Steel structure plate fault detection method and device and computer readable storage medium

Abstract

The invention discloses a steel structure plate fault detection method and device and a computer readable storage medium, and relates to the technical field of fault monitoring. The method comprises the following steps: constructing a data set based on steel structure plate picture data, wherein the data set comprises a training set and a verification set; the training set and the test set are sequentially used for training and testing a fault detection model, and the fault detection model is based on LSTM and is matched with an attention module through an attention mechanism; and inputting to-be-detected steel structure plate picture data into the trained fault detection model, and outputting a steel structure plate fault detection result. The scheme provided by the invention not only has efficient and accurate detection capability, but also can adapt to detection requirements in different environments, and ensures the stability and reliability of a detection result.

Description

Technical Field

[0001] The present invention belongs to the technical field of fault detection, and in particular relates to a steel structure plate fault detection method, equipment and computer-readable storage medium. Background Art

[0002] During the manufacturing and use process, steel structure plates are often eroded by various factors, resulting in cracks, pits, damage and other faults on the surface. These faults are often caused by multiple factors such as unreasonable design, substandard construction quality, and poor material quality. These faults will not only reduce the overall performance of the steel structure, but also pose a serious threat to its safety and reliability. Therefore, it is crucial to conduct comprehensive and detailed inspections on the steel structure before it leaves the factory to ensure that its quality meets relevant standards and requirements.

[0003] Traditional steel structure plate fault detection methods mainly rely on manual inspection. However, this method has many shortcomings. On the one hand, manual inspection is inefficient, the inspection speed is difficult to improve, and it is difficult to meet the inspection needs of large-scale production; on the other hand, manual inspection is easily affected by subjective factors and is prone to miss some potential defects, thus bringing safety hazards to subsequent use. In addition, manual inspection can usually only find more obvious defects, and is often powerless for some subtle and difficult-to-detect faults.

[0004] With the advancement of science and technology, there are some technologies for defect recognition based on images. These technologies realize the automatic recognition and detection of surface defects of steel structure plates through image processing. However, these technologies are not perfect. In practical applications, they are often limited by factors such as the clarity of defect boundaries and lighting conditions, resulting in unstable detection results and difficulty in ensuring accuracy.

[0005] In summary, the current field of steel structure plate fault detection urgently needs a new solution that can overcome the defects of existing technologies. This new solution not only needs to have efficient and accurate detection capabilities, but also needs to be able to adapt to the detection needs in different environments and ensure the stability and reliability of the detection results. Therefore, it is of great significance and value for research and development in this field.

[0006] The information disclosed in this background technology section is only intended to enhance the understanding of the overall background of the invention and should not be regarded as an acknowledgment or any form of suggestion that the information constitutes the prior art already known to a person skilled in the art. Summary of the invention

[0007] The object of the present invention is to solve the above-mentioned deficiencies in the prior art and to provide a steel structure plate fault detection method, device and computer-readable storage medium.

[0008] A first aspect of the present invention provides a method for detecting a steel structure plate fault, comprising the following steps:

[0009] The following steps are involved:

[0010] Building a data set based on steel structure plate image data, the data set includes a training set and a validation set;

[0011] The fault detection model is trained and tested using the training set and the test set in sequence, wherein the fault detection model is based on LSTM and is equipped with an attention module through an attention mechanism;

[0012] The image data of the steel structure plate to be detected is input into the trained fault detection model, and the steel structure plate fault detection results are output.

[0013] In one embodiment of the present invention, the LSTM in the fault detection model processes data information, and the architecture input x of the LSTM t 、output h t , Update gate i t 、Forget gate t , output gate o t ,in:

[0014] Forget Gate t The expression of is as follows:

[0015] f t =σ(W f *[h t-1 , x t ]+b f )

[0016] Update Gate t The expression of is as follows:

[0017] i t =σ(W i *[h t-1 , x t ]+b i )

[0018] c′ t =tanh(W c *[h t-1 , x t ]+b c )

[0019] c t =f t *c t-1 +i t *c′ t

[0020] Output gate o tThe expression of is as follows:

[0021] o t =σ(W o *[h t-1 , x t ]+b o )

[0022] h t =o t *tanh(c t )

[0023] Among them, W f , W i , W c , W o Represents the weight matrix b f 、b i 、b c 、b o represents the bias term; σ represents the sigmoid function, which maps the value to the interval [0,1]; tanh() represents the hyperbolic tangent function, which is used to standardize the data range to [-1,1]; c′ t Represents the candidate cell state, calculates the new information content at the current moment, and updates gate i t Combine to update the cell state.

