Fan blade fault diagnosis model construction and fault diagnosis method and device

Abstract

The invention relates to the technical field of fan blade fault diagnosis, and discloses fan blade fault diagnosis model construction and fault diagnosis methods and devices, and the fan blade fault diagnosis model construction method comprises the steps: constructing a fan blade finite element model according to geometric data, material data and load boundary data of a target fan blade; simulating a blade fault by using the fan blade finite element model to obtain a fault simulation vibration signal and fault strain data; the fault and actual simulation vibration signals and the fault and actual strain data are input into a fan blade fault diagnosis model for multi-scale fault feature extraction, and multi-scale fault features are obtained; based on the local maximum mean value difference loss and the classification loss, the multi-scale fault features are input to a classification module of the fan blade fault diagnosis model for training, the fault diagnosis model is trained through simulation data, strain data and multi-scale feature extraction, and the accuracy of the fault diagnosis model is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fan blade fault diagnosis, and particularly relates to a fan blade fault diagnosis model construction method, a fault diagnosis method and a device. BACKGROUND

[0002] With the rapid development of wind power generation, the running reliability of fan blades, as core components of wind power generators, is directly related to the safety and economic benefits of wind farms. However, fan blades are prone to faults such as cracks and icing during long-term operation, and the downtime loss caused by faults accounts for 15% to 25% of the operation and maintenance cost of the wind farm. Therefore, it is of great significance to establish an efficient and accurate fault diagnosis system to ensure the safe operation of equipment and reduce operation and maintenance costs.

[0003] In related technologies, the method for diagnosing faults of fan blades is to extract features according to spectrum analysis or time domain statistics, combine a machine learning model for classification, and thus diagnose faults of fan blades. However, the deep neural network of this method needs a large amount of labeled data for training, and actual fault samples are scarce, which leads to overfitting of the model and insufficient cross-domain generalization ability, the feature extraction process is difficult to capture multi-scale fault information, and is prone to noise interference, thus leading to inaccurate results of diagnosing faults of fan blades and affecting the stable operation of fan blades. SUMMARY

[0004] The present application provides a fan blade fault diagnosis model construction method and a fault diagnosis method and device to solve the problem of inaccurate results of diagnosing faults of fan blades in related technologies.

[0005] In a first aspect, the present application provides a fan blade fault diagnosis model construction method, comprising: constructing a fan blade finite element model according to geometric data, material data and load boundary data of a target fan blade; the fan blade finite element model is used to simulate the performance of the target fan blade under different working conditions; using the fan blade finite element model, simulating different types of blade faults to obtain fault simulation vibration signals and fault strain data; inputting the fault simulation vibration signals, the fault strain data, actual fault vibration signals and actual strain data into a feature extraction module of a fan blade fault diagnosis model for multi-scale fault feature extraction to obtain multi-scale fault features; based on local maximum mean difference loss and classification loss, inputting the multi-scale fault features into a classification module of the fan blade fault diagnosis model for training to obtain a target fan blade fault diagnosis model.

[0006] The fan blade fault diagnosis model construction method of the application, according to the geometric data, material data and load boundary data of the target fan blade, constructs a fan blade finite element model for simulating the performance of the target fan blade under different working conditions, accurately and comprehensively restores the actual working state of the fan blade, and lays a precise model foundation for subsequent fault simulation and diagnosis. The application uses the fan blade finite element model to simulate different types of blade faults, obtains fault simulation vibration signals and fault strain data, uses the fault simulation vibration signals as source domain data, effectively overcomes the problem of scarcity of actual fault data in related technologies, can systematically and comprehensively obtain the vibration characteristics of various faults, and can master the signal performance under different faults in advance without relying on the occurrence of actual faults. The application inputs the fault simulation vibration signals, the fault strain data, the actual fault vibration signals and the actual strain data into the feature extraction module of the fan blade fault diagnosis model for multi-scale fault feature extraction, obtains multi-scale fault features, captures different scale fault features, avoids the limitation of single scale feature extraction, makes the extracted fault features more comprehensive and more representative, and helps to improve the accuracy of subsequent fault classification. The application inputs the multi-scale fault features into the classification module of the fan blade fault diagnosis model for training to obtain the target fan blade fault diagnosis model, introduces the local maximum mean difference loss and the classification loss, breaks through the limitation of global domain adaptation, improves the cross-domain generalization ability of the model through fine-grained alignment, and improves the precision and reliability of fault diagnosis. Compared with related technologies, the application generates rich source domain data through simulation vibration signals, reduces the dependence on physical test benches, realizes accurate extraction of multi-scale fault features, can efficiently and accurately realize fan blade fault diagnosis, provides strong technical support for fault warning and maintenance decision of the target fan blade, helps to improve the reliability and service life of the fan blade, and ensures the stable operation of the wind turbine generator.

[0007] In an optional embodiment, constructing a fan blade finite element model according to geometric data, material data and load boundary data of a target fan blade comprises: constructing a fan blade three-dimensional model according to the geometric data of the target fan blade; defining blade material properties of the fan blade three-dimensional model according to the material data to obtain a fan blade three-dimensional model with blade material properties; performing mesh division on the fan blade three-dimensional model with blade material properties and adding constraint conditions according to the load boundary data to obtain a discrete numerical model; and performing solving parameter configuration on the discrete numerical model to obtain the fan blade finite element model.

[0008] In an optional implementation, the finite element model of the fan blade is used to simulate different types of blade faults to obtain fault simulation vibration signals and fault strain data, including: inputting fault features of each type of blade fault into the finite element model of the fan blade, simulating each type of blade fault to obtain a target fan blade finite element model corresponding to each type of blade fault; collecting simulation signals and stress changes at key detection points of the target fan blade finite element model corresponding to each type of blade fault; obtaining fault simulation vibration signals and fault strain data according to the simulation signals and stress changes corresponding to the multiple types of blade faults.

[0009] In an optional implementation, the feature extraction module includes a multi-scale convolution unit and a channel attention unit; the fault simulation vibration signals, the fault strain data, the actual fault vibration signals, and the actual strain data are input into the feature extraction module of the fan blade fault diagnosis model for multi-scale fault feature extraction to obtain multi-scale fault features, including: inputting the fault simulation vibration signals, the fault strain data, the actual fault vibration signals, and the actual strain data into the multi-scale convolution unit for multi-scale analysis to obtain a multi-scale feature map; inputting the multi-scale feature map into the channel attention unit to extract features from the multi-scale feature map to obtain multi-scale fault features.

[0010] In an optional implementation, the multi-scale feature map is input into the channel attention unit to extract features from the multi-scale feature map to obtain multi-scale fault features, including: compressing global information of the multi-scale feature map to obtain a channel descriptor vector; performing nonlinear transformation on the channel descriptor vector to obtain a nonlinear vector; performing normalization processing on the nonlinear vector to obtain a channel scaling coefficient vector; multiplying the multi-scale feature map and the channel scaling coefficient vector channel by channel to obtain multi-scale fault features.

