand, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the steps of the above-mentioned artificial intelligence-based wing structure damage detection method.

[0015] The technical solution provided by the present invention can at least bring the following beneficial effects:

[0016] The use of artificial intelligence to predict the damage location of the wing structure can accurately predict the internal and surface damage locations. Taking into account that when the damage types in the wing structure are different, the damage type will affect the prediction result of the damage location, the classification model is pre-trained and used to determine the damage type based on the strain data. The damage type is then fed back to the first detection model. The first detection model combines the strain data and the damage type to jointly predict the damage location, thereby improving the prediction accuracy of the damage location of the wing structure. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 This is a flow chart of a wing structure damage detection method based on artificial intelligence provided by one embodiment of the present invention;

[0019] Figure 2 is a signal characteristic schematic diagram provided by an embodiment of the present invention;

[0020] Figure 3It is a structural diagram of a wing structure damage detection device based on artificial intelligence provided by one embodiment of the present invention;

[0021] Figure 4 It is a hardware architecture diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0023] Please refer to Figure 1 , an embodiment of the present invention provides a wing structure damage detection method based on artificial intelligence, the method comprising:

[0024] Step 100, obtaining N groups of strain data respectively detected by M sensors on the wing to be detected; the N groups of strain data correspond to N signal features one by one; the signal generator and the sensor are placed at the set position of the wing to be detected, and after the signal generator transmits a signal in sequence according to each signal feature, the M sensors detect each transmitted signal to obtain M strain data under the corresponding signal feature; M and N are both integers not less than 2;

[0025] Step 102, determining the damage type of the wing structure using a classification model based on the N groups of strain data; the damage type includes structural fracture and connection problem;

[0026] Step 104 , using the trained first detection model to predict the N groups of strain data and damage types, to obtain a damage location prediction result.

[0027] In an embodiment of the present invention, an artificial intelligence method is used to predict the damage position of a wing structure, and the internal and surface damage positions can be accurately predicted. In addition, considering that when the damage type in the wing structure is different, the damage type will affect the prediction result of the damage position, a classification model is pre-trained, and the classification model is used to first determine the damage type based on strain data, and then the damage type is fed back to the first detection model. The first detection model combines the strain data and the damage type to jointly predict the damage position, thereby improving the prediction accuracy of the damage position of the wing structure.

[0028] Below Figure 1 Each step shown is described separately.

[0029] First, for step 100, N groups of strain data are obtained by respectively detecting the wing to be detected by M sensors; the N groups of strain data correspond to N signal features one by one; the signal generator and the sensor are placed at the set position of the wing to be detected, and after the signal generator transmits a signal in turn according to each signal feature, the M sensors detect each transmitted signal to obtain M strain data under the corresponding signal feature; M and N are both integers not less than 2.

[0030] In an embodiment of the present invention, when detecting damage to a wing structure, a detection device is required to detect strain data. The detection device includes M sensors and a signal generator. During detection, the signal generator and the sensor are placed on the wing structure, the signal generator transmits a signal with specific signal characteristics, and the sensor receives strain data. The sensor may be a piezoelectric ceramic sensor. Preferably, the detection device includes two sensors and a signal generator. After transmitting the signal, the signal generator can also be used as a sensor to receive strain data.

[0031] It should be noted that when the wing structure type is different, the placement position of the sensor and the signal generator of the detection device can be different. After determining the wing structure type, it can be placed at the set position of the wing structure type according to the prompt, and the placement position remains unchanged during the entire detection process.

[0032] In the embodiment of the present invention, N signal characteristics are pre-set, and the signal characteristics may include: waveform, frequency, voltage, pulse duration and modulation method.

[0033] In one embodiment, N is equal to 30, and the 30 signal features are referred to Figure 2 As shown. The signal generator transmits signals of corresponding signal characteristics one by one according to the 30 signal characteristics, and the sensor receives the signals of corresponding signal characteristics to detect and obtain strain data. If the number of sensor pairs is M, then each of the N groups of strain data includes M strain data.

