Aircraft engine fault data generation method and device and computer equipment

Abstract

The application relates to an aero-engine fault data generation method, device and computer equipment. The method comprises the following steps: acquiring normal engine data, fault engine data, sample fault data and fault data requirement information of an aero-engine, and generating a fault data generation model of the aero-engine; based on the sample fault data, fine-tuning the fault data generation model to obtain a multi-scenario fault data generation model, and identifying requirement scenario information of the aero-engine based on the fault data requirement information of the aero-engine; and based on the requirement scenario information, generating fault data corresponding to the requirement scenario information of the aero-engine through the multi-scenario fault data generation model. The method can improve the generation accuracy of the fault data of the aero-engine.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data generation, in particular to an aero-engine fault data generation method, device and computer equipment. BACKGROUND

[0002] In recent years, data-driven methods have received extensive attention in the field of aero-engine fault diagnosis. Through data-driven methods, different fault modes can be more flexibly adapted, and early prediction and accurate diagnosis of aero-engine faults are expected to be achieved. However, due to the extremely high reliability and stability of aero-engines, the probability of fault occurrence in actual operation is extremely low, so it is difficult to obtain sufficient fault data to support model training and verification. In addition, the fault of the aero-engine has high randomness and complexity, so it is almost impossible to completely reproduce all possible fault scenarios through experiments. The lack of sufficient number and diversity of fault data directly limits the development of data-driven fault diagnosis models, becoming a bottleneck problem in the development of this field. Therefore, how to improve the generation efficiency of aero-engine fault data is the current research focus.

[0003] The traditional generation method of aero-engine fault data is to rely on fault simulation technology to obtain fault data. For example, in the fault simulation of gas path components, researchers generate fault data by injecting fault factors into the performance model of each component of the engine; in the fault simulation of the sensing system, the performance model output containing noise is usually used as the sensor signal, and step or ramp signals are added to simulate fault scenarios. However, since the mechanism-based fault simulation relies on idealized assumptions, the occurrence process of the fault is often oversimplified, and it is difficult to fully reflect the complex fault characteristics in reality, thereby resulting in low precision of the generated aero-engine fault data. SUMMARY

[0004] Therefore, it is necessary to provide an aero-engine fault data generation method, device, computer equipment, computer readable storage medium and computer program product in view of the above technical problems.

[0005] In a first aspect, the present application provides an aero-engine fault data generation method, comprising:

[0006] obtaining normal engine data of an aero-engine, fault engine data of the aero-engine, each sample fault data of the aero-engine, and fault data requirement information of the aero-engine, and generating a fault data generation model of the aero-engine based on the normal engine data and the fault engine data;

[0007] Based on the sample fault data, the fault data generation model is subjected to model adjustment processing to obtain a multi-scenario fault data generation model, and based on the fault data requirement information of the aero-engine, requirement scenario information of the aero-engine is identified;

[0008] Based on the requirement scenario information, the multi-scenario fault data generation model is used to generate fault data corresponding to the requirement scenario information of the aero-engine.

[0009] Optionally, the fault data generation model of the aero-engine is generated based on the normal engine data and the fault engine data, comprising:

[0010] Based on the normal engine data and the fault engine data, first engine air path parameters of the aero-engine in a normal state and second engine air path parameters of the aero-engine in an abnormal state are identified;

[0011] The first engine air path parameters are divided into first sub-air path parameters corresponding to each working condition by a preset working condition division strategy, and the second engine air path parameters are divided into second sub-air path parameters corresponding to each working condition by the preset working condition division strategy;

[0012] Based on the first sub-air path parameters corresponding to each working condition and the second sub-air path parameters corresponding to each working condition, normal data distribution information of the aero-engine in a normal state and fault data distribution information of the aero-engine in a fault state are identified;

[0013] Based on the normal data distribution information and the fault data distribution information, the fault data generation model of the aero-engine is constructed.

[0014] Optionally, the fault data generation model of the aero-engine is constructed based on the normal data distribution information and the fault data distribution information, comprising:

[0015] Based on the normal data distribution information, an initial data generation model is trained to obtain a data generation model, and an encoder parameter in the data generation model is identified;

[0016] The encoder parameter is migrated to an initial fault generation model to obtain an initial fault data generation model, and an output layer of the initial fault data generation model is trained based on the fault data distribution information to obtain a fault data generation model.

[0017] Optionally, the multi-scenario fault data generation model is obtained by subjecting the fault data generation model to model adjustment processing based on the sample fault data, comprising:

[0018] identify the fault type label corresponding to each sample fault data, and divide each sample fault data into a sample fault data group of each fault type label;

[0019] query the fault scene corresponding to each fault type label, and adjust the output layer parameters of the fault data generation model based on the sample fault data group of each fault type label, to obtain a sub-fault data generation model of the fault scene corresponding to each fault type label;

[0020] all sub-fault data generation models of fault scenes are taken as a multi-scene fault data generation model.

[0021] Optionally, the demand scene information of the aero-engine is identified based on the fault data demand information of the aero-engine, comprising:

[0022] Based on the fault data demand information, the demand fault information of the aero-engine is identified, and based on the demand fault information, the target fault type required by the fault data demand information is identified;

[0023] Based on the correspondence between each fault type and each fault scene, the target fault scene corresponding to the target fault type is identified;

[0024] The target fault scene is taken as the demand scene information of the aero-engine.

