Method and system for predicting residual life of aero-engine

Abstract

The invention provides a prediction method and system for the residual life of an aero-engine, and belongs to the technical field of life prediction, and the method comprises the steps: obtaining an aero-engine sample data set of a source domain and a target domain; dividing original target domain sample data into a small amount of labeled data and a large amount of unlabeled data; combining the target domain unlabeled data with the source domain data, inputting the combined data into an evolution hierarchical long-short-term memory network for feature extraction, and determining degradation trend features; reducing the feature distribution difference of the degradation trend features by using a domain adaptive method based on dual-distribution migration alignment, and constructing a feature-migration model; training the feature migration model by using a small amount of labeled data of the target domain; and inputting real-time target domain data into the trained feature-migration model, and determining a prediction result of the aero-engine. According to the method, high-precision prediction of the residual life of the aero-engine under the condition of large data distribution difference is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of life prediction, and in particular to a method and system for predicting the remaining life of an aircraft engine. Background Art

[0002] The remaining useful life prediction of an aircraft engine refers to the prediction of how long the engine can continue to operate safely in the future, i.e. its "remaining useful life", based on the historical operating data of the aircraft engine, sensor monitoring data and the degradation pattern of the equipment. RUL prediction can help airlines, aviation maintenance units and manufacturers understand the health status of the engine in advance and take appropriate repair, replacement or maintenance measures, thereby reducing the risk of failure and improving safety and operational efficiency.

[0003] As the aviation industry's demand for intelligent and digital management increases, data-driven RUL prediction methods can improve operational efficiency, optimize resource scheduling, ensure equipment reliability and compliance, meet increasingly stringent aviation safety regulations, identify potential equipment failure risks in advance, effectively improve flight safety, avoid sudden failures and excessive maintenance, reduce maintenance costs and reduce downtime.

[0004] However, traditional life prediction methods cannot effectively obtain labeled sample data sets. A large number of target domain labels are unknown, which makes the prediction error of life prediction too high and limits the performance of the model. The feature mapping of traditional methods cannot effectively narrow the feature distribution differences of the data, resulting in insufficient feature alignment accuracy, thus affecting the accuracy of the prediction. Summary of the invention

[0005] In order to solve the technical problems that traditional life prediction methods cannot effectively obtain labeled sample data sets, a large number of target domain labels are unknown, which makes the prediction error of life prediction too high, the performance of the model is limited, and the feature mapping of traditional methods cannot effectively reduce the feature distribution differences of data, resulting in insufficient feature alignment accuracy, thereby affecting the accuracy of prediction, the present invention provides a method and system for predicting the remaining life of an aircraft engine.

[0006] The technical solution provided by the embodiment of the present invention is as follows: First aspect: An embodiment of the present invention provides a method for predicting the remaining life of an aircraft engine, comprising: S1: Obtain a sample data set of aircraft engines, where the sample data set includes source domain data and target domain data; S2: Divide the target domain data into target domain labeled data and target domain unlabeled data; S3: Combine the target domain unlabeled data with the source domain data to generate mixed data; S4: Input the mixed data into the evolutionary hierarchical long short-term memory network for feature extraction to determine the degradation trend characteristics; S5: Use a domain adaptation method based on dual distribution transfer alignment to reduce the feature distribution differences of degradation trend features; S6: Construct feature-transfer model based on feature distribution differences; S7: Input the labeled data of the target domain into the feature-transfer model for training; S8: Acquire real-time target domain data; S9: Input the real-time target domain data into the trained feature-transfer model to determine the prediction results of the aircraft engine.

[0007] Second aspect: An embodiment of the present invention provides a system for predicting the remaining life of an aircraft engine, comprising: processor; A memory having computer-readable instructions stored therein, wherein when the computer-readable instructions are executed by the processor, the method for predicting the remaining life of an aircraft engine according to the first aspect is implemented.

[0008] The third aspect: An embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for predicting the remaining life of an aircraft engine as described in the first aspect.