[0034] It should be noted that after the N signals are preset, the same N signals are used for the entire training process and the entire detection process of the model.

[0035] Next, step 102 "determine the damage type of the wing structure using a classification model based on N groups of strain data; the damage type includes structural fracture and connection problems" and step 104 "predict the N groups of strain data and damage types using the trained first detection model to obtain damage location prediction results" are explained simultaneously.

[0036] In the embodiment of the present invention, the classification model and the first detection model need to be trained in advance, and then the classification model and the first detection model are used to detect the damage of the wing structure.

[0037] The classification model, the first detection model and their training process are described below.

[0038] In an embodiment of the present invention, the classification model is used to determine the damage type of the wing structure according to the detected N groups of strain data, and the first detection model is used to predict the damage location of the wing structure according to the N groups of strain data and the damage type.

[0039] In the embodiment of the present invention, the classification model can be implemented in at least the following two ways:

[0040] The first method is: the classification model includes a second detection model and a random forest classifier, the second detection model is used to predict the damage location of the N sets of input strain data, and the damage type is determined according to the damage location combined with the random forest classifier.

[0041] The second method is that the classification model directly predicts the damage type based on the N sets of input strain data.

[0042] The above two methods are described below respectively.

[0043] In the first method, the first detection model, the second detection model and the random forest classifier need to be trained. The training methods include:

[0044] S11: Acquire multiple training samples; the training samples include: sample damage position coordinate values, damage types, and N groups of sample strain data; different training samples have different sample damage position coordinate values; the N groups of sample strain data correspond one-to-one to the N signal features, and the acquisition method of the N groups of sample strain data is the same as the acquisition method of the N groups of strain data;

[0045] S12: using N groups of sample strain data as input and sample damage position coordinate values ​​as output, and using multiple training samples to train the second detection model;

[0046] S13: screening multiple target samples whose damage positions are within the connection area from multiple training samples according to the coordinate values ​​of the damage positions of the samples; taking N groups of sample strain data as input and damage types as output, and using the multiple target samples to train a random forest classifier;

[0047] S14: Taking N groups of sample strain data and damage types as input and sample damage position coordinate values ​​as output, and using multiple training samples to train the first detection model.

[0048] In step S11, the method for obtaining training samples may include: damaging the wing structure with corresponding damage types at the positions of multiple sample damage position coordinate values ​​on the wing structure to form a one-to-one corresponding sample structure, and then for each sample structure, using a detection device to detect and obtain corresponding N groups of sample strain data, thereby obtaining multiple training samples.

[0049] In addition, considering that the model training process requires a large number of training samples, for training samples with connection damage, the training samples can be obtained in the above manner. However, for training samples with structural fracture damage, if the wing structure is directly damaged by structural fracture to obtain training samples, the cost is relatively high.

[0050] Based on this, in the embodiment of the present invention, the method for obtaining the training samples whose damage type is structural fracture may be:

[0051] B1: Based on several wing structures with real structural fractures, multiple real strain data are obtained; and based on the real damage positions corresponding to the multiple real strain data, multiple simulated strain data corresponding to each other are obtained by simulation; the simulated strain data are used as input, and the real strain data with the same damage position as the input simulated strain data are used as output to form a training data set;

[0052] B2: using the training data set to train the simulation mapping model, using the currently trained simulation mapping model to obtain predicted strain data at multiple different simulated damage positions, using the particle swarm optimization algorithm to select the optimal predicted strain data from the multiple predicted strain data, using the optimal predicted strain data as the real strain data, and for the simulated damage position corresponding to the real strain data, using the simulation method to obtain the corresponding simulated strain data, adding the newly added real strain data and the corresponding simulated strain data to the training data set, and continuing to perform step B2 using the current training data set until the currently trained simulation mapping model meets the end condition, thereby obtaining a trained simulation mapping model;

[0053] B3: Predict multiple real strain data using the trained simulation mapping model and the simulation method, and obtain training samples with damage type of structural fracture based on the multiple real strain data obtained in step B1 and the predicted multiple real strain data.