[0025] Optionally, the demand scene information of the aero-engine is generated based on the demand scene information through the multi-scene fault data generation model, comprising:

[0026] Based on the target fault scene, each initial fault data corresponding to the target fault scene is generated through the multi-scene fault data generation model, and a fault classifier corresponding to the target fault type is obtained;

[0027] Based on the fault classifier, data classification verification processing is performed on each initial fault data to obtain fault data generation accuracy, and in the case that the fault data generation accuracy is lower than the preset accuracy, the step of performing model adjustment processing on the fault data generation model based on each sample fault data is returned to obtain a multi-scene fault data generation model until the fault data generation accuracy is not lower than the preset accuracy. When the accuracy is not lower than the preset accuracy, each initial fault data obtained by the last iteration is taken as the fault data corresponding to the demand fault information of the aero-engine.

[0028] In a second aspect, the present application also provides an aero-engine fault data generation device, comprising:

[0029] an acquisition module, configured to acquire normal engine data of an aero-engine, fault engine data of the aero-engine, sample fault data of the aero-engine, and fault data requirement information of the aero-engine, and generate a fault data generation model of the aero-engine based on the normal engine data and the fault engine data;

[0030] a recognition module, configured to perform model adjustment processing on the fault data generation model based on the sample fault data, to obtain a multi-scenario fault data generation model, and recognize requirement scenario information of the aero-engine based on the fault data requirement information of the aero-engine;

[0031] a generation module, configured to generate fault data corresponding to the requirement scenario information of the aero-engine by using the multi-scenario fault data generation model based on the requirement scenario information.

[0032] Optionally, the acquisition module is specifically configured to:

[0033] identify first engine air path parameters of the aero-engine in a normal state and second engine air path parameters of the aero-engine in an abnormal state based on the normal engine data and the fault engine data;

[0034] divide the first engine air path parameters into first sub-air path parameters corresponding to each working condition by using a preset working condition division strategy, and divide the second engine air path parameters into second sub-air path parameters corresponding to each working condition by using the preset working condition division strategy;

[0035] identify normal data distribution information of the aero-engine in the normal state and fault data distribution information of the aero-engine in the fault state based on the first sub-air path parameters corresponding to each working condition and the second sub-air path parameters corresponding to each working condition;

[0036] construct the fault data generation model of the aero-engine based on the normal data distribution information and the fault data distribution information.

[0037] Optionally, the acquisition module is specifically configured to:

[0038] train an initial data generation model based on the normal data distribution information to obtain a data generation model, and identify an encoder parameter in the data generation model;

[0039] The encoder parameters are migrated to an initial fault data generation model to obtain an initial fault data generation model, and an output layer of the initial fault data generation model is trained based on the fault data distribution information to obtain a fault data generation model.

[0040] Optionally, the identification module is specifically configured to:

[0041] Each sample fault data is identified to correspond to a fault type label, and each sample fault data is divided into a sample fault data group of each fault type label.

[0042] Each fault type label is queried to correspond to a fault scene, and the output layer parameters of the fault data generation model are adjusted based on the sample fault data group of each fault type label to obtain a sub-fault data generation model of each fault type label corresponding to the fault scene.

[0043] All sub-fault data generation models of the fault scenes are used as a multi-scene fault data generation model.

[0044] Optionally, the identification module is specifically configured to:

[0045] The demand fault information of the aero-engine is identified based on the fault data demand information, and the target fault type required by the fault data demand information is identified based on the demand fault information.

[0046] The target fault scene corresponding to the target fault type is identified based on the corresponding relationship between each fault type and each fault scene.

[0047] The target fault scene is used as the demand scene information of the aero-engine.

[0048] Optionally, the generation module is specifically configured to:

[0049] Each initial fault data corresponding to the target fault scene is generated by the multi-scene fault data generation model based on the target fault scene, and a fault classifier corresponding to the target fault type is obtained.

[0050] Each initial fault data is subjected to data classification verification processing based on the fault classifier to obtain a fault data generation accuracy, and in a case where the fault data generation accuracy is lower than a preset accuracy, the step of performing model adjustment processing on the fault data generation model based on each sample fault data is returned to obtain a multi-scene fault data generation model until the fault data generation accuracy is not lower than the preset accuracy, and each initial fault data obtained through the last iteration is used as fault data corresponding to the demand fault information of the aero-engine.

[0051] Thirdly, this application provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the method described in any one of the first aspects.

[0052] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the method described in any one of the first aspects.

[0053] Fifthly, this application provides a computer program product. The computer program product includes a computer program that, when executed by a processor, implements the steps of the method described in any one of the first aspects.

[0054] The aforementioned method, apparatus, and computer equipment for generating aero-engine fault data acquires normal engine data, faulty engine data, sample fault data of the aero-engine, and fault data requirement information of the aero-engine. Based on the normal engine data and the faulty engine data, a fault data generation model for the aero-engine is generated. Based on each sample fault data, the fault data generation model is adjusted to obtain a multi-scenario fault data generation model. Based on the fault data requirement information of the aero-engine, the required scenario information of the aero-engine is identified. Based on the required scenario information, the fault data corresponding to the required scenario information of the aero-engine is generated through the multi-scenario fault data generation model. This solution generates the aero-engine fault data generation model by utilizing the correlation information between normal engine data and faulty engine data. Since normal engine data has a large number of data samples, this method can combine samples from normal engine data to train the fault data generation model, thereby effectively solving the problems of scarce and imbalanced aero-engine fault data. This approach then combines the fault data from various samples to adjust the model, enabling the generated fault data model to accurately generate fault data for different fault types and scenarios. This ensures the model can effectively learn and generate data for specific fault types. This method not only improves the accuracy of fault data generation but also provides a reliable data source for the diagnosis and classification of different fault types. Finally, based on the multi-scenario fault data generation model trained above, this approach can acquire a large amount of targeted fault data using only the required scenario information, thus improving the accuracy of generating fault data for aero-engines. Attached Figure Description

[0055] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related art, the drawings needed to be used in the embodiments or related art description will be briefly introduced. Obviously, the drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can be obtained based on these drawings without creative labor.

