Complex equipment data processing method based on CITIN

Abstract

The present invention relates to the technical field of complex equipment data processing, and more specifically, to a complex equipment data processing method based on a Constrained Interval-aware Temporal Imputation Network (CITIN) that applies a Constrained Interval-aware Temporal Imputation Network (CITIN) to fill in complex industrial equipment data, thereby reducing the possibility of a neural network generating unreasonable values ​​during the filling process during subsequent data processing, and improving the accuracy and stability of industrial equipment data processing. The method not only considers the impact of engine performance degradation caused by the different operating stages and durations of each flight on the engine, but also more accurately fills in missing parts in the engine performance data. Furthermore, in response to the problem that unreasonable values ​​sometimes occur in neural networks due to their unexplainability, PCHIP is used during the training phase of the filling network to constrain the range of filling results generated by the neural network, thereby reducing the possibility of the ITIN generating unreasonable values.

Description

Technical field:

[0001] The present invention relates to the technical field of complex equipment data processing, and more specifically, to a complex equipment data processing method based on CITIN, which applies a constrained interval-aware temporal imputation network (CITIN) to fill in complex industrial equipment data, reduces the possibility of the neural network generating unreasonable values ​​during the filling process in subsequent data processing, and improves the accuracy and stability of industrial equipment data processing. Background technology:

[0002] Accurately capturing environmental and machine status through sensors facilitates condition monitoring and health management for complex equipment. By adding numerous sensors during the design process, researchers can monitor the status of complex equipment during operation, providing a basis for subsequent maintenance and support. However, due to potential sensor failures, manual oversight, and interdepartmental information sharing regulations, the acquired status data may be incomplete. This incomplete data presents challenges for subsequent analysis.

[0003] There are many ways to handle data with missing values. Directly deleting some of the data containing missing values ​​is the simplest and easiest to do, but this is not applicable in all scenarios, especially for time series data such as aircraft engine operating data. Because the data characteristics of multivariate time series are manifested as continuity in time and correlation between different variables, if some data is directly deleted, this continuous relationship may be interrupted, resulting in large deviations in subsequent work. For aircraft engine data, the lack of complete data for certain flights may make subsequent anomaly detection, fault diagnosis, and life prediction difficult to carry out. Therefore, filling in data containing missing values ​​is considered a better way to deal with incomplete performance data of aircraft engines.

[0004] Statistical methods are the most common method for imputing missing values ​​in industry due to their interpretability and computational simplicity. The most basic statistical imputation methods use descriptive statistics to fill in missing values, such as forward imputation, backward imputation, mean imputation, and mode imputation. These imputation methods are fast and can ensure that the final data contains no missing values. However, these methods fail to consider the contextual relationships between time and feature dimensions, resulting in suboptimal imputation results in most cases.

[0005] In the prior art, methods based on recurrent neural networks (RNNs) are commonly used for imputation. This approach first uses simple statistical methods (such as mean imputation, forward imputation, and cubic spline interpolation) to preliminarily impute missing values. The imputed data is then used to train the RNN. Current methods, when applied to aircraft engines, suffer from two major issues. First, the performance of current neural networks is unstable with real-world industrial data, and sometimes imputed values ​​exceed the normal range, resulting in unreasonable imputed values. While it is possible to uniformly handle these unreasonable values ​​after imputation, this approach does not address the imputation issue; instead, it creates a new feature value to address the unreasonable value. Second, the operating time of aircraft engines varies from flight to flight, and multiple operating conditions of varying durations accumulate within each flight. For civil aircraft, a complete flight process primarily includes takeoff, climb, cruise, descent, and landing. In some special cases, it also includes go-arounds, holding maneuvers, and ground tests. In these states, various differences lead to inconsistencies in the actual running time and performance degradation of two adjacent flight data. However, the traditional filling method does not handle the same situation when processing equally spaced sampling data. Summary of the invention:

[0006] In view of the shortcomings and deficiencies in the prior art, the present invention proposes a complex equipment data processing method based on CITIN that can effectively reduce the possibility of unreasonable values.

[0007] The present invention is achieved by the following measures:

[0008] A complex equipment data processing method based on CITIN, characterized by comprising the following steps:

[0009] Step 1: Get the original data;

[0010] Step 2: Data preprocessing: Split the data set to obtain the test set and training set, splice the original data sequence according to the training set, validation set, and test set, use the training set data as the standard, standardize and normalize, weaken the data differences in the feature dimension, improve the filling effect of the neural network, and use a sliding window to cut the data of different lengths into the same length to facilitate input into the neural network;

[0011] Step 3: Establish and train a filling network model. Based on the GRU, the filling network model introduces a Transformer to process the time interval information. The results are combined with the missing label matrix to adjust the hidden state and input weights of the missing part, forming a time interval-aware temporal interpolation network. Secondly, the piecewise cubic Hermite interpolation polynomial PCHIP is introduced into the training process of the time interval-aware temporal interpolation network to constrain the range of the filling results.

[0012] During training, the incomplete training set is first input into two time interval-aware temporal interpolation networks to obtain the reconstructed data and temporary filled data output by the network. Then, the MAE of the reconstructed data and the original data is used as part of the loss function, and the MAE of the consistency of the filling results of the two networks is used as part of the loss function. The two sets of filling results are averaged and used as the temporary filling results in the training stage. Subsequently, the incomplete data is input into the PCHIP unit to form temporary filling data. The MSE of the temporary filling result and the filling result of the PCHIP unit are used as part of the loss function to jointly optimize the network parameters of the two time interval-aware temporal interpolation networks. Step 4: Application of the filling network: First, the test data is input into the two time interval-aware temporal interpolation networks that have been trained to obtain the reconstructed data and temporary filled data. Then, the temporary filling results in the two networks are averaged to serve as the final filling result of the test set.

[0013] The input of the filling network model of the present invention includes incomplete data X, missing label matrix M, time interval matrix Δ and absolute time s, multivariate time series is incomplete data, where T is the length of the data in the time dimension, D is the characteristic dimension of the data, and the missing marker matrix M is the position where the data of the original data X is missing. The shape is the same as X. For the data of the dth dimension at time t, have

[0014]

[0015] Δ is represented as the time interval matrix of X, which has the same shape as X. The time interval matrix Δ is calculated separately for each dimension, and the calculation rules for the corresponding position include three: First, the time interval δ1 of the first timestamp is set to 0; Second, for the subsequent timestamp t, if there is data in the sample sequence X of the dimension corresponding to the previous timestamp of the timestamp, the current absolute time is subtracted from the absolute time of t-1 as the value corresponding to Δ; Finally, if the data of X at the corresponding position of t-1 is empty, on the basis of subtracting the current absolute time from the absolute time of t-1, the time interval value of the previous timestamp is also added. The calculation of Δ in the dth dimension at time t is expressed as

[0016] Among them, s t represents the absolute time at the tth moment; Represents the time interval matrix of the dth dimension at time t.