[0011] In an optional implementation, the multi-scale fault features are input into the classification module of the fan blade fault diagnosis model based on the local maximum mean difference loss and the classification loss for training to obtain a target fan blade fault diagnosis model, including: constructing a first target function according to the maximum mean difference loss and the classification loss; the first target function is used to minimize the distribution difference between the source domain fault features and the target domain features in the multi-scale fault features; using the first target function to perform overall domain alignment processing on the multi-scale fault features; constructing a second target function according to the local maximum mean difference loss and the classification loss; the second target function is used to represent the distribution difference between the same type of sub-domain of the source domain fault features and the target domain features in the multi-scale fault features; based on the second target function, the parameters of the feature extraction module and the classification module are iteratively optimized until a preset iteration termination condition is reached to obtain the target fan blade fault diagnosis model.

[0012] In a second aspect, the present application provides a method for diagnosing faults of a fan blade, comprising: obtaining a target fault vibration signal and target strain data of a target fan blade; the target fault vibration signal is a fault vibration signal obtained by signal collection on a fault part of the target fan blade, and the target strain data is data of force deformation of the target fan blade obtained by monitoring the fault part of the target fan blade; inputting the target fault vibration signal and the target strain data into a target fan blade fault diagnosis model constructed by the method for constructing a fan blade fault diagnosis model according to any one of the first aspect and the first aspect of the embodiment of the present application to obtain a fault diagnosis result; the input of the target fan blade fault diagnosis model is the target fault vibration signal, and the output of the target fan blade fault diagnosis model is the fault diagnosis result.

[0013] The method for diagnosing faults of a fan blade of the present application obtains a fault diagnosis result through a target fan blade fault diagnosis model, and realizes efficient and accurate diagnosis of faults of the target fan blade.

[0014] In a third aspect, the present application provides a device for constructing a fan blade fault diagnosis model, comprising: a model construction module, configured to construct a fan blade finite element model according to geometric data, material data and load boundary data of a target fan blade; the fan blade finite element model is used to simulate the performance of the target fan blade under different working conditions; a fault simulation module, configured to simulate different types of blade faults by using the fan blade finite element model to obtain fault simulation vibration signals and fault strain data; a feature extraction module, configured to input the fault simulation vibration signals, the fault strain data, actual fault vibration signals and actual strain data into a feature extraction module of a fan blade fault diagnosis model to extract multi-scale fault features; and a model training module, configured to input the multi-scale fault features into a classification module of the fan blade fault diagnosis model based on local maximum mean difference loss and classification loss to train the fan blade fault diagnosis model and obtain a target fan blade fault diagnosis model.

[0015] In a fourth aspect, the present application provides a device for diagnosing faults of a fan blade, comprising: a data acquisition module, configured to obtain a target fault vibration signal and target strain data of a target fan blade; the target fault vibration signal is a fault vibration signal obtained by signal collection on a fault part of the target fan blade, and the target strain data is data of force deformation of the target fan blade obtained by monitoring the fault part of the target fan blade; and a fault diagnosis module, configured to input the target fault vibration signal and the target strain data into a target fan blade fault diagnosis model constructed by the method for constructing a fan blade fault diagnosis model according to any one of the first aspect and the first aspect of the embodiment of the present application to obtain a fault diagnosis result; the input of the target fan blade fault diagnosis model is the target fault vibration signal, and the output of the target fan blade fault diagnosis model is the fault diagnosis result.

[0016] In a fifth aspect, the present application provides an electronic device, comprising a memory and a processor, the memory and the processor being communicatively connected with each other, the memory storing computer instructions, and the processor executing the computer instructions to perform the fan blade fault diagnosis model construction method of the first aspect or any of the corresponding embodiments thereof or the fan blade fault diagnosis method of the second aspect.

[0017] In a sixth aspect, the present application provides a computer readable storage medium, the computer readable storage medium storing computer instructions, the computer instructions being used to make a computer execute the fan blade fault diagnosis model construction method of the first aspect or any of the corresponding embodiments thereof or the fan blade fault diagnosis method of the second aspect.

[0018] In a seventh aspect, the present application provides a computer program product, comprising computer instructions, the computer instructions being used to make a computer execute the fan blade fault diagnosis model construction method of the first aspect or any of the corresponding embodiments thereof or the fan blade fault diagnosis method of the second aspect. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the drawings needed in the specific embodiments or prior art description will be briefly introduced as follows. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0020] Figure 1 is a schematic diagram of an application scenario according to an embodiment of the present application; Figure 2 is a first flowchart of the fan blade fault diagnosis model construction method according to an embodiment of the present application; Figure 3 is a second flowchart of the fan blade fault diagnosis model construction method according to an embodiment of the present application; Figure 4 is a first flowchart of the fan blade fault diagnosis method according to an embodiment of the present application; Figure 5 is a second flowchart of the fan blade fault diagnosis method according to an embodiment of the present application; Figure 6 is a structural block diagram of the fan blade fault diagnosis model construction device according to an embodiment of the present application; Figure 7 is a structural block diagram of the fan blade fault diagnosis device according to an embodiment of the present application; Figure 8 is a schematic diagram of a hardware structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some, but not all, of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts should fall within the scope of the present application.

[0022] It can be understood that, before using the technical solutions disclosed in the embodiments of the present application, the type, use range, use scenario, and the like of personal information involved in the present application should be informed to users and the authorization of the users should be obtained in a proper manner according to relevant laws and regulations.

[0023] The terms "first", "second", and the like are used only for the purpose of description, and should not be understood as indicating or implying relative importance or implicitly indicating the number of the technical features indicated. Therefore, the features defined with "first", "second" can explicitly or implicitly include one or more of the features. In the description of the present application, the meaning of "multiple" is two or more, unless otherwise explicitly and specifically limited.

[0024] As an optional application scenario of the embodiments of the present application, as shown in Figure 1 the fan blade fault diagnosis model construction system or the fan blade fault diagnosis system can include at least one terminal device and at least one server, Figure 1 the system includes a computer 101, a mobile terminal 102, and a server 103, and the terminal devices such as the computer 101 and the mobile terminal 102 are connected to the server 103 through a network 110.

[0025] The terminal device can be specifically a smart phone, a tablet computer, a notebook computer, a palm computer, and can also be a desktop computer, a game console, a smart television, a smart wearable device, a vehicle-mounted terminal, a VR (Virtual Reality) device, an AR (Augmented Reality) device, and the like. The server 103 can be an independent physical server, or a server cluster or a distributed system, or a cloud server providing cloud services. The network 110 can be a wired network or a wireless network, and examples thereof include but are not limited to the Internet, an enterprise intranet, a local area network, a wide area network, a mobile communication network, and combinations thereof.