[0009] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least: In the present invention, feature extraction is performed through evolutionary hierarchical long short-term memory networks to capture specific degradation information, making the prediction results more accurate and comprehensive. The dual distribution migration alignment method is used to reduce the feature distribution differences of degradation trend features, improve the accuracy and robustness of the feature migration model, and make full use of the labeled sample data set through the feature migration model to improve the prediction accuracy of the remaining life of aircraft engines. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0011] Figure 1 A schematic flow chart of a method for predicting the remaining life of an aircraft engine provided by an embodiment of the present invention; Figure 2 A schematic diagram of the structure of an evolutionary hierarchical long short-term memory network provided by an embodiment of the present invention; Figure 3 A schematic diagram of the alignment condition distribution provided by an embodiment of the present invention; Figure 4 A schematic diagram of a feature-model transfer learning framework based on DCTA provided in an embodiment of the present invention; Figure 5 A schematic diagram of the structure of a system for predicting the remaining life of an aircraft engine provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0012] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0013] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "example" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "example" is intended to present the concept in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either of the two.

[0014] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same. "of", "corresponding, relevant" and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same.

[0015] In the embodiments of the present invention, sometimes the subscripts such as W 1 It may be written in non-subscript form such as W1. When the difference is not emphasized, the meaning is the same.

[0016] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0017] Reference Manual Attached Figure 1 , showing a flow chart of a method for predicting the remaining life of an aircraft engine provided by an embodiment of the present invention.

[0018] An embodiment of the present invention provides a method for predicting the remaining life of an aircraft engine, the method comprising: S1: Obtain a sample data set of aircraft engines, where the sample data set includes source domain data and target domain data.

[0019] In a possible implementation manner, the sample data set includes: source domain data and target domain data.

[0020] The source domain data is the degradation data of historical aircraft engines, and the target domain data is specifically the operating data of new aircraft engines.

[0021] It should be noted that the combination of source domain data and target domain data provides rich training data for the model, which effectively improves the prediction accuracy of the model on new engines.

[0022] Reference Manual Attached Figure 2 , showing a schematic diagram of the structure of an evolutionary hierarchical long short-term memory network provided by an embodiment of the present invention.

[0023] S2: Divide the target domain data into target domain labeled data and target domain unlabeled data; Specifically, the target domain data is divided into a small amount of target domain labeled data and a large amount of target domain unlabeled data. In the present invention, the ratio of target domain labeled data to target domain unlabeled data is specifically 3:7, which is conducive to transfer learning.

[0024] S3: Combine the target domain unlabeled data with the source domain data to generate mixed data.

[0025] S4: Input the mixed data into the evolutionary hierarchical long short-term memory network for feature extraction to determine the degradation trend characteristics.

[0026] Among them, the Evolutionary Graded Long Short-Term Memory (EG-LSTM) network processes data of different importance levels in a hierarchical manner by introducing multi-head attention mechanism, dynamic gate control and residual connection, so as to more accurately capture the degradation trend of the equipment.

[0027] It should be noted that EG-LSTM can automatically learn and extract multi-level features in the equipment degradation process, accurately capture long-term, medium-term and short-term degradation trends, and improve the comprehensiveness and accuracy of equipment health status modeling.

[0028] In a possible implementation, the evolutionary hierarchical long short-term memory network includes a hierarchical division module, a multi-unit module and a gating module, and the hierarchical division module includes a multi-head attention mechanism and a Softmax function.

[0029] Among them, the multi-head attention mechanism is a technology for calculating attention weights. It weights the input data through multiple independent attention heads to capture the data dependencies of different parts. Softmax is a commonly used normalization function that converts the input value into a probability distribution so that the sum of the weighted values ​​of each input is 1.

[0030] S4 specifically includes: S401: Use the multi-head attention mechanism to calculate the attention weight of mixed data: ; ; ; ; Among them, Attention represents the multi-head attention mechanism function, Softmax represents the Softmax activation function, Indicates a query, represents the learnable weights of the query matrix, x t express t The input data vector at time instant, K Indicates the key, W K represents the learnable weights of the bond matrix, h t-1 express t -1 moment of hidden state, V Indicates the value, W V represents the learnable weights of the value matrix, Indicates the dimension size of the key vector.