[0017] The input data of the network model in the present invention is marked as The positions of these data are marked using the indicator matrix I. The shape of the indicator matrix is ​​the same as the original data. The data positions where the observed values ​​exist but are hidden in actual use for prediction are represented as 1, and the rest are represented as 0. The indicator matrix I at time t is represented by the dth dimension as follows:

[0018]

[0019] The filling network model of the present invention includes two time interval perception time series interpolation networks, one of which is used to capture regularities from the forward direction of training data, and the other to capture regularities from the reverse direction. The two networks each provide a temporary filling result, and are further provided with a PCHIP unit for calculating the filling result using a statistical method. The inputs of the two time interval perception time series interpolation networks include the original data X, the time interval matrix Δ, and the missing mark matrix M. The first time interval perception time series interpolation network INIT1 starts from the first timestamp and processes all the features of each timestamp in sequence until the last one, representing the process of forward capturing data relationships in the time dimension, and passes through the hidden layer H of the INIT structure. t Input to the next layer, and convert the output c1~c T According to the time dimension, Temporary imputed data1 is Expressed as

[0020] C=Concat(c1,c2,…,c T ),

[0021]

[0022] Where C represents the matrix after the output of each step of the GRU model is spliced; Concat(·) represents concatenating vectors into matrices, and ⊙ represents multiplication of corresponding positions of matrices of the same size. The second time interval perception temporal interpolation network INIT2 has the opposite processing process to the first one. It learns the time series from back to front in the time dimension to capture the impact of future data on the filling results of past data. The filling result of the second time interval perception temporal interpolation network INIT2 is expressed as The input of the PCHIP unit processing flow includes the original data X and the absolute time vector s corresponding to each timestamp. The PCHIP unit needs to process each dimension of X separately. When processing the data of the dth dimension, Calculate the derivative at each point, then calculate each coefficient of the cubic function based on the derivative and the slope, and finally convert the position s containing the missing data into t Substitute the Hermite polynomial composed of multiple functions and solve the corresponding position After filling all missing values, you can get the filling results

[0023] The method for processing all feature dimension data at each moment by the time interval perception temporal interpolation network in the present invention includes, in addition to the original input of each timestamp, the information transmitted by the historical timestamp, where H t-1 Represents the hidden features passed by the previous timestamp. When processing the first timestamp, the H passed by the historical timestamp t-1 is set to 0. The following is the specific process for processing each timestamp data: Step ①: The hidden variables of the previous layer are converted to x after a linear layer dimension reduction. h , x h The size of is the same as the input x;

[0024] Step 2: Use the existing value in x to replace the x after signal attenuation of the features of the hidden layer of the previous layer h The corresponding position of the reconstructed data forms the input x c Then, after a linear layer transformation, the output z of the network is obtained h , expressed as

[0025] x c =x⊙m+x h ⊙(1-m)

[0026] z h =Linear(x c )

[0027] Step ③: Generate weight γ according to the current time interval δ x , used to balance the input z h and historical information x h The important weights of γ x The splicing missing marker matrix m is transformed into the same matrix as x through a linear transformation and a Sigmoid (·) activation function. h and z h The same size, denoted as β, is used to merge the features of Step ① and Step ②. The merging formula is

[0028] c h =β·x h +(1-β)·z h ,

[0029] Step ④:Replace the c in Step ③h After observation value replacement, it is expressed as c c , c c Concatenate with the missing token m in the feature dimension as the input X of the GRU i , and the H of the previous layer t-1 Input into GRU to get H of the current timestamp t .

[0030] In Step ③ of the present invention, the overall input is δ, which is the time interval vector of multi-sensor data for each timestamp; the output is the time attenuation factor γ x , the self-attention mechanism is introduced to enhance the network input's perception of time intervals and improve the ability to weigh the hidden variables and input weights of GRU. The left side of the self-attention mechanism-enhanced network is a basic linear layer used to generate a temporary γ, which is calculated as follows:

[0031] γ=ReLU(δW d ⊙I+b d )

[0032] Where W d and b d is the weight and bias of the network used to provide the attenuation factor; I is the unit matrix, which is used to maintain the existence of time interval information; ReLU (@) is the activation function. The right side of the self-attention mechanism enhanced network is the self-attention mechanism. First, the time interval δ is input into three linear layers in sequence, and the outputs of the three linear layers are used as Query, Key and Value respectively. Then Q and K are dot-producted and divided by the scaling factor To avoid the dot product result being too large, it is normalized by Softmax(·). The result obtained is the weight parameter of the self-attention mechanism. The weight parameter is dot-producted with V and converted to the same size as the output of the basic linear layer through linear transformation. Finally, the results on the left and right sides are added together, and the distribution range of the results is adjusted by an exponential function to form the final γ x .

[0033] In Step ③ of the present invention, in GRU, the hidden state H t-1 and input X i Going a step further, GRU includes a reset gate R t and an update gate Z t , the purpose of the reset gate is to calculate the candidate hidden state The purpose of the update gate is to act as a weight to adjust the t In, H t-1 and The weights of both are two fully connected layers using the Sigmoid(·) activation function, expressed as

[0034] R t =Sigmoid(X i W xr +H t-1 W hr +b r )

[0035] Z t =Sigmoid(X i W xz +H t-1 W hz +b z )

[0036] Where W xr 、W hr and b r is the network parameter of the reset gate, W xz 、W hz and b z It is the network parameter of the update gate, which integrates the reset gate with the hidden state and calculates the candidate hidden state The formula is

[0037]

[0038] Where W xh 、W hh and b h is the network parameter for calculating the candidate hidden state; ⊙ is the element-wise product operator, and the calculation method for the GRU output hidden state is

[0039]

[0040] The PCHIP unit described in the present invention is used for smooth interpolation processing of data, specifically: constructing a cubic polynomial in the interval between every two observation time stamps, and making these polynomials have the same numerical value and derivative on the observation value, including four steps: splitting the original input to provide incomplete sequence data; calculating the derivative of each existing point in the data, ensuring that the derivatives of the data point on the left and right sides are consistent; constructing a cubic Hermite polynomial in each interval, and ensuring that the position corresponding to each numerical value has the same function value and derivative value; using the cubic polynomial to calculate the interpolation result at any position.

