Method for predicting residual service life of aero-engine

Abstract

The invention discloses an aero-engine residual service life prediction method, belongs to the field of aero-engine residual service life prediction, and aims to solve the problems of an existing aero-engine RUL prediction model. According to the method, prediction is carried out through a parallel space-time Transform model, and the parallel space-time Transform model comprises a time attention module TAM, a space attention module SAM, a space-time feature fusion device and a support vector machine MLP; the TAM and the SAM are arranged in parallel, and input samples are respectively input into the TAM and the SAM; obtaining a time sequence enhancement feature through TAM; obtaining spatial enhancement features through SAM; fusing the time sequence enhancement features and the spatial enhancement features by using a spatial-temporal feature fusion device; and the support vector machine MLP predicts the remaining service life of the aero-engine according to the fusion features output by the spatial-temporal feature fusion device.

Description

Technical Field

[0001] The invention relates to a method for predicting the remaining service life of an aero-engine, and belongs to the field of predicting the remaining service life of an aero-engine. Background Art

[0002] As the main power source of aircraft, the operating status and health of aircraft engines are crucial to the normal operation of aircraft. Remaining useful life (RUL) prediction aims to provide information about the health status of the engine based on the collected monitoring data, helping decision makers to formulate reasonable use and maintenance plans, which can effectively prevent the engine from experiencing performance deterioration or even catastrophic failure. Therefore, accurately estimating the RUL of aircraft engines helps to improve the safety and reliability of engine operation.

[0003] The current mainstream RUL prediction methods can be roughly divided into three categories: model-based methods, data-driven methods, and hybrid model-based methods. First, model-based methods attempt to reveal the degradation or failure mechanism based on physical knowledge, which is quite difficult for aircraft engines. One of the main reasons is that aircraft engines are complex electromechanical hydraulic systems composed of tens of thousands of parts, and their state changes involve the coupling of many different components. So far, the degradation or failure mechanisms of many aircraft engines have not been clearly revealed. Similarly, although hybrid methods, model-based methods, and data-driven methods have high accuracy and reliability for RUL prediction, the complexity of modeling leads to certain inadaptability in practical engineering applications.

[0004] In contrast, data-driven methods are based on historically collected monitoring data and make RUL predictions with almost no domain knowledge, with good accuracy. In addition, data-driven methods can also be regularly updated with collected monitoring data, which enhances their adaptability to non-stationary operating conditions. More importantly, there are many sensors in aircraft engines and the amount of data collected is huge, which provides sufficient raw data for RUL estimation, making data-driven methods stand out among the three methods.

[0005] Aircraft engines go through thousands of cycles during their service life. Due to differences in operating levels and operating conditions, the collected monitoring data is often highly nonlinear. Compared with traditional data-driven methods, deep learning has a powerful nonlinear feature extraction capability. It can automatically extract useful and informative features from raw nonlinear monitoring data, and then use these features to predict the RUL of aircraft engines. The entire process does not require a complex modeling process, nor does it require engineers to specifically select statistical features, which makes the RUL prediction task simpler.

[0006] So far, many deep learning models, such as convolutional neural networks (CNNs), recurrent neural networks (RNNs), long short-term memory (LSTMs), temporal convolutional networks (TCNs), and deep residual networks, have been widely used in the field of aircraft engine RUL and have been proven to effectively improve the performance of life prediction. For the prediction task of short sequence data, CNNs and RNNs have demonstrated their excellent ability in life prediction. However, facing the challenge of long sequence data, traditional CNNs and RNNs are unable to cope with the lack of ability to establish long-term dependencies. Fortunately, with the emergence of LSTMs and TCNs, this problem has been alleviated to a certain extent. As an improved version of RNN, LSTM can effectively save and transmit important long-time span information by introducing the mechanism of input gates, forget gates, and output gates. This structure enables LSTM to capture key time dependencies when processing long sequence data, thus playing an important role in aircraft engine RUL prediction. TCN significantly enhances the model's ability to capture long-term dependencies by adopting scalable dilated convolution technology. This technique enables TCN to handle long-distance temporal dependencies with fewer parameters and higher efficiency, thus performing well in prediction tasks of long sequence data.

[0007] Although LSTM and TCN have improved the ability to capture long-term dependencies, there are still challenges when processing very long sequences. For example, when dealing with tasks that require capturing long-distance dependencies, TCN usually needs to enhance its capabilities by increasing the number of layers in the network or expanding the size of the convolution kernel. However, although these methods have improved the performance of the model to a certain extent, they have also inevitably increased the complexity of the model. More importantly, as the complexity of the model increases, the risk of overfitting also increases, which may affect the generalization ability of the model in practical applications. Although LSTM's unique gating mechanism enables it to capture long-term dependencies in time series well, it still faces the problem of gradient vanishing or exploding when processing long sequence data, which often affects the model's training process and the final prediction performance. In addition, LSTM's step-by-step training mechanism also limits its computational efficiency, making it incapable of processing large-scale data. The Transformer model has been proposed to solve the above problems well. Its unique self-attention mechanism can establish the dependency between any two data in a long sequence without being affected by the length of the sequence. In addition, the parallel computing capability of Transformer greatly reduces the time complexity of the training and testing process and alleviates problems such as gradient explosion. Although the above methods have achieved good prediction results in the field of aircraft engine RUL prediction, their prognostic performance is still limited by the following three inherent problems:

[0008] (1) There are many types of sensors used to monitor the health status of aircraft engines, such as temperature sensors, pressure sensors, etc. However, different types of monitoring parameters have different effects on RUL prediction, and only some parameters have a significant impact on RUL prediction. On the one hand, redundant parameters will lead to inefficient data analysis; on the other hand, it may also cause the deep learning model to over-focus on irrelevant data, ultimately leading to increased RUL prediction errors and increased time consumption.

[0009] (2) During the operation of aircraft engines, different flight states will cause the signals collected by sensors to contain different degradation information. For aircraft, some violent movements such as lifting and diving will make the engine more susceptible to fatigue damage, while stable flight is relatively safe. If we do not focus on violent movements during feature extraction, we may capture some useless information and amplify noise, which will have a certain impact on RUL prediction.

[0010] RUL is the result of the performance degradation of aircraft engines during their service life, which is reflected in the cumulative effect of various sensor parameters over time. In other words, the historical operating status of the engine has a significant impact on the remaining life prediction at the current moment. Therefore, when making RUL predictions, it is necessary to obtain the cumulative contribution of sensor parameters to performance degradation at different moments and enhance the contribution of those historical moments that have a more important impact on RUL.

[0011] (3) In the prediction of the remaining life of aircraft engines, if the sensor parameter dimension is taken as the spatial dimension and the change of sensor parameters over time is taken as the temporal dimension, the above three inherent problems require that the established prediction model needs to be able to extract both temporal dependency and spatial dependency at the same time. In recent years, the attention mechanism has attracted widespread attention due to its potential in identifying key information in spatial and temporal dimensions. Many researchers have integrated the attention mechanism into deep learning models to extract more representative features, thereby improving the accuracy of remaining life prediction. In this way, the model can more accurately identify and utilize degradation patterns in sensor data, thereby optimizing the prediction results.

[0012] However, most of the current attention-based methods only highlight important information from the temporal dimension or the spatial dimension. For aircraft engine RUL prediction, information in both spatial and temporal dimensions has an impact on the remaining life prediction. Therefore, in order to further improve the RUL prediction performance, some studies have introduced both temporal attention mechanisms and spatial attention mechanisms into deep learning models, paying attention to important information in both spatial and temporal dimensions. Compared with using temporal attention mechanisms or spatial attention mechanisms alone, the mixed use of temporal attention mechanisms and spatial attention mechanisms can further improve the prediction performance of the model because they can capture both temporal and spatial dependencies at the same time.