[0054] Among them, abaqus or croe software can be used to obtain simulation strain data.

[0055] In an embodiment of the present invention, when training a simulation mapping model, after each training, the optimal predicted strain data is selected from the multiple predicted strain data output and added to the training sample set, thereby increasing the amount of training data, and each training enables the simulation mapping model to converge toward high accuracy once, thereby gradually improving the prediction accuracy of the simulation mapping model, and further ensuring the accuracy of the acquired training samples.

[0056] It should be noted that, when training the simulation mapping model, one simulation mapping model may be trained for multiple signal features, or different simulation mapping models may be trained for different signal features.

[0057] After obtaining a plurality of training samples in S11, S12-S14 may be used to respectively train and obtain a first detection model, a second detection model, and a random forest classifier.

[0058] Specifically, in step S13, after the model of the wing structure is determined, the connection area of ​​the wing structure is known. If the coordinate value of the sample damage position is located in the connection area, it can be screened out as a target sample. Therefore, after determining that the damage position is located in the connection area, the scope of type judgment can be narrowed, and then N groups of strain data can be used to determine the damage type, and the judgment result is more accurate.

[0059] In one implementation, the first detection model and the second detection model may be GRU models.

[0060] In the second method, the first detection model and the classification model need to be trained, and the training methods include:

[0061] S21: Acquire multiple training samples; the training samples include: sample damage position coordinate values, damage types, and N groups of sample strain data; different training samples have different sample damage position coordinate values; the N groups of sample strain data correspond to the N signal features one by one, and the acquisition method of the N groups of sample strain data is the same as the acquisition method of the N groups of strain data;

[0062] S22: using N groups of sample strain data as input and damage types as output, and using multiple training samples to train the classification model;

[0063] S23: Taking N groups of sample strain data and damage types as input and sample damage position coordinate values ​​as output, and using multiple training samples to train the first detection model.

[0064] The method for obtaining the training samples in S21 is the same as that in S11, and will not be described in detail here.

[0065] Since the signal is transmitted to the damage location, if the damage type is a structural fracture, the strain data fluctuation will be greatly offset. When the classification model is directly used to predict the damage type of the strain data, it is easy to judge the connection problem at a certain damage location as a structural fracture at another damage location, and the accuracy is poor. Therefore, in an embodiment of the present invention, the second detection model is first used to determine the initial prediction value of the damage location based on N groups of strain data, and then the initial prediction value of the damage location is used to determine whether the damage location is located in the connection area. If it is located, the random forest classifier will be used to determine the damage type of the wing structure. Otherwise, it can be directly determined that the damage type is a structural fracture. It can be seen that when the classification model adopts the first method, the prediction result will be more accurate.

[0066] It should be noted that in order to improve the detection accuracy, corresponding models can be trained separately for different wing structure types when training the model.

[0067] In the embodiment of the present invention, if the classification model adopts the first method above, then this step 102 may specifically include:

[0068] Step 1020: using the second detection model to predict the N groups of strain data to obtain an initial prediction value of the damage position;

[0069] Step 1022: Determine whether the damage position is located within the connection area based on the initial prediction value of the damage position and the predetermined connection area of ​​the wing structure; if so, determine the damage type of the wing structure based on the N groups of strain data using the random forest classifier; if not, directly determine the damage type of the wing structure as structural fracture.

[0070] It should be noted that the wing structure is made up of multiple parts connected in various ways, including but not limited to screw connection and mortise and tenon connection. When the damage type is a connection problem, it may be a screw falling off, a mortise and tenon connection falling off, etc. When the damage type is a structural fracture, it means that the internal structure has fracture damage in addition to the connection problem.

[0071] Furthermore, considering that the range of strain data and damage location is relatively large, the accuracy of the prediction results is affected to a certain extent. In order to further improve the prediction accuracy, in one embodiment of the present invention, the data can be standardized in a standardized manner so that the data is mapped to a smaller standard range to further improve the accuracy of data prediction.