[0024] In one embodiment of the present invention, the attention mechanism adopts the transformer attention mechanism.

[0025] In one embodiment of the present invention, the fault detection model uses an Adam optimizer.

[0026] In one embodiment of the present invention, the loss function of the fault detection model is as follows:

[0027]

[0028] Where N is the length of the input sequence, is the predicted value of the model, y i is the true value.

[0029] A second aspect of the present invention provides a device, comprising a processor and a memory, wherein the processor is electrically connected to the memory, and the memory is provided with a computer program, and when the computer program is executed by the processor, the steel structure plate fault detection method as described in any one of claims 1-5 is implemented.

[0030] A third aspect of the present invention provides a computer-readable storage medium, comprising a computer program, which, when executed, implements the steel structure plate fault detection method as described in any one of claims 1 to 5.

[0031] Compared with the prior art, the technical effects achieved by the present invention are as follows:

[0032] 1. The structural design of the steel structure plate fault detection model based on LSTM in the present invention effectively alleviates the gradient problem of traditional RNN when processing long sequence data, so that the model can better process longer and more complex defect image sequences. Therefore, this scheme has stronger adaptability and robustness when detecting continuous and complex defect features of steel structure plates;

[0033] 2. The present invention adds the transformer attention mechanism, which can effectively process long sequence data and capture long-distance dependencies. In order to optimize the computational efficiency of the self-attention mechanism, the input embedding is usually quantized and pruned to reduce the computational complexity.

[0034] 3. The present invention is committed to solving the existing technical challenges in the field of steel structure plate fault detection. By constructing a comprehensive and detailed data set (which is divided into a training set and a validation set), it provides a rich learning and validation resource for the model. Multiple rounds of training and validation processes are used to make full use of these data resources, aiming to significantly improve the generalization ability of the model. This means that the model that has undergone such rigorous training will be able to adapt to various data sets and application scenarios more robustly, and will show good detection performance regardless of different defect types, complex background environments, or changing lighting conditions;

[0035] 4. By inputting the fault image of the steel structure plate to be detected into a pre-trained model, the system can automatically output the fault detection result quickly and accurately. This automated detection process greatly improves the detection efficiency and makes the detection process faster and more convenient. Compared with the traditional manual detection method, the present invention significantly reduces the need for manual intervention. This not only reduces labor costs, but also avoids detection errors that may be caused by human factors, and improves the accuracy and reliability of detection. Especially when facing large-scale or high-frequency detection tasks, the automated detection method of the present invention has shown incomparable advantages. It can quickly process large amounts of image data to ensure that every detail is not missed, thereby effectively ensuring the quality and safety of steel structure plates.

[0036] 5. The present invention has demonstrated the ability to significantly improve the efficiency and accuracy of steel structure plate fault detection in practical applications. Through its advanced detection technology and algorithms, the present invention effectively reduces the risk of missed detection and false detection, ensuring the stability and reliability of product quality. This means that in the production process, any tiny defect will hardly escape its keen "eyes", thereby ensuring that each product meets high quality standards. More importantly, the present invention can not only detect defects in a timely and accurate manner, but also provide strong support for quality control in subsequent production. Once a defect is discovered, the production department can quickly take measures to regulate and improve it to avoid similar problems from happening again. This timely feedback mechanism helps to improve production efficiency and quality control capabilities, making the production process more efficient and orderly. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 is a flow chart of a steel structure plate fault detection method according to an embodiment of the present invention;

[0038] Figure 2 is a block diagram of an LSTM unit structure according to an embodiment of the present invention;

[0039] Figure 3 is a structural block diagram of a transformer module according to an embodiment of the present invention. DETAILED DESCRIPTION

[0040] The present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the relevant invention, rather than to limit the invention. It is also necessary to explain that, for ease of description, only the parts related to the invention are shown in the accompanying drawings.