[0013] It is worth noting that in the current RUL prediction methods based on the hybrid attention mechanism, temporal attention and spatial attention are used in series, which will lead to a new challenge, that is, temporal attention and spatial attention will interfere with each other. At present, relevant studies have proved that temporal attention and spatial attention mechanisms will achieve different effects when they are in different front and back positions in the series structure, see Figure 1 As shown, the features first pass through the temporal attention mechanism and then the spatial attention mechanism, or first pass through the spatial attention mechanism and then the spatial attention mechanism. In other words, different arrangements in the series structure will cause the features strengthened by the temporal attention and the features strengthened by the spatial attention to influence each other, which may cause some important degenerate information to be lost. For example, if the temporal attention mechanism is applied first, it will strengthen the multidimensional time series from the time dimension, but the subsequent spatial attention mechanism focuses on the spatial features strengthened by the temporal attention, which may cause some of the original spatial relationships of the multidimensional time series to be lost. Similarly, applying the spatial attention mechanism first and then the temporal attention mechanism will also produce similar problems. On the contrary, if the temporal attention and spatial attention mechanisms are connected in parallel, see Figure 2 As shown in the figure, the features are independently subjected to the temporal attention and spatial attention mechanisms. The two attention mechanisms can independently extract temporal enhancement features and spatial enhancement features from the input multidimensional time series without affecting each other, and extract the temporal dependency and spatial dependency of the original sequence to the greatest extent. In addition, as mentioned earlier, RUL is the result of the degradation of aircraft engines during their service life as the operating time increases. The time accumulation effect of parameters is usually more important for RUL prediction. In the parallel connection structure, the features enhanced by temporal attention can also be given a greater weight to pay more attention to the role of temporal attention. The current difficulty is how to effectively integrate the features enhanced by temporal attention with the features enhanced by spatial attention so that they can fully benefit RUL prediction. Summary of the invention

[0014] In view of the problems existing in the existing aircraft engine RUL prediction model, the present invention provides an aircraft engine remaining useful life prediction method.

[0015] The invention discloses a method for predicting the remaining useful life of an aircraft engine. The remaining useful life of the aircraft engine is predicted by using a parallel spatiotemporal Transformer model. The parallel spatiotemporal Transformer model includes a temporal attention module TAM, a spatial attention module SAM, a spatiotemporal feature fuser and a support vector machine MLP.

[0016] The temporal attention module TAM and the spatial attention module SAM are set in parallel, and the input samples are input into the temporal attention module TAM and the spatial attention module SAM respectively; the temporal enhancement feature is obtained through the temporal attention module TAM; the spatial enhancement feature is obtained through the spatial attention module SAM;

[0017] The temporal enhancement feature and the spatial enhancement feature are fused by using the spatiotemporal feature fusion device. The spatiotemporal feature fusion device takes the temporal enhancement feature as the main body and the spatial enhancement feature as the auxiliary. The multi-head attention mechanism is used to transform the spatial enhancement feature into an attention matrix, and the attention matrix is ​​multiplied by the temporal enhancement feature to highlight the dominant position of the temporal enhancement feature.

[0018] The support vector machine MLP predicts the remaining useful life of aircraft engines based on the fusion features output by the spatiotemporal feature fuser.

[0019] Preferably, the process of obtaining the input sample is:

[0020] Collect different types of sensor parameters of various aircraft engine components under different operating conditions, such as temperature, pressure and speed, and construct original multi-dimensional time series with these sensor data;

[0021] A series of preprocessing operations are performed on the original multidimensional time series, including normalization, exponential smoothing, and sliding time window division;

[0022] After the above preprocessing operation, the input sample is obtained, which is a time series of T×m dimensions. Where T represents the length of the time series and m represents the number of sensors.

[0023] Preferably, the process of obtaining the temporal enhancement feature through the temporal attention module TAM includes:

[0024] Step 11: Perform average pooling along the spatial dimension to convert it into a time series of T×1 dimensions

[0025] Where GAP1D(·) represents one-dimensional global average pooling, express The i-th column in the figure represents the serial number of the sensor, i=1,2,...,m,

[0026] Step 12: Use multi-scale convolution to pool features Processing to obtain the temporal dependencies between different time steps:

[0027]

[0028] In the formula represents the output feature of the pth convolution branch in the multi-scale convolution layer, (k p ,b p ) represents the convolution kernel of the pth convolution branch and the corresponding bias vector, p = 1, 2, ..., n, n represents the number of convolution branches, * represents a one-dimensional convolution operation;

[0029] Step 13: Use the Sigmoid operator to process the output features of each convolution branch in the multi-scale convolution layer to obtain the importance of each time step at different time scales:

[0030]

[0031] In the formula represents the weight information of each time step at the pth time scale, and f(·) represents the Sigmoid activation function;

[0032] Step 14: Obtain the timing enhancement feature F according to the following formula TAM :

[0033]

[0034] In the formula, represents matrix multiplication, and n represents the number of branches in the multi-scale convolutional layer.

[0035] Preferably, the process of obtaining spatial enhancement features through the spatial attention module SAM includes:

[0036] Step 21: Input sample Get a multidimensional time series X∈R with m×T dimensions m×T ;

[0037] Step 22: Get the correlation matrix S:

[0038]

[0039] Where X = [α1, α2, ..., α m ] T , α i ∈R 1×T represents the row vector of X, is the inner product between the i-th and j-th sensor parameter vectors;

[0040] Step 23: Obtain the spatial enhancement feature F according to the following formula SAM :

[0041]

[0042] In the formula, Represents matrix addition.

[0043] Preferably, the process of fusing the temporal enhancement feature and the spatial enhancement feature using the spatiotemporal feature fuser includes:

[0044] Step 31: Enhance the time series features F TAM and spatial enhancement feature F SAM Perform position encoding to obtain fusion features:

[0045]

[0046] in, are the outputs of temporal enhancement features and spatial enhancement features after position encoding respectively;

[0047] POS t,a represents the position encoding with a time step of t, a is an even-odd option, a=2s or a=2s+1, and s represents the dimension;

[0048] When a=2s,

[0049] When a=2s+1,

[0050] Where D model represents the embedding dimension;

[0051] Step 32: As K and Q matrices respectively, As the V matrix, it is input into the multi-head attention mechanism to fuse the temporal features and spatial features to obtain the spatial attention matrix W:

[0052]

[0053] Where d represents K T The dimension size, K T stands for K transpose.

[0054] Step 33: Use the spatial attention matrix W to adjust the spatial correlation of V and obtain the output feature F of the multi-head attention mechanism out

[0055]

[0056] In the formula, F fusion is the result of using the spatial attention matrix to strengthen the temporal features. and γ j Denotes the matrix F fusion The row vector of represents the time series feature vector of the i-th sensor variable after spatial dependency weighting, F outrepresents the output features of the multi-head attention mechanism, W proj is the learnable weight matrix, Represents the output features of the qth attention head in the multi-head attention, q = 1, 2, ..., H, H represents the number of attention heads;

[0057] Step 34: Use normalization and forward propagation to process the output feature F of the multi-head attention out Processing to obtain the fusion features output by the spatiotemporal feature fuser

[0058]

[0059] In the formula, represents the output of the first residual and normalization layer in the spatiotemporal feature fusion, b represents the bias vector, It is the output of the first residual and normalization layer in the spatiotemporal feature fuser and serves as the fused feature output by the spatiotemporal feature fuser.