[0056] Figure 1 Flowchart of the method for generating aero-engine fault data in an embodiment;

[0057] Figure 2 Flowchart of the method for generating fault samples based on GMM-VAE and transfer learning in an embodiment;

[0058] Figure 3 Comparison diagram of the precision of the improved and unimproved generation model in an embodiment;

[0059] Figure 4 Diagram of the normal sample generation area based on GMM-VAE in an embodiment;

[0060] Figure 5 Diagram of the precision of the LPC fault generation sample when the F-N ratio is 0.1% in an embodiment;

[0061] Figure 6 Diagram of the precision of the XNLC fault generation sample when the F-N ratio is 0.1% in an embodiment;

[0062] Figure 7 Diagram of the coverage rate of the LPC fault generation sample when the F-N ratio is 0.1% in an embodiment;

[0063] Figure 8 Diagram of the coverage rate of the XNLC sensor fault generation sample when the F-N ratio is 0.1% in an embodiment;

[0064] Figure 9 Flowchart of the generation example of aero-engine fault data in an embodiment;

[0065] Figure 10 Structural block diagram of the generation device of aero-engine fault data in an embodiment;

[0066] Figure 11 Internal structure diagram of the computer device in an embodiment. DETAILED DESCRIPTION

[0067] In order to make the purposes, technical solutions and advantages of the present application clearer, further detailed description will be given below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application and not to limit the present application.

[0068] The method for generating the engine fault data provided by the embodiments of the present application can be applied to the application environment of generating the engine fault data. The method can be applied to a terminal, a server, or a system including the terminal and the server, and is realized through the interaction of the terminal and the server. The terminal can be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, etc. The terminal generates a fault data generation model of the engine by associating the normal engine data and the fault engine data, and since the normal engine data has a large number of data samples, the method can train the fault data generation model in combination with the samples of the normal engine data, thereby effectively solving the problem of the scarcity and unbalanced samples of the engine fault data. Then the present application adjusts the model in combination with the sample fault data, so that the generated fault data generation model can accurately generate fault data for different fault types and different fault scenarios, thereby ensuring that the model can effectively learn and generate for specific fault types. The method not only improves the accuracy of fault data generation, but also provides a reliable data source for the diagnosis and classification of different fault types. Finally, based on the multi-scene fault data generation model trained above, the present application can obtain a large number of targeted fault data only by the demand scene information, thereby improving the accuracy of generating the engine fault data.

[0069] In one exemplary embodiment, as shown in Figure 1 A method for generating engine fault data is provided. Taking the terminal as an example, the method includes the following steps S101 to S103. Wherein:

[0070] Step S101, obtaining normal engine data of an engine, fault engine data of the engine, sample fault data of the engine, and fault data demand information of the engine, and generating a fault data generation model of the engine based on the normal engine data and the fault engine data.

[0071] In this embodiment, the terminal obtains normal engine data of the aero-engine and fault engine data of the aero-engine, wherein the normal engine data is each gas path parameter of the aero-engine in a normal state, and the fault engine data is each gas path parameter of the aero-engine in a fault state. The gas path parameter includes a gas path parameter of the aero-engine in each working condition. Then, the aero-engine obtains engine data of different fault types, obtains each sample fault data, and obtains demand information of the aero-engine for the generated fault data in response to an information uploading operation of a worker, to obtain fault data demand information of the aero-engine. Finally, the terminal generates a fault data generation model of the aero-engine based on the normal engine data and the fault engine data. The fault data generation model is a GMM-VAE (Gaussian Mixture Model Variational AutoEncoder) model. The model is a data generation model combining a Gaussian Mixture Model (GMM) and a Variational Autoencoder (VAE). The specific training process will be described in detail later.

[0072] In step S102, the fault data generation model is adjusted based on each sample fault data to obtain a multi-scenario fault data generation model, and demand scenario information of the aero-engine is identified based on the fault data demand information of the aero-engine.

[0073] In this embodiment, the terminal adjusts the fault data generation model based on each sample fault data to obtain a multi-scenario fault data generation model, and identifies demand scenario information of the aero-engine based on the fault data demand information of the aero-engine. The fine-tuning process is to train the fault data generation model through sample fault data of different fault types, so that the fault data generation model can generate high-quality fault simulation data suitable for different fault scenarios according to different fault types. The specific fine-tuning process will be described in detail later. The fault types include but are not limited to gas path faults and sensor faults. The fine-tuning method is a transfer learning method. The demand scenario information is a target fault scenario required by the fault data demand information.

[0074] In step S103, the multi-scenario fault data generation model is used to generate fault data corresponding to the demand scenario information of the aero-engine based on the demand scenario information.

[0075] In this embodiment, the terminal generates fault data corresponding to the demand scenario information of the aero-engine based on the demand scenario information through a multi-scenario fault data generation model. During the generation process, a fault classifier of different fault types is also needed to verify the generated fault samples, so as to optimize and adjust the multi-scenario fault data generation model to improve the generation accuracy of the fault data. The specific optimization process will be described in detail later.