[0031] S402: According to the attention weight, the Softmax function is used to obtain the level weight vector of the mixed data: ; in, A 1 represents the level weight vector of mixed data, A MHA represents the output of the multi-head attention mechanism, , , , and They respectively represent the weight values ​​of high-level mixed data, medium-high-level mixed data, medium-level mixed data, medium-low-level mixed data and low-level mixed data.

[0032] S403: Based on the level weight vector, determine the multi-level degradation trend through the multi-unit module.

[0033] It should be noted that weighting the degradation trends at different levels through the grade weight vector can ensure that the model can accurately capture the multi-level degradation characteristics according to the importance of different degradation stages, thereby improving the accuracy and reliability of the remaining life prediction.

[0034] S404: According to the output of the multi-level degradation trend multi-cell unit, the degradation trend characteristics are determined through the gating module: ; ; ; ; ; ; in, i t represents the input gate, f t represents the forget gate, o t represents the output gate, express t The candidate memory state at time, c t express t The state of the memory cell is determined at every moment. σ represents the Sigmoid function, w xi Represents input data x t to the weight matrix of the input gate, w hi Indicates hidden state ht -1 to the weight matrix of the input gate, b i represents the bias vector of the input gate, Represents input data x t To the weight matrix of the forget gate, Indicates hidden state ht -1 to the weight matrix of the forget gate, represents the bias vector of the forget gate, Represents input data x t The weight matrix to the output gate, Indicates hidden state ht -1 to the weight matrix of the output gate, represents the bias vector of the output gate, Represents input data x t to the weight matrix of the candidate memory state, Indicates hidden state ht -1 to the weight matrix of the candidate memory state, b crepresents the bias vector of the candidate memory state, tanh represents the tanh activation function, ht represents the hidden state at time t, s1 and s2 represent the gating signals, represents the increment operator, Represents an element-wise multiplication operation.

[0035] Among them, the input gate determines the current candidate memory state to be written into the memory unit The ratio of the forget gate to the memory unit c t-1 The proportion of information in the memory that is forgotten, the output gate determines the state of the memory unit c t After the activation function, it is output to the hidden state h t The ratio of candidate memory states Combine the current input and hidden state to generate new candidate memory content, hidden state h t The memory cell state c t The content in is output to the hidden state through the output gate control h t .

[0036] In one possible implementation, the multi-unit module includes high-level units, low-level units, medium-level units, medium-high-level units, and medium-low-level units.

[0037] S403 is specifically: S4031: Determine the degradation trend of high-level units using long-term trend update rules.

[0038] S4032: Determine the degradation trend of low-level units using short-term trend update rules.

[0039] S4033: Determine the degradation trend of the mid-level units using the mid-term trend update rule.

[0040] S4034: Determine the degradation trend of medium-high level units and the degradation trend of medium-low level units using the proportional update rule.

[0041] S4035: Determine a multi-level degradation trend by combining the degradation trend of high-level units, the degradation trend of low-level units, the degradation trend of medium-level units, the degradation trend of medium-high-level units, and the degradation trend of medium-low-level units.

[0042] In a possible implementation, the long-term trend update rule is specifically: ; ; in, ct ( m ) represents the memory state of the high-level unit, ct -1 indicates the time t -1 memory state, Δ c t represents the increment of the long-term trend, tanh represents the tanh activation function, Represents input data x t To the weight matrix of the long-term trend increment, Indicates hidden state ht A weight matrix ranging from -1 to long-term trend increments, Indicates the bias of the long-term trend increment; The short-term trend update rules are as follows: ; in, c t ( i ) represents a low-level unit, Indicates time t Candidate memory states of The specific rules for updating the mid-term trend are: ; in, c t ( k ) represents the middle level unit, represents the forget gate, i t represents the input gate; The specific proportion update rules are: ; ; in, c t ( l ) indicates medium and high level units, c t ( j ) indicates low- to medium-level units; The multi-level degradation trends are as follows: ; in, c t express t The state of the memory cell is determined at every moment. A 1 Represents the class weight vector of the sample dataset.

[0043] Reference Manual Attached Figure 3, showing a schematic diagram of the alignment condition distribution provided by an embodiment of the present invention.