[0026] The fault diagnosis method of the fan blade in the related art is usually based on signal analysis, which depends on spectrum analysis or time domain statistics to extract features, and combines a machine learning model for classification. Such a method performs well in a scenario where data is sufficient and working conditions are unchanged, but faces severe challenges in actual application: a deep neural network needs a large amount of labeled data for training, and actual fault samples are scarce, which leads to overfitting of the model and insufficient cross-domain generalization capability; the feature extraction method is difficult to capture multi-scale fault information and is susceptible to noise interference; and an unsupervised domain adaptation method focuses on global distribution alignment and ignores sub-domain fine-grained differences.

[0027] The embodiment of the present application provides a fan blade fault diagnosis model construction method, which trains the fault diagnosis model through simulation data and multi-scale feature extraction, so as to improve the accuracy of the fault diagnosis model.

[0028] According to the embodiment of the present application, a fan blade fault diagnosis model construction method embodiment is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that here.

[0029] In the present embodiment, a fan blade fault diagnosis model construction method is provided, which can be used in a computer device, Figure 2 The first flowchart of the fan blade fault diagnosis model construction method according to the embodiment of the present application is shown in FIG. 1, which includes the following steps: Figure 2 Step S201, constructing a fan blade finite element model according to the geometric data, material data and load boundary data of the target fan blade; the fan blade finite element model is used to simulate the performance of the target fan blade under different working conditions.

[0030] Wherein, the target fan blade is a fan blade that needs to be diagnosed for faults; the geometric data is data used to describe the geometric features of the shape, size and structure of the fan blade, for example, the geometric data includes the length, width, thickness and curved surface shape of the fan blade; the material data is data about the performance parameters of the material used for the fan blade, for example, the material data includes the elastic modulus, density, Poisson's ratio and strength of the material; the load boundary data is data about the external force and boundary constraint conditions of the target fan blade when it is working, for example, wind load, centrifugal force load, gravity load and constraints at the connection part of the blade and the hub; the fan blade finite element model is a digital model constructed by using the finite element analysis method in combination with the geometric data, material data and load boundary data of the target fan blade, which can be used to simulate the mechanical properties and vibration characteristics of the blade under different working conditions.

[0031] ​In step S202, the finite element model of the fan blade is used to simulate different types of blade faults to obtain fault simulation vibration signals and fault strain data.

[0032] The different types of blade faults include crack faults, damage faults, aerodynamic imbalance faults, icing faults, and the like.

[0033] In some optional embodiments, by simulating different types of blade faults, vibration signals of key detection points are collected to obtain fault simulation vibration signals, and the target fan blade at the key detection point is monitored to obtain fault strain data.

[0034] In step S203, the fault simulation vibration signals, the fault strain data, actual fault vibration signals, and actual strain data are input into a feature extraction module of a fan blade fault diagnosis model to perform multi-scale fault feature extraction to obtain multi-scale fault features.

[0035] The actual fault vibration signals are vibration signals collected by a sensor when the target fan blade is tested in a real experimental environment, and the actual strain data are fan strain data collected by a sensor when the target fan blade is tested in a real experimental environment.

[0036] In some optional embodiments, the fault simulation vibration signals and the fault strain data are source domain data, and the actual fault vibration signals and the actual strain data are target domain data.

[0037] In some optional embodiments, the fan blade fault diagnosis model includes a feature extraction module and a classification module, and the feature extraction module includes a multi-scale convolution unit and a channel attention unit.

[0038] In step S204, based on a local maximum mean discrepancy loss and a classification loss, the multi-scale fault features are input into a classification module of the fan blade fault diagnosis model to perform training to obtain a target fan blade fault diagnosis model.

[0039] The local maximum mean discrepancy (LMMD) loss is a loss used to measure the difference between different data distributions, and is used to reduce the distribution difference between the source domain (such as fault simulation data) and the target domain (such as actual fault data); the classification loss is used to measure the difference between the model classification result and the true label, and the core function is to guide the fan blade fault diagnosis model to learn accurate classification ability.

[0040] The fan blade fault diagnosis model construction method provided by the embodiment is used for simulating the performance of the target fan blade under different working conditions, accurately and comprehensively restores the actual working state of the fan blade, and lays a precise model foundation for subsequent fault simulation and diagnosis. The fan blade finite element model is used to simulate different types of blade faults, obtain fault simulation vibration signals and fault strain data, and the fault simulation vibration signals are used as source domain data, so that the problem of actual fault data scarcity in the related art is effectively overcome, vibration characteristics of various faults can be systematically and comprehensively obtained, and the signal performance under different faults can be grasped in advance without relying on the occurrence of actual faults. The fault simulation vibration signals, the fault strain data, the actual fault vibration signals and the actual strain data are input into the feature extraction module of the fan blade fault diagnosis model to perform multi-scale fault feature extraction, multi-scale fault features are obtained, different scale fault features are captured, the limitation of feature extraction under a single scale is avoided, the extracted fault features are more comprehensive and more representative, and the accuracy of subsequent fault classification is improved. The multi-scale fault features are input into the classification module of the fan blade fault diagnosis model for training to obtain the target fan blade fault diagnosis model. The local maximum mean difference loss and the classification loss are introduced, the limitation of global domain adaptation is broken, the model cross-domain generalization capability is improved through fine-grained alignment, and the precision and reliability of fault diagnosis are improved. Compared with the related art, the simulation vibration signals are used to generate rich source domain data, the dependence on a physical test bed is reduced, multi-scale fault features are accurately extracted, the fan blade fault diagnosis can be efficiently and accurately implemented, strong technical support is provided for fault warning and maintenance decision of the target fan blade, the reliability and service life of the fan blade are improved, and stable operation of the wind turbine generator is ensured.

[0041] A fan blade fault diagnosis model construction method is provided in the embodiment, which can be used for a computer device, Figure 3 A second flowchart of the fan blade fault diagnosis model construction method according to the embodiment of the application is shown in Figure 3 The flowchart includes the following steps. In step S301, a fan blade finite element model is constructed according to geometric data, material data and load boundary data of a target fan blade. The fan blade finite element model is used to simulate the performance of the target fan blade under different working conditions.

[0042] Specifically, step S301 includes the following steps. In step S3011, a three-dimensional model of the fan blade is constructed according to the geometric data of the target fan blade.

[0043] In some optional embodiments, the geometric data of the target wind turbine blade is input into three-dimensional software to draw a three-dimensional model of the wind turbine blade. For example, for a scaled wind turbine blade, standard airfoil data is imported into the three-dimensional software, the initial airfoil coordinates are proportionally enlarged according to the ratio of the chord length of the blade cross-section airfoil to the standard airfoil chord length, the airfoil coordinates are translated and rotated based on the aerodynamic center, and then the airfoil coordinates are translated along the chord line in the vertical direction. Each airfoil profile is drawn, the leading edge / trailing edge points are connected by a spline curve, and the wind turbine blade three-dimensional shell model is generated by segmented sweeping, i.e., the wind turbine blade three-dimensional model.