[0076] Based on the above scheme, the fault data generation model of the aero-engine is generated from the association information between the normal engine data and the fault engine data. Since the normal engine data has a large number of data samples, the fault data generation model can be trained in combination with the samples of the normal engine data, thereby effectively solving the problem of the scarcity and sample imbalance of the aero-engine fault data. Then, the model is adjusted and processed in combination with each sample fault data, so that the generated fault data generation model can accurately generate fault data for different fault types and different fault scenarios, thereby ensuring that the model can effectively learn and generate for a specific fault type. This method not only improves the accuracy of fault data generation, but also provides a reliable data source for diagnosis and classification of different fault types. Finally, based on the multi-scenario fault data generation model trained above, a large amount of targeted fault data can be obtained only by demand fault information, thereby improving the generation accuracy of the fault data of the aero-engine.

[0077] Optionally, the fault data generation model of the aero-engine is generated based on the normal engine data and the fault engine data, including: identifying the first engine air path parameters of the aero-engine in the normal state and the second engine air path parameters of the aero-engine in the abnormal state based on the normal engine data and the fault engine data; dividing the first engine air path parameters into first sub-air path parameters corresponding to each working condition by a preset working condition division strategy, and dividing the second engine air path parameters into second sub-air path parameters corresponding to each working condition by the preset working condition division strategy; identifying normal data distribution information of the aero-engine in the normal state and fault data distribution information of the aero-engine in the fault state based on the first sub-air path parameters corresponding to each working condition and the second sub-air path parameters corresponding to each working condition; and constructing the fault data generation model of the aero-engine based on the normal data distribution information and the fault data distribution information.

[0078] In this embodiment, the terminal identifies the first engine air path parameters of the aero-engine in the normal state and the second engine air path parameters of the aero-engine in the abnormal state based on the normal engine data and the fault engine data.

[0079] Then, the terminal, using a preset operating condition division strategy, divides the first engine air path parameters into first sub-air path parameters corresponding to each operating condition, and similarly divides the second engine air path parameters into second sub-air path parameters corresponding to each operating condition. (The preset strategy is used to further divide the parameters into sub-air path parameters.)

[0080] Then, based on the first sub-gas path parameters corresponding to each operating condition and the second sub-gas path parameters corresponding to each operating condition, the terminal identifies the normal data distribution information of the aero-engine under normal conditions and the fault data distribution information of the aero-engine under fault conditions. The aforementioned distribution information, such as the engine data under normal conditions, follows a distribution... Engine data under fault conditions follows a distribution Furthermore, there are similarities between the two. By leveraging the multimodal processing capabilities of the GMM, the model can cover the engine's operating states under a wide range of conditions, thus achieving a normal state generation model. The output should approximate the desired distribution as closely as possible. As shown below.

[0081] (1)

[0082] Next, the terminal constructs a fault data generation model for the aero-engine based on normal data distribution information and fault data distribution information. The specific construction process of the fault data generation model will be explained in detail later.

[0083] Based on the above scheme, a fault data generation model is constructed by examining the correlation between normal engine data and faulty engine data. This allows a large amount of faulty engine data to be generated from normal engine data, effectively solving the problems of scarce and imbalanced samples in aero-engine fault data and increasing the amount of fault data generated.

[0084] Optionally, based on normal data distribution information and fault data distribution information, a fault data generation model for aero-engines is constructed, including: training an initial data generation model based on normal data distribution information to obtain a data generation model, and identifying encoder parameters in the data generation model; transferring the encoder parameters to the initial fault generation model to obtain an initial fault data generation model, and training the output layer of the initial fault data generation model based on fault data distribution information to obtain a fault data generation model.

[0085] In this embodiment, the terminal trains the initial data generation model based on the normal data distribution information, obtains the data generation model, and identifies the encoder parameters in the data generation model. Then, the terminal migrates the encoder parameters to the initial fault generation model to obtain the initial fault data generation model, and trains the output layer of the initial fault data generation model based on the fault data distribution information to obtain the fault data generation model. Specifically, the first stage is the training process of the GMM-VAE generation model, wherein the input data is the air path parameters of the engine, and the model processes the data through the preset working condition. The introduction of GMM improves the expression ability of the model in the processing of multi-modal data and expands the coverage range of the fault sample generation.

[0086] The second stage is the model migration learning process. The GMM-VAE encoder parameters obtained by the first stage are migrated to the fault generation model, and the parameters are frozen to keep the mapping relationship between the normal state and the fault state unchanged. By adjusting the output of the generation model to approximate the distribution of the fault state , as follows:

[0087] (2)

[0088] Based on the above scheme, after training the data generation model, the model parameters of the model are migrated and converted, and then the model is adjusted according to the fault data distribution information, thereby improving the accuracy of the model in generating fault data.

[0089] Optionally, based on the fault data of each sample, the fault data generation model is adjusted and processed to obtain a multi-scene fault data generation model, including: identifying the fault type label corresponding to each sample fault data, dividing each sample fault data into a sample fault data group of each fault type label; querying the fault scene corresponding to each fault type label, and adjusting the output layer parameters of the fault data generation model based on the sample fault data group of each fault type label to obtain a sub-fault data generation model of the fault scene corresponding to each fault type label; and taking all the sub-fault data generation models of the fault scenes as the multi-scene fault data generation model.