[0044] like Figure 3 The input includes the conditional distribution of the source domain and the target domain. The goal is to reduce the distribution difference by embedding it into the Reproducing Kernel Hilbert Space (RKHS), performing first-order matching and second-order matching, and calculating the conditional feature distribution and covariance matrix of the source domain and the target domain respectively. Finally, the conditional distribution of the source domain and the target domain is made as consistent as possible through optimization to achieve cross-domain alignment.

[0045] S5: Use a domain adaptation method based on dual distribution transfer alignment to reduce the feature distribution differences of degradation trend features.

[0046] Among them, the domain adaptation method based on dual distribution transfer alignment (DCTA) realizes the joint alignment of marginal distribution and conditional distribution by combining the domain adaptation methods of kernel Gaussian Wasserstein distance (KGW) and conditional operator discrepancy (COD), ensuring that the model can be effectively generalized under different operating conditions.

[0047] It should be noted that the domain adaptation method using dual distribution transfer alignment can effectively narrow the feature distribution differences between the source domain and the target domain, ensure more accurate predictions in the target domain, and improve the cross-domain adaptability of the model. In the case of large differences in data distribution, it can significantly reduce prediction errors and improve the accuracy of remaining life prediction.

[0048] In a possible implementation, S5 is specifically: The domain adaptation technology based on dual distribution migration alignment is used to perform marginal distribution alignment and conditional distribution alignment on the mixed data to narrow the feature distribution differences of the degradation trend characteristics.

[0049] In a possible implementation, edge distribution alignment of mixed data is specifically performed as follows: ; ; in, Indicates the KGW distance, Represents source domain data, represents the target domain data, and denote the mean embedding of source domain data and target domain data respectively, tr denotes the trace of the matrix, and Represent the covariance operators of source domain data and target domain data respectively, represents the covariance mixing term of the source domain data and the target domain data, Indicates i The weights of the basic kernel functions, Indicates i The basic kernel function, i =1,2,···, M , M Represents the total number of kernel functions.

[0050] Specifically, the kernel function Used to map input data to high-dimensional feature space, the choice of different kernel functions will affect the feature mapping effect. To adapt to complex nonlinear distributions, KGW combines multiple kernel functions (such as RBF kernel, polynomial kernel, etc.) and dynamically adjusts the weights of different kernels through weighted combination.

[0051] The specific steps for conditional distribution alignment of mixed data are: ; ; ; ; ; in, The distance representing the difference of conditional operators, represents the conditional distribution of source domain data under a given label Y, represents the conditional distribution of the target domain data under a given label Y, represents the difference in conditional mean between source domain data and target domain data, represents the difference in conditional covariance, represents the conditional mean operator, represents the conditional covariance operator, represents the cosquare matrix between X and Y, represents the cosquare matrix inside Y, Represents the reverse operation, represents the covariance within X, represents the covariance between Y and X, Representation Tags y The weight of y c Tag values ​​representing key lifecycle stages, y i represents all labels that need to be weighted during conditional distribution alignment, α represents the hyperparameter controlling weight decay, represents the loss of conditional distribution alignment and exp represents the exponential function.

[0052] Reference Manual Attached Figure 4 , showing a schematic diagram of a DCTA-based feature-model transfer learning framework provided in an embodiment of the invention.

[0053] like Figure 4 The source domain (labeled data) and target domain (labeled and unlabeled data) adapt the source domain knowledge to the target domain through feature migration and model migration. The network consists of an EG-LSTM layer, a fully connected layer, and an output layer. The migration process is optimized in combination with a domain adaptation module. The optimization objectives include mean square error, domain alignment, and aligned marginal distribution (KGW) and conditional distribution (COD) constraints, aiming to improve the accuracy and robustness of cross-domain predictions.

[0054] S6: Construct a feature-transfer model based on feature distribution differences.

[0055] S7: Input the labeled data of the target domain into the feature-transfer model for training.

[0056] Among other things, the label refers to the remaining useful life of the aircraft engine.

[0057] It should be noted that by inputting labeled data from the target domain into the feature transfer model for training, a small amount of labeled data in the target domain can be used to fine-tune the model, which can improve the model's adaptability to the target domain characteristics, better learn the target domain-specific degradation patterns, optimize its prediction accuracy in the target domain, and improve the accuracy and robustness of the remaining life prediction.