[0041] The present invention proposes a constrained time interval-aware temporal interpolation network for complex equipment, such as aircraft engine performance data. This network not only takes into account the impact of engine performance degradation at different stages and durations of operation on each flight, but also more accurately fills in the missing parts in the engine performance data. At the same time, in order to address the problem that neural networks sometimes produce unreasonable values ​​due to their unexplainable nature, PCHIP is used during the training phase of the interpolation network to constrain the range of interpolation results generated by the neural network, thereby reducing the possibility of ITIN generating unreasonable values. Description of the drawings:

[0042] Figure 1 It is a schematic diagram of the model in the present invention.

[0043] Figure 2 It is an internal flow chart of the Interval-aware Temporal Imputation Network in the present invention.

[0044] Figure 3 This is a schematic diagram of the self-attention mechanism enhanced network in the present invention.

[0045] Figure 4 is a flow chart of an embodiment of the present invention.

[0046] Figure 5 It is a schematic diagram of the basic structure of a turbofan engine in an embodiment of the present invention.

[0047] Figure 6 These are the results of ten experiments on a civil aviation engine dataset using different methods in an embodiment of the present invention. (a) is a schematic diagram of the MSE indicator, and (b) is a schematic diagram of the MAE indicator.

[0048] Figure 7 3 is a comparison chart of experimental results of CITIN and BRITS under different signal-to-noise ratios in an embodiment of the present invention.

[0049] Figure 8 These are the results of ten experiments on a civil aviation engine dataset using different methods in an embodiment of the present invention, where (a) is a schematic diagram of the MSE indicator and (b) is a schematic diagram of the MAE indicator.

[0050] Figure 9 1 is a histogram of MSE and MAE and their standard deviations at different signal-to-noise ratios on a real civil aviation dataset in an embodiment of the present invention. Specific implementation method:

[0051] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0052] The present invention includes a constrained time interval-aware temporal series interpolation network unit, a PCHIP (Piecewise Cubic Hermite Interpolating Polynomial) unit, and a redesigned loss function, wherein the network input mainly includes incomplete data X, a missing label matrix M, a time interval matrix Δ, and an absolute time s. Multivariate time series is incomplete data, where T is the length of the data in the time dimension and D is the feature dimension of the data. The missing marker matrix M is the position where the data of the original data X is missing, and its shape is the same as X. For the data of the dth dimension at time t, have

[0053]

[0054] Δ represents the time interval matrix of X, and its shape is the same as X. Assume that there is a sample sequence X as shown on the left side of Table 1, x1~x6 are the split vectors of matrix X in the time dimension, " / " represents missing data, otherwise it represents the existence of data corresponding to the position, the absolute time s of x1~x6 = [0,3,7,12,18,20], and the corresponding time interval Δ is shown on the right side of Table 1. The time interval matrix Δ is calculated separately for each dimension, and the calculation rules for the corresponding position include three: first, the time interval δ1 of the first timestamp is set to 0; second, for the subsequent timestamp t, if there is data in the sample sequence X of the dimension corresponding to the previous timestamp of the timestamp, the current absolute time is subtracted from the absolute time of t-1 as the value corresponding to Δ; finally, if the data of X at the corresponding position of t-1 is empty, on the basis of subtracting the current absolute time from the absolute time of t-1, the time interval value of the previous timestamp must also be added. The calculation of Δ in the dth dimension at time t can be expressed as

[0055]

[0056] Among them, s t represents the absolute time at the tth moment; Represents the time interval matrix of the dth dimension at time t.

[0057] Table 1 An example of constructing time intervals based on missing data locations and absolute times

[0058]

[0059] In addition to M and Δ derived from X, a portion of observations in the validation and test sets need to be set as missing to verify and test the ability of the neural network for filling in the missing values ​​and to evaluate the filling effects of different methods. The manually masked data are marked as The positions of these data are marked using the indicator matrix I. The shape of the indicator matrix is ​​the same as the original data. The data positions where the observed values ​​exist but are hidden in actual use for prediction are represented as 1, and the rest are represented as 0. The dth dimension of the indicator matrix I at time t is represented as

[0060] Attachment Figure 1 The main components of the proposed method model are presented, and the use of the input data introduced above is demonstrated. Figure 1 In the upper part, the middle is the input of each module, including the four types introduced above; the upper and lower parts are a time interval perception time series interpolation network, one for capturing the rules from the forward direction of the training data, and the other for capturing the rules from the reverse direction, and each network provides a temporary filling result; the right side is the PCHIP unit, which is used to calculate the filling results of the statistical method. Figure 1 The rectangular block in the middle of the upper half is the input of the two Interval-aware Temporal Imputation Networks, including the original data X, the time interval matrix Δ and the missing mark matrix M. The figure has been split into x for each timestamp in the time dimension. t ,δ t ,m t Interval-aware Temporal ImputationNetwork1 (INIT1) starts from the first timestamp and processes all the features of each timestamp in sequence until the last one, which represents the process of positively capturing data relationships in the time dimension. t Input to the next layer, and convert the output c1~c T According to the time dimension, Temporary imputed data1 (i.e. ) can be expressed as

[0061] C=Concat(c1,c2,…,c T )

[0062]

[0063] Where C represents the matrix after the output of each step of the GRU model is concatenated; Concat(·) represents concatenating vectors into matrices, and ⊙ represents multiplying corresponding positions of matrices of the same size. Interval-aware Temporal ImputationNetwork2 (INIT2) is the opposite of the processing process of the first Cell. It learns the time series from back to front in the time dimension to capture the impact of future data on the filling results of past data. The filling result of Interval-aware Temporal ImputationNetwork2 (INIT2) is expressed as

[0064] The input of the PCHIP unit processing flow in the present invention includes the original data X and the absolute time vector s corresponding to each timestamp. The PCHIP unit needs to process each dimension of X separately. When processing the data of the dth dimension, Calculate the derivative at each point, then calculate each coefficient of the cubic function based on the derivative and the slope, and finally convert the position s containing the missing data into t Substitute the Hermite polynomial composed of multiple functions and solve the corresponding position After filling all missing values, you can get the filling results

[0065] Figure 1 The lower part shows the training and validation process for the upper part. The error metrics for the two sets of temporary imputation results and reconstruction results obtained in the upper part, as well as the error metrics for the PCHIP algorithm, are combined to form a loss function to train two interval-aware temporal imputation networks. After training both networks, the imputation results are averaged as the final result, which is the imputed data marked by the dotted line in the figure.