[0090] In this embodiment, the terminal identifies the fault type label corresponding to each sample fault data, and divides each sample fault data into a sample fault data group of each fault type label. Each fault type label corresponds to a fault type of the aero-engine. Each fault type corresponds to a fault scene. The terminal acquires the fault scene corresponding to each fault type label by querying the fault scene database, and adjusts the output layer parameters of the fault data generation model based on the sample fault data group of each fault type label, to obtain a sub-fault data generation model of the fault scene corresponding to each fault type label. Finally, the terminal takes the sub-fault data generation models of all fault scenes as the multi-scene fault data generation model.

[0091] Based on the above scheme, the model is fine-tuned respectively based on the sample fault data corresponding to each fault data type, and the transfer learning technology is used to reduce the large amount of data and iterations required for neural network training, so that the model can quickly adapt to fault data. And ensure that the generated fault samples can be used for multiple fault types such as air path faults and sensor faults. In order to adapt to different fault scenes, generate high-quality fault simulation data.

[0092] Optionally, based on the fault data demand information of the aero-engine, the demand scene information of the aero-engine is identified, including: identifying the demand fault information of the aero-engine based on the fault data demand information, and identifying the target fault type required by the fault data demand information based on the demand fault information; identifying the target fault scene corresponding to the target fault type based on the correspondence between each fault type and each fault scene; and taking the target fault scene as the demand scene information of the aero-engine.

[0093] In this embodiment, the terminal identifies the demand fault information of the aero-engine based on the fault data demand information, and identifies the target fault type required by the fault data demand information based on the demand fault information. The number of the target fault type can be one or more. The demand content of each fault type is included in the fault data demand information, the terminal identifies each demand fault type in the fault data demand information based on a semantic recognition network based on natural language processing technology, and takes all demand fault types as demand fault information. Then, the terminal identifies the target fault scene corresponding to the target fault type based on the correspondence between each fault type and each fault scene in the fault scene database. Finally, the terminal takes the target fault scene as the demand scene information of the aero-engine.

[0094] Based on the above scheme, the fault data demand information is split into each fault type, which improves the comprehensiveness and accuracy of the identification of the demand scene information.

[0095] Optionally, based on the demand scenario information, the fault data corresponding to the demand scenario information of the aero-engine is generated through the multi-scene fault data generation model, including: based on the target fault scene, generating each initial fault data corresponding to the target fault scene through the multi-scene fault data generation model, and obtaining the fault classifier corresponding to the target fault type; based on the fault classifier, performing data classification verification processing on each initial fault data to obtain fault data generation accuracy, and in the case that the fault data generation accuracy is lower than the preset accuracy, returning to execute the model adjustment processing on the fault data generation model based on each sample fault data to obtain the multi-scene fault data generation model step, until the fault data generation accuracy is not lower than the preset accuracy, and each initial fault data obtained in the last iteration is taken as the fault data corresponding to the demand fault information of the aero-engine.

[0096] In the embodiment, the terminal generates each initial fault data corresponding to the target fault scene through the multi-scene fault data generation model based on the target fault scene, and obtains the fault classifier corresponding to the target fault type. Each fault classifier is a classifier neural network obtained by training sample fault data of each fault type.

[0097] The terminal performs data classification verification processing on each initial fault data based on the fault classifier to identify the proportion of initial fault data belonging to the target fault type in all initial fault data, and then takes the proportion as the fault data generation accuracy.

[0098] The terminal presets the accuracy, and in the case that the fault data generation accuracy is lower than the preset accuracy, returns to execute the model adjustment processing on the fault data generation model based on each sample fault data to obtain the multi-scene fault data generation model step, until the fault data generation accuracy is not lower than the preset accuracy, and each initial fault data obtained in the last iteration is taken as the fault data corresponding to the demand fault information of the aero-engine.

[0099] In the actual experimental verification process, as shown in Figure 2 The flowchart of the fault sample generation based on GMM-VAE and transfer learning is shown in FIG. 1. A reasonable similarity standard between samples is extremely important for evaluating the quality of generated signals and the generation ability of the proposed method. In fact, the fault sample points generated by the fault data generation method proposed in the present application need to be compared with the fault sample points in the ideal case (various sample fault data sets in the ideal case are obtained by simulating fault samples under different amplitudes and different working conditions). Two comparison dimensions of accuracy and coverage rate are proposed to analyze and verify the effectiveness of the method proposed in the present application. Table 1 shows the sensor measuring points and symbol representation used in the subsequent examples.

[0100] Table 1. Each sensor measuring point and symbol representation

[0101]

[0102] wherein the Euclidean distance is the most commonly used similarity measure index between numerical samples. The calculation method of is shown in equation (3).

[0103] (3)

[0104] In the formula is the ideal fault data set of each type of sample, is the generated fault sample, , respectively represent the , dimensional features of the sample By traversing the ideal fault data set of each type of sample (including normal samples, ideal fault samples of each type of gas path and sensor fault samples) and the generated sample , the minimum value is selected as the accuracy of each generated sample. The ideal fault sample fault data set is obtained by Monte Carlo simulation of the fault model multiple times.

[0105] The coverage rate of the sample is considered to be an important indicator of the generalization ability of the generated model, and is also a key measure of the fusion state between the working condition information and the fault information of the aero-engine. Due to the scarcity and discontinuity of data samples, it is difficult for the generated model to achieve full coverage of the learning samples. The actual generated sample is compared with the ideal fault sample to obtain the coverage rate of the generated model. Assuming is the coverage range of a working condition under ideal conditions, is the coverage range of the generated model actually obtained for this working condition, and the coverage rate can be calculated by equation (4). In the specific implementation process, the variation range of the low-pressure rotor speed is used as the measurement standard.