[0058] In a possible implementation manner, S7 specifically includes: The labeled data of the target domain is input into the feature transfer model for training until the total loss function is less than the preset loss function.

[0059] In a possible implementation, the calculation formula of the total loss function is specifically: ; ; in, Loss represents the total loss function, L src represents the loss function of the source domain data, λ 1 and λ 2 Both represent trade-off parameters, and They represent the losses of conditional distribution alignment and marginal distribution alignment, Indicates the source domain datai The true value of the data, Indicates the source domain data i The predicted value of data, n s Represents the total number of samples in the source domain data, represents the square of the 2-norm.

[0060] S8: Acquire real-time target domain data; S9: Input the real-time target domain data into the trained feature-transfer model to determine the prediction results of the aircraft engine.

[0061] The present invention combines the unlabeled data in the target domain with the labeled data in the source domain, and inputs them into the evolutionary hierarchical long short-term memory network to extract features. Then, the domain adaptation technology based on dual distribution migration alignment is used to reduce the difference in their feature distributions and obtain a feature migration model. On the basis of this model, a small amount of labeled data in the target domain is used to fine-tune the model to adapt it to the target domain tasks, ultimately achieving high-precision RUL prediction.

[0062] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least: In the present invention, feature extraction is performed through evolutionary hierarchical long short-term memory networks to capture specific degradation information, making the prediction results more accurate and comprehensive. The dual distribution migration alignment method is used to reduce the feature distribution differences of degradation trend features, improve the accuracy and robustness of the feature migration model, and make full use of the labeled sample data set through the feature migration model to improve the prediction accuracy of the remaining life of aircraft engines.

[0063] Reference Manual Attached Figure 5 , showing a schematic structural diagram of a system for predicting the remaining life of an aircraft engine provided by the present invention.

[0064] The present invention further provides a system 20 for predicting the remaining life of an aircraft engine, which is applied to the above-mentioned method for predicting the remaining life of an aircraft engine, and comprises: Processor 201.

[0065] The memory 202 stores computer-readable instructions. When the computer-readable instructions are executed by the processor 201, the method for predicting the remaining life of an aircraft engine as in the method embodiment is implemented.

[0066] The aircraft engine remaining life prediction system 20 provided by the present invention can execute the above-mentioned aircraft engine remaining life prediction method and achieve the same or similar technical effects. To avoid repetition, the present invention will not go into details.

[0067] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least: In the present invention, feature extraction is performed through evolutionary hierarchical long short-term memory networks to capture specific degradation information, making the prediction results more accurate and comprehensive. The dual distribution migration alignment method is used to reduce the feature distribution differences of degradation trend features, improve the accuracy and robustness of the feature migration model, and make full use of the labeled sample data set through the feature migration model to improve the prediction accuracy of the remaining life of aircraft engines.

[0068] It should be understood that the processor in the embodiment of the present invention may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0069] It should also be understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0070] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When a computer instruction or computer program is loaded or executed on a computer, a process or function according to an embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0071] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0072] In the present invention, "at least one" means one or more, and "plurality" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0073] It should be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0074] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0075] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described equipment, devices and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0076] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0077] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0078] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0079] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0080] An embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the method for predicting the remaining life of an aircraft engine as described in the method embodiment is implemented.

[0081] A computer-readable storage medium provided by the present invention can implement the steps and effects of the method for predicting the remaining life of an aircraft engine in the above method embodiment. To avoid repetition, the present invention will not go into details.

[0082] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least: In the present invention, feature extraction is performed through evolutionary hierarchical long short-term memory networks to capture specific degradation information, making the prediction results more accurate and comprehensive. The dual distribution migration alignment method is used to reduce the feature distribution differences of degradation trend features, improve the accuracy and robustness of the feature migration model, and make full use of the labeled sample data set through the feature migration model to improve the prediction accuracy of the remaining life of aircraft engines.

[0083] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art who is familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0084] There are a few points to note: (1) The drawings of the embodiments of the present invention only involve structures related to the embodiments of the present invention. Other structures may refer to conventional designs.