[0060] Preferably, the process of the support vector machine MLP predicting the remaining useful life of the aircraft engine according to the fusion features output by the spatiotemporal feature fuser is as follows:

[0061] The support vector machine MLP has one hidden layer and two fully connected layers. The outputs of the two fully connected layers are obtained according to the following equations:

[0062]

[0063] in, and Represent the weight matrices corresponding to the two fully connected layers, and b2∈R 1 Represent the bias matrices corresponding to W1 and W2, respectively, X out2 Represents the predicted RUL value.

[0064] Beneficial effects of the present invention: The present invention proposes a parallel spatiotemporal Transformer for aircraft engine RUL prediction. A key point is the design of a novel parallel spatiotemporal attention mechanism, which connects the temporal attention mechanism and the spatial attention mechanism in parallel, and describes the importance of data in time and space dimensions at the same time. TAM (temporal attention module) generates the importance weights of each time step at different time scales through multi-scale local interactions, and then redistributes the importance of each moment of the multi-dimensional time series data through adaptive weighting to obtain temporal enhancement features. SAM (spatial attention module) uses a multi-head attention mechanism to capture the global importance of each sensor parameter, and then gives each sensor parameter of the multi-dimensional time series data the importance in the spatial dimension through adaptive weighting to obtain spatial enhancement features. Specifically, there are the following advantages:

[0065] 1. Parallel extraction of spatial dependency and temporal dependency of multidimensional time series. The temporal attention mechanism and the spatial attention mechanism are connected in parallel to extract the temporal features and spatial features of the multidimensional time series at the same time, effectively avoiding the mutual interference of temporal and spatial information, extracting the real temporal dependency and spatial dependency of the original multidimensional time series to the greatest extent, and improving the modeling ability of complex spatiotemporal relationships of multidimensional time series.

[0066] 2. A novel task-oriented spatiotemporal feature fuser is designed to achieve adaptive fusion of temporal enhancement features and spatial enhancement features guided by the RUL prediction task, so that spatial information and temporal information can complement and fully utilize each other, thereby improving the RUL prediction performance.

[0067] 3. Comprehensive experiments were conducted on the C-MAPSS public dataset. The proposed method achieved state-of-the-art results in the aircraft engine RUL prediction task, proving the effectiveness of the proposed method. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 It is a schematic diagram of the serial structure of temporal attention and spatial attention mechanisms;

[0069] Figure 2 It is a schematic diagram of the parallel structure of temporal attention and spatial attention mechanisms;

[0070] Figure 3 It is a schematic diagram of the parallel spatiotemporal Transformer model of the present invention;

[0071] Figure 4 It is a schematic diagram of the data preprocessing process for generating input samples;

[0072] Figure 5 This is a schematic diagram of the parallel spatiotemporal attention mechanism;

[0073] Figure 6 This is the principle diagram of spatiotemporal feature fusion and RUL prediction;

[0074] Figure 7 This is the training flow chart of the parallel spatiotemporal Transformer model;

[0075] Figure 8 It is a diagram of the sliding time window construction process;

[0076] Fig. 9 is the RMSE and score distribution of different sliding window lengths on the FD001 test dataset, where Fig. 9 (a) is the distribution of scores of different sliding window lengths on FD001. Fig. 9 (b) is the RMSE distribution diagram of different sliding window lengths on FD001;

[0077] Fig.10 is the RUL prediction result of all engines in the four test data sets, where Fig.10 (a) is the RUL prediction result of all engines in the FD001 test dataset. Fig.10 (b) is the RUL prediction result of all engines in the FD002 test dataset. Fig.10 (c) is the RUL prediction result of all engines in the FD003 test dataset. Fig.10 (d) is the RUL prediction results of all engines in the FD004 test dataset;

[0078] Fig.11 This is a comparison chart of the RUL prediction results of the three methods in four test data sets. Fig.11 (a) is the RUL prediction results of all engines in the FD001 test dataset using three methods. Fig.11 (b) is the RUL prediction results of all engines in the FD002 test dataset using three methods. Fig.11 (c) is the RUL prediction results of all engines in the FD003 test dataset using three methods. Fig.11 (d) is the RUL prediction results of all engines in the FD004 test dataset using three methods. DETAILED DESCRIPTION

[0079] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0080] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0081] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but they are not intended to limit the present invention.

[0082] Specific implementation method 1: The following is combined Figures 1 to 11 The present embodiment is described. The present embodiment describes a method for predicting the remaining useful life of an aircraft engine. The remaining useful life of the aircraft engine is predicted using a parallel spatiotemporal Transformer model. The parallel spatiotemporal Transformer model includes a temporal attention module TAM, a spatial attention module SAM, a spatiotemporal feature fuser, and a support vector machine MLP.

[0083] The temporal attention module TAM and the spatial attention module SAM are set in parallel, and the input samples are input into the temporal attention module TAM and the spatial attention module SAM respectively; the temporal enhancement feature is obtained through the temporal attention module TAM; the spatial enhancement feature is obtained through the spatial attention module SAM;

[0084] The temporal enhancement feature and the spatial enhancement feature are fused by using the spatiotemporal feature fusion device. The spatiotemporal feature fusion device takes the temporal enhancement feature as the main body and the spatial enhancement feature as the auxiliary. The multi-head attention mechanism is used to transform the spatial enhancement feature into an attention matrix, and the attention matrix is ​​multiplied by the temporal enhancement feature to highlight the dominant position of the temporal enhancement feature.

[0085] The support vector machine MLP predicts the remaining useful life of aircraft engines based on the fusion features output by the spatiotemporal feature fuser.

[0086] This implementation has designed a novel parallel spatiotemporal attention mechanism, which connects the temporal attention mechanism and the spatial attention mechanism in parallel, and describes the importance of data in both temporal and spatial dimensions. TAM generates the importance weights of each time step at different time scales through multi-scale local interactions, and then redistributes the importance of each moment of the multidimensional time series data through adaptive weighting to obtain temporal enhancement features. SAM uses a self-attention mechanism to capture the global importance of each sensor parameter, and then gives each sensor parameter of the multidimensional time series data the importance in the spatial dimension through adaptive weighting to obtain spatial enhancement features.

[0087] The parallel spatiotemporal Transformer for aircraft engine remaining life prediction consists of three main parts. The first part is data preprocessing, see Figure 4 As shown in the figure, it mainly includes data preprocessing and sliding window sample construction; the second part is to extract temporal enhancement features and spatial enhancement features in parallel through the parallel spatiotemporal attention mechanism, see Figure 5 As shown in Figure 2, the third part is to adaptively fuse spatiotemporal features through a task-oriented feature fuser. Finally, the fused features are input into a simple regression network to achieve RUL prediction, see Figure 6 As shown in the figure. The parallel spatiotemporal attention mechanism uses the temporal attention mechanism to enhance the degradation effect of key timestamps in the multidimensional time series, and uses the spatial attention mechanism to enhance the importance of key sensor parameters in the multidimensional time series, helping the model to pay more attention to important time steps and sensor parameters. In addition, in order to improve the prediction results, the feature fusion device adaptively fuses the two types of features proposed according to the RUL prediction requirements to obtain high-level features that fully enhance the spatiotemporal key information, which can more effectively characterize the engine degradation process.