[0041] See also Figures 1 to 3 The present invention proposes a steel structure plate fault detection method, comprising the following steps:

[0042] S1. Construct a data set based on steel structure plate image data, wherein the data set includes a training set and a validation set.

[0043] In some instances, the steel structure plate image data may be derived from pictures or videos taken on site, or from an image database. The faults shown in the steel structure plate images may be caused by the process or by human factors, or even by adding image blocks to the fault-free steel structure plate images to indicate the existence of faults.

[0044] S2. Use the training set and the test set to train and test the fault detection model in sequence, wherein the fault detection model is based on LSTM and is equipped with an attention module through an attention mechanism.

[0045] By adjusting the parameters and function configuration, an efficient network training process can be constructed. This process continuously fine-tunes the parameters of the steel structure plate fault detection model according to the preset training rounds (epochs) until the predetermined performance threshold is reached. This means that even if the original training rounds are not reached, once the model performance reaches the threshold, the training process can be terminated early. In addition, if the loss function is not significantly optimized after 3 rounds of training, the training of the steel structure plate fault detection model will be terminated.

[0046] Long Short Term Memory Network (LSTM) is a recurrent neural network used to process data information. The architecture of LSTM includes input x t 、output h t , Update gate i t 、Forget gate t , output gate o t ,in:

[0047] Forget Gate t The expression of is as follows:

[0048] f t =σ(W f *[h t-1 , x t ]+b f )

[0049] Update Gate t The expression of is as follows:

[0050] i t =σ(W i *[h t-1 , x t ]+b i )

[0051] c′ t =tanh(W c *[h t-1 , x t ]+b c )

[0052] c t =f t *c t-1 +i t *c′ t

[0053] Output gate o t The expression of is as follows:

[0054] o t =σ(W o *[ht-1 , x t ]+b o )

[0055] h t =o t *tanh(c t )

[0056] Among them, W f , W i , W c , W o Represents the weight matrix b f 、b i 、b c 、b o represents the bias term; σ represents the sigmoid function, which maps the value to the interval [0, 1]; tannhh() represents the hyperbolic tangent function, which is used to standardize the data range to [-1, 1]; c′ t Represents the candidate cell state, calculates the new information content at the current moment, and updates gate i t Combine to update the cell state.

[0057] Finally, the output value of the unit is determined, and the output gate calculates h t The sigmoid layer determines which part of the cell state to output, and the tanh layer processes it to obtain a value between -1 and 1, and finally determines the output.

[0058] Candidate cell state c′ t The introduction of t Dynamically update the current cell state c t , realizing the fusion of new information and historical information. In this way, the model can not only retain long-term feature information, but also adapt to real-time changes in input, improving adaptability and accuracy to different defect types.

[0059] The structural design of LSTM effectively alleviates the gradient problem of traditional RNN when processing long sequence data, enabling the model to better process longer and more complex defect image sequences. Therefore, this solution has stronger adaptability and robustness when detecting continuous and complex defect features of steel structure plates.

[0060] Through the σ activation function sigmoid and the hyperbolic tangent function tanh, the model can finely control the output range of each layer gate (i.e. [0, 1] and [-1, 1]) to ensure the standardization of input and output data. This fine control reduces the risk of gradient disappearance or explosion, ensuring the stability of the model and the training effect.

[0061] The attention mechanism uses the transformer attention mechanism. The Transformer architecture mainly consists of an input part, an encoder, a decoder, and an output part: Figure 3 In this attention mechanism, Source is the input and Query represents an element in the output. Each element in Source is considered to be composed of a key-value pair of data. First, the correlation between the Query value and each Key value is calculated to obtain the weight coefficient. This weight coefficient represents the contribution of each Value value to the output. Next, the Value values ​​are weighted and summed to finally obtain the attention value corresponding to the Query. Therefore, the essence of the attention mechanism is to perform a weighted sum operation on the Value value, where the weight of the Value value is calculated through the interaction between the Query and the Key.

[0062] It is observed that the attention mechanism can selectively increase the weight of important information, allowing the model to focus on key content while ignoring less relevant information. The value here refers to the value of certain content or attributes in the input data.

[0063] S3, inputting the image data of the steel structure plate to be detected into the trained fault detection model, and outputting the steel structure plate fault detection result.