[0085] (2) For the sake of clarity, in the drawings used to describe the embodiments of the present invention, the thickness of layers or regions is exaggerated or reduced, that is, these drawings are not drawn according to the actual scale. It is understood that when an element such as a layer, film, region or substrate is referred to as being "on" or "under" another element, the element may be "directly" "on" or "under" the other element or there may be intermediate elements.

[0086] (3) In the absence of conflict, the embodiments of the present invention and the features therein may be combined with each other to obtain new embodiments.

[0087] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. The protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for predicting the remaining life of an aircraft engine, characterized in that: include: S1: Acquire a sample data set of aircraft engines, wherein the sample data set includes source domain data and target domain data; S2: dividing the target domain data into target domain labeled data and target domain unlabeled data; S3: Combining the target domain unlabeled data with the source domain data to generate mixed data; S4: inputting the mixed data into an evolutionary hierarchical long short-term memory network for feature extraction to determine degradation trend characteristics; S5: using a domain adaptation method based on dual distribution transfer alignment to reduce the feature distribution difference of the degradation trend feature; S6: constructing a feature-migration model based on the feature distribution differences; S7: inputting the target domain labeled data into the feature-transfer model for training; S8: Acquire real-time target domain data; S9: Inputting the real-time target domain data into the trained feature-transfer model to determine the prediction result of the aircraft engine.

2. The method for predicting the remaining life of an aircraft engine according to claim 1, characterized in that: The source domain data is degradation data of historical aircraft engines, and the target domain data is specifically operation data of new aircraft engines.

3. The method for predicting the remaining life of an aircraft engine according to claim 1, characterized in that: The evolutionary hierarchical long short-term memory network includes a hierarchical division module, a multi-unit module and a gating module, wherein the hierarchical division module includes a multi-head attention mechanism and a Softmax function; The S4 specifically includes: S401: Calculate the attention weight of the mixed data using the multi-head attention mechanism: ； ； ； ； Among them, Attention represents the multi-head attention mechanism function, Softmax represents the Softmax activation function, Q Indicates a query, W Q represents the learnable weights of the query matrix, x t express t The input data vector at time instant, K Indicates the key, W K represents the learnable weights of the bond matrix, h t-1 express t -1 moment of hidden state, V Indicates the value, W V represents the learnable weights of the value matrix, d k Indicates the dimension size of the key vector; S402: According to the attention weight, the Softmax function is used to obtain the level weight vector of the mixed data: ; in, A 1 represents the level weight vector of mixed data, A MHA represents the output of the multi-head attention mechanism, , , , and Respectively represent the weight values ​​of high-level mixed data, medium-high-level mixed data, medium-level mixed data, medium-low-level mixed data and low-level mixed data; S403: Determine multi-level degradation trends through the multi-unit module based on the level weight vector; S404: Determine the degradation trend feature through the gating module according to the output of the multi-level degradation trend multi-cell unit: ; ; ; ; ； ； Among them, it represents the input gate, represents the forget gate, ot represents the output gate, express t The candidate memory state at time, c t express t The state of the memory cell is determined at every moment. σ represents the Sigmoid function, wxi represents the weight matrix from the input data xt to the input gate, and whi represents the hidden state ht -1 is the weight matrix of the input gate, bi represents the bias vector of the input gate, Represents input data x t To the weight matrix of the forget gate, Indicates hidden state ht -1 to the weight matrix of the forget gate, represents the bias vector of the forget gate, Represents input data x t The weight matrix to the output gate, Indicates hidden state ht -1 to the weight matrix of the output gate, represents the bias vector of the output gate, Represents input data x t to the weight matrix of the candidate memory state, Indicates hidden state ht -1 to the weight matrix of the candidate memory state, b c represents the bias vector of the candidate memory state, tanh represents the tanh activation function, ht represents the hidden state at time t, s1 and s2 represent the gating signals, represents the increment operator, Represents an element-wise multiplication operation.