[0088] More importantly, in different flight mission scenarios, the data collected by sensors varies greatly, so the degradation information carried by each sensor parameter is unknown and changing, which brings great difficulties to traditional RUL prediction. Fortunately, the parallel spatiotemporal Transformer proposed in this patent combines all the advantages of the parallel spatiotemporal attention mechanism and the task-oriented feature fuser. When the operating conditions or human factors change, the parallel spatiotemporal Transformer can recalibrate the importance of different sensor parameters and the importance of different time steps according to the changed monitoring data, and always extract key information based on RUL requirements, so that it still has good generalization for changing multi-dimensional time series sensor data, and can perform accurate RUL predictions under different flight conditions.

[0089] See also Figure 4 , the process of obtaining input samples:

[0090] Collect different types of sensor parameters of various aircraft engine components under different operating conditions, such as temperature, pressure and speed, and construct original multi-dimensional time series with these sensor data;

[0091] A series of preprocessing operations are performed on the original multidimensional time series, including normalization, exponential smoothing, and sliding time window division;

[0092] After the above preprocessing operation, the input sample is obtained, which is a time series of T×m dimensions. Where T represents the length of the time series and m represents the number of sensors.

[0093] Due to changes in operating conditions, flight missions, etc., the importance of different types of sensor parameters (such as temperature, pressure, fuel flow) often changes. In addition, for the same sensor parameter, the degradation information it contains is also different at different times during the engine operation process. Therefore, in order to pay more attention to important time steps and sensor parameters at the same time, a parallel spatiotemporal attention mechanism is designed to extract spatiotemporal features from the input multidimensional time series data by setting two different attention mechanisms: the temporal attention module weights different times, and the spatial attention module weights different sensor parameters.

[0094] About the temporal attention module TAM. During operation, when the aircraft performs some intense actions such as lifting and diving, the engine will be more prone to fatigue damage, while the cruising process consumes very little engine performance. That is to say, in the collected multi-dimensional time series monitoring data, it is necessary to focus on those moments that consume a lot of engine energy, because these moments are the key to affecting the remaining life of the engine. To solve this problem, the present invention adopts a temporal attention module as the basis to obtain the time dependency in multi-dimensional time series data. The temporal attention module consists of a global average pooling layer, a multi-scale convolution layer, and a nonlinear activation function, among which multi-scale convolution is the core. For a given multi-dimensional time series The detailed process of temporal dependency extraction using the temporal attention module is described as follows:

[0095] See also Figure 5 ,The process of obtaining temporal enhanced features through the temporal attention module TAM includes:

[0096] Step 11: Perform average pooling along the spatial dimension to convert it into a time series of T×1 dimensions Its purpose is to obtain a global average information representation of all sensor parameters at each time step.

[0097]

[0098] Where GAP1D(·) represents one-dimensional global average pooling, express The i-th column in the figure, i represents the serial number of the sensor, i=1 , 2,...,m, The value of each time step in represents the global average information of all sensor parameters at this time step.

[0099] Step 12: During the operation of the aircraft, different important actions have different durations, which may cause the engine performance to show different degradation patterns over time. Some features are more significant in a short time scale, while other features may take a longer time scale to appear. Therefore, multi-scale convolution is used to pool features. Processing is performed to realize the interaction of information at different time steps from different time scales, thereby obtaining the temporal dependency between different time steps.

[0100]

[0101] In the formula represents the output feature of the pth convolution branch in the multi-scale convolution layer, (k p ,b p ) represents the convolution kernel of the pth convolution branch and the corresponding bias vector, p = 1, 2, ..., n, n represents the number of convolution branches, * represents a one-dimensional convolution operation;

[0102] Step 13: Use the Sigmoid operator to process the output features of each convolution branch in the multi-scale convolution layer to obtain the importance of each time step at different time scales:

[0103]

[0104] In the formula represents the weight information of each time step at the pth time scale, and f(·) represents the Sigmoid activation function;

[0105] Step 14: Obtain the timing enhancement feature F according to the following formula TAM :

[0106]

[0107] In the formula, represents matrix multiplication, and n represents the number of branches in the multi-scale convolutional layer.

[0108] These weights Perform matrix multiplication with the original time series, give the original time series time attention, obtain the time series strengthened based on time dependence, and aggregate the strengthened time series at different time scales together to obtain the final time series strengthened by the time attention mechanism.

[0109] Regarding the spatial attention module SAM, in actual processes, different sensors provide various types of parameters, such as temperature, pressure, speed, etc. However, the effect of the engine degradation information contained in these sensor data on RUL prediction is still unknown. If equal attention is paid to these multi-sensor information, the model will not be able to focus on important features, resulting in a waste of attention. Therefore, analyzing the importance of different sensors has become an indispensable part of RUL prediction. To this end, this patent designs a spatial attention module so that the model can automatically establish the importance relationship between sensors under different working conditions, thereby enhancing sensor signals containing more degradation information and suppressing sensor signals containing less degradation information. For a given multidimensional time series X∈R m×T The specific process of using the spatial attention module to assign spatial weights to input samples is as follows:

[0110] See also Figure 5 ,The process of obtaining spatial enhanced features through the spatial attention module SAM includes:

[0111] Step 21: Input sample Get a multidimensional time series X∈R with m×T dimensions m×T ;

[0112] Step 22: Get the correlation matrix S:

[0113]

[0114] Where X = [α1, α2, ..., α m ] T , α i ∈R 1×T represents the row vector of X, is the inner product between the i-th and j-th sensor parameter vectors, which is used to measure their correlation. It can be observed that S is a correlation map between all different sensor parameter vectors, and the i-th row vector in S can represent the correlation between the i-th sensor parameter vector and all other sensor parameter vectors, that is, the global spatial dependency of the i-th sensor parameter vector.

[0115] Step 23. Finally, the correlation matrix S is used to adjust the parameter vectors in the original multidimensional time series X to enhance the important sensor parameters while suppressing the unimportant sensor parameters. Specifically, each row vector in S is processed using Softmax to obtain the spatial attention matrix, which is then multiplied by X to obtain the time series F after spatial dependency enhancement. SAM In addition, in order to ensure the stability of the neural network and avoid gradient disappearance and loss of original information, residual connections are introduced. The spatial enhancement feature F is obtained as follows: SAM :

[0116]

[0117] In the formula, Represents matrix addition.

[0118] For more information about the spatiotemporal feature fuser, see Figure 6 As shown in FIG. 1 , the process of fusing the temporal enhancement features and the spatial enhancement features using the spatiotemporal feature fuser includes:

[0119] Step 31: Enhance the time series features F TAM and spatial enhancement feature F SAM Perform position encoding to obtain fusion features:

[0120]

[0121] in, are the outputs of temporal enhancement features and spatial enhancement features after position encoding respectively;

[0122] POS t,a represents the position encoding with a time step of t, a is an even-odd option, a=2s or a=2s+1, and s represents the dimension;

[0123] When a=2s,

[0124] When a=2s+1,

[0125] Where D model represents the embedding dimension;

[0126] Step 32: As K and Q matrices respectively, As the V matrix, it is input into the multi-head attention mechanism to fuse the temporal features and spatial features to obtain the spatial attention matrix W:

[0127]

[0128] Where d represents K TThe dimension size, K T stands for K transpose.

[0129] The K, Q, and V matrices are three important parameters of the multi-head attention mechanism. Q is the query quantity of the word, K is the "queried" vector, and V is the content vector.