[0044] In step S3012, the blade material properties of the wind turbine blade three-dimensional model are defined according to the material data, and a wind turbine blade three-dimensional model with blade material properties is obtained.

[0045] In some optional embodiments, the wind turbine blade three-dimensional model and the material data are input into analysis software according to the analysis type and calculation requirements, the blade material properties are defined, and a wind turbine blade three-dimensional model with blade material properties is obtained.

[0046] In step S3013, the wind turbine blade three-dimensional model with blade material properties is meshed, and constraint conditions are added according to the load boundary data, and a discrete numerical model is obtained.

[0047] In some optional embodiments, since the blade is a thin-walled curved surface structure with equal thickness, shell elements (such as first-order reduced integration elements) are selected, and a free meshing strategy with a quadrilateral as the main element and a size of about 10 mm is adopted to perform meshing under the premise of ensuring calculation accuracy and efficiency.

[0048] In some optional embodiments, the blade is subjected to dynamically changing aerodynamic force, gravity, centrifugal force, and non-steady state load during operation. During analysis, simplification is performed based on the blade element momentum theory and the Saint-Venant principle. According to the load boundary data, the continuous load distribution is equivalent to the surface load applied by the blade element in segments, and full constraint boundary conditions are applied at the blade root to obtain a discrete numerical model.

[0049] In step S3014, the discrete numerical model is configured with solving parameters to obtain a wind turbine blade finite element model.

[0050] In some optional embodiments, large wind turbine blades are made of fiber-reinforced polymer matrix composites, which have orthogonal anisotropic properties. The blade layup structure is formed by stacking in a specific order (such as ±45° double symmetric laying, combined with glass fiber / epoxy skin and polyvinyl chloride foam core), and a wind turbine blade finite element model is obtained to optimize stiffness, reduce mass, and improve pressure stability.

[0051] In step S302, the finite element model of the fan blade is used to simulate different types of blade faults to obtain fault simulation vibration signals and fault strain data.

[0052] Specifically, the step S302 includes: In step S3021, the fault features of each type of blade fault are input into the finite element model of the fan blade, each type of blade fault is simulated, and the target finite element model of the fan blade corresponding to each type of blade fault is obtained.

[0053] In some optional embodiments, based on the established finite element grid model of the fan blade, features are introduced for different fault types and modeling analysis is performed.

[0054] In some optional embodiments, the plurality of types of blade faults include crack faults, damage faults, aerodynamic imbalance faults, and icing faults.

[0055] For example, for the crack fault, separate grid modeling is adopted, the severity is represented in three levels according to the crack length, and 20 positions are set on the windward / backwind surface and the leading / trailing edge of the blade root, middle, and tip. Each position is set with normal and three crack working conditions for modeling analysis to obtain the target finite element model of the fan blade corresponding to the crack fault; for the damage fault, the damage degree is simulated by reducing the local element elastic modulus (reduced to 1 / 5, 1 / 25, and 1 / 125 of the initial value respectively) (the lower the modulus, the more serious the damage), and the element selection position is referred to the crack fault for modeling analysis to obtain the target finite element model of the fan blade corresponding to the damage fault; for the aerodynamic imbalance fault, it is manifested as abnormal pitch angle of a single blade (deflection of 2.5°, 5°, or 7.5°), which leads to inconsistent aerodynamic load with other blades, and modeling analysis is performed to obtain the target finite element model of the fan blade corresponding to the aerodynamic imbalance fault; for the icing fault, low stiffness mass elements (thicknesses of 5 mm, 10 mm, and 15 mm representing the severity of icing) are added for modeling, and 5 positions of the blade root, middle, tip, leading edge, and trailing edge are mainly considered for modeling analysis to obtain the target finite element model of the fan blade corresponding to the icing fault.

[0056] In step S3022, simulation signals and stress changes are collected at key detection points of the target finite element model of the fan blade corresponding to each type of blade fault.

[0057] In some optional embodiments, a vibration sensor is used to collect simulation signals reflecting the actual fan blade fault state at the key detection points at a sampling frequency of 1024 Hz, including displacement and acceleration signals of the output point, and a stress sensor is used to collect the stress changes of the target fan blade at the key detection points.

[0058] Step S3023, according to the simulation signals corresponding to the stress changes of the multiple types of blade faults, fault simulation vibration signals and fault strain data are obtained.

[0059] In some optional embodiments, the fault simulation vibration signals are composed of the simulation signals corresponding to the multiple types of blade faults, and the fault strain data are composed of the stress changes corresponding to the multiple types of blade faults.

[0060] Step S303, the fault simulation vibration signals, the fault strain data, the actual fault vibration signals and the actual strain data are input into a feature extraction module of a fan blade fault diagnosis model to perform multi-scale fault feature extraction, and multi-scale fault features are obtained.

[0061] Specifically, the above step S303 includes: Step S3031, the fault simulation vibration signals, the fault strain data, the actual fault vibration signals and the actual strain data are input into a multi-scale convolution unit to perform multi-scale analysis, and a multi-scale feature map is obtained.

[0062] In some optional embodiments, the fan blade fault diagnosis model includes a feature extraction module and a classification module, and the feature extraction module includes a multi-scale convolution unit and a channel attention unit.

[0063] In some optional embodiments, the multi-scale convolution unit is used for multi-scale analysis of the fault simulation vibration signals and the actual fault vibration signals. For example, the multi-scale convolution unit includes three parallel branches, each branch is connected in series to form a cavity convolution layer, a batch normalization layer and a ReLU (Rectified Linear Unit) activation function layer. The cavity convolution layer expands the receptive field by inserting cavities (zero elements) into the convolution kernel, efficiently captures the implicit information in the input features, cooperates with the zero padding strategy, adopts different cavity rates for different branches, which are set to 2, 3 and 4 respectively, to balance the receptive field expansion and local information preservation, and analyze the input data from multiple scales. Through the multi-scale feature extraction structure, the input signal is analyzed from different scales, and the richness of the overall features is effectively improved.

[0064] For example, the multi-scale feature map can be represented as:

[0065] Wherein, is a feature map obtained by the multi-scale feature extraction structure, is a feature map of the i th channel, C is a dimension, is a channel number of the feature map, C is a width of the feature map. W ​

[0066] Step S3032: Input the multi-scale feature map into the channel attention unit, extract features from the multi-scale feature map, and obtain multi-scale fault features.

[0067] In some alternative implementations, the channel attention unit includes a global average pooling layer, two fully connected layers, a ReLU activation function, and a Sigmoid (normalized) activation function.

[0068] In some optional implementations, step S3032 above includes: Step a1: Compress the multi-scale feature map globally to obtain the channel descriptor vector.