[0064] In some embodiments, the loss function of the fault detection model is as follows:

[0065]

[0066] Where N is the length of the input sequence, is the predicted value of the model, y i is the true value.

[0067] The loss function design of the steel structure plate fault detection model was optimized to accurately evaluate and minimize the prediction error of the model. In order to improve the efficiency of model error correction, the weights of large error terms were particularly enhanced. This approach ensures the continuous differentiability of the model training process and simplifies the optimization steps. In addition, by accelerating the convergence of the model, higher accuracy can be achieved in regression tasks. At the same time, this optimization strategy also effectively prevents the model from overfitting, thereby improving its generalization performance.

[0068] In some embodiments, an Adam optimizer is applied to the fault detection model. The Adam optimizer is a gradient-based optimization algorithm that can effectively optimize weight parameters and bias parameters in the steel structure plate fault detection model.

[0069] Optimizing parameters is mainly achieved through the following four key steps:

[0070] 1. First-order moment estimation (momentum mechanism): The Adam optimizer tracks the first-order momentum of the gradient of each parameter, i.e. momentum. This step helps speed up the optimization path, especially when encountering many flat areas, it can effectively push the parameters to move towards the optimal solution faster.

[0071] 2. Second-order moment estimation: At the same time, Adam also calculates the exponentially weighted average of the square of the gradient, that is, the second-order momentum. This step allows the optimizer to dynamically adjust the learning rate of each parameter according to the size of the gradient, ensuring that a cautious update step is taken in the direction of the larger gradient, and a larger update step is taken in the direction of the smaller gradient, thereby improving the efficiency of the optimization.

[0072] 3. Adaptive learning rate: Based on the estimation of first-order and second-order momentum, Adam can independently calculate an adaptive learning rate for each parameter. This mechanism enables the learning rate to be dynamically adjusted according to historical gradient information, ensuring the stability and efficiency of the update process.

[0073] 4. Bias correction: In the early stages of training, since the initial values ​​of the first-order and second-order momentum may be close to zero, Adam introduces a bias correction technique to eliminate this effect. This step ensures that the model can be updated with the correct step size in the early stages of training.

[0074] In steel structure plate fault detection, the Adam optimizer shows the following advantages:

[0075] 1. Adaptive learning rate: The Adam optimizer has the unique ability to dynamically adjust the learning rate based on the gradient history of each parameter. This feature is particularly important when dealing with highly complex and sophisticated tasks such as steel structure plate fault detection. By dynamically adjusting the learning rate, Adam can not only effectively respond to the different learning needs of different parameters during the training process, but also significantly reduce the risk of training instability that may be caused by a fixed learning rate. This flexible learning rate adjustment mechanism enables the model to converge more robustly during the training process, thereby improving the accuracy and reliability of steel structure plate fault detection.

[0076] 2. Effective convergence: The Adam optimizer demonstrates its powerful ability to quickly converge to a better solution in the early stages of training by cleverly combining momentum and squared weighted average of gradients. This feature enables Adam to perform better than traditional gradient descent methods in dealing with complex and changeable training tasks, especially in scenarios where sparse gradients or gradients change dramatically. The introduction of momentum helps the model maintain historical gradient information when updating parameters, thereby accelerating the convergence process; while the squared weighted average of the gradient effectively adjusts the learning rate, allowing the model to remain stable when the gradient changes greatly, avoiding oscillations during training. Therefore, the Adam optimizer can significantly improve the training efficiency and detection accuracy of the model when dealing with complex tasks such as steel structure plate fault detection.

[0077] 3. Bias correction mechanism: The Adam optimizer demonstrates its unique advantages in the early stages of training by introducing a bias correction mechanism. This mechanism ensures that the model can converge to the global optimal solution faster during the parameter update process, thereby significantly improving the efficiency and accuracy of training. Bias correction reduces the bias caused by inaccurate estimates of the momentum term and the squared gradient term in the early stages of training, thereby promoting a smooth transition of the model during training. This feature of the Adam optimizer is particularly important when dealing with complex and variable training tasks, such as steel structure plate fault detection. In the face of sparse gradients or drastic gradient changes, traditional gradient descent methods may cause unstable training because they cannot effectively cope with these challenges. The Adam optimizer can maintain the stability of training in these scenarios by dynamically adjusting the learning rate and combining the bias correction mechanism to prevent the model from falling into a local optimal solution or oscillation. Therefore, the Adam optimizer, with its bias correction mechanism, dynamic learning rate adjustment, and the combination of momentum and squared gradient weighted average, has become the preferred method for handling complex training tasks, providing strong support for improving model performance, accelerating training speed, and improving training stability.