[0066] Interval-aware Temporal Imputation Network is a network based on GRU. The typical feature of GRU is that it processes each timestamp step by step. In essence, it assumes that there is a regression relationship between the data of the current timestamp and the data of the processed timestamp. At time t, the original input contains x t , δ t , m t , representing the vector at time t in the original data, time interval, and missing marker matrix, respectively.

[0067] Figure 2It shows the processing method for all feature dimension data at each moment, including the original input of each timestamp and the information transmitted by the historical timestamp. t-1 Represents the hidden features passed by the previous timestamp. When processing the first timestamp, the H passed by the historical timestamp t-1 Set to 0. The following is the specific process for processing each timestamp data:

[0068] Step ①: The hidden variables of the previous layer are converted to x after a linear layer dimension reduction. h , x h The dimensions of are the same as the input x.

[0069] Step 2: Use the existing value in x to replace the x after signal attenuation of the features of the hidden layer of the previous layer h The corresponding position of the reconstructed data forms the input x c Then, after a linear layer transformation, the output z of the network is obtained h This step can be expressed as

[0070] x c =x⊙m+x h ⊙(1-m)

[0071] z h =Linear(x c )

[0072] Step ③: Generate weight γ according to the current time interval δ x , used to balance the input z h and historical information x h Then γ x The splicing missing marker matrix m is transformed into the same matrix as x through a linear transformation and a Sigmoid (·) activation function. h and z h The same size, denoted as β, is used to merge the features of Step ① and Step ②. The merging formula is

[0073] c h =β·x h +(1-β)·z h

[0074] Step ④:Replace the c in Step ③ h After observation value replacement, it is expressed as c c . c Concatenate with the missing token m in the feature dimension as the input X of the GRU i , and the H of the previous layer t-1 Input into GRU to get H of the current timestamp t .

[0075] Next, we will introduce the generation of time decay factor and GRU Cell in detail:

[0076] The time decay factor generation process can be used to calculate the time decay factor γ based on the time interval matrix δ x , the time decay vector is introduced into the input update process, so that the weight of features closer to the current timestamp is increased and the weight of features farther away is reduced according to the time interval δ from the previous timestamp x (Che et al, 2018). This invention uses the self-attention mechanism to enhance the network input's perception of time intervals and improve the ability to balance the hidden variables and input weights of the GRU. The basic structure is as follows Figure 3 As shown, Figure 3 The overall input is δ, which is the time interval vector of multi-sensor data for each timestamp; the output is the time decay factor γ x . Figure 3 There are two parts in it. The left side is a basic linear layer, which is used to generate a temporary γ. The calculation method is

[0077] γ=ReLU(δW d ⊙I+b d )

[0078] Where W d and b d are the weights and biases of the network used to provide the attenuation factor; I is the identity matrix used to maintain the time interval information; ReLU(·) is the activation function. Figure 3 The right side of the self-attention mechanism first inputs the time interval δ into three linear layers in sequence, and uses the outputs of the three linear layers as Query, Key, and Value respectively. Then Q and K are dot-producted and divided by the scaling factor Avoid dot product results that are too large. Then normalize it through Softmax(·), and the result obtained is the weight parameter of the self-attention mechanism. The weight parameter is dot-producted with V and converted to the same size as the output of the basic linear layer through linear transformation. Finally, the results on the left and right sides are added together, and the distribution range of the results is adjusted by an exponential function to form the final γ x .

[0079] In GRU, the hidden state H t-1 and input X i Got further processing. GRU contains a reset gate R t and an update gate Z t The purpose of the reset gate is to calculate the candidate hidden state The purpose of the update gate is to act as a weight to adjust the t In, H t-1and The structures of both are two fully connected layers using Sigmoid(·) activation function, expressed as

[0080] R t =Sigmoid(X i W xr +H t-1 W hr +b r )

[0081] Z t =Sigmoid(X i W xz +H t-1 W hz +b z )

[0082] Where W xr 、W hr and b r is the network parameter of the reset gate, W xz 、W hz and b z is the network parameter of the update gate. Then, the reset gate is integrated with the hidden state to calculate the candidate hidden state The formula is

[0083]

[0084] Where W xh 、W hh and b h is the network parameter for calculating candidate hidden states; ⊙ is the element-wise product operator. The calculation method for the GRU output hidden state is

[0085]

[0086] The PCHIP algorithm can be used for smooth interpolation of data. The implementation idea of ​​the algorithm is to construct a cubic polynomial in the interval between every two observation timestamps, and make these polynomials have the same numerical value and derivative on the observations. The algorithm includes four steps: split the original input to provide incomplete sequence data; calculate the derivative of each existing point in the data, and ensure that the derivatives of the data points on the left and right sides are consistent; construct a cubic Hermite polynomial in each interval, and ensure that the same function value and derivative value are at the position corresponding to each numerical value; use the cubic polynomial to calculate the interpolation result at any position. The PCHIP unit is further explained below: (1) Input data construction: Take the original data and real time interval data x, s provided above as input, calculate the corresponding PCHIP interpolation result for each dimension, and you can get another set of complete numerical values. For the data of the dth dimension, first give a set of incomplete data Xd , and the real time interval are sequentially spliced ​​to obtain the input of PCHIP, that is, (2) Calculating derivatives: Calculating Although the derivative of the value can be directly calculated by d = △y / △x, this calculation method cannot guarantee the smoothness and monotonicity of the result for the PCHIP interpolation method. In PCHIP, the following method is used to determine the derivative d of any k-th segment (i.e., between k and k+1): k .

[0087] First calculate the basic slope m k Let h k =s k+1 -s k , then in s k The basic slope at Then calculate d k When k≠1 and k≠T, if m k =0, m k-1 =0 or SGN(m k )≠SGN(m k-1 ) holds true, then let d k =0; otherwise according to h k and h k-1 To m k and m k-1 Perform weighted harmonic mean calculation d k , the calculation method is

[0088] w1=2h k +h k-1

[0089] w2=h k +2h k-1

[0090] 1 / d k =1 / (w1+w2)*(w1 / m k +w2 / m k-1 )

[0091] Where SGN(·) represents the sign of the value, which can be -1, 0, or 1.

[0092] When k = 1 or k = T, it is necessary to use the one-sided three-point derivative estimate method to determine the boundary point derivative value. When k = 1, use h k , h k+1 , m k , m k+1 Calculate d kFirst get d k A preliminary estimate of

[0093] d k =G(h k , h k+1 , m k , m k+1 )=((2×h k +h k+1 )×m k ) / (h k +h k+1 )

[0094] In order to ensure that the derivative does not destroy the shape and monotonicity of the interpolation function, it is necessary to adjust the derivative value under some conditions. First, determine SGN(d k )=SGN(m k ), if not satisfied, directly set d k =0; otherwise, judge SGN(m k )≠SGN(m k+1 ) and |d k |>3×|m k |Is it true? If so, then d k =3×m k When k=T, d k =G(h k-1 , h k-2 , m k-1 , m k-2 ).