[0069] In some optional implementations, a global average pooling layer is used to compress global information from multi-scale feature maps to obtain channel descriptor vectors. For example, the channel descriptor vector can be represented as:

[0070] in, For channel descriptor vectors, For the first C A channel descriptor vector for each channel. For dimensions.

[0071] In some alternative implementations, the first C The channel descriptor vector of each channel can be represented as:

[0072] in, For the first A channel descriptor vector for each channel. The width of the feature map. For the number of channels, For the first The first channel, the... A feature map of width 1.

[0073] Step a2: Perform a nonlinear transformation on the channel descriptor vector to obtain a nonlinear vector.

[0074] In some alternative implementations, the nonlinear vector can be represented as:

[0075] in, It is a non-linear vector. This is the weight matrix of the second fully connected layer. This is the weight matrix of the first fully connected layer. It is the ReLU activation function. A channel descriptor vector is described.

[0076] Step a3, normalizing the non-linear vector to obtain a channel scaling coefficient vector.

[0077] In some optional embodiments, the non-linear vector is normalized by using a Sigmoid activation function to obtain a channel scaling coefficient vector, which can be expressed as:

[0078] wherein, is the channel scaling coefficient vector, is the dimension, is the channel scaling coefficient vector of the i-th channel. C

[0079] In some optional embodiments, the channel scaling coefficient vector of the i-th channel can be expressed as: C wherein,

[0080] is the channel scaling coefficient vector of the i-th channel, is the non-linear vector of the i-th channel.

[0081] Step a4, multiplying the multi-scale feature map and the channel scaling coefficient vector channel by channel to obtain a multi-scale fault feature.

[0082] In some optional embodiments, the multi-scale fault feature can be expressed as:

[0083] wherein, is the multi-scale fault feature of the i-th channel, is the channel scaling coefficient vector of the i-th channel, is the feature map of the i-th channel. In some optional embodiments, the multi-scale fault feature is composed of multi-scale fault features of multiple channels, which can be expressed as , which can also be referred to as domain-invariant feature.

[0084] Step S304, based on the local maximum mean difference loss and the classification loss, inputting the multi-scale fault feature into a classification module of the fan blade fault diagnosis model for training to obtain a target fan blade fault diagnosis model.

[0085] Step S304, based on the local maximum mean difference loss and the classification loss, inputting the multi-scale fault feature into a classification module of the fan blade fault diagnosis model for training to obtain a target fan blade fault diagnosis model.

[0086] ​​​​​​Specifically, the step S304 includes: In step S3041, a first objective function is constructed according to the maximum mean discrepancy loss and the classification loss; the first objective function is used to minimize the distribution difference between the source domain fault feature and the target domain fault feature in the multi-scale fault feature.

[0087] The maximum mean discrepancy (MMD) loss is used to measure the difference between two probability distributions.

[0088] In some optional embodiments, the multi-scale fault feature includes the source domain fault feature and the target domain fault feature. Due to the domain shift caused by the change of wind conditions, the distribution difference exists between the source domain fault feature and the target domain fault feature, so that the wind turbine blade fault diagnosis model trained on the source domain is difficult to generalize to the target domain. Therefore, the overall domain alignment processing is performed based on the first objective function.

[0089] For example, the maximum mean discrepancy loss can be represented as:

[0090] wherein, is the maximum mean discrepancy loss, is the multi-scale fault feature of the source domain fault feature is the multi-scale fault feature of the target domain fault feature is the total number of source domain fault features, is the total number of target fault features, is the reproducing kernel Hilbert space, is a nonlinear mapping function for transforming the original sample to space, is a kernel function (usually a Gaussian kernel function), is the i-th fault feature in the source domain fault feature is the i-th fault feature in the target domain fault feature .

[0091] For example, the first objective function can be represented as:

[0092] wherein, is the first objective function, is the minimum value, is the total number of source domain fault features, is the classification loss, is the source domain fault feature​​​​​ the first failure feature in the source domain failure feature the first failure feature in the source domain failure feature the first non-failure feature in the source domain failure feature the classifier network the maximum mean discrepancy loss the first failure feature in the source domain failure feature the multi-scale failure feature of the source domain failure feature the multi-scale failure feature of the target domain failure feature the hyper-parameter for balancing the weights of different loss terms.

[0093] In some optional embodiments, the classification loss can be represented as:

[0094] wherein, the classification loss, the first failure feature in the source domain failure feature the first non-failure feature in the source domain failure feature the classifier network L the total number of classes of the source domain failure feature the i-th component output by the classifier, the binary discriminant function.

[0095] Step S3042, using the first objective function, performing overall domain alignment processing on the multi-scale failure feature.

[0096] wherein, the input to the first objective function, performing overall domain alignment processing on the multi-scale failure feature.

[0097] Step S3043, constructing a second objective function according to the local maximum mean discrepancy loss and the classification loss; the second objective function is used to represent the distribution difference of the same type of sub-domain between the source domain failure feature and the target domain feature in the multi-scale failure feature.

[0098] In some optional embodiments, after the first objective function analysis, the global domain adaptation can make the overall distribution of the two domains approximate, but it is easy to cause the different sub-domains to overlap too much and cause misjudgment, and the sub-domain adaptation method is introduced to learn the local domain offset, so as to realize the high consistency of the local distribution between the source domain failure feature and the target domain feature, and the approximate alignment of the global distribution, and the sub-domain adaptation target is set to directly align the same type of sub-domain between the source domain failure feature and the target domain feature. ​​​​​​​​

[0099] Specifically, a second objective function is constructed by introducing the local maximum mean difference loss and the classification loss to measure the distribution difference between the source domain fault features and the target domain features related subdomains. For example, the local maximum mean difference loss can be expressed as:

[0100] in, This represents the loss due to the local maximum mean difference. Source domain fault characteristics Multi-scale fault characteristics. Fault characteristics of the target domain Multi-scale fault characteristics. This represents the total number of source domain fault characteristics. The total number of target fault characteristics. For the regenerating nucleus Hilbert space, To transform the original sample to Nonlinear mapping function of space, The kernel function is used (Gaussian kernel function is usually chosen). Source domain fault characteristics The first in One fault characteristic, Fault characteristics of the target domain The first in One fault characteristic, This belongs to the fault category of The weight, This belongs to the fault category of The weight.

[0101] In some alternative implementations, It can be represented as:

[0102]

[0103] in, This belongs to the fault category of The weight, Source domain fault characteristics The first in The first of the non-fault characteristics One element, Source domain fault characteristics The first in One fault characteristic, Source domain fault characteristics The first in A non-faulty feature, For source domain fault feature dataset, This represents the total number of source domain fault characteristics.