[0072] Specifically, after obtaining a plurality of training samples, before the training, the method further includes: calculating the mean and variance of the sample damage position coordinate values ​​in the plurality of training samples, and calculating the mean and variance of the sample strain data in the plurality of training samples; using the mean and variance of the sample damage position coordinate values, standardizing the sample damage position coordinate values ​​in each training sample; using the mean and variance of the sample strain data, standardizing the sample strain data in each training sample; and using the standardized training samples for training;

[0073] Then, in the actual detection process, after obtaining the N groups of strain data and before determining the damage type of the wing structure, it may also include: standardizing the N groups of strain data using the mean and variance of the sample strain data.

[0074] It should be noted that when the classification model adopts the first method mentioned above, the second detection model is used to predict the N groups of strain data after standardization. After obtaining the initial prediction value of the damage position, it is also necessary to perform quasi-standardization on the initial prediction value of the damage position, and then use the initial prediction original value of the loss position obtained by the quasi-standardization to determine whether the damage position is located in the connection area.

[0075] In order to verify that the embodiment of the present invention has a high damage detection accuracy, the scheme of the embodiment of the present invention and the comparative scheme were tested. The comparative scheme directly uses the second detection model to predict the damage location of N groups of strain data, while the embodiment of the present invention uses the first detection model to predict the damage location of N groups of strain data and damage types after the classification model first determines the damage type. The errors of the two are as follows:

[0076] The mean square error in the comparison scheme is 2.7686 and the mean absolute error is 1.3354;

[0077] In the embodiment of the present invention, the mean square error is 1.4615 and the mean absolute error is 0.9398.

[0078] This shows that the embodiment of the present invention has higher loss detection accuracy.

[0079] Please refer to Figure 3 , an embodiment of the present invention provides a wing structure damage detection device based on artificial intelligence, the device comprising:

[0080] The acquisition unit 300 is used to acquire N groups of strain data respectively detected by M sensors on the wing to be detected; the N groups of strain data correspond to N signal features one by one; the signal generator and the sensor are both placed at the set position of the wing to be detected, and after the signal generator transmits a signal in sequence according to each signal feature, the M sensors detect each transmitted signal to obtain M strain data under the corresponding signal feature; M and N are both integers not less than 2;

[0081] A classification unit 302 is used to determine the damage type of the wing structure using a classification model according to the N groups of strain data; the damage type includes structural fracture and connection problem;

[0082] The detection unit 304 is used to predict the N groups of strain data and damage types using the trained first detection model to obtain a damage position prediction result.

[0083] In one embodiment of the present invention, the classification model includes: a second detection model and a random forest classifier;

[0084] The classification unit is specifically used to: use the second detection model to predict the N groups of strain data to obtain an initial prediction value of the damage position; determine whether the damage position is located in the connection area based on the initial prediction value of the damage position and a predetermined wing structure connection area; if so, use the random forest classifier to determine the damage type of the wing structure for the N groups of strain data; if not, directly determine that the damage type of the wing structure is a structural fracture.

[0085] In one embodiment of the present invention, the training methods of the first detection model, the second detection model and the random forest classifier include:

[0086] Acquire multiple training samples; the training samples include: sample damage position coordinate values, damage types and N groups of sample strain data; different training samples have different sample damage position coordinate values; the N groups of sample strain data correspond to the N signal features one by one, and the acquisition method of the N groups of sample strain data is the same as the acquisition method of the N groups of strain data;

[0087] Taking N groups of sample strain data as input and sample damage position coordinate values ​​as output, the second detection model is trained using multiple training samples;

[0088] According to the coordinate values ​​of the sample damage positions, multiple target samples with damage positions within the connection area are selected from multiple training samples; N groups of sample strain data are used as input and damage types are used as output, and a random forest classifier is trained using multiple target samples;

[0089] Taking N groups of sample strain data and damage types as input and sample damage position coordinate values ​​as output, the first detection model is trained using multiple training samples.