[0095] In this way, the derivative corresponding to each point can be obtained, and a sequence is formed according to the timestamps of different observations, which is represented by d.

[0096] (3) Constructing cubic Hermite polynomial

[0097] First, the interpolation parameters are calculated, taking into account the relationship between the derivative and the slope.

[0098] t k =(d k +d k+1 -2×m k ) / h k

[0099] The Hermite interpolation method also uses a cubic polynomial for interpolation. For any [x i , x i+1 ], the polynomial is in the form of

[0100] F k (x) = pk (xx i ) 3 +q k (xx i ) 2 +r k (xx i )+s k

[0101] The calculation method is p k =t k / h k ,q k =(m k -d k ) / h k -t,r k =d k , s k =y k .

[0102] 1) Calculate interpolation results

[0103] Finally, according to the provided position in a corresponding interval such as x p ∈[x k , x k+1 ], select F k (x) can be based on the provided x p Find the corresponding interpolation result.

[0104] In the traditional RNN-based infilling method training process, the error measurement between the neural network's reconstruction results and the observed values ​​is usually used as the loss function. The loss function of the model training process of the method proposed in the present invention includes four parts: the reconstruction error of each of the two Interval-aware Temporal Imputation Networks, the error of the infilling results between the two Interval-aware Temporal Imputation Networks, and the error between the average of the infilling results of the two Interval-aware Temporal Imputation Networks and the infilling results of the PCHIP unit. After the parameter training of the two models is completed, the test data is input into the model to obtain two sets of forward and reverse reconstruction results, and the reconstruction results are added and averaged to obtain the final infilling result.

[0105] First, the reconstruction error of the two Interval-aware Temporal Imputation Networks. For each model, the reconstruction error consists of three parts: x h,z h ,c h Compared with the MAE of x, the loss needs to be calculated at each time stamp and accumulated after all time stamps are calculated. The calculation at each time stamp can be expressed as

[0106]

[0107] Where SUM(·) represents the sum of all numbers in the matrix. The reconstruction loss of each timestamp is added as the loss function of the final unidirectional RNN network, which is expressed as The forward filling network can be expressed as The backfill network can be expressed as The second is the filling result error between the two Interval-aware Temporal Imputation Networks, that is, the error in the consistency of the filling results. This error needs to be scaled to 10% of the original value due to its large value. The calculation process can be expressed as Finally, the error between the filling result and the PCHIP filling result can be expressed as The final loss function of the network training can be expressed as

[0108] Example:

[0109] Taking aircraft engines as an example, a time interval-aware network with piecewise cubic Hermite interpolation constraints is proposed for interpolating degradation data of aviation equipment lifecycle:

[0110] The first step is to obtain the original data. This step is used to extract part of the data from the engine's operating data to form a data set. This example uses two data sets, namely C-MAPSS data and real civil aviation engine data. The data format of the two data sets is the same, both are multiple sets of multivariate time series data of unequal length and containing missing values. The data is first collected by the sensors installed on the aircraft engine, and then the aircraft's Quick Access Recorder (QAR) records the complete data of all stages, and uses ACARS as a channel to send the key data of the key stages to the ground in real time. Since the data of each flight will form a long data containing various stages, first select a key state that can reflect the engine performance, and then extract the characteristic values ​​under this state to form a timestamp data, and then splice all flights of a single engine in chronological order to obtain the operating status data of each engine;

[0111] The next step is the data preprocessing phase: The first step is to split the dataset, dividing multiple engines into different datasets according to engine number. 20% of the entire dataset is used as the test set, and 20% of the training set is used as the validation set. The second step is to manually mask 10% of the data to verify the model effect and test the filling effect of different methods. The third step is to splice the original data sequence according to the training set, validation set, and test set. Using the training set data as the standard, standardization and normalization are performed to weaken the data differences in the feature dimension and improve the filling effect of the neural network. The fourth step is to use a sliding window to cut the data of different lengths into the same length to facilitate input into the neural network.

[0112] The third step is to train the imputation network. During the training process, the incomplete training set is first input into two Interval-aware Temporal Imputation Networks to obtain the reconstructed data and temporary imputed data output by the network. The MAE of the reconstructed data and the original data is then used as part of the loss function. The MAE of the consistency of the imputation results of the two networks is also used as part of the loss function. The two sets of imputation results are averaged and used as the temporary imputed data in the training phase. Subsequently, the incomplete data is input into the PCHIP unit to form temporary imputed data. The MSE of the temporary imputed data and the imputation results of the PCHIP unit are used as part of the loss function to jointly optimize the network parameters of the two Interval-aware Temporal Imputation Networks.

[0113] Finally, the application of the filling network: first, the test data is input into two trained interval-aware temporal imputation networks to obtain reconstructed data and temporary filling data, and then the temporary filling results of the two networks are averaged to be the final filling result of the test set. Finally, the difference between the filling result of the test set and the original data can be measured. The measurement indicators are MSE and MAE, and the calculation methods are

[0114]

[0115] Next is the conduction of the experiment and analysis of the experimental results. It mainly consists of three parts. The first part introduces the dataset and experimental platform; the second and third parts respectively present the filling results and analysis of the C-MAPSS dataset and the real aircraft engine dataset.

[0116] The C-MAPSS dataset is a publicly available dataset related to engine health, generated by the C-MAPSS simulator and used to simulate the degradation trends of turbofan engines (Saxena et al., 2008). The dataset consists of four subsets, FD001 to FD004, with details shown in Table 2. Each dataset contains the engine number, cycle number, three operating states, and 21 sensor data points.

[0117] Table 2 Details of the C-MAPSS dataset

[0118]

[0119] To mitigate the impact of varying operating conditions, this experiment selected FD001 and FD003. Furthermore, to mitigate the impact of data with low correlation to engine state changes on the experimental results, the Pearson correlation coefficient method was used to remove sensor data with low correlation to the experimental results. The remaining data is shown in Table 3. The original missing data rate was set at 10%. Based on this, a 10% infill was set for network training and validation.