[0078] 4. Applicable to large-scale data sets: Steel structure plate fault detection is a highly complex and sophisticated task, which usually involves the processing and analysis of a large amount of image data. In this process, the Adam optimizer demonstrates its excellent performance and unique advantages. First, the Adam optimizer is able to efficiently process these large-scale data sets. By dynamically adjusting the learning rate and combining the square weighted average of momentum and gradient, Adam can significantly improve the convergence speed of the model while ensuring training stability. This means that in the same amount of time, more image data can be processed using the Adam optimizer, thereby speeding up the training process and shortening the model development cycle. Secondly, the Adam optimizer also performs well in reducing the risk of overfitting. Overfitting is a common problem when dealing with large-scale data sets, which may cause the model to perform well on the training set but degrade on the test set. The Adam optimizer, through its unique bias correction mechanism and dynamic learning rate adjustment strategy, can effectively balance the complexity and generalization ability of the model during the training process, thereby reducing the risk of overfitting. Therefore, when dealing with tasks involving large amounts of image data, such as steel structure plate fault detection, the Adam optimizer can not only significantly reduce training time and improve training efficiency, but also reduce the risk of overfitting and improve the generalization performance of the model. This makes the Adam optimizer an ideal choice for dealing with such complex tasks.

[0079] 5. High stability: Due to its combination of momentum and adaptive adjustment mechanism, Adam shows higher stability and reliability in variable gradients and complex loss spaces. This makes it perform well in the training process of steel structure plate fault detection model.

[0080] According to an embodiment of the present invention, the target object can be evaluated at any time during the life or operation stage of any steel structure plate. In some embodiments, one or more systems and methods described herein can be used to implement quality assurance testing on one or more target objects, such as selecting a representative sample for quality testing after a large number of such target objects are manufactured or assembled to evaluate whether there are defects on their surface. For example, when mass-producing target objects according to a general process, a certain number of target objects can be randomly selected for photographing to verify whether there are defects on the surface of the target objects produced under the process. In addition, the target object can also be inspected regularly or irregularly, or after a specific event. For example, when the target object is in use, images of one or more aspects of it can be taken and processed to determine whether the target object is in good condition, or whether maintenance, repair or further inspection is required. If necessary, the target object can also be repeatedly inspected multiple times according to the specific situation, such as after each use, between any two uses, or according to a predetermined schedule.

[0081] According to another embodiment of the present invention, the device comprises a processor and a memory, wherein the processor is electrically connected to the memory, and the memory is provided with a computer program, and when the computer program is executed by the processor, the steel structure plate fault detection method is implemented.

[0082] A computer-readable storage medium according to another embodiment of the present invention includes a computer program, and when the computer program is executed, the steel structure plate fault detection method is implemented.

[0083] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example" or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In addition, the specific features, structures, materials or characteristics described may be combined in an appropriate manner in any one or more embodiments or examples. In addition, those skilled in the art may connect and combine the various embodiments or examples described in the specification, as well as the features of the various embodiments or examples, without contradiction.

[0084] In addition, the terms "first", "second", etc. are used for description only and should not be interpreted as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one feature. In the description of the present invention, unless otherwise clearly and specifically defined, the meaning of "multiple" is two or more.

[0085] Any process or method description in the flowchart or other descriptions herein can be understood as a module, segment or code portion representing executable instructions, including one or more steps for implementing a specific logical function or process. And the scope of the preferred embodiment of the present invention includes additional implementations, wherein, depending on the functions involved, these functions can be performed in a substantially simultaneous manner or in a reverse order relative to the order shown or discussed, which should be understood by those skilled in the art to which the embodiments of the present invention relate.