[0090] In one embodiment of the present invention, the device may further include: a standardization processing unit, used to calculate the mean and variance of the sample damage position coordinate values ​​in multiple training samples, and calculate the mean and variance of the sample strain data in multiple training samples; using the mean and variance of the sample damage position coordinate values, standardize the sample damage position coordinate values ​​in each training sample; using the mean and variance of the sample strain data, standardize the sample strain data in each training sample; and use the standardized training samples for training;

[0091] After acquiring the N groups of strain data and before determining the damage type of the wing structure, the standardization processing unit is further used to: perform standardization processing on the N groups of strain data using the mean and variance of the sample strain data.

[0092] In one embodiment of the present invention, the signal characteristics include: waveform, frequency, voltage, pulse duration and modulation method.

[0093] It should be noted that: the wing structure damage detection device based on artificial intelligence provided in the above embodiment is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the wing structure damage detection device based on artificial intelligence provided in the above embodiment and the wing structure damage detection method based on artificial intelligence belong to the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0094] The embodiment of the present application also provides a computer device, please refer to Figure 4 The computer device includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, at least one program, a code set or an instruction set is loaded and executed by the processor to implement the artificial intelligence-based wing structure damage detection method provided by the above-mentioned method embodiments.

[0095] An embodiment of the present application also provides a computer-readable storage medium, on which is stored at least one instruction, at least one program, code set or instruction set, and the at least one instruction, at least one program, code set or instruction set is loaded and executed by a processor to implement the artificial intelligence-based wing structure damage detection method provided by the above-mentioned method embodiments.

[0096] An embodiment of the present application also provides a computer program product, which includes a computer program. A processor of a computer device reads the computer program from a computer-readable storage medium, and the processor executes the computer program, so that the computer device executes any of the artificial intelligence-based wing structure damage detection methods described in the above embodiments.

[0097] For the convenience of description, the above system or device is described as being divided into various modules or units according to their functions. Of course, when implementing the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0098] It can be known from the description of the above implementation methods that those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application can be essentially or partly contributed to the prior art in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present application or certain parts of the embodiments.

[0099] Finally, it should be noted that, in this article, relational terms such as first, second, third and fourth are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0100] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A wing structure damage detection method based on artificial intelligence, characterized in that: The method comprises: N groups of strain data are obtained by respectively detecting the wing to be detected by M sensors; the N groups of strain data correspond to N signal features one by one; the signal generator and the sensor are placed at the set position of the wing to be detected, and after the signal generator transmits a signal in sequence according to each signal feature, the M sensors detect each transmitted signal to obtain M strain data under the corresponding signal feature; M and N are both integers not less than 2; According to the N groups of strain data, a classification model is used to determine the damage type of the wing structure; the damage type includes structural fracture and connection problem; The trained first detection model is used to predict N groups of strain data and damage types to obtain damage location prediction results; The classification model includes: a second detection model and a random forest classifier; Determining the damage type of the wing structure using a classification model based on N sets of strain data includes: Using the second detection model to predict the N groups of strain data, to obtain an initial prediction value of the damage position; According to the initial predicted value of the damage position and the predetermined connection area of ​​the wing structure, determining whether the damage position is located in the connection area; if so, determining the damage type of the wing structure according to the N groups of strain data using the random forest classifier; if not, directly determining the damage type of the wing structure as structural fracture; The training methods of the first detection model, the second detection model and the random forest classifier include: Acquire multiple training samples; the training samples include: sample damage position coordinate values, damage types and N groups of sample strain data; different training samples have different sample damage position coordinate values; the N groups of sample strain data correspond to the N signal features one by one, and the acquisition method of the N groups of sample strain data is the same as the acquisition method of the N groups of strain data; Taking N groups of sample strain data as input and sample damage position coordinate values ​​as output, the second detection model is trained using multiple training samples; According to the coordinate values ​​of the sample damage positions, multiple target samples with damage positions within the connection area are selected from multiple training samples; N groups of sample strain data are used as input and damage types are used as output, and a random forest classifier is trained using multiple target samples; Taking N groups of sample strain data and damage types as input and sample damage position coordinate values ​​as output, the first detection model is trained using multiple training samples.