[0130] Step 33: Use the spatial attention matrix W to adjust the spatial correlation of V and obtain the output feature F of the multi-head attention mechanism out

[0131]

[0132]

[0133] In the formula, F fusion is the result of using the spatial attention matrix to strengthen the temporal features. and γ j Denotes the matrix F fusion The row vector of Represents the time series feature vector of the i-th sensor variable after spatial dependency weighting. It can be observed that The weighted sum of all original sensor parameter time series feature vectors, the influence of the remaining m-1 parameters on the mth parameter is considered, that is, the spatial dependence between different parameters is established, which is more conducive to extracting degradation features that can characterize the engine state. out represents the output features of the multi-head attention mechanism, W proj is the learnable weight matrix, Represents the output features of the qth attention head in the multi-head attention, q = 1, 2, ..., H, H represents the number of attention heads;

[0134] Step 34: Use normalization and forward propagation to process the output feature F of the multi-head attention out The degraded features are processed to make them more obvious, and in order to avoid the gradient vanishing or gradient exploding problem, the residual connection is introduced into the calculation process. In order to highlight the important degraded features and suppress the unimportant features without changing the time dependency, the fusion features output by the spatiotemporal feature fusion are obtained as follows:

[0135]

[0136] In the formula, represents the output of the first residual and normalization layer in the spatiotemporal feature fusion, b represents the bias vector, It is the output of the first residual and normalization layer in the spatiotemporal feature fuser and serves as the fused feature output by the spatiotemporal feature fuser.

[0137] The process of predicting the remaining useful life of an aircraft engine by using the support vector machine MLP based on the fusion features output by the spatiotemporal feature fuser is as follows:

[0138] The support vector machine MLP has one hidden layer and two fully connected layers. The outputs of the two fully connected layers are obtained according to the following equations:

[0139]

[0140] in, and Represent the weight matrices corresponding to the two fully connected layers, and b2∈R 1 Represent the bias matrices corresponding to W1 and W2, respectively, X out2 Represents the predicted RUL value.

[0141] See also Figure 7 The flowchart of the proposed parallel spatiotemporal Transformer model for aircraft engine RUL prediction is shown. First, the original multidimensional time series is preprocessed by normalization, exponential smoothing, and sliding time window partitioning. Secondly, the designed parallel spatiotemporal attention mechanism is used to simultaneously extract temporal enhancement features and spatial enhancement features, and the designed task-oriented feature fuser is used to fuse the two extracted features to obtain representative engine degradation features. Thirdly, the RUL prediction task is implemented using a multilayer perceptron, and the Adam optimizer is used to optimize the weights of the parallel spatiotemporal Transformer. Finally, the test sample is input into the trained parallel spatiotemporal Transformer, and the predicted RUL value is output.

[0142] Verification of the implementation example. To verify the effectiveness of the parallel spatiotemporal Transformer in aircraft engine RUL prediction, the parallel spatiotemporal Transformer model was verified on the CMAPSS dataset, and the results were compared with other current advanced methods. All experiments in this chapter were conducted on the same workstation, which has an Intel Core i5-9300H CPU, an NVIDIA GeForce GTX 1650 GPU, Windows 10 operating system, and Python 3.8 with PyTorch 1.7.1 + CU110 compilation language.

[0143] Experimental dataset description:

[0144] NASA obtained simulated data from 21 sensors during the operation of aircraft engines by developing the Commercial Aerospace Propulsion System Simulation (C-MAPSS). The C-MAPSS dataset contains four subsets, each of which is divided into a training set and a test set consisting of different operating conditions and fault types. The training set covers the data information of the complete life cycle of each engine from the start of operation to the final failure, while the test set contains the data information of each engine from the start of operation to a certain point in time before the failure occurs. The purpose is to predict the remaining operating cycle of each engine in the test set before the end of its service life. The detailed information of each dataset is shown in Table 1.

[0145] Table 1 Introduction to C-MAPSS dataset

[0146]

[0147] Dataset preprocessing:

[0148] (1) Sensor signal selection

[0149] Although the C-MAPSS dataset provides monitoring parameters of 21 sensors, the data collected by some of these sensors remain basically stable throughout the life cycle of the engine, and such data cannot reflect the degradation trend of the engine. Therefore, in order to reduce the time complexity of data processing and model learning, these data should be discarded. After screening, 14 groups of sensor signals with obvious increasing or decreasing trends were finally selected as the original input of the model. These sensors are S2, S3, S4, S7, S8, S9, S11, S12, S13, S14, S15, S17, S20, and S21.

[0150] (2) Marking URL label

[0151] The input of the model is multidimensional time series data, and the output is RUL. To this end, RUL must be used as a supervisory label during the training phase. However, the original data in the C-MAPSS training set does not contain this label itself. Therefore, the data must be annotated with the corresponding RUL values ​​to facilitate accurate model training.

[0152] The C-MAPSS training dataset records in detail the entire operational life of the engine from initial deployment to failure. In this dataset, the total service life of each engine is defined by the total number of cycles accumulated up to its final operational time point. The RUL at each time point is calculated by subtracting the cumulative number of cycles recorded at that specific time point from the total service life.

[0153]

[0154] In the formula represents the RUL of the nth engine at time t, r n represents the total service life of the nth engine, and t represents the number of cycles that occur.

[0155] For aircraft engines, the relationship between their RUL and the number of operating cycles is not linear. The RUL does not change significantly during the initial time of engine operation. However, once the number of cycles exceeds a critical threshold, the RUL will drop rapidly as the number of cycles increases. Therefore, the present invention uses a piecewise linear programming method to represent the RUL of the engine, and sets the calculated RUL upper limit to 125. After exceeding this value, the RUL will remain at a fixed value.

[0156]

[0157] Where r max Indicates the RUL threshold during the initial operation of the engine, and the threshold is set to 125.

[0158] (3) Data normalization

[0159] After labeling the RUL, the next step is to normalize the raw data. Since the raw data comes from different sensors, their dimensions and ranges may be different. This difference often leads to a large number of dimensional data dominating the gradient calculation, thereby affecting the experimental results and hindering the convergence of the model. By normalizing the data, the accuracy and convergence speed of the model can be improved to a certain extent. The present invention first classifies the raw data according to the engine operating conditions, and then uses the Z-Score formula to normalize the data under each operating condition. The Z-Score formula is as follows:

[0160]

[0161] In the formula and They represent the normalized data and original data of the jth sensor under the working mode k, respectively. and They represent the mean and standard deviation of the j-th sensor data under operating mode k respectively.

[0162] (4) Exponential smoothing

[0163] In order to solve the problems of difficulty in extracting degradation trend features and inability to remove noise, the present invention uses the exponential moving average (EMA) method to process the data. EMA reduces the influence of extreme values ​​in the process of smoothing data, making it easier to capture long-term degradation trends. The calculation formula of the EMA method is as follows:

[0164]

[0165] Where α represents the attenuation value (exponential smoothing factor), 0<α≤1, and They represent the data of the j-th sensor of the n-th engine after exponential smoothing processing and the data before processing respectively.

[0166] Exponential smoothing combines current data with historical data, making full use of extended time series information. This approach enables the model to effectively identify deep degradation patterns in time series data while minimizing the destructive effects of outliers. After debugging, the present invention sets the attenuation value of the FD001 training set to 1 and the attenuation value of the remaining training sets to 0.1.

[0167] (5) Sliding window setting

[0168] Since the degradation of aircraft engines is a continuous time process, inputting multiple continuous time nodes into the model simultaneously is conducive to capturing more useful time series information. Therefore, after the above processing steps, the present invention uses a sliding time window method to batch input data into the model.