[0086] The logic and / or steps represented in the flowchart or other content described herein. For example, it can be considered as an ordered list of executable instructions for implementing logical functions, which can be contained in any computer-readable medium for use by an instruction execution system, device or apparatus (a computer-based system, a system including a processor, or other system that can obtain instructions from an instruction execution system and execute the instructions), device or apparatus), or in conjunction with these instructions to execute the system, device or apparatus. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transport the program for use in an instruction execution system, device or apparatus, or in conjunction with such an instruction execution system, device or apparatus. More specific examples of computer-readable media (non-exhaustive list) include: electrical connections with one or more wires (electronic devices), portable computer disk cartridges (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), fiber optic devices, and portable read-only memory (CDROM). Furthermore, the computer-readable medium may even be a paper or other suitable medium on which the program may be printed, since the program may be obtained electronically, for example by optically scanning the paper or other medium, and then edited, interpreted or processed in another suitable manner if necessary, and then stored in a computer memory.

[0087] It should be understood that parts of the present invention may be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods may be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or combination of the following known in the art may be used to implement: a discrete logic circuit having logic gates for implementing logic functions on data signals, an application-specific integrated circuit having appropriate combination logic gates, a programmable gate array (PGA), a field programmable gate array (FPGA), and the like.

[0088] Those skilled in the art will appreciate that all or part of the steps carried by the method for implementing the above-mentioned embodiment can be completed by program instructions related to hardware, and the program can be stored in a computer-readable storage medium. When executed, it includes one or a combination of the steps of the method embodiment.

[0089] In addition, the functional units in the embodiments of the present invention may be integrated into one processing module, or each unit may exist physically independently, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may be stored in a computer-readable storage medium. The storage medium may be a read-only memory, a disk, or an optical disk, etc.

[0090] The above is only a specific implementation of the present invention and does not limit the protection scope of the present invention. Any person skilled in the art can easily think of fabrications and changes within the technical scope of the present invention, all of which should be included in the protection scope of the present invention. Therefore, the protection scope of the claims should be taken as the protection scope of the present invention.

Claims

1. A method for detecting steel structure plate faults, characterized in that: The following steps are involved: Building a data set based on steel structure plate image data, the data set includes a training set and a validation set; The fault detection model is trained and tested using the training set and the test set in turn, wherein the fault detection model is based on LSTM and is equipped with an attention module through an attention mechanism; The image data of the steel structure plate to be detected is input into the trained fault detection model, and the steel structure plate fault detection results are output.

2. The steel structure plate fault detection method according to claim 1, characterized in that: In the fault detection model, LSTM processes data information. The architecture input x t 、output h t , Update gate i t 、Forget gate t , output gate o t ,in: Forget Gate t The expression of is as follows: f t =σ(W f *[h t-1 ,x t ]+b f ) Update Gate t The expression of is as follows: i t =σ(W i *[h t-1 ,x t ]+b i ) c' t =tanh(W c *[h t-1 ,x t ]+b c ) c t =f t *c t-1 +i t *c′ t Output gate o t The expression of is as follows: the t =σ(W o *[h t-1 ,x t ]+b o ) h t =o t *tanh(c t ) Among them, W f , W i , W c , W o represents the weight matrix; b f , b i , b c , b o represents the bias term; σ represents the sigmoid function, which maps the value to the interval [0, 1]; tanh() represents the hyperbolic tangent function, which is used to standardize the data range to [-1, 1]; c′ t Represents the candidate cell state, calculates the new information content at the current moment, and updates gate i t Combine to update the cell state.

3. The steel structure plate fault detection method according to claim 1, characterized in that: The attention mechanism adopts the transformer attention mechanism.

4. The steel structure plate fault detection method according to claim 1, characterized in that: The fault detection model uses the Adam optimizer.

5. The steel structure plate fault detection method according to claim 1, characterized in that: The loss function of the fault detection model is shown as follows: Where N is the length of the input sequence, is the predicted value of the model, y i is the true value.

6. A device, characterized in that: It comprises a processor and a memory, the processor is electrically connected to the memory, the memory is provided with a computer program, and when the computer program is executed by the processor, the steel structure plate fault detection method as described in any one of claims 1 to 5 is implemented.

7. A computer-readable storage medium, comprising a computer program, characterized in that: When the computer program is executed, the steel structure plate fault detection method as described in any one of claims 1 to 5 is implemented.

Images