[0104] In some alternative implementations, It can be represented as:

[0105]

[0106] in, This belongs to the fault category of The weight, Fault characteristics of the target domain The first in The first of the non-fault characteristics One element, Source domain fault characteristics The first in One fault characteristic, Source domain fault characteristics The first in A non-faulty feature, For the target domain fault feature dataset, The total number of target fault characteristics.

[0107] In some alternative implementations, the second objective function can be expressed as:

[0108] in, The second objective function is... To obtain the minimum value, This represents the total number of source domain fault characteristics. For classification loss, Source domain fault characteristics The first in One fault characteristic, Source domain fault characteristics The first in A non-faulty feature, For classifier networks, This represents the loss due to the local maximum mean difference. Source domain fault characteristics Multi-scale fault characteristics. Fault characteristics of the target domain Multi-scale fault characteristics. This is a hyperparameter used to balance the weights of different loss terms.

[0109] In step S3044, the parameters of the feature extraction module and the classification module are iteratively optimized based on the second target function until a preset iteration termination condition is reached, and a target wind turbine blade fault diagnosis model is obtained.

[0110] In some optional embodiments, the second target loss is calculated by forward propagation, the parameters of the feature extraction module and the classification module are optimized by back propagation, the iterative training process is repeated until the model converges, and a trained target wind turbine blade fault diagnosis model is obtained.

[0111] In the embodiments of the present application, the contribution of each scale feature to the model performance is considered, the correlation between feature channels is focused on, a channel-level attention mechanism is introduced as the core component of the feature extractor, the fused feature map is directly regulated, the information-rich channels are enhanced, and the redundant channels are suppressed, so that the contribution weight of each scale feature is finely modeled at the channel level. In the embodiments of the present application, the local maximum mean discrepancy loss is used to replace the maximum mean discrepancy loss to form a new second target function, so as to make the model reduce the distance between the source domain and the related subdomain of the target domain, and thus learn the discriminative features with domain invariance. The reason why the vibration signal is selected in the embodiments of the present application is that the vibration signal can effectively capture the fault features such as structural damage and aerodynamic imbalance of the blade. The vibration signal is generated by finite element simulation, which can efficiently construct the source domain data with complete labels and rich fault modes, and significantly reduce the dependence on the physical test bench. The multi-scale convolution unit is used in the embodiments of the present application to solve the problem that the fixed convolution kernel is difficult to extract diversified fault features. The multi-scale convolution expands the receptive field through different hole rates to capture subtle and macroscopic fault information. The channel attention mechanism adaptively weights the channels to enhance the effective information and suppress the noise. The reason why the LMMD loss (Local Maximum Mean Discrepancy) is used in the embodiments of the present application is that the existing blade fault diagnosis methods based on unsupervised domain adaptation mainly focus on aligning the global distribution of features to make the learned features have cross-domain invariance, but ignore the alignment of the subdomains (same label sample set) of two domains, which makes it difficult to classify the mixed data of different subdomains after adaptation. The LMMD loss can realize the alignment of the subdomains of the simulation source domain and the experimental target domain, and break through the limitation of global domain adaptation.

[0112] In the embodiments of the present application, a fault diagnosis method for a wind turbine blade is provided, which can be used for a computer device, Figure 4 The flowchart of the fault diagnosis method for a wind turbine blade according to the embodiments of the present application is shown in Figure 4 The flowchart of the fault diagnosis method for a wind turbine blade according to the embodiments of the present application is shown in Step S401, obtaining a target fault vibration signal and target strain data of a target fan blade; the target fault vibration signal is a fault vibration signal obtained by signal collection on a fault position of the target fan blade, and the target strain data is data of force deformation of the target fan blade obtained by monitoring on the fault position of the target fan blade.

[0113] Step S402, inputting the target fault vibration signal and the target strain data into a target fan blade fault diagnosis model constructed by the fan blade fault diagnosis model construction method of the embodiment of the application to obtain a fault diagnosis result; the input of the target fan blade fault diagnosis model is the target fault vibration signal, and the output of the target fan blade fault diagnosis model is the fault diagnosis result.

[0114] The fan blade fault diagnosis method of the embodiment of the application obtains the fault diagnosis result through the target fan blade fault diagnosis model, and realizes efficient and accurate diagnosis of the target fan blade fault.

[0115] In the embodiment, a fan blade fault diagnosis method is provided, which can be used for a computer device, Figure 5 is a flowchart of the fan blade fault diagnosis method according to the embodiment of the application, as Figure 5 shown, the flowchart includes the following steps: The actual fault vibration signal is collected, a fan blade finite element model containing a fault is established to perform simulation calculation, a fault simulation vibration signal is obtained, convolution layers, pooling layers, activation functions, learning rates, iteration numbers and model parameters are set, weights and biases of the model are randomly initialized, a fan blade fault diagnosis model is obtained, the actual fault vibration signal and the fault simulation vibration signal are input into the fan blade fault diagnosis model to generate a prediction label, an optimizer updates parameters of the fan blade fault diagnosis model through a back propagation algorithm, whether the maximum iteration number is reached is determined, if the maximum iteration number is reached, the fan blade fault diagnosis model trained is obtained, the fan blade fault diagnosis model is used for blade fault diagnosis to obtain a blade fault diagnosis result, and if the maximum iteration number is not reached, the iteration is circular until the maximum iteration number is reached.

[0116] In the embodiment, a fan blade fault diagnosis model construction device is also provided, which is used for implementing the above-mentioned embodiments and preferred embodiments, and will not be described herein. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, implementation of hardware, or a combination of software and hardware, is also possible and contemplated.

[0117] The embodiment provides a fan blade fault diagnosis model construction device, as Figure 6 shown, which includes: The model construction module 601 is configured to construct a fan blade finite element model according to geometric data, material data, and load boundary data of a target fan blade; and the fan blade finite element model is configured to simulate performance of the target fan blade under different working conditions.

[0118] The fault simulation module 602 is configured to simulate different types of blade faults by using the fan blade finite element model to obtain fault simulation vibration signals and fault strain data.

[0119] The feature extraction module 603 is configured to input the fault simulation vibration signals, the fault strain data, actual fault vibration signals, and actual strain data into a feature extraction module of a fan blade fault diagnosis model to perform multi-scale fault feature extraction to obtain multi-scale fault features.

[0120] The model training module 604 is configured to input the multi-scale fault features into a classification module of the fan blade fault diagnosis model based on a local maximum mean difference loss and a classification loss to perform training to obtain a target fan blade fault diagnosis model.

[0121] In some optional embodiments, the model construction module 601 includes: A three-dimensional model construction unit is configured to construct a fan blade three-dimensional model according to geometric data of a target fan blade.

[0122] A material attribute definition unit is configured to define blade material attributes of the fan blade three-dimensional model according to material data to obtain a fan blade three-dimensional model with blade material attributes.

[0123] A mesh division unit is configured to divide the fan blade three-dimensional model with blade material attributes into meshes and add constraint conditions according to load boundary data to obtain a discrete numerical model.