[0169] The schematic diagram of using the sliding window method to segment data is as follows Figure 8 As shown, T represents the window size and S represents the sliding step size.

[0170] Evaluation indicators: Two common indicators are used to evaluate the prediction performance of the model on the C-MAPSS dataset, namely score and root mean square error (RMSE). The calculation formulas of the two indicators are as follows:

[0171]

[0172] In the formula and Y i They represent the RUL predicted value and true value of the i-th sample respectively, and N represents the number of samples.

[0173] The RMSE curve is symmetrical, which means that it evaluates early predictions and late predictions equally. The Score curve is asymmetrical because for aircraft engines, if the RUL prediction is too large, it may lead to major accidents. Therefore, the Score function increases the penalty for late predictions to make the evaluation criteria more reasonable.

[0174] Hyperparameter settings:

[0175] In order to determine the optimal hyperparameters of the proposed method, a grid search method is used to determine the hyperparameters of the parallel spatiotemporal Transformer. The relevant training hyperparameters and model hyperparameters are detailed in Tables 2 and 3, respectively. According to existing studies, before the sliding window length selection experiment, the initial sliding window sizes of the four datasets FD001 to FD004 are tentatively set to 30, 20, 35, and 15, respectively. To ensure that the model can capture comprehensive degradation information, the sliding step size is uniformly set to 1. This study adopts an early stopping strategy to stop training when the validation set loss begins to increase, thereby reducing the risk of overfitting. Therefore, there is no fixed number of epochs for each training, and the maximum number of training iterations and the waiting rounds of the early stopping mechanism are set to 200 and 3, respectively. The Adam optimizer is used to update the network parameters, and the learning rate is set to 0.001. In order to further reduce the complexity of the model and the possibility of overfitting, a dropout rate of 0.1 is used. In addition, the training data and validation data are randomly selected from the training set in a ratio of 4:1 to the number of engines.

[0176] Table 2 Training hyper parameters

[0177]

[0178] Table 3 Model hyperparameters

[0179]

[0180]

[0181] Sliding window length selection:

[0182] In the time series prediction task, identifying the relationship between adjacent time nodes is crucial for RUL prediction. In order to study the impact of different sliding window lengths on the performance of the parallel spatiotemporal Transformer on each sub-dataset and reveal the optimal configuration of the model on different sub-datasets, this paper conducted a sliding window length selection experiment.

[0183] According to existing research, a sliding window that is too short may not be able to effectively capture degradation information that is critical to model performance, while a window that is too long may introduce irrelevant interference. Therefore, experiments with different sliding window lengths should be conducted for each subset to determine the optimal sliding window length suitable for the parallel spatiotemporal Transformer. The four subsets in the C-MAPSS dataset have different maximum and minimum life cycles, respectively. The minimum life cycle number of each subset should be selected as the upper limit of the sliding window length. Table 4 shows the maximum and minimum life cycles of the engines in the four subsets, respectively.

[0184] Table 4 Maximum and minimum number of cycles for each subset

[0185]

[0186] Table 4 shows that the minimum number of life cycles of the test set is always lower than that of the training set. Therefore, the upper limit of the sliding window length of each subset should be determined by the minimum life cycle of the test set. The minimum number of life cycles of the test set in each subset is 31, 21, 38, and 19, respectively. Therefore, the upper limit of the sliding window length is set to 30, 20, 35, and 15, respectively, and experiments are conducted with 5 sliding window lengths as increments. The hyperparameters of these experiments are shown in Tables 2 and 3. To ensure the reliability of the results, each experiment is repeated ten times, and the average value is taken as the final result. The experimental results are shown in Table 5. Fig. 9 The box plot of the experimental results of the FD001 dataset shows that setting the sliding window length to 30 not only makes the Score and RMSE reach the lowest possible values, but also minimizes the variance. This optimal setting shows that the prediction process has strong stability. Therefore, the optimal window length of FD001 is set to 30. As the sliding window length decreases, the Score and RMSE gradually increase. This is because the smaller the sliding window length, the less likely it is that the features will be completely captured by the model. Similarly, the same method is used for the FD002, FD003, and FD004 subsets, and the optimal sliding window lengths are determined to be 20, 15, and 15, respectively.

[0187] Table 5 Experimental performance of the method using different window sizes on four subsets.

[0188]

[0189] Comparison between the method of the present invention and other existing methods:

[0190] The performance of the proposed parallel spatiotemporal Transformer is verified using four sub-datasets of C-MAPSS and compared with other advanced RUL prediction methods to highlight the advantages of the parallel spatiotemporal Transformer. The model is trained with the hyperparameters in Table 3 and the sliding window length in Table 4, and then the test data is input into the trained model for RUL prediction. The prediction results are shown in Table 6. In order to reduce the impact of randomness, all experiments were performed ten times, and the average value of these ten experiments was taken as the final result. The experimental results show that the model proposed in the present invention performs well on the C-MAPSS dataset.

[0191] Table 6. Parallel spatiotemporal Transformer performance on the C-MAPSS dataset.

[0192]

[0193] Fig.10The prediction results of the parallel spatiotemporal Transformer model in four test sets are shown. The true RUL of each engine in each dataset is sorted in an easy-to-observe way. As can be seen from the figure, the parallel spatiotemporal Transformer performs well on the C-MAPSS dataset, and the prediction results are very close to the true RUL values. Among all datasets, FD001 and FD003 have the most accurate prediction results, followed by FD002, while FD004 has the least ideal prediction results. This difference can be attributed to the complexity and diversity of the operating conditions and failure modes of FD002 and FD004. Specifically, FD001 contains 1 operating condition and 1 failure mode, and FD003 contains 1 operating condition and 2 failure modes. FD002 contains 6 operating conditions and 1 failure mode, and FD004 contains 6 operating conditions and 2 failure modes. Therefore, the FD002 and FD004 datasets are more complex, which poses a greater challenge for in-depth analysis of engine degradation trends. In addition, engines with smaller true RUL values ​​have higher prediction accuracy than engines with larger true RUL values. The reason for this phenomenon is that engines near the end of their service life have accumulated a large amount of degradation information, and the model can capture more degradation information, so the predicted value is closer to the true value.

[0194] In order to further demonstrate the superiority of the performance of the parallel spatiotemporal Transformer model, the parallel spatiotemporal Transformer was compared with some of the more advanced methods. These advanced methods include mainstream models such as DCNN, LSTM, CNN-LSTM, AGCNN, BiGRU-AS, DSAN, STRUL, and Transformer, as well as their variants, such as PAOTransformer, TATFA-Transformer, and TF-SCN. All methods in the comparative experiment are used for RUL prediction of aircraft engines, which is consistent with the method objectives introduced in this invention. The experimental results are shown in Table 7.

[0195] The model proposed in the present invention has a significant improvement over traditional models such as DCNN, LSTM, CNN-LSTM, and its performance is also better than some of the latest Transformer improved models, such as PAOTransformer, TF-SCN, etc. It is worth noting that the parallel spatiotemporal Transformer model has the best performance on the FD002 and FD004 datasets. Compared with the second best model TF-SCN, the RMSE is reduced by 11.6% and 11.8%, and the Score is reduced by 35.4% and 6.1%, respectively. On the FD001 dataset, TF-SCN achieved the best results, and the parallel spatiotemporal Transformer followed closely, ranking second in both RMSE and Score indicators. On the FD003 dataset, TATFA-Transformer achieved the best results, and the parallel spatiotemporal Transformer ranked second and seventh in RMSE and Score indicators, respectively.