[0120] Table 3 Details of sensor data used in the experiment

[0121]

[0122] The real-world aircraft engine dataset is derived from turbofan engine operating data from an Asian airline. The basic structure of a turbofan engine is shown in Figure 5, primarily comprising high- and low-pressure turbines (HPT and LPT), a combustion chamber (CC), high- and low-pressure compressors (HPC and LPC), and a fan. The data used in this paper contains four dimensions: N1 represents the low-pressure rotor speed, N2 represents the high-pressure rotor speed, and FF represents fuel consumption. Thermocouples on the T49.5 wiring harness assembly are used to collect engine exhaust gas temperature (EGT).

[0123] The actual data contains faulty data that may affect the data patterns. Since this paper does not involve fault diagnosis, to avoid the influence of faulty data, only the 10 or more samples before the fault occurred were collected as the dataset. The preset missing rate and manually masked data percentage of this dataset are consistent with those of the C-MAPSS dataset, both at 10%.

[0124] The experiments used Python version 3.10 and PyTorch version 2.1.1, an Intel Core i7-9700K CPU, and 16GB of DDR4 memory. In this experiment, the stride for both datasets was 48, the number of features for the C-MAPSS dataset was 17, and the number of features for the real civil aircraft engine dataset was 4. The GRU hidden layer size was 128, and the mini-batch size was 32. Training lasted for a maximum of 30 epochs, with training terminated early if the loss did not decrease within three consecutive epochs. Adam was used as the network optimizer, with an initial learning rate of 0.001. The experimental results on the real aircraft engine dataset are analyzed as follows:

[0125] Considering that different datasets require different hyperparameters, experiments based on real aircraft engine data were conducted with a step size of 0.6, with α ∈ [0, 6]. Ten experiments were conducted for each α value, and the mean and median values ​​were calculated. The MSE and MAE of the experimental results are shown in Tables 4 and 5. Table 4 shows that the MSE achieved a minimum mean value of 0.2590 at α = 0.6 and a minimum median value of 0.2456 at α = 1.2. Table 5 shows that the MAE achieved a minimum mean value of 0.3951 at α = 0.6 and a minimum median value of 0.3857 at α = 1.2. Since this example prioritizes the rationality of the infill results and prefers results with a lower overall loss, α = 0.6 was selected as the hyperparameter for the real aircraft engine dataset.

[0126] Table 4 Relationship between α and MSE on the civil aviation engine dataset

[0127]

[0128] Table 5 Relationship between α and MAE on the civil aviation engine dataset

[0129]

[0130] In order to evaluate the effectiveness of the proposed method, CITIN was compared with different methods proposed in recent years, such as CSDI, GP-VAE, and BRITS. The error of ten repeated experiments is as follows: Figure 6 As shown. Figure 6 As can be seen in (a), the overall MSE value of the CSDI result is relatively large. Figure 6 As can be seen in (b), GP-VAE has a larger value under the MAE evaluation indicator. In the 10 experimental results, CITIN outperforms the other three methods in terms of both MSE and MAE error evaluation.

[0131] Tables 6 and 7 show the mean, standard deviation, and median of the MSE and MAE for the ten experimental results. Table 6 shows that the proposed method has the smallest mean MSE and MAE, at 0.0717 and 0.1949, respectively. The standard deviations of the MSE and MAE for the ten repeated experiments are also small, at 0.0022 and 0.0029, respectively. This indicates that the proposed method has better overall filling performance and more stable filling results across multiple experiments. Table 7 shows that the proposed method has the best MSE and MAE, at 0.0722 and 0.1951, respectively, which is the best among the methods of this invention. This indicates that the proposed method performs more consistently in most cases.

[0132] Comparative experiments of different methods:

[0133] Table 6 Average results on the C-MAPSS dataset

[0134]

[0135] Table 7 Median results on the C-MAPSS dataset

[0136]

[0137] Hypothesis Validity Experiment:

[0138] As mentioned earlier, the basic principle of the proposed method is to reduce the likelihood of generating unreasonable values ​​by compressing the deviation between the imputed values ​​and the original data distribution. Therefore, it is necessary to verify the deviation between the imputed values ​​and the actual values ​​compared to the method without using statistical constraints. For ease of comparison, BRITS, which performed well among the comparison methods, was used as a control for the proposed method. Table 8 shows the results of ten experiments, including the range of differences and deviations between the imputed values ​​and the actual values ​​for BRITS and the proposed method, as well as the average values ​​of these results. The left side of the table shows the left and right bounds for BRITS, while the right side shows the left and right bounds for the proposed method. As shown in Table 8, the average left bound for the difference between the imputed values ​​and the actual values ​​for BRITS in the ten experiments is -1.4613, and the right bound is 1.9997, with an average interval of 3.4610. In comparison, the left and right bounds for the proposed method are -1.2999 and 1.8036, respectively, with an interval length of 3.1035. Therefore, the absolute values ​​of both the left and right bounds of the proposed method are smaller than those of BRITS. Experimental results show that the error range between the filled value and the actual value of the proposed method is smaller.

[0139] Table 8. Deviation left and right boundaries of BRITS and the proposed method on the C-MAPSS dataset

[0140]

[0141] Experimental results under different signal-to-noise ratios

[0142] Figure 7 The comparison of experimental results of CITIN and BRITS under different signal-to-noise ratios is shown. Figure 7 (a) shows a comparison of MSEs at different SNRs. As can be seen from the figure, when the SNR is 10dB, the MSEs of both methods exceed 0.1, and remain below 0.1 at all other SNRs. However, CITIN's MSE is consistently lower than BRITS's across various SNRs, and the standard deviation across the ten experiments is also smaller. Figure 7 The MAE comparison experiment in (b) shows that CITIN's error is smaller than BRITS under various signal-to-noise ratio conditions, and CITIN's standard deviation is significantly smaller than BRITS under various noise conditions in the experiment. These results show that CITIN performs better and is more stable under different noise levels.

[0143] Analysis of experimental results on real aircraft engine dataset

[0144] Selecting Effective Weight Coefficients: Considering that different datasets require different hyperparameters, experiments based on real aircraft engine data were conducted with a step size of 0.6, with α ∈ [0, 6]. Ten experiments were conducted for each α, and the mean and median values ​​were calculated. The MSE and MAE of the experimental results are shown in Tables 9 and 10. Table 9 shows that the MSE achieved a minimum mean value of 0.2590 at α = 0.6 and a minimum median value of 0.2456 at α = 1.2. Table 10 shows that the MAE achieved a minimum mean value of 0.3951 at α = 0.6 and a minimum median value of 0.3857 at α = 1.2. Since this example prioritizes the rationality of the infill results and therefore prefers results with a lower overall loss, α = 0.6 was selected as the hyperparameter for the real aircraft engine dataset.