[0124] A parameter solving unit is configured to perform parameter configuration on the discrete numerical model to obtain a fan blade finite element model.

[0125] In some optional embodiments, the fault simulation module 602 includes: A fault simulation unit is configured to input fault features of each type of blade fault into the fan blade finite element model to simulate each type of blade fault to obtain a target fan blade finite element model corresponding to each type of blade fault.

[0126] A signal acquisition unit is configured to acquire simulation signals and stress changes at key detection points of the target fan blade finite element model corresponding to each type of blade fault.

[0127] The signal determination unit is configured to obtain the fault simulation vibration signal and the fault strain data according to the simulation signal corresponding to the multiple types of blade faults and the stress change condition.

[0128] In some optional embodiments, the feature extraction module 603 comprises: The signal analysis unit is configured to input the fault simulation vibration signal, the fault strain data, the actual fault vibration signal, and the actual strain data into the multi-scale convolution unit for multi-scale analysis to obtain a multi-scale feature map.

[0129] The feature extraction unit is configured to input the multi-scale feature map into the channel attention unit to perform feature extraction on the multi-scale feature map to obtain a multi-scale fault feature.

[0130] In some optional embodiments, the feature extraction unit comprises: The global compression sub-unit is configured to perform global information compression on the multi-scale feature map to obtain a channel descriptor vector.

[0131] The nonlinear transformation sub-unit is configured to perform nonlinear transformation on the channel descriptor vector to obtain a nonlinear vector.

[0132] The normalization processing sub-unit is configured to perform normalization processing on the nonlinear vector to obtain a channel scaling coefficient vector.

[0133] The feature determination sub-unit is configured to multiply the multi-scale feature map and the channel scaling coefficient vector channel by channel to obtain the multi-scale fault feature.

[0134] In some optional embodiments, the model training module 604 comprises: The first target function construction unit is configured to construct a first target function according to the maximum mean difference loss and the classification loss; the first target function is used to minimize the distribution difference between the source domain fault feature and the target domain feature in the multi-scale fault feature.

[0135] The alignment processing unit is configured to perform overall domain alignment processing on the multi-scale fault feature by using the first target function.

[0136] The second target function construction unit is configured to construct a second target function according to the local maximum mean difference loss and the classification loss; the second target function is used to represent the distribution difference between the source domain fault feature and the same-class sub-domain feature of the target domain feature in the multi-scale fault feature.

[0137] The model training unit is configured to perform optimization iteration on the parameters of the feature extraction module and the classification module based on the second target function until a preset iteration termination condition is reached to obtain the target fan blade fault diagnosis model.

[0138] The embodiment provides a fan blade fault diagnosis device, which comprises a fan blade fault diagnosis modelFigure 7 As shown, comprising: The data acquisition module 701 is configured to acquire target fault vibration signals of the target fan blade and target strain data; the target fault vibration signals are fault vibration signals obtained by signal collection on a fault position of the target fan blade, and the target strain data is data of force deformation of the target fan blade obtained by monitoring on the fault position of the target fan blade.

[0139] The fault diagnosis module 702 is configured to input the target fault vibration signals and the target strain data into a target fan blade fault diagnosis model constructed by the fan blade fault diagnosis model construction method according to the embodiment of the application to obtain a fault diagnosis result; the input of the target fan blade fault diagnosis model is the target fault vibration signals, and the output of the target fan blade fault diagnosis model is the fault diagnosis result.

[0140] The fan blade fault diagnosis model construction device provided by the embodiment of the application can execute the fan blade fault diagnosis model construction method provided by any of the embodiments of the application, and the fan blade fault diagnosis device can execute the fan blade fault diagnosis method provided by any of the embodiments of the application, and has the corresponding function modules and beneficial effects of the execution method. The further function description of each module and unit is the same as that of the corresponding embodiment, and will not be described here.

[0141] Figure 8 A structural schematic diagram of an electronic device provided by the embodiment of the application.

[0142] The following will be specifically described Figure 8 which shows a structural schematic diagram of an electronic device suitable for implementing the electronic device in the embodiment of the application. The electronic device can include a processor (for example, a central processor, a graphics processor, etc.) 801, which can perform various appropriate actions and processes according to programs stored in a read-only memory (ROM) 802 or programs loaded from a storage 808 into a random access memory (RAM) 803. In the RAM 803, various programs and data required for operation of the electronic device are also stored. The processor 801, the ROM 802, and the RAM 803 are connected to each other through a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0143] Generally, the following devices can be connected to the I / O interface 805: an input device 806 including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 807 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage 808 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 809. The communication device 809 can allow the electronic device to communicate with other devices wirelessly or by wire to exchange data. Although Figure 8Electronic devices are shown with various apparatuses, but it should be understood that not all of the illustrated apparatuses are required, and that more or fewer apparatuses can alternatively be implemented.

[0144] In particular, the processes described above with reference to the flowcharts can be implemented as a computer software program according to embodiments of the present application. For example, embodiments of the present application include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for executing the methods illustrated by the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via the communication apparatus 809, or installed from the memory 808, or installed from the ROM 802. When the computer program is executed by the processor 801, the above-mentioned functions defined in the fan blade fault diagnosis model construction method and the fan blade fault diagnosis method of embodiments of the present application are performed.

[0145] Figure 8 The electronic device shown is merely an example and should not impose any limitation on the functions and use range of embodiments of the present application.

[0146] Embodiments of the present application also provide a computer-readable storage medium, the above-mentioned method according to embodiments of the present application can be implemented in hardware or firmware, or be implemented as computer code that can be recorded on a storage medium, or be implemented by downloading and originally stored in a remote storage medium or non-transitory machine-readable storage medium and then stored in a local storage medium through a network, so that the method described herein can be processed by such software using a general-purpose computer, a special-purpose processor, or programmable or special-purpose hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory, a random access memory, a flash memory, a hard disk or a solid-state disk, etc.; further, the storage medium can also include a combination of the above-mentioned types of memories. It can be understood that the computer, processor, microprocessor controller or programmable hardware includes a storage component that can store or receive software or computer code, when the software or computer code is accessed and executed by the computer, processor or hardware, the fan blade fault diagnosis model construction method and the fan blade fault diagnosis method shown in the above embodiments are implemented.

[0147] Part of the present application can be applied as a computer program product, for example, computer program instructions, when executed by a computer, through the operation of the computer, can invoke or provide the method and / or technical solutions according to the present application. Those skilled in the art should understand that the form of computer program instructions in computer readable medium includes but is not limited to source files, executable files, installation package files and the like, and accordingly, the way of computer program instructions executed by computer includes but is not limited to: the computer directly executes the instructions, or the computer compiles the instructions and then executes the corresponding compiled program, or the computer reads and executes the instructions, or the computer reads and installs the instructions and then executes the corresponding installed program. Here, the computer readable medium can be any available computer readable storage medium or communication medium accessible to the computer.