[0106] (4)

[0107] In the design and evaluation process of the generated model, the selection of hyperparameters has a crucial influence on the accuracy and diversity of the generated samples of the model. In the training stage, the framework of the network, the activation function and the number of components of the GMM need to be designed, as shown in Table 2.

[0108] Table 2. GMM-VAE model hyperparameter settings

[0109]

[0110] After the model training is completed, in order to ensure that the sampling process fully covers the GMM distribution, 1000 simulated normal running state data samples are generated by the normal data generator. In order to compare the accuracy of the two generation models before and after the improvement of GMM-VAE, the overall accuracy of the two is compared in Figure 3 , both of which meet the expected accuracy. Compared with the traditional VAE, GMM-VAE enhances the sample generation ability, so that the expression ability of the model in extreme cases is improved, which is manifested as the maximum accuracy error and the minimum accuracy error are significantly higher than VAE, which meets the research expectation.

[0111] Because the representation of GMM-VAE model in the hidden space has high continuity, when the number of generated samples is sufficient, the distribution of the samples will show a band-shaped feature, which can cover multiple areas of the hidden space, rather than being limited to some local, Figure 4 shows the distribution of GMM-VAE generated samples in each dimension.

[0112] The quality and accuracy of the generated samples are crucial for their application in the fault diagnosis process. Due to space limitations, Figure 5 and Figure 6 only show the accuracy performance of the LPC (Low Pin Count, bus) fault and XNLC sensor failure (NOx sensor failure) generation model under the index when the proportion of fault samples and normal samples is 0.1% (i.e. the number of fault samples of each type is 100). Figure 5 In , D-Normal indicates the calculation result between the generated fault samples (1000 samples in total) and the ideal state normal samples. Figure 6 D-LPC, D-HPC, D-HPT, D-LPT respectively represent the calculation results between the generated fault samples and the ideal state of each type of air path fault samples.

[0113] Figure 7 and Figure 8The XNLC distribution of 1000 generated sample points is shown under the condition of F-N ratio of 0.1%. In the process of transfer learning, the model tends to fit the distribution of fault samples, which may affect the coverage of the generated samples. In order to accurately evaluate the coverage of the generated samples, the maximum and minimum values of XNLC in the 1000 generated samples are compared with the corresponding parameters in the actual fault samples. The specific results are shown in Table 2. Due to the scarcity and poor continuity of fault samples, it is difficult for the GMM-VAE model to learn the overall picture of the fault distribution. As shown in Table 3, as the F-N ratio decreases, the transfer ability of the model is limited, resulting in a gradual decrease in the coverage of each model. However, compared with the VAE model before improvement, the model still has more excellent expression ability.

[0114] Table 3 Coverage of generated model under different F-N ratio

[0115]

[0116] Based on the above scheme, the generated fault samples are verified by a mature multi-class fault classifier. Without optimizing the fault classification algorithm, the generated samples can significantly improve the accuracy of fault diagnosis, especially in the case of sample imbalance or fault data scarcity.

[0117] The application also provides a generation example of aero-engine fault data, as shown in Figure 9 The specific processing process includes the following steps:

[0118] Step S901, obtaining normal engine data of an aero-engine, fault engine data of the aero-engine, each sample fault data of the aero-engine, and fault data requirement information of the aero-engine.

[0119] Step S902, identifying first engine air path parameters of the aero-engine in a normal state and second engine air path parameters of the aero-engine in an abnormal state based on the normal engine data and the fault engine data.

[0120] Step S903, dividing the first engine air path parameters into first sub-air path parameters corresponding to each working condition by a preset working condition division strategy, and dividing the second engine air path parameters into second sub-air path parameters corresponding to each working condition by the preset working condition division strategy.

[0121] Step S904, identifying normal data distribution information of the aero-engine in a normal state and fault data distribution information of the aero-engine in a fault state based on the first sub-air path parameters corresponding to each working condition and the second sub-air path parameters corresponding to each working condition.

[0122] Step S905, based on the normal data distribution information, training the initial data generation model to obtain the data generation model, and identifying the encoder parameters in the data generation model.

[0123] Step S906, migrating the encoder parameters to the initial fault generation model to obtain the initial fault data generation model, and training the output layer of the initial fault data generation model based on the fault data distribution information to obtain the fault data generation model.

[0124] Step S907, identifying the fault type label corresponding to each sample fault data, and dividing the sample fault data into sample fault data groups of each fault type label.

[0125] Step S908, querying the fault scene corresponding to each fault type label, and adjusting the output layer parameters of the fault data generation model based on the sample fault data group of each fault type label to obtain a sub-fault data generation model corresponding to each fault type label. Fault scene.

[0126] Step S909, taking all sub-fault data generation models of fault scenes as a multi-scene fault data generation model.

[0127] Step S910, identifying the demand fault information of the aero-engine based on the fault data requirement information, and identifying the target fault type required by the fault data requirement information based on the demand fault information.

[0128] Step S911, identifying the target fault scene corresponding to the target fault type based on the correspondence between each fault type and each fault scene.

[0129] Step S912, taking the target fault scene as the demand scene information of the aero-engine.

[0130] Step S913, based on the target fault scene, generating each initial fault data corresponding to the target fault scene through the multi-scene fault data generation model, and obtaining the fault classifier corresponding to the target fault type.

[0131] Step S914, based on the fault classifier, performing data classification verification processing on each initial fault data to obtain the fault data generation accuracy, and in the case that the fault data generation accuracy is lower than the preset accuracy, returning to execute step S908 until the fault data generation accuracy is not lower than the preset accuracy. The initial fault data obtained by the last iteration is taken as the fault data corresponding to the demand fault information of the aero-engine.