[0145] Table 9 Relationship between α and MSE on the civil aviation engine dataset

[0146]

[0147] Table 10 Relationship between α and MAE on the civil aviation engine dataset

[0148]

[0149] Comparative experiments of different methods

[0150] Figure 8The experimental results of the four methods on a real civil aircraft engine dataset are compared. The figure shows that CSDI has the largest error overall among the four methods, both in terms of MSE and MAE. This may be due to the small amount of real aircraft engine data and the fact that CSDI failed to fully converge due to inherent issues with the method. In most experimental results, CITIN's MSE and MAE were smaller than those of GP-VAE and BRITS. In the third experiment, CITIN's MAE was larger than CSDI's; in the sixth experiment, CITIN's MSE and MAE were both larger than BRITS's. One possible reason for this is that the training samples used in these experiments were not representative of the general engine degradation process.

[0151] Tables 11 and 12 show the mean and standard deviation of the MSE and MAE errors, as well as the corresponding medians, for the four methods across ten repeated experiments. Table 11 shows that the mean MSE and MAE values ​​for CITIN over ten experiments are 0.2590 and 0.3951, while the standard deviations are 0.0612 and 0.0515, respectively, which are slightly larger than the BRITS values ​​of 0.0631 and 0.0455. This is likely due to the greater noise in the C-MAPSS dataset, which results in greater variability in CITIN results across multiple experiments. Table 12 shows that the median of the proposed method is also the smallest among all compared methods, with an MSE and MAE of 0.2615 and 0.4101, respectively.

[0152] Table 11 Average results on the civil aviation engine dataset

[0153]

[0154] Table 12 Median results on the civil aviation engine dataset

[0155]

[0156] Hypothesis Validation Experiment: To verify CITIN's ability to reduce the likelihood of generating unreasonable values, it is necessary to compare its effectiveness with other advanced methods. For convenience, the experiment directly uses BRITS as a control. The deviations between the imputed values ​​and the actual values ​​are shown in Table 13. Similar to the experiments with the CMAPSS dataset, the imputed values ​​obtained using different methods are subtracted from the actual values, and the table records the left and right bounds, as well as the deviation range, for multiple experiments. As shown in Table 13, the average absolute values ​​of the left and right bounds for CITIN's ten experiments are 1.6312 and 2.0151, respectively, which are smaller than the absolute values ​​of the left and right bounds of BRITS (1.6418 and 2.1647). The average overall deviation range is also smaller.

[0157] Table 13. Left and right margins of deviations of BRITS and the proposed method on the real civil aviation dataset

[0158]

[0159] Experimental results under different signal-to-noise ratios:

[0160] Figure 9 The MSE and MAE data of the infill values ​​and actual values ​​of real civil aviation data at different signal-to-noise ratios are shown. It can be seen from the figure that although the standard deviations of the two methods at different signal-to-noise ratios are close, CITIN has a smaller error than BRITS at the same signal-to-noise ratio.

[0161] This paper proposes a constrained, time-interval-aware temporal interpolation network for complex equipment, such as aircraft engine performance data. This network not only considers the impact of engine performance degradation during different flight phases and durations, but also more accurately fills in missing data. Furthermore, to address the issue of irrational values ​​sometimes occurring in neural networks due to their uninterpretability, the network uses PCHIP (Physical Handling Interpolation) to constrain the range of interpolation results generated by the neural network during training, reducing the likelihood of irrational ITIN (Information Processing) (ITIN) values. Experiments using the C-MAPSS dataset and a real-world civil aviation engine dataset demonstrate that the proposed method achieves superior results compared to currently state-of-the-art methods. Furthermore, the proposed method discusses hyperparameter selection and conducts stability experiments on the proposed method under varying noise conditions. The results demonstrate that the proposed method exhibits superior stability compared to state-of-the-art RNN-based methods. The proposed method can be applied not only to more accurately model the missing data interpolation process for aircraft engines, but is also applicable to interpolating data from other types of complex equipment degradation. Furthermore, the proposed paradigm of using statistical methods to constrain neural networks can be applied to combinations of other neural networks and statistical methods.

Claims

1. A complex equipment data processing method based on CITIN, characterized in that: The following steps are involved: Step 1: Get the original data; Step 2: Data preprocessing: Split the data set to obtain the test set and training set, splice the original data sequence according to the training set, validation set, and test set, use the training set data as the standard, standardize and normalize, weaken the data differences in the feature dimension, improve the filling effect of the neural network, and use a sliding window to cut the data of different lengths into the same length to facilitate input into the neural network; Step 3: Establish and train a filling network model. Based on the GRU, the filling network model introduces a Transformer to process the time interval information. The results are combined with the missing label matrix to adjust the hidden state and input weights of the missing part, forming a time interval-aware temporal interpolation network. Secondly, the piecewise cubic Hermite interpolation polynomial PCHIP is introduced into the training process of the time interval-aware temporal interpolation network to constrain the range of the filling results. During training, the incomplete training set is first input into two time interval-aware temporal interpolation networks to obtain reconstructed data and temporary filled data output by the networks. The MAE of the reconstructed data and the original data is then used as part of the loss function. The MAE of the consistency of the two network filling results is also used as part of the loss function. The two sets of filling results are averaged and used as the temporary filling result in the training phase. Subsequently, the incomplete data is input into the PCHIP unit to form temporary filling data. The MSE of the temporary filling result and the filling result of the PCHIP unit are used as part of the loss function to jointly optimize the network parameters of the two time interval-aware temporal interpolation networks. Step 4: Application of the filling network: First, input the test data into the two trained time interval-aware time series interpolation networks to obtain reconstructed data and temporary filling data. Then, the temporary filling results of the two networks are averaged to obtain the final filling result of the test set.

2. The complex equipment data processing method based on CITIN according to claim 1 is characterized in that: The input of the filling network model includes incomplete data X, missing label matrix M, time interval matrix Δ and absolute time s, multivariate time series is incomplete data, where T is the length of the data in the time dimension, D is the characteristic dimension of the data, and the missing marker matrix M is the position where the data of the original data X is missing. The shape is the same as X. For the data of the dth dimension at time t, have θ is represented as the time interval matrix of X, with the same shape as X. The time interval matrix θ is calculated separately for each dimension, and the calculation rules for the corresponding position include three: first, the time interval δ1 of the first timestamp is set to 0; Secondly, for the subsequent timestamp t, if there is data in the sample sequence X of the dimension corresponding to the previous timestamp of the timestamp, the current absolute time is subtracted from the absolute time of t-1 as the value corresponding to Δ; finally, if the data of x at the corresponding position of t-1 is empty, on the basis of subtracting the current absolute time from the absolute time of t-1, the time interval value of the previous timestamp is also added. The calculation of Δ in the dth dimension at time t is expressed as Among them, s t represents the absolute time at the tth moment; Represents the time interval matrix of the dth dimension at time t.