[0148] Although the embodiments of the present application are described in conjunction with the drawings, various modifications and changes can be made by those skilled in the art without departing from the spirit and scope of the present application, and such modifications and changes fall within the scope defined by the appended claims.

Claims

1. A method for constructing a wind turbine blade fault diagnosis model, characterized in that, The method comprises: constructing a fan blade finite element model according to geometric data, material data, and load boundary data of a target fan blade; the fan blade finite element model is used to simulate performance of the target fan blade under different working conditions; simulating different types of blade faults by using the fan blade finite element model to obtain fault simulation vibration signals and fault strain data; inputting the fault simulation vibration signals, the fault strain data, actual fault vibration signals, and actual strain data into a feature extraction module of a fan blade fault diagnosis model to perform multi-scale fault feature extraction to obtain multi-scale fault features; training the multi-scale fault features based on local maximum mean difference loss and classification loss by inputting the multi-scale fault features into a classification module of the fan blade fault diagnosis model to obtain a target fan blade fault diagnosis model.

2. The method of claim 1, wherein, The method comprises: constructing a fan blade finite element model according to geometric data, material data, and load boundary data of a target fan blade; the fan blade finite element model is used to simulate performance of the target fan blade under different working conditions; constructing a fan blade three-dimensional model according to the geometric data of the target fan blade; defining blade material properties of the fan blade three-dimensional model according to the material data to obtain the fan blade three-dimensional model with blade material properties; dividing the fan blade three-dimensional model with blade material properties into grids and adding constraint conditions according to the load boundary data to obtain a discrete numerical model; 3. The method according to claim 1 or 2, characterized in that, configuring solving parameters for the discrete numerical model to obtain the fan blade finite element model. The method comprises: inputting fault features of each type of blade fault into the fan blade finite element model to simulate each type of blade fault to obtain a target fan blade finite element model corresponding to each type of blade fault; collecting simulation signals and stress changes at key detection points of the target fan blade finite element model corresponding to each type of blade fault; 4. The method according to claim 1 or 2, characterized in that, obtaining the fault simulation vibration signals and the fault strain data according to the simulation signals and the stress changes corresponding to a plurality of types of blade faults. The feature extraction module comprises a multi-scale convolution unit and a channel attention unit; the method comprises: inputting the fault simulation vibration signals, the fault strain data, actual fault vibration signals, and actual strain data into the multi-scale convolution unit to perform multi-scale analysis to obtain multi-scale feature maps; 5. The method of claim 4, wherein, inputting the multi-scale feature maps into the channel attention unit to perform feature extraction on the multi-scale feature maps to obtain the multi-scale fault features. The method comprises: The multi-scale feature map is globally compressed to obtain a channel descriptor vector; The channel descriptor vector is subjected to a nonlinear transformation to obtain a nonlinear vector; The nonlinear vector is subjected to a normalization process to obtain a channel scaling coefficient vector; The multi-scale feature map is multiplied with the channel scaling coefficient vector channel by channel to obtain the multi-scale fault feature.

6. The method of claim 1 or 2, wherein, The multi-scale fault feature is input into a classification module of the fan blade fault diagnosis model based on the local maximum mean difference loss and the classification loss for training to obtain a target fan blade fault diagnosis model, including: A first target function is constructed according to the maximum mean difference loss and the classification loss; the first target function is used to minimize the distribution difference between the source domain fault feature and the target domain feature in the multi-scale fault feature; The first target function is used to perform overall domain alignment processing on the multi-scale fault feature; A second target function is constructed according to the local maximum mean difference loss and the classification loss; the second target function is used to represent the distribution difference between the source domain fault feature and the target domain feature in the same type of sub-domain in the multi-scale fault feature; Based on the second target function, the parameters of the feature extraction module and the classification module are iteratively optimized until a preset iteration termination condition is reached to obtain the target fan blade fault diagnosis model.

7. A method of diagnosing a failure of a wind turbine blade, characterized by, The method includes: Obtaining a target fault vibration signal and target strain data of a target fan blade; the target fault vibration signal is a fault vibration signal obtained by signal collection on a fault part of the target fan blade, and the target strain data is data of stress deformation of the target fan blade obtained by monitoring on the fault part of the target fan blade; The target fault vibration signal and the target strain data are input into a target fan blade fault diagnosis model constructed by the fan blade fault diagnosis model construction method according to any one of claims 1 to 6 to obtain a fault diagnosis result; the input of the target fan blade fault diagnosis model is the target fault vibration signal, and the output of the target fan blade fault diagnosis model is the fault diagnosis result. 8.A wind turbine blade failure diagnosis model construction device characterized by comprising: The device includes: A model construction module is configured to construct a fan blade finite element model according to geometric data, material data and load boundary data of a target fan blade; the fan blade finite element model is used to simulate the performance of the target fan blade under different working conditions; A fault simulation module is configured to simulate different types of blade faults by using the fan blade finite element model to obtain fault simulation vibration signals and fault strain data; A feature extraction module is configured to input the fault simulation vibration signals, the fault strain data, actual fault vibration signals and actual strain data into a feature extraction module of a fan blade fault diagnosis model to extract multi-scale fault features to obtain multi-scale fault features; A model training module is configured to input the multi-scale fault features into a classification module of the fan blade fault diagnosis model based on a local maximum mean difference loss and a classification loss for training to obtain a target fan blade fault diagnosis model.

9. A failure diagnosis device for a wind turbine blade, characterized by, The device comprises: a data acquisition module configured to acquire a target fault vibration signal of a target fan blade and target strain data; the target fault vibration signal is a fault vibration signal obtained by signal collection on a fault site of the target fan blade, and the target strain data is data of force deformation of the target fan blade obtained by monitoring the fault site of the target fan blade; a fault diagnosis module configured to input the target fault vibration signal and the target strain data into a target fan blade fault diagnosis model constructed by the fan blade fault diagnosis model construction method according to any one of claims 1 to 6 to obtain a fault diagnosis result; the input of the target fan blade fault diagnosis model is the target fault vibration signal, and the output of the target fan blade fault diagnosis model is the fault diagnosis result.

10. An electronic device, comprising: comprise: a memory and a processor, which are in communication connection with each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the fan blade fault diagnosis model construction method according to any one of claims 1 to 6 or the fan blade fault diagnosis method according to claim 7.

11. A computer readable storage medium, characterized in that, The computer readable storage medium stores computer instructions, and the computer instructions are used to make the computer execute the fan blade fault diagnosis model construction method according to any one of claims 1 to 6 or the fan blade fault diagnosis method according to claim 7.

12. A computer program product, characterised in that, comprise computer instructions, and the computer instructions are used to make the computer execute the fan blade fault diagnosis model construction method according to any one of claims 1 to 6 or the fan blade fault diagnosis method according to claim 7.

Images