[0132] It should be understood that although each step in the flowchart involved in the embodiments described above is shown in sequence according to the direction of the arrow, these steps are not necessarily executed in the order indicated by the arrow. Unless otherwise specified herein, there is no strict order limitation for the execution of these steps, and these steps can be executed in other orders. Moreover, at least part of the steps in the flowchart involved in the embodiments described above can include multiple steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily sequential, but can be alternately executed with at least part of other steps or steps or stages in other steps.

[0133] Based on the same inventive concept, the embodiments of the present application also provide an aero-engine fault data generation device for implementing the aero-engine fault data generation method described above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more aero-engine fault data generation device embodiments provided below can refer to the limitations of the aero-engine fault data generation method described above, which will not be repeated here.

[0134] In one exemplary embodiment, as shown in Figure 10 An aero-engine fault data generation device is provided, comprising: an acquisition module 1010, an identification module 1020, and a generation module 1030, wherein:

[0135] The acquisition module 1010 is configured to acquire normal engine data of an aero-engine, fault engine data of the aero-engine, each sample fault data of the aero-engine, and fault data requirement information of the aero-engine, and generate a fault data generation model of the aero-engine based on the normal engine data and the fault engine data;

[0136] The identification module 1020 is configured to perform model adjustment processing on the fault data generation model based on each sample fault data to obtain a multi-scenario fault data generation model, and identify requirement scenario information of the aero-engine based on the fault data requirement information of the aero-engine;

[0137] The generation module 1030 is configured to generate fault data corresponding to the requirement scenario information of the aero-engine by the multi-scenario fault data generation model based on the requirement scenario information.

[0138] Optionally, the acquisition module 1010 is specifically configured to:

[0139] identify first engine air path parameters of the aero-engine in a normal state and second engine air path parameters of the aero-engine in an abnormal state based on the normal engine data and the fault engine data;

[0140] divide the first engine air path parameters into first sub-air path parameters corresponding to each working condition by a preset working condition division strategy, and divide the second engine air path parameters into second sub-air path parameters corresponding to each working condition by the preset working condition division strategy;

[0141] identify normal data distribution information of the aero-engine in a normal state and fault data distribution information of the aero-engine in a fault state based on the first sub-air path parameters corresponding to each working condition and the second sub-air path parameters corresponding to each working condition;

[0142] construct a fault data generation model of the aero-engine based on the normal data distribution information and the fault data distribution information.

[0143] Optionally, the obtaining module 1010 is specifically configured to:

[0144] train an initial data generation model based on the normal data distribution information to obtain a data generation model, and identify an encoder parameter in the data generation model;

[0145] migrate the encoder parameter to an initial fault generation model to obtain an initial fault data generation model, and train an output layer of the initial fault data generation model based on the fault data distribution information to obtain a fault data generation model.

[0146] Optionally, the identifying module 1020 is specifically configured to:

[0147] identify a fault type label corresponding to each sample fault data, and divide each sample fault data into a sample fault data group of each fault type label;

[0148] query a fault scene corresponding to each fault type label, and respectively adjust an output layer parameter of the fault data generation model based on the sample fault data group of each fault type label to obtain a sub-fault data generation model of the fault scene corresponding to each fault type label;

[0149] take all sub-fault data generation models of fault scenes as a multi-scene fault data generation model.

[0150] Optionally, the identifying module 1020 is specifically configured to:

[0151] Based on the fault data demand information, demand fault information of the aero-engine is identified, and based on the demand fault information, a target fault type required by the fault data demand information is identified;

[0152] Based on the correspondence between each fault type and each fault scene, a target fault scene corresponding to the target fault type is identified;

[0153] The target fault scene is taken as demand scene information of the aero-engine.

[0154] Optionally, the generation module 1030 is specifically configured to:

[0155] Based on the target fault scene, each initial fault data corresponding to the target fault scene is generated through the multi-scene fault data generation model, and a fault classifier corresponding to the target fault type is obtained;

[0156] Based on the fault classifier, data classification verification processing is performed on each initial fault data to obtain fault data generation accuracy, and in a case where the fault data generation accuracy is lower than a preset accuracy, the step of performing model adjustment processing on the fault data generation model based on each sample fault data is returned to perform until the fault data generation accuracy is not lower than the preset accuracy, each initial fault data obtained through the last iteration is taken as fault data corresponding to the demand fault information of the aero-engine.

[0157] Each module in the above-mentioned aero-engine fault data generation device can be realized by software, hardware and a combination thereof in whole or in part. Each module can be embedded in or independent of a processor in a computer device in a hardware form, or can be stored in a memory in a computer device in a software form, so as to be called and executed by a processor to perform operations corresponding to each module.

[0158] In one exemplary embodiment, a computer device is provided, which can be a terminal, and an internal structure diagram thereof can be as shown in Figure 11As shown in the figure. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. Among them, the processor, the memory and the input / output interface are connected through the system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capability. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The input / output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used to communicate with the external terminal in a wired or wireless manner. The wireless manner can be realized through WIFI, mobile cellular network, NFC (near field communication) or other technologies. The computer program is executed by the processor to realize an aero-engine fault data generation method. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0159] Those skilled in the art can understand that, Figure 11 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0160] In one exemplary embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to realize the steps corresponding to the aero-engine fault data generation method.

[0161] In one embodiment, a computer readable storage medium is provided, which stores a computer program, and the computer program is executed by the processor to realize the steps corresponding to the aero-engine fault data generation method.