3. The complex equipment data processing method based on CITIN according to claim 1 is characterized in that: The input data of the padded network model is labeled after the manual masking. The positions of these data are marked using the indicator matrix I. The shape of the indicator matrix is ​​the same as the original data. The data positions where the observed values ​​exist but are hidden in actual use for prediction are represented as 1, and the rest are represented as 0. The indicator matrix I at time t is represented by the dth dimension as follows:

4. The complex equipment data processing method based on CITIN according to claim 3 is characterized in that: The filling network model includes two time interval-aware temporal interpolation networks, one of which is used to capture regularities from the forward direction of the training data, and the other to capture regularities from the reverse direction. The two networks each provide a temporary filling result, and are further provided with a PCHIP unit for calculating the filling result using a statistical method. The inputs of the two time interval-aware temporal interpolation networks include the original data X, the time interval matrix Δ, and the missing marker matrix M. The first time interval-aware temporal interpolation network INIT1 starts from the first timestamp and processes all the features of each timestamp in sequence until the last one, representing the process of forward capturing the data relationship in the time dimension, and passes through the hidden layer H of the INIT structure. t Input to the next layer, and convert the output c1~c T According to the time dimension, Temporary imputed data1 is Expressed as C=Concat(c1,c2,…,c T ), Where C represents the matrix after the output of each step of the GRU model is spliced; Concat(·) represents concatenating vectors into matrices, and ⊙ represents multiplication of corresponding positions of matrices of the same size. The second time interval perception temporal interpolation network INIT2 has the opposite processing process to the first one. It learns the time series from back to front in the time dimension to capture the impact of future data on the filling results of past data. The filling result of the second time interval perception temporal interpolation network INIT2 is expressed as The input of the PCHIP unit processing flow includes the original data X and the absolute time vector s corresponding to each timestamp. The PCHIP unit needs to process each dimension of X separately. When processing the data of the dth dimension, Calculate the derivative at each point, then calculate each coefficient of the cubic function based on the derivative and the slope, and finally convert the position s containing the missing data into t Substitute the Hermite polynomial composed of multiple functions and solve the corresponding position After filling all missing values, you can get the filling results 5. The complex equipment data processing method based on CITIN according to claim 4 is characterized in that: The time interval-aware temporal interpolation network processes all feature dimension data at each moment. In addition to the original input of each timestamp, it also includes the information transmitted by the historical timestamps, where H t-1 Represents the hidden features passed by the previous timestamp. When processing the first timestamp, the H passed by the historical timestamp t-1 is set to 0. The following is the specific process for processing each timestamp data: Step ①: The hidden variables of the previous layer are converted to x after a linear layer dimension reduction. h , x h The size of is the same as the input x; Step 2: Use the existing value in x to replace the x after signal attenuation of the features of the hidden layer of the previous layer h The corresponding position of the reconstructed data forms the input x c Then, after a linear layer transformation, the output z of the network is obtained h , expressed as x c =x☉m+x h ☉(1-m) z h =Linear(x c ) Step ③: Generate weight γ according to the current time interval δ x , used to balance the input z h and historical information x h The important weights of γ x The splicing missing marker matrix m is transformed into the same matrix as x through a linear transformation and a Sigmoid (·) activation function. h and z h The same size, denoted as β, is used to merge the features of Step ① and Step ②. The merging formula is c h =β·x h +(1-β)·z h , Step ④:Replace the c in Step ③ h After observation value replacement, it is expressed as c c , c c Concatenate with the missing token m in the feature dimension as the input X of the GRU i , and the H of the previous layer t-1 Input into GRU to get H of the current timestamp t .

6. The complex equipment data processing method based on CITIN according to claim 5 is characterized in that: In Step ③, the overall input is δ, which is the time interval vector of multi-sensor data for each timestamp; the output is the time decay factor γ x , the self-attention mechanism is introduced to enhance the network input's perception of time intervals and improve the ability to weigh the GRU's hidden variables and input weights. The left side of the self-attention mechanism-enhanced network is a basic linear layer used to generate a temporary γ, which is calculated as follows: γ=ReLU(δW d ☉Ⅰ+b d ) Where W d and b d are the weights and biases of the network used to provide the decay factor; I is the unit matrix, which is used to maintain the existence of time interval information; ReLU(·) is the activation function. The right side of the self-attention mechanism enhanced network is the self-attention mechanism. First, the time interval δ is input into three linear layers in sequence. The outputs of the three linear layers are used as Query, Key and Value respectively. Then Q and K are dot-producted and divided by the scaling factor To avoid the dot product result being too large, it is normalized by Softmax(·). The result obtained is the weight parameter of the self-attention mechanism. The weight parameter is dot-producted with V and converted to the same size as the output of the basic linear layer through linear transformation. Finally, the results on the left and right sides are added together, and the distribution range of the results is adjusted by an exponential function to form the final γ x .

7. The complex equipment data processing method based on CITIN according to claim 6 is characterized in that: In Step ③, in GRU, the hidden state H t-1 and input X i Going a step further, GRU includes a reset gate R t and an update gate Z t , the purpose of the reset gate is to calculate the candidate hidden state The purpose of the update gate is to act as a weight to adjust the t In, H t-1 and The weights of both are two fully connected layers using the Sigmoid(·) activation function, expressed as R t =Sigmoid(X i W xr +H t-1 W hr +b r ) Z t =Sigmoid(X i W xz +H t-1 W hz +b z ) Where W xr 、W hr and b r is the network parameter of the reset gate, W xz 、W hz and b z It is the network parameter of the update gate, which integrates the reset gate with the hidden state and calculates the candidate hidden state The formula is Where W xh 、W hh and b h is the network parameter for calculating the candidate hidden state; ⊙ is the element-wise product operator, and the calculation method for the GRU output hidden state is 8. The complex equipment data processing method based on CITIN according to claim 7 is characterized in that: The PCHIP unit is used for smooth interpolation processing of data, specifically: constructing a cubic polynomial in the interval between every two observation time stamps, and making these polynomials have the same numerical value and derivative on the observation value, including four steps: splitting the original input to provide incomplete sequence data; calculating the derivative of each existing point in the data, ensuring that the derivatives of the data point on the left and right sides are consistent; constructing a cubic Hermite polynomial in each interval, and ensuring that the position corresponding to each numerical value has the same function value and derivative value; using the cubic polynomial to calculate the interpolation result at any position.

Images