[0196] In general, the parallel spatiotemporal Transformer has consistently achieved commendable results on the C-MAPSS dataset, especially on the FD002 and FD004 subsets, where it significantly outperforms other models. The advantage of the parallel spatiotemporal Transformer stems from its innovative architecture, which integrates a spatial attention module and a temporal attention module, the former for evaluating spatial importance from a holistic perspective and the latter for evaluating temporal importance at different scales. The two modules are connected in parallel and cleverly fuse the spatiotemporal features through a multi-head attention mechanism. It is worth noting that in the RUL prediction task of the present invention, the temporal attention features, which are regarded as the most critical, are input into the multi-head attention module as V to adapt to the specific requirements of the task. This approach ensures that the temporal critical features are significantly retained during the feature fusion process. In addition, the model also adopts an exponential smoothing method to embed historical degradation data into time nodes, thereby mitigating the impact of noise and improving the accuracy of RUL prediction. In essence, the sophisticated design and strategic data processing of the parallel spatiotemporal Transformer contribute to its superior performance. The experimental results verify that the proposed method is superior to most existing methods.

[0197] Table 7. Comparison of RMSE and scores of different methods.

[0198]

[0199] Ablation experiment:

[0200] The ablation experiment of the present invention is used to prove the role of the proposed parallel spatiotemporal attention mechanism and feature fuser, and two different situations are studied. The first situation is that the network only integrates the parallel spatiotemporal attention mechanism without adding the feature fuser. The output characteristics of the parallel spatiotemporal attention mechanism are merged through a summation process, and this specific model is named ST-N. The second scheme is characterized in that the network only contains a feature fuser, in which the Q, K and V of the multi-head attention mechanism are obtained by linear transformation of the same input information. This model is called TN. The above two models and the parallel spatiotemporal Transformer model proposed in the present invention were tested on the test set, and RMSE and Score were used as evaluation indicators. The experiment was repeated 10 times and the average value was taken as the final result. The prediction results of the three models are compared in Table 8. It can be seen from the table that TN is better than ST-N, because except for the Score value of the experimental results on FD003, the other experimental values ​​of TN are lower than ST-N. It is worth noting that TN performs well on datasets with complex operating conditions and diverse fault types such as FD002, FD003 and FD004 because its RMSE value is always low. In addition, the parallel spatiotemporal Transformer has lower RMSE and Score values. Specifically, the performance of ST-N lags behind that of the parallel spatiotemporal Transformer by a large margin, where the RMSE values ​​of the parallel spatiotemporal Transformer on the four subsets are reduced by 20.4%, 17.4%, 18.1%, and 24.1% respectively compared with ST-N; the Score is reduced by 45.6%, 18.5%, 16.3%, and 39.0% respectively. This difference highlights the advantages of the feature fusion method introduced in this invention, which can select and integrate the most significant spatial features and fuse them with temporal features while discarding irrelevant data. The performance of TN also lags behind that of the parallel spatiotemporal Transformer. On the four subsets, the RMSE values ​​of the parallel spatiotemporal Transformer are reduced by 15.8%, 10.2%, 8.0%, and 4.4% respectively compared with TN; the Score is reduced by 41.0%, 18.3%, 20.0%, and 17.4% respectively. This further highlights the applicability of the parallel spatiotemporal Transformer in the RUL prediction task. The multi-head attention mechanism in the Transformer can only evaluate the importance of features from a single dimension. In contrast, after integrating the spatiotemporal feature extraction module with the Transformer decoder, the importance of spatiotemporal features can be considered simultaneously, so that the model can capture more comprehensive degradation information. The experiment verifies the effectiveness of the spatiotemporal feature extraction method proposed in the present invention.

[0201] Table 8 RMSE and scores of the three RUL prediction models on the test set in the ablation study.

[0202]

[0203] To intuitively display the results of the ablation experiment, Figure 8 The prediction results of the three methods on four data sets are shown. As can be seen from the figure, the prediction results of the parallel spatiotemporal Transformer on the four data sets are closer to the true RUL value than the other two methods, which means that the parallel spatiotemporal Transformer has a higher prediction accuracy.

[0204] In summary, in order to solve the problems of complex working conditions, noise interference of collected signals, difficulty in capturing important degradation information in time series information, and unknown importance of different sensors for RUL prediction in aircraft engine RUL prediction tasks, the present invention proposes a new architecture of parallel spatiotemporal Transformer. First, the temporal features and spatial features are captured independently by the parallel connected temporal feature extractor and spatial feature extractor. Secondly, the temporal and spatial features are deeply fused through the improved Transformer decoder. Finally, the high-dimensional features obtained by feature extraction and fusion are mapped to the final RUL value through the fully connected layer.

[0205] The model was tested several times to determine its optimal hyperparameters. Then, the model was subjected to an ablation experiment to determine that the model had no redundant structure. Finally, the proposed method was compared with the current mainstream methods. The experimental results show that the proposed method has obvious advantages over the mainstream methods in terms of RUL prediction accuracy, and also performs well on FD002 and FD004, which are more difficult to predict.

[0206] The RUL prediction of aircraft engines is an important and evolving research issue. In the future, we will focus on studying the impact of factors such as engine materials, structure, and maintenance on RUL, hoping to achieve more accurate RUL prediction by considering more comprehensive factors.

[0207] Specific implementation method 2. This implementation method is further optimized on the basis of implementation method 1, proposes a second key point, and designs a novel task-oriented spatiotemporal feature fuser, which realizes the adaptive fusion of temporal enhancement features and spatial enhancement features guided by the RUL prediction task, so that spatial information and temporal information can complement and fully utilize each other, thereby improving the RUL prediction performance.

[0208] The temporal features obtained by learning the temporal dependency of multivariate time series focus on describing the trend of time series changes over time, the correlation between different time steps, and other information, which is more helpful for the model to understand and predict the dynamic behavior and evolution in the time series. The spatial features obtained by learning the spatial dependency of multivariate time series reflect the correlation and interaction between different variables and how they jointly affect the behavior of the system. This means that for different downstream tasks, the roles played by these two types of features are different. For example, in RUL prediction, the temporal features play the main role, while the spatial features play the auxiliary role; on the contrary, in some classification tasks, the spatial features play the main role, while the temporal features play the auxiliary role. However, many current RUL prediction methods based on deep learning extract spatial and temporal features by embedding different attention mechanisms. However, on the one hand, most of them aggregate the two types of features through simple concat or matrix addition, which greatly reduces the final effect; on the other hand, they are all designed for specific tasks. When applied to other fields or tasks, these models often need to be redesigned and trained, wasting a lot of time and resources. Therefore, how to flexibly use spatial features and temporal features, and adaptively integrate spatial features and temporal features in downstream tasks to achieve the best effect.

[0209] Using a task-oriented spatiotemporal feature fuser can make a model multi-purpose.