[0162] In one embodiment, a computer program product is provided, including a computer program, and the computer program is executed by the processor to realize the steps corresponding to the aero-engine fault data generation method.

[0163] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant regulations.

[0164] It can be understood by those skilled in the art that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing related hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, it can include the processes of the above-mentioned embodiments of each method. Any reference to memory, database or other medium used in the embodiments provided by the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The database involved in the embodiments provided by the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided by the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without being limited thereto.

[0165] The technical features of the above embodiments can be combined in any way. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described, but as long as the combination of the technical features does not exist contradictory, it should be considered as the scope of the present application.

[0166] The above-described embodiments are merely illustrative of several embodiments of the present application, and the description is relatively specific and detailed, but should not be understood as a limitation on the scope of the patent. It should be noted that for those skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are all within the scope of the present application. Therefore, the scope of protection of the present application should be subject to the appended claims.

Claims

1. A method of generating aircraft engine fault data, the method comprising: The method comprises: normal engine data of an aero-engine, failure engine data of the aero-engine, each sample failure data of the aero-engine, and failure data requirement information of the aero-engine are acquired, and a failure data generation model of the aero-engine is generated based on the normal engine data and the failure engine data; model adjustment processing is performed on the failure data generation model based on each sample failure data, a multi-scenario failure data generation model is obtained, and requirement scenario information of the aero-engine is identified based on the failure data requirement information of the aero-engine; the failure data corresponding to the requirement scenario information of the aero-engine is generated through the multi-scenario failure data generation model based on the requirement scenario information.

2. The method of claim 1, wherein, The failure data generation model of the aero-engine is generated based on the normal engine data and the failure engine data, comprising: first engine air path parameters of the aero-engine in a normal state and second engine air path parameters of the aero-engine in an abnormal state are identified based on the normal engine data and the failure engine data; the first engine air path parameters are divided into first sub-air path parameters corresponding to each working condition through a preset working condition division strategy, and the second engine air path parameters are divided into second sub-air path parameters corresponding to each working condition through the preset working condition division strategy; normal data distribution information of the aero-engine in a normal state and failure data distribution information of the aero-engine in a failure state are identified based on the first sub-air path parameters corresponding to each working condition and the second sub-air path parameters corresponding to each working condition; the failure data generation model of the aero-engine is constructed based on the normal data distribution information and the failure data distribution information.

3. The method of claim 2, wherein, The failure data generation model of the aero-engine is constructed based on the normal data distribution information and the failure data distribution information, comprising: an initial data generation model is trained based on the normal data distribution information to obtain a data generation model, and an encoder parameter in the data generation model is identified; the encoder parameter is migrated to an initial failure generation model to obtain an initial failure data generation model, and an output layer of the initial failure data generation model is trained based on the failure data distribution information to obtain a failure data generation model.

4. The method of claim 1, wherein, The failure data generation model is subjected to model adjustment processing based on each sample failure data to obtain a multi-scenario failure data generation model, comprising: each sample failure data is divided into a sample failure data group of each failure type label by identifying a failure type label corresponding to each sample failure data; a failure scenario corresponding to each failure type label is queried, and an output layer parameter of the failure data generation model is adjusted based on the sample failure data group of each failure type label to obtain a sub-failure data generation model of a failure scenario corresponding to each failure type label; The sub-failure data generation model of all failure scenarios is generated as a multi-scenario failure data generation model.

5. The method of claim 1, wherein, The demand information of the failure data of the aero-engine is used to identify the demand scenario information of the aero-engine, including: Based on the demand information of the failure data, the demand failure information of the aero-engine is identified, and based on the demand failure information, the target failure type required by the demand information of the failure data is identified; Based on the correspondence between each failure type and each failure scenario, the target failure scenario corresponding to the target failure type is identified; The target failure scenario is used as the demand scenario information of the aero-engine.

6. The method of claim 5, wherein, Based on the target failure scenario, the multi-scenario failure data generation model is used to generate the failure data corresponding to the target failure scenario, and the failure classifier corresponding to the target failure type is obtained; Based on the failure classifier, the data classification verification processing is performed on each initial failure data to obtain the failure data generation accuracy, and in the case that the failure data generation accuracy is lower than the preset accuracy, the model adjustment processing based on each sample failure data on the failure data generation model is performed to obtain the multi-scenario failure data generation model step, until the failure data generation accuracy is not lower than the preset accuracy, each initial failure data obtained by the last iteration is used as the failure data corresponding to the demand failure information of the aero-engine. The device comprises:

7. An apparatus for generating engine fault data, comprising: An acquisition module is configured to acquire normal engine data of an aero-engine, failure engine data of the aero-engine, each sample failure data of the aero-engine, and demand information of failure data of the aero-engine, and generate a failure data generation model of the aero-engine based on the normal engine data and the failure engine data; An identification module is configured to perform model adjustment processing on the failure data generation model based on each sample failure data to obtain a multi-scenario failure data generation model, and identify demand scenario information of the aero-engine based on the demand information of failure data of the aero-engine; A generation module is configured to generate failure data corresponding to the demand scenario information of the aero-engine based on the demand scenario information through the multi-scenario failure data generation model. The processor executes the computer program to realize the steps of the method of any one of claims 1 to 6.

8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that, The computer program is executed by the processor to realize the steps of the method of any one of claims 1 to 6.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to realize the steps of the method of any one of claims 1 to 6.

10. A computer program product comprising a computer program, characterized in that, ​

Images