4. The method for predicting the remaining life of an aircraft engine according to claim 3, characterized in that: The multi-unit module includes high-level units, low-level units, medium-level units, medium-high-level units and medium-low-level units; The S403 is specifically as follows: S4031: Determine the degradation trend of high-level units using long-term trend update rules; S4032: Determine the degradation trend of low-level units using short-term trend update rules; S4033: Determine the degradation trend of the mid-level units using the mid-term trend update rule; S4034: Determine the degradation trend of the medium-high level unit and the degradation trend of the medium-low level unit by using the proportional update rule; S4035: Determine a multi-level degradation trend by combining the high-level unit degradation trend, the low-level unit degradation trend, the medium-level unit degradation trend, the medium-high-level unit degradation trend and the medium-low-level unit degradation trend.

5. The method for predicting the remaining life of an aircraft engine according to claim 4, characterized in that: The long-term trend update rule is specifically: ; ; in, c t ( m ) represents the memory state of the high-level unit, ct -1 indicates the time t -1 memory state, Δ c t represents the increment of the long-term trend, tanh represents the tanh activation function, Represents input data x t To the weight matrix of the long-term trend increment, Indicates hidden state ht A weight matrix ranging from -1 to long-term trend increments, Indicates the bias of the long-term trend increment; The short-term trend update rules are specifically as follows: ; in, c t ( i ) indicates a low-level unit, Indicates time t Candidate memory states of The mid-term trend update rules are specifically: ; in, c t ( k ) represents the middle level unit, represents the forget gate, i t represents the input gate; The proportion update rule is specifically: ; ; in, c t ( l ) indicates medium and high level units, c t ( j ) indicates low- to medium-level units; The multi-level degradation trend is specifically: ; in, c t express t The state of the memory cell is determined at every moment. A 1 represents the class weight vector for mixed data.

6. The method for predicting the remaining life of an aircraft engine according to claim 1, characterized in that: The S5 is specifically: The mixed data is aligned with the edge distribution and the conditional distribution by using the domain adaptation technology based on the dual distribution migration alignment, so as to reduce the feature distribution difference of the degradation trend feature.

7. The method for predicting the remaining life of an aircraft engine according to claim 6, characterized in that: The edge distribution alignment of the mixed data is specifically as follows: ; ; in, Indicates the KGW distance, Represents source domain data, represents the target domain data, and denote the mean embedding of source domain data and target domain data respectively, tr denotes the trace of the matrix, and Represent the covariance operators of source domain data and target domain data respectively, represents the covariance mixing term of the source domain data and the target domain data, Indicates i The weights of the basic kernel functions, Indicates i The basic kernel function, i =1,2,···, M , M Represents the total number of kernel functions; The conditional distribution alignment of the mixed data is specifically as follows: ; ; ; ; ; in, The distance representing the difference of conditional operators, represents the conditional distribution of source domain data under a given label Y, represents the conditional distribution of the target domain data under a given label Y, represents the difference in conditional mean between source domain data and target domain data, represents the difference in conditional covariance, represents the conditional mean operator, represents the conditional covariance operator, represents the cosquare matrix between X and Y, represents the cosquare matrix inside Y, Represents the reverse operation, represents the covariance within X, represents the covariance between Y and X, Representation Tags y The weight of y c Tag values ​​representing key lifecycle stages, y i represents all labels that need to be weighted during conditional distribution alignment, α represents the hyperparameter controlling weight decay, represents the loss of conditional distribution alignment and exp represents the exponential function.

8. The method for predicting the remaining life of an aircraft engine according to claim 1, characterized in that: The S7 is specifically: The labeled data of the target domain is input into the feature-transfer model for training until the total loss function is less than the preset loss function.

9. The method for predicting the remaining life of an aircraft engine according to claim 8, characterized in that: The calculation formula of the total loss function is specifically: ; ; in, Loss represents the total loss function, L src represents the loss function of the source domain data, λ 1 and λ 2 represents the trade-off parameter, and They represent the losses of conditional distribution alignment and marginal distribution alignment, Indicates the source domain data i The true value of the data, Indicates the source domain data i The predicted value of data, n s Represents the total number of samples in the source domain data, represents the square of the 2-norm.

10. A system for predicting the remaining life of an aircraft engine, characterized in that: include: processor; A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the method for predicting the remaining life of an aircraft engine as claimed in any one of claims 1 to 9 is implemented.

Images