[0210] Most current methods mainly use simple matrix addition or splicing to aggregate spatiotemporal features, which may lead to problems such as loss or redundancy of key information. In order to avoid this situation, the present invention proposes a novel task-oriented spatiotemporal feature fuser, which flexibly defines Q, K, and V in the multi-head attention mechanism according to downstream tasks, and uses the multi-head attention mechanism to achieve adaptive fusion of spatiotemporal features, so that temporal feature information and spatial feature information are fully integrated and complement each other under the guidance of downstream tasks. Specifically, unlike the traditional transformer, which uses linear transformation to obtain Q, K, and V matrices for any downstream task, the proposed spatiotemporal feature fuser directly assigns different types of features output by the parallel spatiotemporal attention mechanism to Q, K, and V for different downstream tasks. For example, for the RUL prediction task (using the existing multi-sensor time series to predict sensor data at future time points, implementation method one is this type of task), the more important temporal features are used as the V matrix, and the spatial features are used as Q and K to calculate the spatial correlation attention weight matrix. On the contrary, for classification tasks (such as fault diagnosis, using algorithms to determine the type of fault in a mechanical part), spatial features may be more important, so the spatial features are used as the V matrix, and the temporal features are used as the Q and K matrices to calculate the temporal correlation attention weight matrix. Based on the above analysis, by combining the parallel spatiotemporal attention mechanism with the spatiotemporal feature fuser, the traditional Transformer model can improve the deficiency of not being able to consider both temporal correlation and spatial correlation at the same time, thus achieving multi-purpose use of one model. Figure 6 As shown, the spatiotemporal feature fusion replaces the encoder of the Transformer with PSTAM, retaining the decoder structure. More importantly, for the RUL prediction task that the present invention needs to solve, the spatiotemporal feature fusion uses the spatially enhanced features extracted by the spatial attention module to represent K and Q in the decoder, and the temporal enhanced features extracted by the temporal feature module to represent V.

[0211] Although the present invention is described with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the present invention. It should therefore be understood that many modifications may be made to the exemplary embodiments and that other arrangements may be devised without departing from the spirit and scope of the present invention as defined by the appended claims. It should be understood that the features described in the various dependent claims and the present invention may be combined in a manner different from that described in the original claims. It should also be understood that the features described in conjunction with the individual embodiments may be used in other described embodiments.

Claims

1. A method for predicting the remaining useful life of an aircraft engine, characterized in that: The remaining useful life of aircraft engines is predicted using a parallel spatiotemporal Transformer model, which includes a temporal attention module TAM, a spatial attention module SAM, a spatiotemporal feature fuser, and a support vector machine MLP. The temporal attention module TAM and the spatial attention module SAM are set in parallel, and the input samples are input into the temporal attention module TAM and the spatial attention module SAM respectively; the temporal enhancement feature is obtained through the temporal attention module TAM; the spatial enhancement feature is obtained through the spatial attention module SAM; The temporal enhancement features and spatial enhancement features are fused using the spatiotemporal feature fusion; The spatiotemporal feature fusion device takes the temporal enhancement feature as the main body and the spatial enhancement feature as the auxiliary. It uses the multi-head attention mechanism to transform the spatial enhancement feature into an attention matrix, and multiplies the attention matrix with the temporal enhancement feature to highlight the dominant position of the temporal enhancement feature. The support vector machine MLP predicts the remaining useful life of aircraft engines based on the fusion features output by the spatiotemporal feature fuser.

2. The method for predicting the remaining useful life of an aircraft engine according to claim 1, characterized in that: The process of obtaining input samples: Collect different types of sensor parameters of various aircraft engine components under different operating conditions, such as temperature, pressure and speed, and construct original multi-dimensional time series with these sensor data; A series of preprocessing operations are performed on the original multidimensional time series, including normalization, exponential smoothing, and sliding time window division; After the above preprocessing operation, the input sample is obtained, which is a time series of T×m dimensions. Where T represents the length of the time series and m represents the number of sensors.

3. The method for predicting the remaining useful life of an aircraft engine according to claim 2, characterized in that: The process of obtaining temporal enhanced features through the temporal attention module TAM includes: Step 11: Perform average pooling along the spatial dimension to convert it into a time series of T×1 dimensions Where GAP1D(·) represents one-dimensional global average pooling, express The i-th column in the figure represents the serial number of the sensor, i=1,2,...,m, Step 12: Use multi-scale convolution to pool features Processing to obtain the temporal dependencies between different time steps: In the formula represents the output feature of the pth convolution branch in the multi-scale convolution layer, (k p ,b p ) represents the convolution kernel of the pth convolution branch and the corresponding bias vector, p = 1, 2, ..., n, n represents the number of convolution branches, * represents a one-dimensional convolution operation; Step 13: Use the Sigmoid operator to process the output features of each convolution branch in the multi-scale convolution layer to obtain the importance of each time step at different time scales: In the formula represents the weight information of each time step at the pth time scale, and f(·) represents the Sigmoid activation function; Step 14: Obtain the timing enhancement feature F according to the following formula TAM : In the formula, represents matrix multiplication, and n represents the number of branches in the multi-scale convolutional layer.

4. The method for predicting the remaining useful life of an aircraft engine according to claim 3, characterized in that: The process of obtaining spatial enhanced features through the spatial attention module SAM includes: Step 21: Input sample Get a multidimensional time series X∈R with m×T dimensions m×T ; Step 22: Get the correlation matrix S: Where X = [α1, α2, ..., α m ] T , α i ∈R 1×T represents the row vector of X, is the inner product between the i-th and j-th sensor parameter vectors; Step 23: Obtain the spatial enhancement feature F according to the following formula SAM : In the formula, Represents matrix addition.

5. A method for predicting the remaining useful life of an aircraft engine according to claim 4, characterized in that: The process of fusing temporal enhancement features and spatial enhancement features using the spatiotemporal feature fuser includes: Step 31: Enhance the time series features F TAM and spatial enhancement feature F SAM Perform position encoding to obtain fusion features: in, are the outputs of temporal enhancement features and spatial enhancement features after position encoding respectively; POS t,a represents the position encoding with a time step of t, a is an even-odd option, a=2s or a=2s+1, and s represents the dimension; When a=2s, When a=2s+1, Where D model represents the embedding dimension; Step 32: As K and Q matrices respectively, As the V matrix, it is input into the multi-head attention mechanism to fuse the temporal features and spatial features to obtain the spatial attention matrix W: Where d represents K T The dimension size, K T stands for K transpose. Step 33: Use the spatial attention matrix W to adjust the spatial correlation of V and obtain the output feature F of the multi-head attention mechanism out In the formula, F fusion is the result of using the spatial attention matrix to strengthen the temporal features. and γ j Denotes the matrix F fusion The row vector of represents the time series feature vector of the i-th sensor variable after spatial dependency weighting, F out represents the output features of the multi-head attention mechanism, W proj is the learnable weight matrix, Represents the output features of the qth attention head in the multi-head attention, q = 1, 2, ..., H, H represents the number of attention heads; Step 34: Use normalization and forward propagation to process the output feature F of the multi-head attention out Processing to obtain the fusion features output by the spatiotemporal feature fuser In the formula, represents the output of the first residual and normalization layer in the spatiotemporal feature fusion, b represents the bias vector, It is the output of the first residual and normalization layer in the spatiotemporal feature fuser and serves as the fused feature output by the spatiotemporal feature fuser.

6. A method for predicting the remaining useful life of an aircraft engine according to claim 5, characterized in that: The process of predicting the remaining useful life of an aircraft engine by using the support vector machine MLP based on the fusion features output by the spatiotemporal feature fuser is as follows: The support vector machine MLP has one hidden layer and two fully connected layers. The outputs of the two fully connected layers are obtained according to the following equations: in, and Represent the weight matrices corresponding to the two fully connected layers, and b2∈R 1 Represent the bias matrices corresponding to W1 and W2, X out2 Represents the predicted RUL value.

Images