2. The method according to claim 1, characterized in that After obtaining a plurality of training samples, before the training is performed, the method further includes: calculating the mean and variance of the sample damage position coordinate values ​​in the plurality of training samples, and calculating the mean and variance of the sample strain data in the plurality of training samples; using the mean and variance of the sample damage position coordinate values, performing standardization processing on the sample damage position coordinate values ​​in each training sample; using the mean and variance of the sample strain data, performing standardization processing on the sample strain data in each training sample; and performing training using the standardized training samples; After acquiring the N groups of strain data and before determining the damage type of the wing structure, the method further includes: performing standardization processing on the N groups of strain data using the mean and variance of the sample strain data.

3. The method according to any one of claims 1-2, characterized in that: The signal characteristics include: waveform, frequency, voltage, pulse duration and modulation method.

4. A wing structure damage detection device based on artificial intelligence, characterized in that: The device comprises: The acquisition unit is used to acquire N groups of strain data respectively detected by M sensors on the wing to be detected; the N groups of strain data correspond to N signal features one by one; the signal generator and the sensor are both placed at the set position of the wing to be detected, and after the signal generator transmits a signal in sequence according to each signal feature, the M sensors detect each transmitted signal to obtain M strain data under the corresponding signal feature; M and N are both integers not less than 2; A classification unit, for determining the damage type of the wing structure using a classification model according to the N groups of strain data; the damage type includes structural fracture and connection problem; A detection unit, used to predict N groups of strain data and damage types using the trained first detection model to obtain a damage location prediction result; The classification model includes: a second detection model and a random forest classifier; The classification unit is specifically used to: predict the N groups of strain data using the second detection model to obtain an initial prediction value of the damage position; determine whether the damage position is located in the connection area according to the initial prediction value of the damage position and a predetermined connection area of ​​the wing structure; if so, determine the damage type of the wing structure based on the N groups of strain data using the random forest classifier; if not, directly determine that the damage type of the wing structure is a structural fracture; The training methods of the first detection model, the second detection model and the random forest classifier include: Acquire multiple training samples; the training samples include: sample damage position coordinate values, damage types and N groups of sample strain data; different training samples have different sample damage position coordinate values; the N groups of sample strain data correspond to the N signal features one by one, and the acquisition method of the N groups of sample strain data is the same as the acquisition method of the N groups of strain data; Taking N groups of sample strain data as input and sample damage position coordinate values ​​as output, the second detection model is trained using multiple training samples; According to the coordinate values ​​of the sample damage positions, multiple target samples with damage positions within the connection area are selected from multiple training samples; N groups of sample strain data are used as input and damage types are used as output, and a random forest classifier is trained using multiple target samples; Taking N groups of sample strain data and damage types as input and sample damage position coordinate values ​​as output, the first detection model is trained using multiple training samples.

5. The device according to claim 4, characterized in that The device also includes: A standardization processing unit is used to calculate the mean and variance of the sample damage position coordinate values ​​in multiple training samples, and calculate the mean and variance of the sample strain data in multiple training samples; use the mean and variance of the sample damage position coordinate values ​​to standardize the sample damage position coordinate values ​​in each training sample; use the mean and variance of the sample strain data to standardize the sample strain data in each training sample; and use the standardized training samples for training; After acquiring the N groups of strain data and before determining the damage type of the wing structure, the standardization processing unit is further used to: perform standardization processing on the N groups of strain data using the mean and variance of the sample strain data.

6. The device according to any one of claims 4-5, characterized in that: The signal characteristics include: waveform, frequency, voltage, pulse duration and modulation method.

7. A computer device, characterized in that: The computer device includes a memory and a processor, the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory to implement the steps of any of the methods described in claims 1-3.

8. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method described in any one of claims 1 to 3 are implemented.

9. A computer program product, characterized in that The method comprises a computer program, which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 3.

Images